Sustainable management and utilisation of the North-west European Shelf (NWS) seas could benefit from reliable forecasts of the marine environment on monthly to seasonal timescales. Recent advances in global seasonal forecast systems and regional marine reanalyses for the NWS allow us to investigate the potential for seasonal forecasts of the state of the NWS.
We identify three possible approaches to address this issue: (A) basing NWS seasonal forecasts directly on output from the Met Office's GloSea5 global seasonal forecast system; (B) developing empirical downscaling relationships between large-scale climate drivers predicted by GloSea5 and the state of the NWS; and (C) dynamically downscaling GloSea5 using a regional model. We show that the GloSea5 system can be inadequate for simulating the NWS directly (approach A). We explore empirical relationships between the winter North Atlantic Oscillation (NAO) and NWS variables estimated using a regional reanalysis (approach B). We find some statistically significant relationships and present a skillful prototype seasonal forecast for English Channel sea surface temperature.
We find large-scale relationships between inter-annual variability in the boundary conditions and inter-annual variability modelled on the shelf, suggesting that dynamic downscaling may be possible (approach C). We also show that for some variables there are opposing mechanisms correlated with the NAO, for which dynamic downscaling may improve on the skill possible with empirical forecasts. We conclude that there is potential for the development of reliable seasonal forecasts for the NWS and consider the research priorities for their development.
The works published in this journal are distributed under the Creative Commons Attribution 4.0 License. This licence does not affect the Crown copyright work, which is re-usable under the Open Government Licence (OGL). The Creative Commons Attribution 4.0 License and the OGL are interoperable and do not conflict with, reduce, or limit each other. © Crown copyright 2018
The North-west European Shelf (NWS) seas are of wide economic,
environmental,
and political importance. They support many ecosystem services and human
activities, including fisheries, energy extraction and transmission (both
renewable and non-renewable), shipping, and waste removal. Most of these
services and activities are sensitive to the variable environmental
conditions under which they operate. Examples include the following:
shipping (transport and industrial) and offshore oil, gas, and renewable
operations are sensitive to wind and wave conditions and currents; the capacity of sea-floor gas distribution networks is sensitive to bottom
temperature (with capacity decreased in cold conditions when demand is
likely to be highest); commercial and recreational fisheries are sensitive to large-scale primary
production and the seasonal evolution of the marine food web, which in turn
are sensitive to surface temperature, salinity, and seasonal
re-stratification (see e.g. Fernandes, 2015); some specific commercial species have a life cycle or food web which is
sensitive to near-bottom temperature (see e.g. Dulvy et al., 2008; Jones
et al., 2015; Pinnegar et al., 2017; Pörtner and Farrell, 2008;
Pörtner and Peck, 2010); and coastal installation operations may be sensitive to local flooding (surge
events), sea temperatures (e.g. ambient temperatures for power station
cooling), and consequential local ecosystem impacts (Brown et al.,
2016; Dawson et al., 2016; Schneider et al., 2015).
Because of these sensitivities, the ability to predict variations in the
marine environment is of great potential value for marine operations,
management, planning, and conservation. Weather and wave forecasting are of
course well-established tools, and more recently operational forecasting of
marine environmental variables such as temperature, salinity, and currents
for a few days ahead has matured to the extent that daily forecasts are now
widely and freely available (e.g. for European regional seas through the
Copernicus Marine Environment Monitoring Service;
One can envisage many potential applications for marine predictions at this extended range. For example, the setting and management of fishing quotas for maximum sustainable yield could take account of likely conditions for recruitment in that year, energy producers and suppliers could anticipate winter seasons with higher than usual demand and/or stress on the offshore supply network, and environmental regulators could use early warning in regions at increased risk of water quality issues to target monitoring resources in those regions. But first we must demonstrate that a useful level of predictive skill is achievable on these timescales. Of course, on the 3–6-month timescale, environmental variations are dominated by the seasonal cycle, and this is normally factored into existing decision-making processes. Here we are asking whether we can predict anomalies relative to the average seasonal cycle (e.g. unusually cold winters or unusually stormy autumns).
Recent progress in climate modelling and prediction has shown that skillful predictions are possible several months ahead for some key elements of European climate. For example, Scaife et al. (2014) show substantial skill in predicting the winter North Atlantic Oscillation (NAO) index 1–4 months ahead. There is even significant skill in predicting the winter NAO index 1 year ahead (Dunstone et al., 2016). The NAO is a key determinant of the nature of the European winter (e.g. Hurrell, 1995) with a positive NAO index indicating mild, wet conditions and a negative NAO index indicating cold, dry conditions. This predictive skill has been exploited to demonstrate predictability in a number of user-relevant variables, e.g. relating the phase of the NAO index to the likely number of transport disruption (road, rail, and air) impacts (Palin et al., 2016), but so far there has been less attention to seasonal predictability of marine variables (e.g. Hobday et al., 2016).
To address the problem of seasonal prediction for the NWS we must consider
how to derive marine variables from seasonal climate predictions. The
seasonal predictions are produced from global ocean–atmosphere climate
models which are initialised with observed conditions and run forward in
time through the 1–6-month forecast period. Because the climate is not
perfectly predictable on these timescales an ensemble of runs is used to
derive a probabilistic prediction. To turn this into a prediction for NWS
marine variables, three possible approaches can be envisaged:
read off the NWS marine variables (temperature, salinity, currents) directly
from the underlying climate model; use observations to derive empirical relationships between large-scale
climate indices (e.g. NAO index) and the marine quantity of interest, for
instance for a fishery metric like cod recruitment
(Engelhard et al., 2014; Stige et al., 2006) or squid abundance
(van der Kooij et al., 2016), then input the forecast climate index
to the empirical model to get a forecast of the quantity of interest
(“empirical downscaling”); and use the marine and atmospheric variables predicted by the climate model as
inputs to a higher-resolution regional model of the NWS (“dynamical
downscaling”).
Approach A is appealingly simple but is unlikely to be successful. The ocean
resolution of the global seasonal forecast system GloSea5 used by Scaife et
al. (2014) is around 20 km in the NWS region, with 18 vertical levels in the
top 50 m. While this is high by the standards of current global seasonal
prediction systems it is still insufficient to resolve many features of the
NWS circulation. Further, in common with most global climate models the
GloSea5 system does not represent key shelf sea processes such as tidal
mixing. While this has been shown to be a successful approach in other
regions (e.g. by Hobday et al., 2016, in Australia) we will
show later that approach A can
be inapplicable to the NWS under some conditions.
A number of scientific and technical developments have recently come together to enable us, for the first time, to assess the potential for seasonal forecasts for the NWS: first is the demonstration of the skill of global seasonal forecasting systems in predicting key European climate indices several months ahead (e.g. Scaife et al., 2014); second is the development of regional oceanographic models of the NWS proven for use in operational prediction (O'Dea et al., 2012); and third is the combination of regional NWS models with historical observations to produce a consistent estimate of the time-varying state of the NWS over recent decades (“regional NWS reanalysis”; Wakelin et al., 2014). The NWS regional model allows us to investigate the prerequisites of the dynamