<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD Journal Publishing with OASIS Tables v3.0 20080202//EN" "journalpub-oasis3.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:oasis="http://docs.oasis-open.org/ns/oasis-exchange/table" dtd-version="3.0"><?xmltex \makeatother\@nolinetrue\makeatletter?><?xmltex \hack{\allowdisplaybreaks}?>
  <front>
    <journal-meta>
<journal-id journal-id-type="publisher">OS</journal-id>
<journal-title-group>
<journal-title>Ocean Science</journal-title>
<abbrev-journal-title abbrev-type="publisher">OS</abbrev-journal-title>
<abbrev-journal-title abbrev-type="nlm-ta">Ocean Sci.</abbrev-journal-title>
</journal-title-group>
<issn pub-type="epub">1812-0792</issn>
<publisher><publisher-name>Copernicus Publications</publisher-name>
<publisher-loc>Göttingen, Germany</publisher-loc>
</publisher>
</journal-meta>

    <article-meta>
      <article-id pub-id-type="doi">10.5194/os-12-1137-2016</article-id><title-group><article-title>A hybrid variational-ensemble data assimilation scheme with systematic error
correction for limited-area ocean models</article-title>
      </title-group><?xmltex \runningtitle{Hybrid variational-ensemble bias-aware data assimilation}?><?xmltex \runningauthor{P.~Oddo et al.}?>
      <contrib-group>
        <contrib contrib-type="author" corresp="yes" rid="aff1">
          <name><surname>Oddo</surname><given-names>Paolo</given-names></name>
          <email>paolo.oddo@cmre.nato.int</email>
        </contrib>
        <contrib contrib-type="author" corresp="no" rid="aff2">
          <name><surname>Storto</surname><given-names>Andrea</given-names></name>
          
        <ext-link>https://orcid.org/0000-0003-3856-8905</ext-link></contrib>
        <contrib contrib-type="author" corresp="no" rid="aff3">
          <name><surname>Dobricic</surname><given-names>Srdjan</given-names></name>
          
        </contrib>
        <contrib contrib-type="author" corresp="no" rid="aff1">
          <name><surname>Russo</surname><given-names>Aniello</given-names></name>
          
        <ext-link>https://orcid.org/0000-0003-3651-8146</ext-link></contrib>
        <contrib contrib-type="author" corresp="no" rid="aff1">
          <name><surname>Lewis</surname><given-names>Craig</given-names></name>
          
        </contrib>
        <contrib contrib-type="author" corresp="no" rid="aff1">
          <name><surname>Onken</surname><given-names>Reiner</given-names></name>
          
        </contrib>
        <contrib contrib-type="author" corresp="no" rid="aff1">
          <name><surname>Coelho</surname><given-names>Emanuel</given-names></name>
          
        <ext-link>https://orcid.org/0000-0002-5368-0375</ext-link></contrib>
        <aff id="aff1"><label>1</label><institution>NATO Science and Technology Organization, Centre for Maritime Research and Experimentation,
Viale San Bartolomeo 400, 19126 La Spezia, Italy</institution>
        </aff>
        <aff id="aff2"><label>2</label><institution>Euro-Mediterranean Centre for Climate Change (CMCC), via Franceschini 31, 40128 Bologna, Italy</institution>
        </aff>
        <aff id="aff3"><label>3</label><institution>European Commission, Joint Research Centre, Via Enrico Fermi 2749, 21027 Ispra (VA), Italy</institution>
        </aff>
      </contrib-group>
      <author-notes><corresp id="corr1">Paolo Oddo (paolo.oddo@cmre.nato.int)</corresp></author-notes><pub-date><day>14</day><month>October</month><year>2016</year></pub-date>
      
      <volume>12</volume>
      <issue>5</issue>
      <fpage>1137</fpage><lpage>1153</lpage>
      <history>
        <date date-type="received"><day>16</day><month>May</month><year>2016</year></date>
           <date date-type="rev-request"><day>23</day><month>May</month><year>2016</year></date>
           <date date-type="rev-recd"><day>15</day><month>September</month><year>2016</year></date>
           <date date-type="accepted"><day>15</day><month>September</month><year>2016</year></date>
      </history>
      <permissions>
<license license-type="open-access">
<license-p>This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit <ext-link ext-link-type="uri" xlink:href="http://creativecommons.org/licenses/by/3.0/">http://creativecommons.org/licenses/by/3.0/</ext-link></license-p>
</license>
</permissions><self-uri xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016.html">This article is available from https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016.html</self-uri>
<self-uri xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016.pdf">The full text article is available as a PDF file from https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016.pdf</self-uri>


      <abstract>
    <p>A hybrid variational-ensemble data assimilation scheme to estimate the
vertical and horizontal parts of the background error covariance matrix for
an ocean variational data assimilation system is presented and tested in a
limited-area ocean model implemented in the western Mediterranean Sea. An
extensive data set collected during the Recognized Environmental Picture
Experiments conducted in June 2014 by the Centre for Maritime Research and
Experimentation has been used for assimilation and validation. The hybrid
scheme is used to both correct the systematic error introduced in the system
from the external forcing (initialisation, lateral and surface open boundary
conditions) and model parameterisation, and improve the representation of
small-scale errors in the background error covariance matrix. An ensemble
system is run offline for further use in the hybrid scheme, generated through
perturbation of assimilated observations. Results of four different
experiments have been compared. The reference experiment uses the classical
stationary formulation of the background error covariance matrix and has no
systematic error correction. The other three experiments account for, or not,
systematic error correction and hybrid background error covariance matrix
combining the static and the ensemble-derived errors of the day. Results show
that the hybrid scheme when used in conjunction with the systematic error
correction reduces the mean absolute error of temperature and salinity misfit
by 55 and 42 % respectively, versus statistics arising from standard
climatological covariances without systematic error correction.</p>
  </abstract>
    </article-meta>
  </front>
<body>
      

      <?xmltex \hack{\newpage}?>
<sec id="Ch1.S1" sec-type="intro">
  <title>Introduction</title>
      <p>The study and the characterisation of the ocean is a complex
discipline involving different aspects of modern science. In order to obtain
a coherent and time-evolving three-dimensional (3-D) picture of the ocean
from historical and present-day observations, as well as be able to predict
the future evolution of the environment, we need to solve theoretical and
technical issues.</p>
      <p>It is not feasible to observe all variables of interest with adequate spatial
and temporal scales. Modern technologies, like satellite remote sensing and
autonomous vehicles, have significantly increased our capability to observe
the environment in general and the ocean in particular. However, the huge
number of degrees of freedom characterising the ocean state still prevents
sampling at the desired resolution. In order to fill the observational gaps
and expand the temporal horizon covered by the observations (both in the past
and in the future), oceanographers combine direct observations with
theoretical studies by means of models and data assimilation.</p>
      <p>A numerical hydrodynamic model is basically the discretised version of the
primitive equations, it is an approximation of nature. Moving from the
continuous to the discrete space, additional approximations are introduced
and should be accounted for when analysing model results. These
approximations affect the model solutions in terms of quality and accuracy
and, more importantly, differences between the numerical solution and the
true state amplify along time, due to the chaotic component of the ocean
dynamic.</p>
      <p><?xmltex \hack{\newpage}?>In order to minimise these differences and improve the quality and accuracy
of model results, data assimilation techniques have been developed during the
past decades. Data assimilation is a technique to correct the model
solution-based on statistical and physical constraints derived from
observations and model simulations.</p>
      <p>Even if different kinds of data assimilation techniques exist, most of them
rely on the same basic principle, the combination of physically based and
statistical approaches to maximise the conditional probability of the model
state given the observation.</p>
      <p>Data assimilation schemes developed for oceanographic studies can be
classified in two categories. The first one is the Kalman filter (KF) type
algorithm with background error covariances
(BECs) matrices usually derived from ensemble statistics (Evensen, 2003). The
second type of assimilation algorithms employ stationary BECs derived from
long-term model integrations (Yin et al., 2011; Weaver and Courtier, 2001;
Pannekoucke and Massart, 2008). A key avenue to improving data assimilation
is accurate specification of the error statistics for the background
forecast, also known as the prior or first guess (Schlatter et al., 1999).</p>
      <p>The ensemble Kalman filter (EnKF)
(Evensen, 1994) consists of a set of short-term forecasts and data
assimilation cycles. In the EnKF, the BECs are estimated from an ensemble of
model simulations. The presumed benefit of utilising these ensemble-based
techniques is their ability to provide a flow-dependent estimate of the BECs.
The traditional EnKF incorporates probabilistic information on analysis
errors in the generation of the ensemble by imposing a set of perturbations
for each ensemble member, generating the individual numerical forecasts from
different sets of initial conditions implied by the different sets of
observations and/or different numerical model configurations. The EnKF is
related to the classic Kalman filter, which
provides the optimal analysis in the case that the forecast dynamics are
linear and both background and observation errors have normal distributions.
The main difference is that the KF explicitly forecasts the evolution of the
complete forecast error covariance matrix using linear dynamics, while the
EnKF estimates this matrix from a sample ensemble of fully non-linear
forecasts. The EnKF also addresses the computational difficulty of
propagating or even storing the forecast error covariance matrix. Using
ensemble simulations also implies that the EnKF does not assume the
covariances to propagate linearly.</p>
      <p>On the other hand, many current and past operational data assimilation
methods use long time series of previous forecasts to develop stationary and
often also spatially homogeneous approximations to BECs. Schemes that use
such statistics include optimum interpolation and three-dimensional
variational data assimilation (3DVAR), and have the advantage of being less
computationally demanding, namely allowing for higher resolution. In reality,
BECs may vary substantially depending on the flow and error of the day. A
four-dimensional variational data assimilation (4DVAR) system implicitly
includes a time-evolving covariance model through the evolution of initial
errors under tangent-linear dynamics (Lorenc, 2003) within the assimilation
time window. However, the time-evolving covariance model may still be limited
by usage of a stationary covariance model at the beginning of each 4DVAR
cycle. Furthermore, like the EnKF, 4DVAR is computationally intensive,
requiring multiple integrations of tangent-linear and adjoint versions of the
forecast model. The specification of flow-dependent statistics is per se a
demanding task, due to the difficulty of retrieving information on errors in
model space.</p>
      <p>The ensemble EnKF provides an alternative to variational data assimilation
systems. Under assumptions of linearity of error growth and normality of
observation and forecast errors, it has been proved that the EnKF scheme
produces the correct BECs as the ensemble size increases (Burgers et
al., 1998). However, for smaller ensembles, the EnKF is rank deficient and
its BEC estimates suffer from a variety of sampling errors, including
spurious correlations between widely separated locations that need to be
removed by means of specific techniques (e.g. covariance filtering or
localisation).</p>
      <p>Assimilation methods using a static type of the BEC have recently gained
considerable attention because of their flexibility (Lorenc, 2003).
Furthermore, present computational resources limit the number of ensemble
members accounted on operational EnKF. Thus, it is appealing to have an
algorithm that could work with smaller-sized ensembles and that could benefit
from whatever flow-dependent information this smaller ensemble provides.</p>
      <p>Recent encouraging results suggest that if ensemble information is used in
the variational data assimilation framework to augment the static BEC,
analyses can be improved. Hereinafter, we call this method a “hybrid”
scheme. Development of hybrid schemes has been an area of active research in
atmospheric data assimilation (Hamill and Snyder, 2000; Etherton and Bishop,
2004; Wang et al., 2007). Several studies have been conducted on the hybrid
schemes. Studies by Hamill and Snyder (2000), Etherton and Bishop (2004) and
Wang et al. (2007) used simple models and simulated observations to suggest
the effectiveness of incorporating ensembles in the 3DVAR to improve the
analyses. Although the recourse to hybrid covariances and the choice of the
relative weights given to them remains empirical in practice, it has been
shown in particular that hybrid models tend to be more robust than
conventional ensemble-based data assimilation schemes, especially when the
model errors are larger than observational ones (Wang et al., 2007, 2008,
2009). This feature is attractive for the regional assimilation problems in
oceanography, where information on the background state is often scant and
incomplete. Promising application of the hybrid scheme in global
oceanographic exercise has been recently provided by Penny et al. (2015).
They compared hybrid, classical 3DVAR and EnKF schemes in an observing system
simulation experiment and also using real data, showing that the hybrid
scheme reduces errors for all prognostic model variables eliminating growth
in biases present in the EnKF and 3DVAR.</p>
      <p>Recent work has also started addressing the issue of multi-scale data
assimilation, where the analyses are a combination of corrections with
different spatial-scale signals, assuming somehow that spatial scales are
separable and that observations may naturally bear information across several
spatial scales. Examples of these schemes range from multi-scale 3DVAR
systems (MS-VAR), sequential applications of horizontal operators with
different correlation length scales (Mirouze et al., 2016), or inclusion of a
large-scale analysis in the analysis formulation as additional constraint
(Guidard and Fischer, 2008). A possible simplification is to assume that
systematic errors are characterised by long length scales, as often occurs to
some extent (Dee, 2005).</p>
      <p>In this study, we develop a hybrid data assimilation system for the REP14-MED
(Mediterranean Recognized Environmental Picture, 2014) NEMO model
implementation, based on the existing 3DVAR system. Section 2 describes the
hybrid variational data assimilation scheme adopted accounting for systematic
error corrections. In Sect. 3 details on the experiments set-up are provided.
In Sect. 4 the results are presented and discussed. Finally, Sect. 5 offer
the summery and conclusions.</p>
</sec>
<sec id="Ch1.S2">
  <title>The hybrid variational-ensemble scheme</title>
      <p>A 3DVAR algorithm has been used to implement and test our hybrid assimilation
scheme. 3DVAR is relatively easy to implement and to expand, it can easily
take into consideration different estimates of BEC, its core is independent
of the primitive equations model core, and it is portable. The cost function
in 3DVAR is defined as

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mi>J</mml:mi><mml:mo>(</mml:mo><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>)</mml:mo><mml:mo>=</mml:mo></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mo mathsize="1.1em">(</mml:mo><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">B</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="Ch1.E1"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold-italic">y</mml:mi><mml:mo>-</mml:mo><mml:mi>H</mml:mi><mml:mfenced close="]" open="["><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced></mml:mfenced><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="bold-italic">y</mml:mi><mml:mo>-</mml:mo><mml:mi>H</mml:mi><mml:mfenced close="]" open="["><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced></mml:mfenced><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          where <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold-italic">x</mml:mi></mml:math></inline-formula> is the analysis state vector at the minimum of <inline-formula><mml:math display="inline"><mml:mi>J</mml:mi></mml:math></inline-formula>,
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the background state vector, <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> is the
background error covariance matrix, <inline-formula><mml:math display="inline"><mml:mi>H</mml:mi></mml:math></inline-formula> is the non-linear observational
operator, <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold-italic">y</mml:mi></mml:math></inline-formula> are the observations and <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">R</mml:mi></mml:math></inline-formula> is the
observational error covariance matrix. The cost function is linearised around
the background state and take the following form:

              <disp-formula id="Ch1.E2" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi>J</mml:mi><mml:mo>(</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>)</mml:mo><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">B</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mo mathsize="1.1em">(</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo>=</mml:mo><mml:mfenced open="[" close="]"><mml:mi mathvariant="bold-italic">y</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mfenced close=")" open="("><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced></mml:mfenced></mml:mrow></mml:math></inline-formula> is the misfit, <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">H</mml:mi></mml:math></inline-formula> is the linearised observational operator
and <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> are the increments.</p>
      <p><?xmltex \hack{\newpage}?>Following Dobricic and Pinardi (2008), the present 3DVAR scheme assumes that
the <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> matrix can be rewritten and thus decomposed as follows:

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mlabeledtr id="Ch1.E3"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mi mathvariant="bold">B</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="bold-italic">V</mml:mi><mml:msup><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr><mml:mlabeledtr id="Ch1.E4"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mo>=</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">D</mml:mi></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mrow><mml:mi>u</mml:mi><mml:mo>,</mml:mo><mml:mspace width="0.125em" linebreak="nobreak"/><mml:mi>v</mml:mi></mml:mrow></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="italic">η</mml:mi></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub><mml:mo>.</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          This also has the advantage of imposing pre-conditioning, as the minimisation
is performed on the control variable <inline-formula><mml:math display="inline"><mml:mi>v</mml:mi></mml:math></inline-formula> (with <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="bold">V</mml:mi><mml:mi>v</mml:mi></mml:mrow></mml:math></inline-formula>), which also serves the purpose of avoiding the inversion of
<inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>.</p>
      <p>Basically, the background error covariance matrix is modelled as a linear
sequence of several <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold-italic">V</mml:mi></mml:math></inline-formula> operators. Each <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold-italic">V</mml:mi></mml:math></inline-formula> defines a specific
error space. From right to left, <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> defines the vertical
covariance computed using multivariate empirical orthogonal functions (EOFs), <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> projects the vertical error to the
horizontal space by means of a recursive filter, <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="italic">η</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> (the
balance operator) is a 2-D barotropic model accounting for sea surface height
adjustments and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mrow><mml:mi>u</mml:mi><mml:mo>,</mml:mo><mml:mi>v</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> force a geostrophic balance between
temperature, salinity and the velocity components. Finally,
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">D</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is a divergence damping operator avoiding spurious
currents close to the coast in the presence of complex coast lines (details
in Dobricic and Pinardi, 2008). It is clear that this <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>
formulation introduces flexibility in the code and allows for the possibility
to test different hypotheses.</p>
      <p>In our static formulation of the 3DVAR, the vertical transformation operator
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> has the form:

              <disp-formula id="Ch1.E5" content-type="numbered"><mml:math display="block"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:msubsup><mml:mi mathvariant="bold">Λ</mml:mi><mml:mi mathvariant="normal">c</mml:mi><mml:mrow><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:mn mathvariant="normal">2</mml:mn></mml:mrow></mml:msubsup><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where columns of <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> contain multivariate eigenvectors and
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">Λ</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is a diagonal matrix with eigenvalues of
EOFs. Promising recently published results (Dobricic et
al., 2015) propose a new method to estimate the vertical part of the
background error covariance matrix for an ocean variational data assimilation
system based on high frequency estimates from a Bayesian hierarchical model.
A general approach to defining hybrid assimilation schemes is to compute
<inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> as a linear combination of the “static” covariance operator,
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, and the flow-dependent operator,
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, derived from the statistics of an ensemble of
analyses/forecast:

              <disp-formula id="Ch1.E6" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="bold">B</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula>

        The relative weighting (<inline-formula><mml:math display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula>) still requires empirical tuning but in
general can be adjusted to the size of the ensemble. Large ensemble size can
provide robust estimate of <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and thus Eq. (6) can be
theoretically implemented with small <inline-formula><mml:math display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> values (Ménétrier and
Auligné, 2015).</p>
      <p>The proposed approach introduces the flow-dependent <bold>B</bold> by defining
the increment as a weighted sum of parts corresponding to climatological and
ensemble covariance matrices:

              <disp-formula id="Ch1.E7" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula>

        It can be demonstrated (see Appendix A for details) by combining Eqs. (2),
(6) and (7) that the following cost function has the minimum for the same
value of <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mrow></mml:math></inline-formula> as the cost function with the background error covariance matrix defined in Eq. (6) (e.g. Wang et al., 2007):

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mi>J</mml:mi><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced><mml:mo>=</mml:mo></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:mo mathsize="1.1em">(</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="Ch1.E8"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mo>.</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          In Appendix A we further demonstrate that by updating each member of the
forecast ensemble with Eq. (8), we obtain the same estimate for the analysis
error covariance matrix as when doing it with Eq. (6).</p>
      <p>By defining the control vector <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold-italic">v</mml:mi></mml:math></inline-formula> consisting of climatological and
ensemble parts <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mo>=</mml:mo><mml:mo>(</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>,</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula> the cost
function becomes

              <disp-formula id="Ch1.E9" content-type="numbered"><mml:math display="block"><mml:mrow><?xmltex \hack{\hbox\bgroup\fontsize{8.5}{8.5}\selectfont$\displaystyle}?><mml:mi>J</mml:mi><mml:mfenced close=")" open="("><mml:mi mathvariant="bold-italic">v</mml:mi></mml:mfenced><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:msubsup><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">c</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:msub><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:msubsup><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">e</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:msub><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mo>,</mml:mo><?xmltex \hack{$\egroup}?></mml:mrow></mml:math></disp-formula>

        and increment <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mrow></mml:math></inline-formula>:

              <disp-formula id="Ch1.E10" content-type="numbered"><mml:math display="block"><mml:mrow><?xmltex \hack{\hbox\bgroup\fontsize{8.5}{8.5}\selectfont$\displaystyle}?><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>=</mml:mo><mml:mfenced close=")" open="("><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">D</mml:mi></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mrow><mml:mi>u</mml:mi><mml:mo>,</mml:mo><mml:mspace width="0.125em" linebreak="nobreak"/><mml:mi>v</mml:mi></mml:mrow></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="italic">η</mml:mi></mml:msub><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mfenced><mml:mspace linebreak="nobreak" width="0.125em"/><mml:mfenced close=")" open="("><mml:msup><mml:mi mathvariant="italic">α</mml:mi><mml:mrow><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:mn mathvariant="normal">2</mml:mn></mml:mrow></mml:msup><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:msubsup><mml:mi mathvariant="bold">Λ</mml:mi><mml:mi mathvariant="normal">c</mml:mi><mml:mrow><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:mn mathvariant="normal">2</mml:mn></mml:mrow></mml:msubsup><mml:msub><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msup><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:mrow><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:mn mathvariant="normal">2</mml:mn></mml:mrow></mml:msup><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msubsup><mml:mi mathvariant="bold">Λ</mml:mi><mml:mi mathvariant="normal">e</mml:mi><mml:mrow><mml:mn mathvariant="normal">1</mml:mn><mml:mo>/</mml:mo><mml:mn mathvariant="normal">2</mml:mn></mml:mrow></mml:msubsup><mml:msub><mml:mi mathvariant="bold-italic">v</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mo>,</mml:mo><?xmltex \hack{$\egroup}?></mml:mrow></mml:math></disp-formula>

        where columns of <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">Λ</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>
can now be computed at any frequency from a relatively small size ensemble.</p>
      <p>Ensemble statistics may also provide estimates of the day of the horizontal
correlation radii to be used in <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>. Using the recursive
filter formulation <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> takes the form:

              <disp-formula id="Ch1.E11" content-type="numbered"><mml:math display="block"><mml:mrow><?xmltex \hack{\hbox\bgroup\fontsize{8.8}{8.8}\selectfont$\displaystyle}?><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi mathvariant="bold">W</mml:mi><mml:mi>y</mml:mi></mml:msub><mml:mfenced close=")" open="("><mml:msubsup><mml:mi>L</mml:mi><mml:mi>y</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>,</mml:mo><mml:mi mathvariant="bold">Δ</mml:mi><mml:mi mathvariant="bold-italic">y</mml:mi></mml:mfenced><mml:msub><mml:mi>G</mml:mi><mml:mi>y</mml:mi></mml:msub><mml:mfenced close=")" open="("><mml:msubsup><mml:mi>L</mml:mi><mml:mi>y</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mi mathvariant="bold">Δ</mml:mi><mml:mi mathvariant="bold-italic">y</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">W</mml:mi><mml:mi>x</mml:mi></mml:msub><mml:mfenced close=")" open="("><mml:msubsup><mml:mi>L</mml:mi><mml:mi>x</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>,</mml:mo><mml:mi mathvariant="bold">Δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced><mml:msub><mml:mi>G</mml:mi><mml:mi>x</mml:mi></mml:msub><mml:mfenced open="(" close=")"><mml:msubsup><mml:mi>L</mml:mi><mml:mi>x</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mi mathvariant="bold">Δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced><mml:mo>,</mml:mo><?xmltex \hack{$\egroup}?></mml:mrow></mml:math></disp-formula>

        where <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>G</mml:mi><mml:mi>x</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>G</mml:mi><mml:mi>y</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> represent the zonal and meridional recursive filter
operators, <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">W</mml:mi><mml:mi>x</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">W</mml:mi><mml:mi>y</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> are the diagonal matrices
with normalisation coefficient, <inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mrow><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi></mml:mrow><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="bold">Δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo mathvariant="bold">,</mml:mo><mml:mi mathvariant="bold-italic">y</mml:mi></mml:mrow></mml:math></inline-formula> are the zonal and meridional length scale and grid spacing respectively.</p>
      <p>According to Belo Pereira and Berre (2006), for any simulated error
<inline-formula><mml:math display="inline"><mml:mi mathvariant="italic">ε</mml:mi></mml:math></inline-formula> it is possible to define a zonal and a meridional length
scale:

              <disp-formula id="Ch1.E12" content-type="numbered"><mml:math display="block"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mrow><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi></mml:mrow><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>=</mml:mo><mml:msqrt><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:msup><mml:mi mathvariant="italic">σ</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>(</mml:mo><mml:mi mathvariant="italic">ε</mml:mi><mml:mo>)</mml:mo></mml:mrow><mml:mrow><mml:msup><mml:mi mathvariant="italic">σ</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mfenced open="(" close=")"><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mrow><mml:mo>∂</mml:mo><mml:mi mathvariant="italic">ε</mml:mi></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi></mml:mrow></mml:mfrac></mml:mstyle></mml:mfenced><mml:mo>+</mml:mo><mml:msup><mml:mfenced open="(" close=")"><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mrow><mml:mo>∂</mml:mo><mml:mi mathvariant="italic">σ</mml:mi><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">ε</mml:mi></mml:mfenced></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi></mml:mrow></mml:mfrac></mml:mstyle></mml:mfenced><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:mfrac></mml:mstyle></mml:msqrt><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where <inline-formula><mml:math display="inline"><mml:mrow><mml:mo>∂</mml:mo><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi></mml:mrow></mml:math></inline-formula> are the derivatives in <inline-formula><mml:math display="inline"><mml:mi>x</mml:mi></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mi>y</mml:mi></mml:math></inline-formula> direction, <inline-formula><mml:math display="inline"><mml:mrow><mml:msup><mml:mi mathvariant="italic">σ</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>(</mml:mo><mml:mi mathvariant="italic">ε</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msup><mml:mi mathvariant="italic">σ</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>(</mml:mo><mml:mo>∂</mml:mo><mml:mi mathvariant="italic">ε</mml:mi><mml:mo>/</mml:mo><mml:mo>∂</mml:mo><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:math></inline-formula> are
the variances of the background error and of the derivative. In the
ensemble-based approach <inline-formula><mml:math display="inline"><mml:mi mathvariant="italic">ε</mml:mi></mml:math></inline-formula> is the ensemble anomalies computed with
respect to the ensemble mean.</p>
      <p>Though most data assimilation methods assume that the model forecast (i.e.
the background) is unbiased, that is rarely the case. Model bias can
systematically cause the model to drift away from the truth, eventually
propagating into the analyses. In limited-area models (LAM) integrated for
relatively short time the systematic errors (bias) may derive from inadequate
model physics and parameterisations as well as inaccurate initialisation and
open boundary conditions, including the atmospheric forcing. An adequate
solution is strictly necessary since the systematic error in the large-scale
forcing field can prevent the right small-scale dynamics from developing
properly and thus can strongly reduce the potential benefits deriving from
the increased resolution and/or improved physics.</p>
      <p>Here, we assume that systematic errors are associated to large-scale errors.
This idea is consistent with the high-resolution model presented in Sect. 3
and with the experimental set-up where the large-scale uncertainties
(initialisation, boundary conditions and surface forcing) are not accounted
for
in the generation of the ensemble members.</p>
      <p>Further expanding the decomposition introduced in Eq. (7) and following
recent studies suggesting the possibility to treat multiple-scale errors
during the analysis steps (Li et al., 2015), we reformulate the analysis
increments as

              <disp-formula id="Ch1.E13" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where the first two terms on right-hand side represent the increments deriving from the
minimisation of Eq. (9) while the last term indicates the increments due to
the large-scale systematic error not sampled either in the climatological or
ensemble-based estimates of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>. Note that the scale decomposition
requires that large-scale and small-scale background error are mutually
uncorrelated. It is worth mentioning that the large-scale systematic error
could be partially accounted in the generation of the ensemble members;
however, this would imply a considerably large number of ensemble members with
clear implications on the computational side and the corresponding
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> would then incorporate error information at different
scales.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F1" specific-use="star"><caption><p>Top-left panel: western Mediterranean Sea. Bottom-right panel: model
domain and collected data during REP14-MED experiment. Green dots indicate
sea level anomaly measurements from
satellite (from 3 to 30 June 2014), red dots CTD positions (from 7 to 24 June
2014), blue dots glider trajectories (surfacing points, from 8 to 23 June
2014). The magenta lines indicate the box used to compute ensemble
statistics. Bathymetric lines are also shown (m).</p></caption>
        <?xmltex \igopts{width=341.433071pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f01.png"/>

      </fig>

      <p>The availability of an ensemble simulation allows us to retrieve estimates of
the model bias or systematic error. Recalling that

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:mfenced open="[" close="]"><mml:mi mathvariant="bold-italic">y</mml:mi><mml:mo>-</mml:mo><mml:mi>H</mml:mi><mml:mfenced close=")" open="("><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced></mml:mfenced><mml:mo>=</mml:mo><mml:mi mathvariant="bold-italic">y</mml:mi><mml:mo>-</mml:mo><mml:mi>H</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>+</mml:mo><mml:mi>H</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>-</mml:mo><mml:mi>H</mml:mi><mml:mfenced close=")" open="("><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="Ch1.E14"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:mi>H</mml:mi><mml:mo>(</mml:mo><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">r</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:mo>)</mml:mo><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          where <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold-italic">d</mml:mi></mml:math></inline-formula> is the misfit, <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the true state of the
ocean, <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the observational error resulting from
the sum of <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">oi</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">or</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> (the
instrumental and representation observational errors respectively),
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">r</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the background random error and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the background systematic error (errors are defined as
departures from the true state). Every assimilation scheme is designed to
correct the random error, which is assumed to have zero mean. The
representation error can be defined based on the knowledge of the dynamics of
the simulated system or as being proportional to the variance of the
measurements (Oke and Sakov, 2008). <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mrow><mml:mi mathvariant="normal">e</mml:mi><mml:mo>,</mml:mo><mml:mspace width="0.125em" linebreak="nobreak"/><mml:mi mathvariant="normal">c</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> and
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">Λ</mml:mi><mml:mrow><mml:mi mathvariant="normal">e</mml:mi><mml:mo>,</mml:mo><mml:mspace linebreak="nobreak" width="0.125em"/><mml:mi mathvariant="normal">c</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> introduced in the vertical covariance
(Eq. 10) provide a multivariate statistical representation of <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">r</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>. To obtain a bias, or systematic error, estimate we can average over the ensemble members and
assuming that the observational error is also unbiased:

              <disp-formula id="Ch1.E15" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mover accent="true"><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo mathvariant="normal">‾</mml:mo></mml:mover><mml:mo>=</mml:mo><mml:mo>-</mml:mo><mml:mover accent="true"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow><mml:mo mathvariant="normal">‾</mml:mo></mml:mover><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        since <inline-formula><mml:math display="inline"><mml:mover accent="true"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">r</mml:mi></mml:msub></mml:mrow><mml:mo mathvariant="normal">‾</mml:mo></mml:mover></mml:math></inline-formula> also is zero by definition. Thus,
analysing the misfit of the ensemble members, we can obtain an estimate of
the bias or systematic error. In other words, the ensemble system is
exploited not only to estimate the flow-dependent components of the
background error covariances, but also to estimate the large-scale bias in
the analysis step. From the previous relationship it is clear that the
large-scale bias is originally defined in observation space and successively
mapped in model space.</p>
      <p>In our formulation we assume that the scales in <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="italic">ε</mml:mi><mml:mi mathvariant="normal">r</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> are significantly different and the estimate of the
ensemble systematic error is used simultaneously to the 3DVAR analysis step
to correct the background fields. The small-scale increments arise from the
classical minimisation of the cost function <inline-formula><mml:math display="inline"><mml:mi>J</mml:mi></mml:math></inline-formula>:

              <disp-formula id="Ch1.E16" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:munder><mml:mo movablelimits="false">min⁡</mml:mo><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mrow></mml:munder><mml:mfenced close=")" open="("><mml:mi>J</mml:mi><mml:mo>(</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>)</mml:mo></mml:mfenced><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        while the large-scale increments due to the systematic error are defined as

              <disp-formula id="Ch1.E17" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">s</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mi>L</mml:mi><mml:mo mathsize="1.5em">(</mml:mo><mml:mfenced close=")" open="("><mml:mover accent="true"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">eb</mml:mi></mml:msub></mml:mrow><mml:mo mathvariant="normal">‾</mml:mo></mml:mover><mml:mo>+</mml:mo><mml:msup><mml:mi>P</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:mo>(</mml:mo><mml:mover accent="true"><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo mathvariant="normal">‾</mml:mo></mml:mover><mml:mo>)</mml:mo></mml:mfenced><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:mo mathsize="1.5em">)</mml:mo><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where <inline-formula><mml:math display="inline"><mml:mover accent="true"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">eb</mml:mi></mml:msub></mml:mrow><mml:mo mathvariant="normal">‾</mml:mo></mml:mover></mml:math></inline-formula> is the background ensemble mean, <inline-formula><mml:math display="inline"><mml:mi>L</mml:mi></mml:math></inline-formula>
is a low-pass filter used to ensure that scales of the two increments do not
overlap and <inline-formula><mml:math display="inline"><mml:mi>P</mml:mi></mml:math></inline-formula> is a generic linearised observational operator.</p>
      <p>Such a scheme thus requires a fairly dense observational network to estimate
the bias, whose availability may in general depend on the simulation area and
period. The method is potentially affected by systematic observational errors
and thus is sensitive to the design of the observational networks. On the
other hand the analysis of the systematic error can provide useful insights
about the error not represented in the ensemble space and thus help in the
definition of the ensemble generation procedure.</p>
      <p>Depending on data availability and ensemble size, the bias estimator can be
constant or spatially or temporal dependent.</p>
</sec>
<sec id="Ch1.S3">
  <title>Experimental set-up</title>
      <p>In June 2014, a REP14-MED sea trial off the west coast of Sardinia was
conducted by CMRE (Centre for Maritime Research and Experimentation),
coordinating efforts of 20 partners from six different nations. Two research
vessels collected a massive amount of data in an area of approximatively
<inline-formula><mml:math display="inline"><mml:mrow><mml:mn>100</mml:mn><mml:mspace width="0.125em" linebreak="nobreak"/><mml:mi mathvariant="normal">km</mml:mi><mml:mo>×</mml:mo><mml:mn>100</mml:mn><mml:mspace linebreak="nobreak" width="0.125em"/><mml:mi mathvariant="normal">km</mml:mi></mml:mrow></mml:math></inline-formula> with various oceanographic instruments
(lowered CTD, undulating
towed vehicle, CTD chain, ship mounted ADCP, shallow and deep underwater
gliders, moorings, surface drifters and profiling floats). Complementary
information has been retrieved, by remote sensing (sea level anomalies and
sea surface temperature both from Copernicus Marine Environmental Monitoring
Service, CMEMS; <uri>http://marine.copernicus.eu/</uri>) and existing results from
atmospheric and oceanographic operational models. In situ CTD and glider data
together with remotely sensed SLA (sea level anomalies) and SST (seas surface
temperature) have been used for assimilation while for the sake of simplicity
only CTD data are used for model validations. Spatial and temporal
distributions of observations are provided in Figs. 1 and 2 respectively.</p>

<?xmltex \floatpos{t}?><table-wrap id="Ch1.T1" specific-use="star"><caption><p>Model configuration details.</p></caption><oasis:table frame="topbot"><oasis:tgroup cols="2">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="left"/>
     <oasis:tbody>
       <oasis:row>  
         <oasis:entry colname="col1">No. points <inline-formula><mml:math display="inline"><mml:mi>x</mml:mi></mml:math></inline-formula>/<inline-formula><mml:math display="inline"><mml:mi>y</mml:mi></mml:math></inline-formula>/<inline-formula><mml:math display="inline"><mml:mi>z</mml:mi></mml:math></inline-formula></oasis:entry>  
         <oasis:entry colname="col2">235/246/91</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1"><inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="normal">Δ</mml:mi><mml:mi>x</mml:mi></mml:mrow></mml:math></inline-formula>/<inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="normal">Δ</mml:mi><mml:mi>y</mml:mi></mml:mrow></mml:math></inline-formula></oasis:entry>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mo>∼</mml:mo></mml:math></inline-formula> 1000 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1"><inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="normal">Δ</mml:mi><mml:mi>T</mml:mi></mml:mrow></mml:math></inline-formula> time-step</oasis:entry>  
         <oasis:entry colname="col2">100 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">s</mml:mi></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Vertical discretisation</oasis:entry>  
         <oasis:entry colname="col2">Hybrid <inline-formula><mml:math display="inline"><mml:mi>z</mml:mi></mml:math></inline-formula>–sigma (Siddorn and Furner, 2013)</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Surface fluxes</oasis:entry>  
         <oasis:entry colname="col2">MFS Bulk (Oddo et al., 2009)</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Atmospheric data</oasis:entry>  
         <oasis:entry colname="col2">Hourly COSMO</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">SST</oasis:entry>  
         <oasis:entry colname="col2">GOS_SST <inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>30 <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="normal">W</mml:mi><mml:mspace width="0.125em" linebreak="nobreak"/><mml:msup><mml:mi mathvariant="normal">m</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">2</mml:mn></mml:mrow></mml:msup><mml:mspace linebreak="nobreak" width="0.125em"/><mml:msup><mml:mi mathvariant="normal">K</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Lateral boundary condition (LBC)</oasis:entry>  
         <oasis:entry colname="col2">No slip</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Open LBC (barotropic/baroclinic/tracers)</oasis:entry>  
         <oasis:entry colname="col2">Flather (1976)/imposition/Neumann</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Bottom friction</oasis:entry>  
         <oasis:entry colname="col2">Non-linear</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">EOS</oasis:entry>  
         <oasis:entry colname="col2">EOS-80</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Tracer advection</oasis:entry>  
         <oasis:entry colname="col2">TVD (Zalesak, 1979)</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Tracer diffusion</oasis:entry>  
         <oasis:entry colname="col2">Laplacian along iso-surface <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>K</mml:mi><mml:mi mathvariant="normal">h</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:mn>10</mml:mn></mml:mrow></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Momentum advection</oasis:entry>  
         <oasis:entry colname="col2">Vector form</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Momentum diffusion</oasis:entry>  
         <oasis:entry colname="col2">Bilaplacian along iso-surface <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>A</mml:mi><mml:mi mathvariant="normal">h</mml:mi></mml:msub><mml:mo>=</mml:mo></mml:mrow></mml:math></inline-formula> <inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>2.5 <inline-formula><mml:math display="inline"><mml:mrow><mml:mo>×</mml:mo><mml:msup><mml:mn>10</mml:mn><mml:mn mathvariant="normal">7</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Vertical turbulence</oasis:entry>  
         <oasis:entry colname="col2">GLS with Canuto et al. (2001) and <inline-formula><mml:math display="inline"><mml:mi>k</mml:mi></mml:math></inline-formula>–eps</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Free surface</oasis:entry>  
         <oasis:entry colname="col2">Filtered (Madec, 2008)</oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup></oasis:table></table-wrap>

      <p>NEMO (Nucleus for European Modelling of the Ocean; Madec, 2008) has been
implemented as the primitive equations dynamical model component of the data
assimilation system. The ocean engine of NEMO is adapted to regional and
global ocean circulation problems. Prognostic variables are the meridional
and zonal velocities, sea surface height, temperature and salinity. In the
horizontal direction, the model uses a curvilinear orthogonal grid and in the
vertical direction, a full or partial step <inline-formula><mml:math display="inline"><mml:mi>z</mml:mi></mml:math></inline-formula> coordinate, a <inline-formula><mml:math display="inline"><mml:mi>s</mml:mi></mml:math></inline-formula> coordinate
or a mixture of the two can be applied. The NEMO ocean engine is very
flexible allowing several choices for discretisation and parameterisations;
details on the present configuration are provided on Table 1 while model
domain and bathymetry are shown in Fig. 1.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F2"><caption><p>Number of observations per day. The colour coding is according to
Fig. 1. The <inline-formula><mml:math display="inline"><mml:mi>y</mml:mi></mml:math></inline-formula> axis indicates the number of vertical profiles for CTD and
gliders and number of points within the model domain for the SLA data. The
<inline-formula><mml:math display="inline"><mml:mi>x</mml:mi></mml:math></inline-formula> axis indicates time.</p></caption>
        <?xmltex \igopts{width=199.169291pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f02.png"/>

      </fig>

      <p>An ensemble of data assimilation system with 14 independent
members with daily assimilation cycles has been performed to generate the
ensemble statistics. All simulation/assimilation experiments presented
hereafter started on 1 June and ended on 30 June 2014, the MED-REP14 period.
All the experiments are initialised and forced at the lateral open boundaries
using the Mercator-Ocean (Drévillon et al., 2008) product in the
Mediterranean Sea, while surface fluxes are computed by means of bulk
formulae using hourly atmospheric data with 7.0 km horizontal resolution
provided by the Italian Meteorological Centre and based on the COSMO-ME model, an
implementation of the COnsortium for Small-scale MOdelling (COSMO). The
ensemble members have been generated simultaneously assimilating perturbed
observations varying the corresponding observational error, and assuming
different horizontal correlation radii in <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>.</p>
      <p>For the observation perturbation, either weak or strong criteria for
retaining observations are used among the ensemble members. Conservative
quality check procedures assume good quality flags in both temperature and
salinity and reduce the total number of assimilated observations. Filters
have been applied horizontally and vertically to reduce the higher spatial
sampling of observation with respect to the model grid. Within the ensemble
members, different vertical cut-off scales have been used in the low-pass
filter resulting in differently smoothed profiles. Horizontal data binning
has been applied to the observations falling in 1 or 2 model grid cells
while keeping the original vertical resolution. When the filtering or
binning procedures are applied the corresponding full resolution profile
standard deviation has been used as an estimate of the observational error.
Similar procedures have been applied to CTD and gliders data.</p>
      <p>The default horizontal correlation radii (<inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mrow><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi></mml:mrow><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula>) have been
computed according to Eq. (12) from the 15-year CMEMS Mediterranean reanalysis.
They correspond to 21 and 12 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> in the meridional and zonal
directions respectively. Two additional sets of correlation radii have been
used in the ensemble generation, they have been defined on the base of
sensitivity experiments and correspond to 12/6 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> (meridional/zonal)
and 6/3 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> respectively. The three sets of correlation radii remain
constant during the simulated period. An example of the observational
perturbation, associated error and horizontal correlation radii is shown in
Fig. 3.</p>
      <p>The ensemble generation method spans the uncertainty linked with the
observational sampling and assimilation formulation, implicitly acting on the
background ensemble spread. This method clearly connects the growth of the
ensemble spread to the function used to perturb the observations and
simultaneously links the ensemble spread to observation availability. For the
time being, the perturbation of surface and lateral boundary conditions is
not considered, assuming that the flow-dependent component of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> is
associated with the small-scale error fluctuations that are reasonably well
reproduced with the observation perturbation only. This approach also ensures
consistency with the assumptions done to derive Eq. (8). Although the
large-scale forcing may act as an attractor for the ensemble perturbations,
especially at the sea surface and in proximity of the boundaries, the goal of
the present implementation is the evaluation of the feasibility of a hybrid
system that simultaneously corrects the model's systematic and random errors.
Further refinements of the ensemble generation strategy may be considered in
the future.</p>
      <p>All the ensemble members use a static and homogeneous <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> where
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">Λ</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> are derived from
multivariate (temperature, salinity and sea surface height) vertical EOF
computed from the anomalies with respect to the long-term mean of a 15-year
CMEMS Mediterranean reanalysis. The incremental analysis update (IAU)
strategy has been used to incorporate analysis increments into the model
integration in a gradual manner (Bloom et al., 1996), spreading the analysis
increments uniformly on a 6 h time window.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F3"><caption><p>Example of perturbed CTD vertical profile with different quality
check procedure and filtering applied. The solid black line indicates the
full resolution CTD profile while horizontal lines are the associated
observational error. The other colours indicate the perturbed profile. In the
middle panel the three tested couples of horizontal correlation length scales
(<inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mrow><mml:mo>(</mml:mo><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi><mml:mo>)</mml:mo></mml:mrow><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula>) are shown. The circles indicate the distance
where the horizontal correlation of a single observation is zero. The green
circle length scales are <inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mi>x</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>=</mml:mo></mml:mrow></mml:math></inline-formula> 12 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> and
<inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mi>y</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>=</mml:mo></mml:mrow></mml:math></inline-formula> 21 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula>, this set has also been used in the
reference experiment. The red circle radii are <inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mi>x</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>=</mml:mo></mml:mrow></mml:math></inline-formula> 6 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mi>y</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>=</mml:mo></mml:mrow></mml:math></inline-formula> 12 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula>. The blues circle
radii are <inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mi>x</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>=</mml:mo></mml:mrow></mml:math></inline-formula> 3 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mi>y</mml:mi><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup><mml:mo>=</mml:mo></mml:mrow></mml:math></inline-formula> 6 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula>.</p></caption>
        <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f03.png"/>

      </fig>

      <p>In the hybrid variational assimilation system, the generated ensemble
information has been projected into <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> through the multivariate
(temperature, salinity and sea surface height) vertical EOFs providing
spatially varying daily estimates of <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">S</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">Λ</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>. Ensemble information has been also used to
compute daily varying horizontal correlation radii, <inline-formula><mml:math display="inline"><mml:mrow><mml:msubsup><mml:mi>L</mml:mi><mml:mrow><mml:mi>x</mml:mi><mml:mo>,</mml:mo><mml:mi>y</mml:mi></mml:mrow><mml:mi mathvariant="italic">ε</mml:mi></mml:msubsup></mml:mrow></mml:math></inline-formula>,
in <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>. Several <inline-formula><mml:math display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> values have been tested.
Sensitivity experiments have shown that the best results were obtained
setting <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:mo>=</mml:mo><mml:mn>0.5</mml:mn></mml:mrow></mml:math></inline-formula>, meaning that 50 % of the vertical error
covariance derives from the climatological statistics, while the remaining
50 % derives from the ensemble statistics. In our hybrid system the
observational representation error is proportional to the variance of the
measures after binning in a 1 km square grid.</p>

<?xmltex \floatpos{t}?><table-wrap id="Ch1.T2"><caption><p>Experiments.</p></caption><oasis:table frame="topbot"><oasis:tgroup cols="4">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="left"/>
     <oasis:colspec colnum="3" colname="col3" align="left"/>
     <oasis:colspec colnum="4" colname="col4" align="left"/>
     <oasis:thead>
       <oasis:row rowsep="1">  
         <oasis:entry colname="col1"/>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula></oasis:entry>  
         <oasis:entry colname="col3">Bias corr.</oasis:entry>  
         <oasis:entry colname="col4">H. Corr Radii</oasis:entry>
       </oasis:row>
     </oasis:thead>
     <oasis:tbody>
       <oasis:row>  
         <oasis:entry colname="col1">Exp-ref</oasis:entry>  
         <oasis:entry colname="col2">Static-Clim</oasis:entry>  
         <oasis:entry colname="col3">No</oasis:entry>  
         <oasis:entry colname="col4">Const</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Exp-Hy1</oasis:entry>  
         <oasis:entry colname="col2">Hybrid</oasis:entry>  
         <oasis:entry colname="col3">No</oasis:entry>  
         <oasis:entry colname="col4">Variable</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Exp-Cl1</oasis:entry>  
         <oasis:entry colname="col2">Static-Clim</oasis:entry>  
         <oasis:entry colname="col3">Yes</oasis:entry>  
         <oasis:entry colname="col4">Const</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">Exp-Hy2</oasis:entry>  
         <oasis:entry colname="col2">Hybrid</oasis:entry>  
         <oasis:entry colname="col3">Yes</oasis:entry>  
         <oasis:entry colname="col4">Variable</oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup></oasis:table></table-wrap>

      <p>The ensemble statistics have also been used to estimate the model systematic
error and a large-scale systematic error correction has been applied. For
every simulated day, <inline-formula><mml:math display="inline"><mml:mover accent="true"><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo mathvariant="normal">‾</mml:mo></mml:mover></mml:math></inline-formula> has been computed using a depth depended
observational window to avoid sampling error in the deep layers. The temporal
window increases linearly from 11 days, at surface, to 25 days in the bottom
layers. The resulting <inline-formula><mml:math display="inline"><mml:mover accent="true"><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo mathvariant="normal">‾</mml:mo></mml:mover></mml:math></inline-formula> has been mapped onto the model grid
(<inline-formula><mml:math display="inline"><mml:mrow><mml:msup><mml:mi>P</mml:mi><mml:mi>T</mml:mi></mml:msup></mml:mrow></mml:math></inline-formula> in Eq. 17) by means of the Barnes (1994) univariate objective
analysis with smoothing length scales of 170 and 90 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> along <inline-formula><mml:math display="inline"><mml:mi>x</mml:mi></mml:math></inline-formula> and
<inline-formula><mml:math display="inline"><mml:mi>y</mml:mi></mml:math></inline-formula> respectively. A length scale of 75 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> has been used in the low-pass filter (<inline-formula><mml:math display="inline"><mml:mi>L</mml:mi></mml:math></inline-formula> in Eq. 17).</p>
      <p>To test bias correction and the impact of the ensemble-based EOFs, results
from four different experiments are compared. Exp-ref uses climatological,
spatially homogeneous <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> and no bias correction; Exp-Hy1 uses the
hybrid, spatially and temporally varying, formulation of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> but no
bias correction; Exp-Cl1 uses the static formulation of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> (as
Exp-ref) but the bias correction is applied; and, finally, Exp-Hy2 uses both the
hybrid formulation of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> and the bias correction. The differences
between the experiments are summarised in Table 2.</p>
</sec>
<sec id="Ch1.S4">
  <title>Result and discussion</title>
      <p>The quality of the ensemble has been evaluated on the base of ensemble spread
values and distributions. The ensemble spread is defined as the standard
deviation across the ensemble members.</p>
      <p>In Fig. 4, the time evolution of temperature and salinity standard deviations
computed from the ensemble members are shown from the surface to 1000 m
depth. At all depths and for both temperature and salinity, the ensemble
spread reaches a stable value on 10/12 June after 2/3 days assimilating the CTD and
glider data from the first cruise leg (Fig. 2). The small spread during the
first days is mostly confined to the surface layers and is due to the SLA
assimilation. Between 13 and 22 June, the ensemble spread is nearly constant
at all depths, probably constrained by the dense observational network,
meaning that only a few days are needed to spin-up our ensemble system. Later
on during the simulated period, the data density decreases and temperature
and salinity ensemble spreads behaviours differ significantly. The salinity
ensemble spread remains nearly constant, whereas the standard deviation of
surface temperature decreases. We can speculate that the decreased ensemble
variability in temperature is due to the surface and lateral forcing shared
among all the ensemble members that rapidly constrain the temperature within
the model domain when no observations are assimilated. On the other hand,
salinity reacts slower to surface forcing. Thus, the methodology used to
generate the ensemble could be improved to also account for errors in the
external forcing (surface and lateral open boundary conditions) and model
parameterisations.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F4"><caption><p>Ensemble standard deviation computed over the observational domain
in the observational box (magenta) shown in Fig. 1. Ordinate indicates ocean
depth in metres while abscissa is time.</p></caption>
        <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f04.png"/>

      </fig>

      <?xmltex \floatpos{t}?><fig id="Ch1.F5"><caption><p>Horizontal maps of ensemble standard deviation for temperature
(left) and salinity (right) at 0.5 m on 12, 18, 24 and 30 June 2014.</p></caption>
        <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f05.png"/>

      </fig>

      <p>The horizontal distribution of the near-surface (0.5 m depth) ensemble
standard deviations for temperature and salinity valid for 12, 18, 24 and
30 June are shown in Fig. 5.</p>
      <p>During the simulated period the two state variables show different behaviours.
Temperature standard deviation maxima are mostly confined within the
observational space and have well-defined small/medium size structures. On
30 June, when no more in situ observations are available, a large-scale
maxima structure is evident close to the north-west domain open boundaries
partially due to SLA assimilation, and the different structures and dynamics
developed by the individual ensemble members approaching the open boundaries.
On the other hand, salinity spread horizontal distributions are significantly
different.</p>
      <p>During the entire simulated period, maxima of the salinity ensemble spread
are evident outside the area sampled by the observational campaign and
structures are generally larger than in temperature. These are probably due
to errors in the salinity content of water masses forcing the simulations at
the lateral open boundaries and conflicting with in situ observations, thus
generating fronts and instabilities. The adopted method to generate the
ensemble members does not account for uncertainties in the forcing (surface
or lateral) or initialisation; further work is necessary in order to assess
the impact of the forcing perturbation. The present work focuses on the
potential benefits of a hybrid approach, rather than on evaluating the
ensemble generation itself.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F6"><caption><p>The top panel shows the static and spatially homogeneous vertical
error correlation matrix, the bottom panel the ensemble estimate on 22 June
2014 at lat 7.0<inline-formula><mml:math display="inline"><mml:msup><mml:mi/><mml:mo>∘</mml:mo></mml:msup></mml:math></inline-formula> N and long 40.0<inline-formula><mml:math display="inline"><mml:msup><mml:mi/><mml:mo>∘</mml:mo></mml:msup></mml:math></inline-formula> E. The numbers on the axis
indicate to the model levels where the first 78 values represent temperature
and the second 78 values represent the salinity levels. The matrix blocks
represent the correlation of <inline-formula><mml:math display="inline"><mml:mi>T</mml:mi></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mi>S</mml:mi></mml:math></inline-formula> and their cross-correlations.</p></caption>
        <?xmltex \igopts{width=199.169291pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f06.png"/>

      </fig>

      <p>Figure 6 shows an example of how the ensemble method changes the estimates of
the salinity and temperature error vertical correlations and
cross-correlations. On 22 June 2014, the ensemble estimates exhibit
correlations of background temperature and salinity significantly different
from the climatological estimate. Clearly, the ensemble method has added
information to the climatological estimates from the variability generated by
the ensemble simulations on particular days. An interesting feature of
temperature and salinity vertical error correlations on 22 June 2014 is the
presence of several local maxima and minima. The similarities between static
and ensemble-based correlations reflects the error in the large-scale
dynamical processes, introduced in our system by lateral open boundaries
conditions. Salinity correlations (top right corner on both Fig. 6 panels)
show the largest differences between climatological/steady and the daily
estimates. While the salinity climatological correlation field is
characterised by generally positive values, in the daily estimate a clear
anti-correlation pattern is observed starting at level 50 and persisting
toward the bottom. This clearly indicates a more complex vertical error
structure probably due to the presence of an intermediate water mass (the
Modified Levantine Intermediate Waters) and deficiencies in the model to
correctly simulate it. Similar patterns, even if less pronounced, are also
observed in the temperature correlation and temperature–salinity
cross-correlations. Furthermore, vertical scales of the correlations differ
significantly. For instance, salinity vertical correlations are longer at the
ocean bottom in case of the ensemble <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>, while an opposite feature
is found for temperature. This suggests that the ensemble simulations lead to
stronger consistency of the vertical cross-correlation at the ocean bottom
between temperature and salinity with respect to the static <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F7" specific-use="star"><caption><p>Left panels: <inline-formula><mml:math display="inline"><mml:mi>y</mml:mi></mml:math></inline-formula> axis indicates depth in metre, <inline-formula><mml:math display="inline"><mml:mi>x</mml:mi></mml:math></inline-formula> axis indicates
time in days, colour are the systematic error correction horizontally
averaged over the whole model domain for temperature (<inline-formula><mml:math display="inline"><mml:msup><mml:mi/><mml:mo>∘</mml:mo></mml:msup></mml:math></inline-formula>C, top) and
salinity (bottom). Right panels: <inline-formula><mml:math display="inline"><mml:mi>z</mml:mi></mml:math></inline-formula> axis indicates depth, <inline-formula><mml:math display="inline"><mml:mi>x</mml:mi></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mi>y</mml:mi></mml:math></inline-formula> are
latitude and longitude respectively, the colours are the systematic error
correction averaged between 12 and 28 June at 100, 350 and 1000 m depth for
temperature (<inline-formula><mml:math display="inline"><mml:msup><mml:mi/><mml:mo>∘</mml:mo></mml:msup></mml:math></inline-formula>C, top) and salinity (bottom).</p></caption>
        <?xmltex \igopts{width=341.433071pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f07.png"/>

      </fig>

      <p>The temperature and salinity corrections due to the systematic error are
shown in Fig. 7. The panels on the left show how the vertical structure of
the systematic error, averaged over the entire domain, evolve during the
simulated period. In the right panels the maps of the systematic error
correction averaged between 12 and 28 June at 100, 350 and 1000 m depth are
shown. During the first 4 days the number of in situ observations increase
and the spatial coverage improves. The systematic error computation and thus
the corresponding correction is strongly affected by this observation
sampling error. The sampling error is particularly evident in the surface and
near-surface corrections (between the surface and 300 m depth), where scale
of horizontal variability is small, that oscillate between positive and
negative values. In the deeper layers, the amplitude of this oscillation is
significantly smaller. However, the overall effect of the correction after
4 days is to decrease the warm bias present in the deep temperature initial
conditions and to increase salinity content at intermediate depths. At the
end of the first cruise leg, 11 June, the systematic error stabilises. After
the initial shock due to the correction of the initial state, the systematic
error correction corrects errors due to the surface forcing, the lateral open
boundary condition and the inadequate model parameterisations.</p>
      <p>The combined analysis of vertical structures and horizontal maps supports some
inferences. A thin layer with negative temperature correction is present
between 5 and 15 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula>, the effect is to increase the stratification
above 15 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula> and decrease below. The model systematic error is clearly
due to the vertical diffusion; however, it is difficult to distinguish between
error in the surface forcing or in the vertical turbulence closure scheme
adopted. At 100 m depth the temperature correction is generally positive.
The corrections map at this depth shows minimum values along the boundaries
indicating that vertical mixing can be the source of the model failure. In
the deeper layer the correction is more stable over the integrated time, and
maps show maximum correction values close to the lateral open boundaries.
Thus, the adopted scheme is mostly acting on the external lateral forcing.</p>
      <p>The surface salinity field reacts slowly to surface forcing. Simulation
errors are mostly due to advection/diffusion processes. The salinity
corrections in the first 100 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula> are characterised by positive and
negative values partially due to the observation sampling and systematic
model error. At 100 m depth the scheme increases the salinity content of the
water masses along the southern open boundary with the exception of the
south-east corner where negative corrections are found. This is probably due
to a misplacement of the water masses present at this depth. At 300 m depth
the systematic error correction is generally positive and acts along the open
boundaries. The simultaneous analysis of the temperature corrections
indicates that warm and salty intermediate waters (the Modified Levantine
Intermediate Waters) are poorly represented in the nesting model. At deeper
layers the salinity correction is negative along the boundaries while it goes
toward zero in the centre of the model domain. We can argue that the vertical
stratification of the nested model is too weak and tends to mix intermediate
and deep water masses. The positive core in the centre of the model domain
suggests also that the choice of NEMO parameters for the vertical mixing is
not optimal, leading to the vertical diffusion of the salt introduced with
the data assimilation scheme.</p>
      <p>The impacts of the daily ensemble-based <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> and the bias correction
on the quality of the simulations are evaluated comparing observations with
model backgrounds.</p>
      <p>In order to fully assess the performance of each experiment the mean squared
error (MSE) is decomposed following Oke et al. (2002) and the single
components analysed:

              <disp-formula id="Ch1.Ex4"><mml:math display="block"><mml:mrow><mml:mtext>MSE</mml:mtext><mml:mo>=</mml:mo><mml:msup><mml:mtext>MB</mml:mtext><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mtext>SDE</mml:mtext><mml:mn mathvariant="normal">2</mml:mn></mml:msup><mml:mo>+</mml:mo><mml:mn mathvariant="normal">2</mml:mn><mml:msub><mml:mi>S</mml:mi><mml:mi mathvariant="normal">m</mml:mi></mml:msub><mml:msub><mml:mi>S</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mtext>CC</mml:mtext></mml:mfenced><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where MB is the model mean bias, SDE the standard deviation error,
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>S</mml:mi><mml:mi mathvariant="normal">m</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>S</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> are the modelled and observed standard
deviations and CC is the cross-correlation between modelled and observed
fields. The skill of each experiment, with respect to a reference experiment
(Exp-ref), is calculated based on the MSE. The skill score (SS; e.g. Murphy
and Epstein, 1989) is defined as

              <disp-formula id="Ch1.Ex5"><mml:math display="block"><mml:mrow><mml:mtext>SS</mml:mtext><mml:mo>=</mml:mo><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mtext>MSE</mml:mtext><mml:mrow><mml:msub><mml:mtext>MSE</mml:mtext><mml:mi mathvariant="normal">r</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where MSE<inline-formula><mml:math display="inline"><mml:msub><mml:mi/><mml:mi mathvariant="normal">r</mml:mi></mml:msub></mml:math></inline-formula> is the MSE of Exp-ref. The normalised (using the
observed standard deviation) root mean square error (RMSE), mean bias and
standard deviation error together with the cross-correlation and Skill Score
vertical profiles for temperature and salinity and for the four experiments
are shown in Figs. 8 and 9 respectively. In Table 3 the non-normalised
statistics vertically integrated at predefined layers are also listed. All
the statistics are computed using data and corresponding model backgrounds
collected during the second leg of the cruise, which started on 12 June and
ended 25 June 2014.</p>

<?xmltex \floatpos{t}?><table-wrap id="Ch1.T3" specific-use="star"><caption><p>Model bias (MB), standard deviation error (SDE), cross-correlation
(CC) and skill score (SS) for the different experiments and integrated
between different layers. For each quantity the best performing models is
highlighted in bold.</p></caption><oasis:table frame="topbot"><?xmltex \begin{scaleboxenv}{.82}[.82]?><oasis:tgroup cols="20">
     <oasis:colspec colnum="1" colname="col1" align="left"/>
     <oasis:colspec colnum="2" colname="col2" align="right"/>
     <oasis:colspec colnum="3" colname="col3" align="right"/>
     <oasis:colspec colnum="4" colname="col4" align="right"/>
     <oasis:colspec colnum="5" colname="col5" align="right"/>
     <oasis:colspec colnum="6" colname="col6" align="left"/>
     <oasis:colspec colnum="7" colname="col7" align="right"/>
     <oasis:colspec colnum="8" colname="col8" align="right"/>
     <oasis:colspec colnum="9" colname="col9" align="right"/>
     <oasis:colspec colnum="10" colname="col10" align="right"/>
     <oasis:colspec colnum="11" colname="col11" align="left"/>
     <oasis:colspec colnum="12" colname="col12" align="right"/>
     <oasis:colspec colnum="13" colname="col13" align="right"/>
     <oasis:colspec colnum="14" colname="col14" align="right"/>
     <oasis:colspec colnum="15" colname="col15" align="right"/>
     <oasis:colspec colnum="16" colname="col16" align="left"/>
     <oasis:colspec colnum="17" colname="col17" align="right"/>
     <oasis:colspec colnum="18" colname="col18" align="right"/>
     <oasis:colspec colnum="19" colname="col19" align="right"/>
     <oasis:colspec colnum="20" colname="col20" align="right"/>
     <oasis:thead>
       <oasis:row>  
         <oasis:entry colname="col1">Layer</oasis:entry>  
         <oasis:entry rowsep="1" namest="col2" nameend="col5" align="center">MB </oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry rowsep="1" namest="col7" nameend="col10" align="center">SDE </oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry rowsep="1" namest="col12" nameend="col15" align="center">CC </oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry rowsep="1" namest="col17" nameend="col20" align="center">SS </oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">  
         <oasis:entry colname="col1"/>  
         <oasis:entry colname="col2">ref</oasis:entry>  
         <oasis:entry colname="col3">Hy1</oasis:entry>  
         <oasis:entry colname="col4">Cl1</oasis:entry>  
         <oasis:entry colname="col5">Hy2</oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7">ref</oasis:entry>  
         <oasis:entry colname="col8">Hy1</oasis:entry>  
         <oasis:entry colname="col9">Cl1</oasis:entry>  
         <oasis:entry colname="col10">Hy2</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">ref</oasis:entry>  
         <oasis:entry colname="col13">Hy1</oasis:entry>  
         <oasis:entry colname="col14">Cl1</oasis:entry>  
         <oasis:entry colname="col15">Hy2</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">ref</oasis:entry>  
         <oasis:entry colname="col18">Hy1</oasis:entry>  
         <oasis:entry colname="col19">Cl1</oasis:entry>  
         <oasis:entry colname="col20">Hy2</oasis:entry>
       </oasis:row>
     </oasis:thead>
     <oasis:tbody>
       <oasis:row rowsep="1">  
         <oasis:entry namest="col1" nameend="col20" align="center">Temperature </oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">0–50</oasis:entry>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.05</oasis:entry>  
         <oasis:entry colname="col3"><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col4">0.11</oasis:entry>  
         <oasis:entry colname="col5">0.16</oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col9"><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.06</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.01</oasis:entry>  
         <oasis:entry colname="col13">0.15</oasis:entry>  
         <oasis:entry colname="col14">0.17</oasis:entry>  
         <oasis:entry colname="col15"><bold>0.22</bold></oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18">0.19</oasis:entry>  
         <oasis:entry colname="col19">0.07</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.21</bold></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">50–110</oasis:entry>  
         <oasis:entry colname="col2">0.10</oasis:entry>  
         <oasis:entry colname="col3">0.13</oasis:entry>  
         <oasis:entry colname="col4"><bold>0.03</bold></oasis:entry>  
         <oasis:entry colname="col5"><bold>0.03</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7">0.10</oasis:entry>  
         <oasis:entry colname="col8">0.16</oasis:entry>  
         <oasis:entry colname="col9">0.05</oasis:entry>  
         <oasis:entry colname="col10"><bold>0.03</bold></oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.19</oasis:entry>  
         <oasis:entry colname="col13"><bold>0.26</bold></oasis:entry>  
         <oasis:entry colname="col14">0.17</oasis:entry>  
         <oasis:entry colname="col15">0.19</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.50</oasis:entry>  
         <oasis:entry colname="col19">0.37</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.47</bold></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">110–215</oasis:entry>  
         <oasis:entry colname="col2">0.09</oasis:entry>  
         <oasis:entry colname="col3">0.08</oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.05</oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.03</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.04</bold></oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.04</bold></oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.06</oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.08</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.24</oasis:entry>  
         <oasis:entry colname="col13">0.17</oasis:entry>  
         <oasis:entry colname="col14"><bold>0.26</bold></oasis:entry>  
         <oasis:entry colname="col15">0.25</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.06</oasis:entry>  
         <oasis:entry colname="col19">0.20</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.29</bold></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">215–470</oasis:entry>  
         <oasis:entry colname="col2">0.06</oasis:entry>  
         <oasis:entry colname="col3"><bold>0.04</bold></oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.06</oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.04</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.12</bold></oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.12</bold></oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.12</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.14</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.07</oasis:entry>  
         <oasis:entry colname="col13">0.18</oasis:entry>  
         <oasis:entry colname="col14">0.21</oasis:entry>  
         <oasis:entry colname="col15"><bold>0.34</bold></oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18">0.11</oasis:entry>  
         <oasis:entry colname="col19">0.10</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.25</bold></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">470–930</oasis:entry>  
         <oasis:entry colname="col2">0.41</oasis:entry>  
         <oasis:entry colname="col3">0.40</oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.03</oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.10</oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.10</oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.04</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.04</oasis:entry>  
         <oasis:entry colname="col13">0.37</oasis:entry>  
         <oasis:entry colname="col14">0.46</oasis:entry>  
         <oasis:entry colname="col15"><bold>0.51</bold></oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18">0.11</oasis:entry>  
         <oasis:entry colname="col19">0.81</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.84</bold></oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">  
         <oasis:entry colname="col1">0—930</oasis:entry>  
         <oasis:entry colname="col2">0.23</oasis:entry>  
         <oasis:entry colname="col3">0.22</oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.03</oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.08</oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.08</oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.05</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.07</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.08</oasis:entry>  
         <oasis:entry colname="col13">0.27</oasis:entry>  
         <oasis:entry colname="col14">0.33</oasis:entry>  
         <oasis:entry colname="col15"><bold>0.39</bold></oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18">0.06</oasis:entry>  
         <oasis:entry colname="col19">0.47</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.55</bold></oasis:entry>
       </oasis:row>
       <oasis:row rowsep="1">  
         <oasis:entry namest="col1" nameend="col20" align="center">Salinity </oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">0–50</oasis:entry>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col3"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.02</oasis:entry>  
         <oasis:entry colname="col4"><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col5"><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.02</oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.02</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.23</oasis:entry>  
         <oasis:entry colname="col13">0.07</oasis:entry>  
         <oasis:entry colname="col14">0.36</oasis:entry>  
         <oasis:entry colname="col15"><bold>0.61</bold></oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.24</oasis:entry>  
         <oasis:entry colname="col19">0.28</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.52</bold></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">50–110</oasis:entry>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.01</oasis:entry>  
         <oasis:entry colname="col3"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.01</oasis:entry>  
         <oasis:entry colname="col4">0.01</oasis:entry>  
         <oasis:entry colname="col5"><bold>0.00</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7">0.02</oasis:entry>  
         <oasis:entry colname="col8">0.02</oasis:entry>  
         <oasis:entry colname="col9"><bold>0.00</bold></oasis:entry>  
         <oasis:entry colname="col10"><bold>0.00</bold></oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.14</oasis:entry>  
         <oasis:entry colname="col13">0.23</oasis:entry>  
         <oasis:entry colname="col14"><bold>0.64</bold></oasis:entry>  
         <oasis:entry colname="col15">0.62</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18">0.10</oasis:entry>  
         <oasis:entry colname="col19">0.64</oasis:entry>  
         <oasis:entry colname="col20"><bold>0.66</bold></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">110–215</oasis:entry>  
         <oasis:entry colname="col2">0.02</oasis:entry>  
         <oasis:entry colname="col3"><bold>0.00</bold></oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.01</oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.02</oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.14</oasis:entry>  
         <oasis:entry colname="col13"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.01</oasis:entry>  
         <oasis:entry colname="col14"><bold>0.19</bold></oasis:entry>  
         <oasis:entry colname="col15">0.09</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.04</oasis:entry>  
         <oasis:entry colname="col19"><bold>0.08</bold></oasis:entry>  
         <oasis:entry colname="col20">0.01</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">215–470</oasis:entry>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.10</oasis:entry>  
         <oasis:entry colname="col3"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.10</oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.04</oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.04</oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.03</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.04</oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.08</oasis:entry>  
         <oasis:entry colname="col13"><bold>0.20</bold></oasis:entry>  
         <oasis:entry colname="col14">0.09</oasis:entry>  
         <oasis:entry colname="col15">0.07</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.02</oasis:entry>  
         <oasis:entry colname="col19"><bold>0.62</bold></oasis:entry>  
         <oasis:entry colname="col20"><bold>0.62</bold></oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">470–930</oasis:entry>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.05</oasis:entry>  
         <oasis:entry colname="col3"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.05</oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.02</oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.02</oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.44</oasis:entry>  
         <oasis:entry colname="col13">0.49</oasis:entry>  
         <oasis:entry colname="col14"><bold>0.52</bold></oasis:entry>  
         <oasis:entry colname="col15">0.46</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18">0.03</oasis:entry>  
         <oasis:entry colname="col19"><bold>0.38</bold></oasis:entry>  
         <oasis:entry colname="col20">0.36</oasis:entry>
       </oasis:row>
       <oasis:row>  
         <oasis:entry colname="col1">0–930</oasis:entry>  
         <oasis:entry colname="col2"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.05</oasis:entry>  
         <oasis:entry colname="col3"><inline-formula><mml:math display="inline"><mml:mo>-</mml:mo></mml:math></inline-formula>0.05</oasis:entry>  
         <oasis:entry colname="col4"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col5"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.01</bold></oasis:entry>  
         <oasis:entry colname="col6"/>  
         <oasis:entry colname="col7"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col8"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col9"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col10"><inline-formula><mml:math display="inline"><mml:mo mathvariant="bold">-</mml:mo></mml:math></inline-formula><bold>0.02</bold></oasis:entry>  
         <oasis:entry colname="col11"/>  
         <oasis:entry colname="col12">0.27</oasis:entry>  
         <oasis:entry colname="col13">0.31</oasis:entry>  
         <oasis:entry colname="col14"><bold>0.36</bold></oasis:entry>  
         <oasis:entry colname="col15">0.33</oasis:entry>  
         <oasis:entry colname="col16"/>  
         <oasis:entry colname="col17">–</oasis:entry>  
         <oasis:entry colname="col18">0.00</oasis:entry>  
         <oasis:entry colname="col19"><bold>0.42</bold></oasis:entry>  
         <oasis:entry colname="col20"><bold>0.42</bold></oasis:entry>
       </oasis:row>
     </oasis:tbody>
   </oasis:tgroup><?xmltex \end{scaleboxenv}?></oasis:table></table-wrap>

      <p>The analysis of the single components of the model error allows us to
identify the effect of the bias correction procedure and the impact of the
daily, ensemble-based estimate of vertical covariance. The two simulations
without the bias correction (Exp-ref and Exp-Hy1) are characterised by a
similar vertical structure and values of temperature and salinity RMSE
(Figs. 8a and 9a for temperature and salinity respectively). Both the
simulations are characterised by a large temperature RMSE below 500 m depth,
while they both show a maximum in salinity RMSE at about 400 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula>. Large
errors are also observed in temperature between 60 and 200 m depth and in
salinity between 40 and 150 m depth. The vertical profiles of the mean bias
(Figs. 8b, 9b) clearly indicates that at intermediate and deep layers this
error structure is due to a large bias characterising the system (initial
state and lateral open boundary conditions). The near-surface RMSE maxima do
not have a clear correspondence in the mean bias structure. The bias nature
of this error is confirmed by the results obtained with the two experiments
where the bias correction has been applied. This also confirms that our
simple systematic error correction procedure is capable of significantly
reducing this bias. Both Exp-Hy2 and Exp-Cl2 RMSE and mean bias are
characterised by a nearly uniform and relatively small values. However, the
systematic error correction increases the temperature mean bias between 20
and 70 m depth, meaning that scales (both spatial and temporal), procedure
or observation sampling used are probably not adequate at these depths. On
the other hand, at similar depths, the systematic error correction reduces
the salinity mean bias (Fig. 9b). We argue that temperature and salinity
systematic errors in these layers have different length scales.</p>

      <?xmltex \floatpos{t}?><fig id="Ch1.F8"><caption><p>Vertical profiles of temperature error components and skill score
for the different experiments: black lines indicate the Exp-ref results, red
lines indicate Exp-Hy1 results, blue lines indicate Exp-Cl1 results and green
lines indicate Exp-Hy2 results. <bold>(a)</bold> Normalised root mean square
error. <bold>(b)</bold> Normalised mean bias. <bold>(c)</bold> Normalised standard
deviation error. <bold>(d)</bold> Cross-correlation. <bold>(e)</bold> Skill score.</p></caption>
        <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f08.png"/>

      </fig>

      <?xmltex \floatpos{t}?><fig id="Ch1.F9"><caption><p>As Fig. 8 but for salinity.</p></caption>
        <?xmltex \igopts{width=241.848425pt}?><graphic xlink:href="https://os.copernicus.org/articles/12/1137/2016/os-12-1137-2016-f09.png"/>

      </fig>

      <p>The standard deviation error indicates the capability of our system to
correctly reproduce the amplitude of the observed spatial/temporal
variability. Differences between climatological and daily estimates of the
background error covariance are evident. The usage of daily hybrid
<inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> without the bias correction introduces in the system a large
temperature standard deviation error between 60 and 150 m depth,
significantly larger than in Exp-Ref. It is interesting to note that the same
vertical error statistic (<inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>) when applied together with the bias
correction procedure (Exp-Hy2) reduces significantly the standard deviation
error at the same depths.</p>
      <p>The differences introduced by the daily, ensemble based, estimates of the
background vertical error covariance are evident analysing the
cross-correlation (Figs. 8d and 9d for temperature and salinity respectively)
and the skill scores (Figs. 8e, 9e). The Exp-Hy2 with systematic error
correction and daily estimate of the vertical error background covariance has
a temperature cross-correlation generally higher than the other experiments.
These differences are at a maximum between 20 and 80 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula> and below
250 m depth. On the other hand, in the salinity field the maximum
differences are observed near the surface (between 0 and 50 m depth) while
in the deeper layers Exp-Cl1 and Epx-Hy2 perform in a similar way. Both the
experiments with the bias correction show a decreased cross-correlation with
observed salinity between 200 and 400 m depth.</p>
      <p>The overall experiment statistics are listed in Table 3. Exp-Hy2 vertically
integrated temperature skill score is 55 %; 47 % is due to the
systematic error correction (Exp-Cl1 SS is 0.47), whereas the remaining part
is due to the introduction of the daily ensemble-based estimates of
<inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>. The simple introduction of the ensemble-based <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>
(Exp-Hy1) produces a worsening of the solution between 50 and 215 m depth.
We argue that the small structures introduced with the assimilation scheme
are not in balance with the surrounding environment and develop incorrect
dynamics; the correction of the systematic error allows the model to
incorporate the information provided.</p>
      <p>The vertically integrated salinity Exp-Hy2 and Exp-Cl1 skill scores are
42 %, although they have different vertical distributions. Exp-Hy2 strongly
outperforms Exp-Cl1 in the surface layers (0–50 m depth), while it does not
significantly improve the model solution between 110 and 215 m depth. In the
other layers the two systems have a similar performance. The improvements and
the worsening are both due to the cross-correlation between observations and
modelled salinities. This can be a consequence of the relatively small
ensemble size that has not adequately sampled the model error.</p>
</sec>
<sec id="Ch1.S5" sec-type="conclusions">
  <title>Summary and conclusions</title>
      <p>During June 2014 an extensive sea-trial
(Recognized Environmental Picture, REP14-MED) off the west coast of Sardinia
was conducted by CMRE (Centre for Maritime Research and Experimentation). Two
research vessels and a glider fleet collected a massive amount of data in an
area of approximatively 10 000 <inline-formula><mml:math display="inline"><mml:mrow><mml:msup><mml:mi mathvariant="normal">km</mml:mi><mml:mn mathvariant="normal">2</mml:mn></mml:msup></mml:mrow></mml:math></inline-formula>. Remote sensing data and
existing products from atmospheric and oceanographic operational models were
also collected as an additional observational data set and boundary
conditions, respectively.</p>
      <p>A Nucleus of European Modeling of the Ocean (NEMO; Madec, 2008)-based model
has been implemented in the area with a horizontal resolution of
approximatively 1 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">km</mml:mi></mml:math></inline-formula> and 91 hybrid vertical levels. The model has
been initialised and forced at the lateral open boundaries using
Mercator-Ocean (Drévillon et al., 2008) daily analyses, while atmospheric
forcing was computed by means of interactive bulk formulae (Oddo et
al., 2009) using the hourly operational products from the COSMO-ME
limited-area atmospheric model.</p>
      <p>In order to address the data assimilation issues characterising ocean
limited-area models with dense observational networks, a 3DVAR assimilation
scheme was implemented and coupled with the NEMO-based code. Following
Dobricic and Pinardi (2008) the present variational scheme decomposes the
background error covariance matrix (<inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>) in a sequence of linear
operators, each of them representing a specific component of the error
structure. Two main issues have been encountered in the present assimilation
exercise. The first is related to the small scales sampled by the dense
observation network, which are poorly represented in the traditionally
stationary vertical component of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula>. The second concerns the large
systematic errors partially introduced by the external forcing (experiment
initialisation, lateral or surface open boundary conditions) and partially
due to inadequate model physics. In order to overcome these limits and
improve the system, a variational-ensemble hybrid assimilation system has
been developed and implemented. A small size ensemble (14 members) has been
created by combining perturbation of observations and background-error
horizontal correlation radii in the <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> matrix. The choice of
creating the ensemble members by perturbing only the analyses is mostly
justified by the nature of the experiment we conducted. In fact a
perturbation in the model initialisation or model parameterisations would
require a relatively long integration time in order to fully develop and
reach a stable condition. On the other hand, perturbing the observations on a
daily assimilation system allows us to quickly generate ensemble statistics
with amplitudes similar to the model error. The statistical information
retrieved from the ensemble members has been used to address both the
small-scale and the systematic error issues. In order to improve the
representation of the small-scale error in the background error covariances,
the climatological-based <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> operator (accounting for the
multivariate vertical background error covariances) has been replaced with a
daily and spatially varying estimate computed by applying multivariate EOFs
analysis to the ensemble members' anomalies. Furthermore, the climatological
estimates of the recursive filter horizontal correlation radii used to model
the <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> operator have been substituted with daily estimates
computed from the ensemble statistics according Belo Pereira and Berre
(2006).</p>
      <p>To correct the systematic error the ensemble members' misfit statistics have
been used. For every simulated day an estimate of the systematic error has
been obtained by averaging the misfit over the ensemble members and assuming
that observational error and random model error have both zero mean. The
results have been mapped onto the model grid using a univariate objective
analysis (Barnes, 1994) and superimposed to the ensemble daily mean. At each
assimilation step the differences between the corrected ensemble mean and the
last available daily average corresponding fields have been filtered with a
low-pass filter with a 75 km length scale and the results superimposed to the
3DVAR corrections.</p>
      <p>The implementation of our strategy suffers from the need to empirically
choose the parameters associated with the combination of stationary and
ensemble-derived covariances and with the scales for the large-scale bias
estimation, which can both benefit from further tuning in the future.
However, these experiments represent a proof-of-concept for including
flow-dependent and large-scale aspects in a variational assimilation
framework.</p>
      <p>In order to test the validity of our hypothesis and to quantitatively
estimate the differences introduced with the hybrid-variational scheme
designed, the results of four different experiments have been compared.
Exp-ref uses the standard 3DVAR scheme with static and homogeneous
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> both computed using a 15-year Mediterranean
CMEMS reanalysis (Adani et al., 2011). In Exp-Hy1 the climatological
<inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">H</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> have been weighted with daily estimates
from the ensemble statistic, with <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>V</mml:mi><mml:mi mathvariant="normal">V</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> also varying spatially.
Exp-Cl1 uses the same <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> formulation of Exp-ref but the systematic
error correction procedure has been applied. Finally, Exp-Hy2 uses the same
<inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> formulation of Exp-Hy1 but the systematic error correction
procedure has also been applied. The simple introduction of the hybrid
estimate of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> does not significantly improve the model results.
This is probably due to the relatively small ensemble size and the amplitude
of the large-scale systematic error characterising our experiments. The
vertically integrated Skill Score of Exp-Hy1 with respect to Exp-ref is 0.06
for temperature and 0 for salinity, indicating improvements in temperature
mean error of 6 % and no improvements in the salinity field. However, a
significant worsening of the model temperature is observed between 50 and
100 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula>. The systematic error correction accounts for a large part of
the improvements and the ensemble-based estimates of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> produce the
best results when used in combination with the systematic error correction
(Exp-Hy2). We can argue that the small-scale corrections introduced with the
new formulation of <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> are not in balance with the surrounding
environment and thus not properly ingested into the model solution, thus
requiring the additional large-scale bias correction. In both Exp-Cl1 and
Exp-Hy2, the systematic error correction correctly reduces the large warm
bias affecting the temperature initial state and lateral open boundary
condition below 500 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula> and simultaneously removes the salinity error
at intermediate depth due to the absence in the external data of the correct
water masses at this depths. The adopted methodology seems to produce
satisfactory results. During the first days, with the observational data
availability increasing, the systematic error oscillates and finally adjusts
the errors associated to the initial conditions of the experiments. The
amplitude of the corrections during this initial phase is relatively large.
After the errors due to the initialisation have been reduced, the amplitude
of the systematic error correction significantly reduces and acts mostly on
the lateral open boundary conditions. The improvements in the Exp-Hy2 are
mostly due to improvements in the cross-correlation and thus to a better
reproduction of horizontal and vertical dynamics and structures. In terms of
SS for temperature, Exp-Hy2 performs best at all the depths, with an overall
improvements of 55 % with respect to Exp-Ref while the Exp-Cl1 vertically
integrated temperature improvement is 47 %. Large parts of differences
between Exp-Hy2 and Exp-Cl1 can be traced back to an improved
cross-correlation coefficient between modelled and observed values. The
salinity statistics show different model behaviours. The vertically
integrated SS are similar for Exp-Cl1 and Exp-Hy2, both improve Exp-Ref
results by about 42 %. However, the distribution of the error differs
significantly along the vertical. Exp-Hy2 outperform Exp-Cl2 in the first
100 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula> of the water column as consequence of larger cross-correlation
coefficient with observations, while Exp-Cl2 perform better in the
intermediate layers (between 110 and 215 <inline-formula><mml:math display="inline"><mml:mi mathvariant="normal">m</mml:mi></mml:math></inline-formula>). It should be noted that
the idea of using the current information from misfits or from the ensemble
to improve the analysis has been recently applied in several other studies
(Wang et al., 2007, 2008; Desroziers et al., 2006; Hamill and Snyder, 2000;
Etherton and Bishop, 2004). However, in all the methods, and also in schemes
presented and adopted in this paper the weights given to the climatological
and ensemble-based <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> estimates are arbitrary. The vertical
dependency of the hybrid system's performance suggests that the empirical
methods used in the estimates of the ensemble size and the relative weights
of static and hybrid <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> require a more objective and formal
approach. The two quantities are clearly correlated. In this study we used a
small ensemble size (14 members) and constant (both spatially and temporally)
weights obtaining, however, encouraging results. Recently, Dobricic et
al. (2015) overcame this issue by proposing a method based on Bayesian
hierarchical model where the
relative weights arise directly from the computations based on Bayes'
theorem.</p>
      <p>There are several possible future improvements of the hybrid variational
scheme method presented for estimating background error covariances.
Ménétrier and Auligné (2015) suggested a theoretical framework
where hybrid weights and parameters for the localisation of ensemble-derived
covariances are jointly optimised as a function of the ensemble size. An
alternative possibility may be to include the <inline-formula><mml:math display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> parameter in the
minimisation of the cost function obtaining an optimised and variable
relative weight.</p>
</sec>
<sec id="Ch1.S6">
  <title>Data availability</title>
      <p>Data used in this publication are property of the North
Atlantic Treaty Organization, who owns all rights, including intellectual
property rights. Data originated during REP14-MED experiment by NATO STO
Centre for Maritime Research and Experimentation (CMRE,
<uri>www.cmre.nato.int</uri>) and by experiment partners. For the distribution of
the data set please contact data@cmre.nato.int.</p><?xmltex \hack{\clearpage}?>
</sec>

      
      </body>
    <back><app-group>

<app id="App1.Ch1.S1">
  <title>Accounting for the hybrid formulation in the cost function</title>
      <p>We start from the cost function:

              <disp-formula id="App1.Ch1.E1" content-type="numbered"><mml:math display="block"><mml:mrow><?xmltex \hack{\hbox\bgroup\fontsize{9.5}{9.5}\selectfont$\displaystyle}?><mml:mi>J</mml:mi><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">B</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mo>.</mml:mo><?xmltex \hack{$\egroup}?></mml:mrow></mml:math></disp-formula>

        To define our hybrid assimilation schemes, we compute <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> as a linear
combination of the “static” covariance operator, <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>,
and the flow-dependent operator, <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>:

              <disp-formula id="App1.Ch1.E2" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="bold">B</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where <inline-formula><mml:math display="inline"><mml:mi mathvariant="italic">α</mml:mi></mml:math></inline-formula> is the relative weight. Substituting Eq. (A2) in Eq. (A1), we
obtain the new hybrid cost function:

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mi>J</mml:mi><mml:mo>(</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>)</mml:mo></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:mo mathsize="1.1em">(</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E3"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mo mathsize="1.1em">(</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo mathsize="1.1em">(</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo mathsize="1.1em">)</mml:mo><mml:mo>.</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          We define now the increment as a weighted sum of parts corresponding to
static and flow-dependent covariance matrices:

              <disp-formula id="App1.Ch1.Ex2"><mml:math display="block"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>=</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula>

        We want to demonstrate that

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mi>J</mml:mi><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mi>T</mml:mi></mml:msup><mml:mo mathsize="1.1em">(</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E4"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mo mathsize="1.1em">(</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo mathsize="1.1em">(</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo mathsize="1.1em">)</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          has the minimum for the same value of <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mrow></mml:math></inline-formula> as Eq. (A3).</p>
      <p>To minimise Eq. (A4), <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and
<inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> must satisfy <inline-formula><mml:math display="inline"><mml:mrow><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mrow><mml:mo>∂</mml:mo><mml:mi>J</mml:mi><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:mstyle displaystyle="false"><mml:mfrac style="text"><mml:mrow><mml:mo>∂</mml:mo><mml:mi>J</mml:mi><mml:mfenced open="(" close=")"><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mfenced></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn></mml:mrow></mml:math></inline-formula>, which gives

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mo>∂</mml:mo><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo mathsize="1.5em">(</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msubsup><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi><mml:mi>T</mml:mi></mml:msubsup><mml:mo mathsize="1.1em">(</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">)</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo mathsize="1.5em">)</mml:mo></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E5"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi>J</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mo mathsize="1.1em">[</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mo>∂</mml:mo><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo mathsize="1.5em">(</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo mathsize="1.5em">)</mml:mo></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E6"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi>J</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          where <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi>J</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> is the observational term. Assuming that
<inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> can be perturbed
independently, both the second terms on the left-hand side of Eqs. (A5) and
(A6) are null:

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mlabeledtr id="App1.Ch1.E7"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mo>∂</mml:mo><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mfenced close=")" open="("><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mi>T</mml:mi></mml:msup><mml:mo mathsize="1.5em">(</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.5em">)</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr><mml:mlabeledtr id="App1.Ch1.E8"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mo>∂</mml:mo><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mfenced close=")" open="("><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mi mathvariant="italic">δ</mml:mi><mml:msup><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mfenced><mml:mo>=</mml:mo><mml:mn mathvariant="normal">0</mml:mn><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          and

              <disp-formula id="App1.Ch1.E9" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi>J</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi>J</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>=</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi>J</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>=</mml:mo><mml:mn mathvariant="normal">2</mml:mn><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi></mml:mfenced><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula>

        This is a reasonable assumption, because the two random values are sampled
from different Gaussians. Although they are defined over the same space, one
is sampled from historical states, and the other from current forecasts.
Pre-multiplying Eq. (A5) by <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> and Eq. (A6) by
<inline-formula><mml:math display="inline"><mml:mrow><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula>, removing the null terms,
summing the two subsequent equations and applying Eq. (A9) yields

              <disp-formula id="App1.Ch1.E10" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mn mathvariant="normal">0</mml:mn><mml:mo>=</mml:mo><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mo>+</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mn mathvariant="normal">1</mml:mn><mml:mn mathvariant="normal">2</mml:mn></mml:mfrac></mml:mstyle><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo mathsize="1.1em">]</mml:mo><mml:mstyle displaystyle="true"><mml:mfrac style="display"><mml:mrow><mml:mo>∂</mml:mo><mml:msub><mml:mi>J</mml:mi><mml:mi mathvariant="normal">o</mml:mi></mml:msub></mml:mrow><mml:mrow><mml:mo>∂</mml:mo><mml:mi mathvariant="bold-italic">x</mml:mi></mml:mrow></mml:mfrac></mml:mstyle><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula>

        Multiplying Eq. (A10) by the inverse of the hybrid covariance:

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mn mathvariant="normal">0</mml:mn></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced open="(" close=")"><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E11"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mfenced open="(" close=")"><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mo>-</mml:mo><mml:mi mathvariant="bold-italic">d</mml:mi><mml:mo mathsize="1.1em">]</mml:mo><mml:mo>.</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          This is also the minimum of Eq. (A3) that we wanted as a proof.</p>
      <p>Furthermore, defining the background and analysis perturbations around the
true state <inline-formula><mml:math display="inline"><mml:mrow><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub></mml:mrow></mml:math></inline-formula> as

              <disp-formula id="App1.Ch1.E12" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub></mml:mrow></mml:math></disp-formula>

        and

              <disp-formula id="App1.Ch1.E13" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        by adding and subtracting the true state (A11) becomes

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mn mathvariant="normal">0</mml:mn></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced close=")" open="("><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub></mml:mfenced></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E14"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:mfenced open="(" close=")"><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub></mml:mfenced><mml:mo>-</mml:mo><mml:mfenced open="(" close=")"><mml:mi>y</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced><mml:mo mathsize="1.1em">]</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          or

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd><mml:mrow><mml:mn mathvariant="normal">0</mml:mn></mml:mrow></mml:mtd><mml:mtd><mml:mrow><mml:mo>=</mml:mo><mml:msup><mml:mfenced open="[" close="]"><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E15"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced close="]" open="["><mml:mi mathvariant="bold">H</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub><mml:mo>-</mml:mo><mml:mfenced open="(" close=")"><mml:mi>y</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub></mml:mfenced></mml:mfenced><mml:mo>,</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          which can also be written as

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mfenced open="{" close="}"><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E16"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace linebreak="nobreak" width="0.25em"/><?xmltex \hack{\hbox\bgroup\fontsize{9}{9}\selectfont$\displaystyle}?><mml:mo>=</mml:mo><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mfenced close="]" open="["><mml:mi>y</mml:mi><mml:mo>-</mml:mo><mml:mi mathvariant="bold">H</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">t</mml:mi></mml:msub></mml:mfenced><mml:mo>.</mml:mo><?xmltex \hack{$\egroup}?></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          Multiplying each side of Eq. (A12) by its transpose, taking the expectation,
assuming that observational errors are independent of background errors,

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mfenced open="{" close="}"><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mi mathvariant="bold">A</mml:mi></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace width="0.25em" linebreak="nobreak"/><mml:msup><mml:mfenced close="}" open="{"><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mi>T</mml:mi></mml:msup></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mo>=</mml:mo><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi>E</mml:mi><mml:mfenced open="{" close="}"><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced><mml:mi>T</mml:mi></mml:msup></mml:mfenced></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E17"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mi>T</mml:mi></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:msup><mml:mi mathvariant="bold">RR</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mo>.</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          Assuming the <inline-formula><mml:math display="inline"><mml:mi mathvariant="bold">B</mml:mi></mml:math></inline-formula> contains the true background error covariances,
i.e. well-specified background errors, and using Eq. (A2):

              <disp-formula id="App1.Ch1.E18" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi>E</mml:mi><mml:mfenced close="}" open="{"><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub><mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">b</mml:mi></mml:msub></mml:mfenced><mml:mi mathvariant="normal">T</mml:mi></mml:msup></mml:mfenced><mml:mo>=</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo>.</mml:mo></mml:mrow></mml:math></disp-formula>

        Thus,

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mfenced close="}" open="{"><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mi mathvariant="bold">A</mml:mi></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace linebreak="nobreak" width="0.25em"/><mml:msup><mml:mfenced close="}" open="{"><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mi>T</mml:mi></mml:msup></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mo>=</mml:mo><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:mo mathsize="1.1em">]</mml:mo></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E19"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mspace linebreak="nobreak" width="0.25em"/><mml:mo mathsize="1.1em">[</mml:mo><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub><mml:msup><mml:mo mathsize="1.1em">]</mml:mo><mml:mrow><mml:mo>-</mml:mo><mml:mi>T</mml:mi></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:msup><mml:mi mathvariant="bold">RR</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          or

              <disp-formula specific-use="align" content-type="numbered"><mml:math display="block"><mml:mtable displaystyle="true"><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mfenced open="{" close="}"><mml:msup><mml:mfenced open="[" close="]"><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mi mathvariant="bold">A</mml:mi></mml:mrow></mml:mtd></mml:mtr><mml:mtr><mml:mtd/><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mspace width="0.25em" linebreak="nobreak"/><mml:msup><mml:mfenced open="{" close="}"><mml:msup><mml:mfenced close="]" open="["><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mi>T</mml:mi></mml:msup></mml:mrow></mml:mtd></mml:mtr><mml:mlabeledtr id="App1.Ch1.E20"><mml:mtd/><mml:mtd/><mml:mtd><mml:mrow><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mspace width="0.25em" linebreak="nobreak"/><mml:mo>=</mml:mo><mml:msup><mml:mfenced close="]" open="["><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mo>.</mml:mo></mml:mrow></mml:mtd></mml:mlabeledtr></mml:mtable></mml:math></disp-formula>

          Dividing by <inline-formula><mml:math display="inline"><mml:mrow><mml:mfenced close="}" open="{"><mml:msup><mml:mfenced close="]" open="["><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced close=")" open="("><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced></mml:mrow></mml:math></inline-formula>,

              <disp-formula id="App1.Ch1.E21" content-type="numbered"><mml:math display="block"><mml:mrow><mml:mi mathvariant="bold">A</mml:mi><mml:mo>=</mml:mo><mml:msup><mml:mfenced close="}" open="{"><mml:msup><mml:mfenced close="]" open="["><mml:mi mathvariant="italic">α</mml:mi><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">c</mml:mi></mml:msub><mml:mo>+</mml:mo><mml:mfenced open="(" close=")"><mml:mn mathvariant="normal">1</mml:mn><mml:mo>-</mml:mo><mml:mi mathvariant="italic">α</mml:mi></mml:mfenced><mml:msub><mml:mi mathvariant="bold">B</mml:mi><mml:mi mathvariant="normal">e</mml:mi></mml:msub></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:msup><mml:mi mathvariant="bold">H</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:msup><mml:mi mathvariant="bold">R</mml:mi><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mi mathvariant="bold">H</mml:mi></mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mn mathvariant="normal">1</mml:mn></mml:mrow></mml:msup><mml:mo>,</mml:mo></mml:mrow></mml:math></disp-formula>

        where <inline-formula><mml:math display="inline"><mml:mrow><mml:mi mathvariant="bold">A</mml:mi><mml:mo>=</mml:mo><mml:mi>E</mml:mi><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub><mml:msup><mml:mfenced close=")" open="("><mml:mi mathvariant="italic">δ</mml:mi><mml:msub><mml:mi mathvariant="bold-italic">x</mml:mi><mml:mi mathvariant="normal">a</mml:mi></mml:msub></mml:mfenced><mml:mi>T</mml:mi></mml:msup></mml:mrow></mml:math></inline-formula> is the analysis error covariance matrix. Equation (A15)
demonstrates that independent forecast updates in each ensemble member by
using Eq. (A4) give the same optimal estimate of updated covariances as
Eq. (A3).</p><?xmltex \hack{\clearpage}?>
</app>
  </app-group><ack><title>Acknowledgements</title><p>CMRE activities have been supported by the NATO Allied Command Transformation
(ACT) through the contract SAC000404. The authors wish to thank masters and
crews of NR/V <italic>Alliance</italic> (CMRE) and R/V <italic>Planet</italic> (German
Ministry of Defence), as well as the many colleagues, both from CMRE and the
Wehrtechnische Dienststelle für Schiffe und Marinewaffen, Maritime
Technologie und Forschung (WTD71; Germany), which contributed to provide the
assimilated CTD and glider (3 from CMRE and 1 from WTD71) data. In particular
the authors wish to thank Heinz-Volker Fiekas, scientist in charge on-board
the R/V <italic>Planet</italic>, and Michaela Knoll, who managed the R/V
<italic>Planet</italic> CTD data set. <?xmltex \hack{\newline}?><?xmltex \hack{\newline}?>
Edited by: J. Chiggiato <?xmltex \hack{\newline}?>
Reviewed by: P. Sakov and one anonymous referee</p></ack><ref-list>
    <title>References</title>

      <ref id="bib1.bib1"><label>1</label><mixed-citation>
Adani, M., Dobricic, S., and Pinardi, N.: Quality assessment of a 1985–2007
Mediterranean Sea reanalysis, J. Atmos. Ocean. Tech., 28, 569–589, 2011.</mixed-citation></ref>
      <ref id="bib1.bib2"><label>2</label><mixed-citation>
Barnes, S. L.: Applications of the Barnes objective analysis scheme. Part I:
effect of undersampling, wave positions, and stations randomness, J. Atmos.
Ocean. Tech., 11, 1433–1448, 1994.</mixed-citation></ref>
      <ref id="bib1.bib3"><label>3</label><mixed-citation>
Belo Pereira, M. and Berre, L.: The use of an ensemble approach to study the
background error covariances in a global NWP model, Mon. Weather Rev., 134,
2466–2489, 2006.</mixed-citation></ref>
      <ref id="bib1.bib4"><label>4</label><mixed-citation>
Bloom, S. C., Takacs, L. L., Da Silva, A. M., and Ledvina, D.: Data
assimilation using incremental analysis updates, Mon. Weather Rev., 124,
1256–1271, 1996.</mixed-citation></ref>
      <ref id="bib1.bib5"><label>5</label><mixed-citation>
Burgers, G., van Leeuwen, P. J., and Evensen, G.: Analysis scheme in the
ensemble Kalman filter, Mon. Weather Rev., 126, 1719–1724, 1998.</mixed-citation></ref>
      <ref id="bib1.bib6"><label>6</label><mixed-citation>
Canuto, V. M., Howard, A., Cheng, Y., and Dubovikov, M. S.: Ocean turbulence.
Part I: One-point closure model momentum and heat vertical diffusivities,
J. Phys. Oceanogr., 31 1413–1426, 2001.</mixed-citation></ref>
      <ref id="bib1.bib7"><label>7</label><mixed-citation>
Dee, D. P.: Bias and data assimilation, Q. J. Roy. Meteor. Soc., 131,
3323–3343, 2005.</mixed-citation></ref>
      <ref id="bib1.bib8"><label>8</label><mixed-citation>
Desroziers, G., Berre, L., Chapnik, B., and Poli, P.: Diagnosis of
observation, background- and analysis-error statistics in observation space,
Q. J. Roy. Meteor. Soc., 131, 3385–3396, 2006.</mixed-citation></ref>
      <ref id="bib1.bib9"><label>9</label><mixed-citation>
Dobricic, S. and Pinardi, N.: An oceanographic three-dimensional variational
data assimilation scheme, Ocean Model., 22, 89–105, 2008.</mixed-citation></ref>
      <ref id="bib1.bib10"><label>10</label><mixed-citation>Dobricic, S., Wikl, C. K., Milliff, R. F., Pinardi, N., and Berliner, L. M.:
Assimilation of oceanographic observations with estimates of vertical
background-error covariances by a Bayesian hierarchical model, Q. J. Roy.
Meteor. Soc., 141, 182–194, <ext-link xlink:href="http://dx.doi.org/10.1002/qj.2348" ext-link-type="DOI">10.1002/qj.2348</ext-link>, 2015.</mixed-citation></ref>
      <ref id="bib1.bib11"><label>11</label><mixed-citation>
Drévillon, M., Bourdallé-Badie, R., Derval, C., Drillet, Y.,
Lellouche, J. M., Rémy, E., Tranchant, B., Benkiran, M., Greiner, E.,
Guinehut, S., Verbrugge, N., Garric, G., Testut, C. E., Laborie, M., Nouel,
L., Bahurel, P., Bricaud, C., Crosnier, L., Dombrosky, E., Durand, E., Ferry,
N., Hernandez, F., Le Galloudec, O., Messal, F., and Parent, L.: The
GODAE/Mercator-Ocean global ocean forecasting system: results, applications
and prospects, Journal of Operational Oceanography, 1, 51–57, 2008.</mixed-citation></ref>
      <ref id="bib1.bib12"><label>12</label><mixed-citation>
Etherton, B. J. and Bishop, C. H.: Resilience of hybrid ensemble/3DVAR
analysis schemes to model error and ensemble covariance error, Mon. Weather
Rev., 132, 1065–1080, 2004.</mixed-citation></ref>
      <ref id="bib1.bib13"><label>13</label><mixed-citation>
Evensen, G.: Sequential data assimilation with a nonlinear quasigeostrophic
model using Monte Carlo methods to forecast error statistics, J. Geophys.
Res., 99, 10143–10162, 1994.</mixed-citation></ref>
      <ref id="bib1.bib14"><label>14</label><mixed-citation>
Evensen, G.: The Ensemble Kalman Filter: Theoretical Formulation and
Practical Implementation, Ocean Dynam., 53, 343–367, 2003.</mixed-citation></ref>
      <ref id="bib1.bib15"><label>15</label><mixed-citation>
Flather, R. A.: A tidal model of the northwest European continental shelf,
Mémoires de la Société Royale des Sciences de Liège, 6,
141–164, 1976.</mixed-citation></ref>
      <ref id="bib1.bib16"><label>16</label><mixed-citation>
Guidard, V. and Fischer, C.: Introducing the coupling information in a
limited-area variational assimilation, Q. J. Roy. Meteor. Soc., 134,
723–735, 2008.</mixed-citation></ref>
      <ref id="bib1.bib17"><label>17</label><mixed-citation>
Hamill, T. M. and Snyder, C.: A hybrid ensemble Kalman filter-3D variational
analysis scheme, Mon. Weather Rev., 128, 2905–2919, 2000.</mixed-citation></ref>
      <ref id="bib1.bib18"><label>18</label><mixed-citation>
Li, Z., McWilliams, J. C., Ide, K., and Fararra, J. D.: A Multi-Scale Data
Assimilation Scheme: Formulation and Illustration, Mon. Weather Rev., 143,
3804–3822, 2015.</mixed-citation></ref>
      <ref id="bib1.bib19"><label>19</label><mixed-citation>
Lorenc, A. C.: The potential of the ensemble Kalman filter for NWP –
A comparison with 4D-VAR, Q. J. Roy. Meteor. Soc., 129, 3183–3203, 2003.</mixed-citation></ref>
      <ref id="bib1.bib20"><label>20</label><mixed-citation>
Madec, G.: NEMO ocean engine, Note du Pole de modélisation, Institut
Pierre-Simon Laplace (IPSL), France, No. 27, 2008.</mixed-citation></ref>
      <ref id="bib1.bib21"><label>21</label><mixed-citation>
Ménétrier, B. and Auligné, T.: Optimized localization and
hybridization to filter ensemble-based covariances, Mon. Weather Rev., 143,
3931–3947, 2015.</mixed-citation></ref>
      <ref id="bib1.bib22"><label>22</label><mixed-citation>Mirouze, I., Blockley, E. W., Lea, D. J., Martin, M. J., and Bell, M. J.: A
multiple length scale correlation operator for ocean data assimilation,
Tellus A, 68, 29744, <ext-link xlink:href="http://dx.doi.org/10.3402/tellusa.v68.29744" ext-link-type="DOI">10.3402/tellusa.v68.29744</ext-link>, 2016.</mixed-citation></ref>
      <ref id="bib1.bib23"><label>23</label><mixed-citation>
Murphy, A. H. and Epstein, E. S.: Skill scores and correlation coefficients
in model verification, Mon. Weather Rev., 119, 572–581, 1989.</mixed-citation></ref>
      <ref id="bib1.bib24"><label>24</label><mixed-citation>Oddo, P., Adani, M., Pinardi, N., Fratianni, C., Tonani, M., and Pettenuzzo,
D.: A nested Atlantic-Mediterranean Sea general circulation model for
operational forecasting, Ocean Sci., 5, 461–473,
<ext-link xlink:href="http://dx.doi.org/10.5194/os-5-461-2009" ext-link-type="DOI">10.5194/os-5-461-2009</ext-link>, 2009.</mixed-citation></ref>
      <ref id="bib1.bib25"><label>25</label><mixed-citation>
Oke, P. R. and Sakov, P.: Representation error of oceanic observations for
data assimilation, J. Atmospheric Ocean. Tech., 25, 1004–1017, 2008.</mixed-citation></ref>
      <ref id="bib1.bib26"><label>26</label><mixed-citation>
Oke, P. R., Allen, J. S., Miller, R. N., Egbert, G. D., Austin, J. A., Barth,
J. A., Boyd, T. J., Kosro, P. M., and Levine, M. D.: A modelling study of the
three-dimensional continental shelf circulation off Oregon. Part I:
Model-Data Comparison, J. Phys. Oceanogr., 32, 1360–1382, 2002.</mixed-citation></ref>
      <ref id="bib1.bib27"><label>27</label><mixed-citation>
Pannekoucke, O. and Massart, S.: Estimation of the local diffusion tensor and
normalization for heterogeneous correlation modelling using a diffusion
equation, Q. J. Roy. Meteor. Soc., 134, 1425–1438, 2008.</mixed-citation></ref>
      <ref id="bib1.bib28"><label>28</label><mixed-citation>Penny, S., Behringer, D., Carton, J., and Kalnay, E.: A Hybrid Global Ocean
Data Assimilation System at NCEP, Mon. Weather Rev., 143, 4660–4677,
<ext-link xlink:href="http://dx.doi.org/10.1175/MWR-D-14-00376.1" ext-link-type="DOI">10.1175/MWR-D-14-00376.1</ext-link>, 2015.</mixed-citation></ref>
      <ref id="bib1.bib29"><label>29</label><mixed-citation>
Siddorn, J. and Furner, R.: An analytical stretching function that combines
the best attributes of geopotential and terrain-following vertical
coordinates, Ocean Model., 66, 1–13, 2013.</mixed-citation></ref>
      <ref id="bib1.bib30"><label>30</label><mixed-citation>Schlatter, T. W., Carr, F. H., Langland, R. H., Carbone, R. E., Crook, N. A.,
Daley, R. W., Derber, J. C., and Mullen, S. L.: A five-year plan for research
related to the assimilation of meteorological data, NCAR Technical Note
NCAR/TN-443, <ext-link xlink:href="http://dx.doi.org/10.5065/D65T3HFX" ext-link-type="DOI">10.5065/D65T3HFX</ext-link>, 1999.</mixed-citation></ref>
      <ref id="bib1.bib31"><label>31</label><mixed-citation>
Wang, X., Hamill, T. M., Whitaker, J. S., and Bishop, C. H.: A comparison of
hybrid ensemble transform Kalman filter-OI and ensemble square-root filter
analysis schemes, Mon. Weather Rev., 135, 1055–1076, 2007.</mixed-citation></ref>
      <ref id="bib1.bib32"><label>32</label><mixed-citation>Wang, X., Barker, D. M., Snyder, C., and Hamill, T. M.: A hybrid ETKF–3DVAR
data assimilation scheme for the WRF Model. Part II: Real observation
experiments, Mon. Weather Rev., 136, 5132–5147, 2008.
 </mixed-citation></ref><?xmltex \hack{\newpage}?>
      <ref id="bib1.bib33"><label>33</label><mixed-citation>
Wang, X., Hamill, T. M., Whitaker, J. S., and Bishop, C. H.: A comparison of
the hybrid and EnSRF analysis schemes in the presence of model error due to
unresolved scales, Mon. Weather Rev., 137, 3219–3232, 2009.</mixed-citation></ref>
      <ref id="bib1.bib34"><label>34</label><mixed-citation>
Weaver, A. and Courtier, P.: Correlation modeling on a sphere using a
generalized diffusion equation, Q. J. Roy. Meteor. Soc., 127, 1815–1846,
2001.</mixed-citation></ref>
      <ref id="bib1.bib35"><label>35</label><mixed-citation>
Yin, Y., Alves, O., and Oke, P.: An ensemble ocean data assimilation system
for seasonal prediction, Mon. Weather Rev., 139, 786–808, 2011.</mixed-citation></ref>
      <ref id="bib1.bib36"><label>36</label><mixed-citation>
Zalesak, S. T.: Fully multidimensional flux corrected transport for fluids,
J. Comput. Phys., 31, 335–362, 1979.</mixed-citation></ref>

  </ref-list><app-group content-type="float"><app><title/>

    </app></app-group></back>
    <!--<article-title-html>A hybrid variational-ensemble data assimilation scheme with systematic error
correction for limited-area ocean models</article-title-html>
<abstract-html><p class="p">A hybrid variational-ensemble data assimilation scheme to estimate the
vertical and horizontal parts of the background error covariance matrix for
an ocean variational data assimilation system is presented and tested in a
limited-area ocean model implemented in the western Mediterranean Sea. An
extensive data set collected during the Recognized Environmental Picture
Experiments conducted in June 2014 by the Centre for Maritime Research and
Experimentation has been used for assimilation and validation. The hybrid
scheme is used to both correct the systematic error introduced in the system
from the external forcing (initialisation, lateral and surface open boundary
conditions) and model parameterisation, and improve the representation of
small-scale errors in the background error covariance matrix. An ensemble
system is run offline for further use in the hybrid scheme, generated through
perturbation of assimilated observations. Results of four different
experiments have been compared. The reference experiment uses the classical
stationary formulation of the background error covariance matrix and has no
systematic error correction. The other three experiments account for, or not,
systematic error correction and hybrid background error covariance matrix
combining the static and the ensemble-derived errors of the day. Results show
that the hybrid scheme when used in conjunction with the systematic error
correction reduces the mean absolute error of temperature and salinity misfit
by 55 and 42 % respectively, versus statistics arising from standard
climatological covariances without systematic error correction.</p></abstract-html>
<ref-html id="bib1.bib1"><label>1</label><mixed-citation>
Adani, M., Dobricic, S., and Pinardi, N.: Quality assessment of a 1985–2007
Mediterranean Sea reanalysis, J. Atmos. Ocean. Tech., 28, 569–589, 2011.
</mixed-citation></ref-html>
<ref-html id="bib1.bib2"><label>2</label><mixed-citation>
Barnes, S. L.: Applications of the Barnes objective analysis scheme. Part I:
effect of undersampling, wave positions, and stations randomness, J. Atmos.
Ocean. Tech., 11, 1433–1448, 1994.
</mixed-citation></ref-html>
<ref-html id="bib1.bib3"><label>3</label><mixed-citation>
Belo Pereira, M. and Berre, L.: The use of an ensemble approach to study the
background error covariances in a global NWP model, Mon. Weather Rev., 134,
2466–2489, 2006.
</mixed-citation></ref-html>
<ref-html id="bib1.bib4"><label>4</label><mixed-citation>
Bloom, S. C., Takacs, L. L., Da Silva, A. M., and Ledvina, D.: Data
assimilation using incremental analysis updates, Mon. Weather Rev., 124,
1256–1271, 1996.
</mixed-citation></ref-html>
<ref-html id="bib1.bib5"><label>5</label><mixed-citation>
Burgers, G., van Leeuwen, P. J., and Evensen, G.: Analysis scheme in the
ensemble Kalman filter, Mon. Weather Rev., 126, 1719–1724, 1998.
</mixed-citation></ref-html>
<ref-html id="bib1.bib6"><label>6</label><mixed-citation>
Canuto, V. M., Howard, A., Cheng, Y., and Dubovikov, M. S.: Ocean turbulence.
Part I: One-point closure model momentum and heat vertical diffusivities,
J. Phys. Oceanogr., 31 1413–1426, 2001.
</mixed-citation></ref-html>
<ref-html id="bib1.bib7"><label>7</label><mixed-citation>
Dee, D. P.: Bias and data assimilation, Q. J. Roy. Meteor. Soc., 131,
3323–3343, 2005.
</mixed-citation></ref-html>
<ref-html id="bib1.bib8"><label>8</label><mixed-citation>
Desroziers, G., Berre, L., Chapnik, B., and Poli, P.: Diagnosis of
observation, background- and analysis-error statistics in observation space,
Q. J. Roy. Meteor. Soc., 131, 3385–3396, 2006.
</mixed-citation></ref-html>
<ref-html id="bib1.bib9"><label>9</label><mixed-citation>
Dobricic, S. and Pinardi, N.: An oceanographic three-dimensional variational
data assimilation scheme, Ocean Model., 22, 89–105, 2008.
</mixed-citation></ref-html>
<ref-html id="bib1.bib10"><label>10</label><mixed-citation>
Dobricic, S., Wikl, C. K., Milliff, R. F., Pinardi, N., and Berliner, L. M.:
Assimilation of oceanographic observations with estimates of vertical
background-error covariances by a Bayesian hierarchical model, Q. J. Roy.
Meteor. Soc., 141, 182–194, <a href="http://dx.doi.org/10.1002/qj.2348" target="_blank">doi:10.1002/qj.2348</a>, 2015.
</mixed-citation></ref-html>
<ref-html id="bib1.bib11"><label>11</label><mixed-citation>
Drévillon, M., Bourdallé-Badie, R., Derval, C., Drillet, Y.,
Lellouche, J. M., Rémy, E., Tranchant, B., Benkiran, M., Greiner, E.,
Guinehut, S., Verbrugge, N., Garric, G., Testut, C. E., Laborie, M., Nouel,
L., Bahurel, P., Bricaud, C., Crosnier, L., Dombrosky, E., Durand, E., Ferry,
N., Hernandez, F., Le Galloudec, O., Messal, F., and Parent, L.: The
GODAE/Mercator-Ocean global ocean forecasting system: results, applications
and prospects, Journal of Operational Oceanography, 1, 51–57, 2008.
</mixed-citation></ref-html>
<ref-html id="bib1.bib12"><label>12</label><mixed-citation>
Etherton, B. J. and Bishop, C. H.: Resilience of hybrid ensemble/3DVAR
analysis schemes to model error and ensemble covariance error, Mon. Weather
Rev., 132, 1065–1080, 2004.
</mixed-citation></ref-html>
<ref-html id="bib1.bib13"><label>13</label><mixed-citation>
Evensen, G.: Sequential data assimilation with a nonlinear quasigeostrophic
model using Monte Carlo methods to forecast error statistics, J. Geophys.
Res., 99, 10143–10162, 1994.
</mixed-citation></ref-html>
<ref-html id="bib1.bib14"><label>14</label><mixed-citation>
Evensen, G.: The Ensemble Kalman Filter: Theoretical Formulation and
Practical Implementation, Ocean Dynam., 53, 343–367, 2003.
</mixed-citation></ref-html>
<ref-html id="bib1.bib15"><label>15</label><mixed-citation>
Flather, R. A.: A tidal model of the northwest European continental shelf,
Mémoires de la Société Royale des Sciences de Liège, 6,
141–164, 1976.
</mixed-citation></ref-html>
<ref-html id="bib1.bib16"><label>16</label><mixed-citation>
Guidard, V. and Fischer, C.: Introducing the coupling information in a
limited-area variational assimilation, Q. J. Roy. Meteor. Soc., 134,
723–735, 2008.
</mixed-citation></ref-html>
<ref-html id="bib1.bib17"><label>17</label><mixed-citation>
Hamill, T. M. and Snyder, C.: A hybrid ensemble Kalman filter-3D variational
analysis scheme, Mon. Weather Rev., 128, 2905–2919, 2000.
</mixed-citation></ref-html>
<ref-html id="bib1.bib18"><label>18</label><mixed-citation>
Li, Z., McWilliams, J. C., Ide, K., and Fararra, J. D.: A Multi-Scale Data
Assimilation Scheme: Formulation and Illustration, Mon. Weather Rev., 143,
3804–3822, 2015.
</mixed-citation></ref-html>
<ref-html id="bib1.bib19"><label>19</label><mixed-citation>
Lorenc, A. C.: The potential of the ensemble Kalman filter for NWP –
A comparison with 4D-VAR, Q. J. Roy. Meteor. Soc., 129, 3183–3203, 2003.
</mixed-citation></ref-html>
<ref-html id="bib1.bib20"><label>20</label><mixed-citation>
Madec, G.: NEMO ocean engine, Note du Pole de modélisation, Institut
Pierre-Simon Laplace (IPSL), France, No. 27, 2008.
</mixed-citation></ref-html>
<ref-html id="bib1.bib21"><label>21</label><mixed-citation>
Ménétrier, B. and Auligné, T.: Optimized localization and
hybridization to filter ensemble-based covariances, Mon. Weather Rev., 143,
3931–3947, 2015.
</mixed-citation></ref-html>
<ref-html id="bib1.bib22"><label>22</label><mixed-citation>
Mirouze, I., Blockley, E. W., Lea, D. J., Martin, M. J., and Bell, M. J.: A
multiple length scale correlation operator for ocean data assimilation,
Tellus A, 68, 29744, <a href="http://dx.doi.org/10.3402/tellusa.v68.29744" target="_blank">doi:10.3402/tellusa.v68.29744</a>, 2016.
</mixed-citation></ref-html>
<ref-html id="bib1.bib23"><label>23</label><mixed-citation>
Murphy, A. H. and Epstein, E. S.: Skill scores and correlation coefficients
in model verification, Mon. Weather Rev., 119, 572–581, 1989.
</mixed-citation></ref-html>
<ref-html id="bib1.bib24"><label>24</label><mixed-citation>
Oddo, P., Adani, M., Pinardi, N., Fratianni, C., Tonani, M., and Pettenuzzo,
D.: A nested Atlantic-Mediterranean Sea general circulation model for
operational forecasting, Ocean Sci., 5, 461–473,
<a href="http://dx.doi.org/10.5194/os-5-461-2009" target="_blank">doi:10.5194/os-5-461-2009</a>, 2009.
</mixed-citation></ref-html>
<ref-html id="bib1.bib25"><label>25</label><mixed-citation>
Oke, P. R. and Sakov, P.: Representation error of oceanic observations for
data assimilation, J. Atmospheric Ocean. Tech., 25, 1004–1017, 2008.
</mixed-citation></ref-html>
<ref-html id="bib1.bib26"><label>26</label><mixed-citation>
Oke, P. R., Allen, J. S., Miller, R. N., Egbert, G. D., Austin, J. A., Barth,
J. A., Boyd, T. J., Kosro, P. M., and Levine, M. D.: A modelling study of the
three-dimensional continental shelf circulation off Oregon. Part I:
Model-Data Comparison, J. Phys. Oceanogr., 32, 1360–1382, 2002.
</mixed-citation></ref-html>
<ref-html id="bib1.bib27"><label>27</label><mixed-citation>
Pannekoucke, O. and Massart, S.: Estimation of the local diffusion tensor and
normalization for heterogeneous correlation modelling using a diffusion
equation, Q. J. Roy. Meteor. Soc., 134, 1425–1438, 2008.
</mixed-citation></ref-html>
<ref-html id="bib1.bib28"><label>28</label><mixed-citation>
Penny, S., Behringer, D., Carton, J., and Kalnay, E.: A Hybrid Global Ocean
Data Assimilation System at NCEP, Mon. Weather Rev., 143, 4660–4677,
<a href="http://dx.doi.org/10.1175/MWR-D-14-00376.1" target="_blank">doi:10.1175/MWR-D-14-00376.1</a>, 2015.
</mixed-citation></ref-html>
<ref-html id="bib1.bib29"><label>29</label><mixed-citation>
Siddorn, J. and Furner, R.: An analytical stretching function that combines
the best attributes of geopotential and terrain-following vertical
coordinates, Ocean Model., 66, 1–13, 2013.
</mixed-citation></ref-html>
<ref-html id="bib1.bib30"><label>30</label><mixed-citation>
Schlatter, T. W., Carr, F. H., Langland, R. H., Carbone, R. E., Crook, N. A.,
Daley, R. W., Derber, J. C., and Mullen, S. L.: A five-year plan for research
related to the assimilation of meteorological data, NCAR Technical Note
NCAR/TN-443, <a href="http://dx.doi.org/10.5065/D65T3HFX" target="_blank">doi:10.5065/D65T3HFX</a>, 1999.
</mixed-citation></ref-html>
<ref-html id="bib1.bib31"><label>31</label><mixed-citation>
Wang, X., Hamill, T. M., Whitaker, J. S., and Bishop, C. H.: A comparison of
hybrid ensemble transform Kalman filter-OI and ensemble square-root filter
analysis schemes, Mon. Weather Rev., 135, 1055–1076, 2007.
</mixed-citation></ref-html>
<ref-html id="bib1.bib32"><label>32</label><mixed-citation>
Wang, X., Barker, D. M., Snyder, C., and Hamill, T. M.: A hybrid ETKF–3DVAR
data assimilation scheme for the WRF Model. Part II: Real observation
experiments, Mon. Weather Rev., 136, 5132–5147, 2008.

</mixed-citation></ref-html>
<ref-html id="bib1.bib33"><label>33</label><mixed-citation>
Wang, X., Hamill, T. M., Whitaker, J. S., and Bishop, C. H.: A comparison of
the hybrid and EnSRF analysis schemes in the presence of model error due to
unresolved scales, Mon. Weather Rev., 137, 3219–3232, 2009.
</mixed-citation></ref-html>
<ref-html id="bib1.bib34"><label>34</label><mixed-citation>
Weaver, A. and Courtier, P.: Correlation modeling on a sphere using a
generalized diffusion equation, Q. J. Roy. Meteor. Soc., 127, 1815–1846,
2001.
</mixed-citation></ref-html>
<ref-html id="bib1.bib35"><label>35</label><mixed-citation>
Yin, Y., Alves, O., and Oke, P.: An ensemble ocean data assimilation system
for seasonal prediction, Mon. Weather Rev., 139, 786–808, 2011.
</mixed-citation></ref-html>
<ref-html id="bib1.bib36"><label>36</label><mixed-citation>
Zalesak, S. T.: Fully multidimensional flux corrected transport for fluids,
J. Comput. Phys., 31, 335–362, 1979.
</mixed-citation></ref-html>--></article>
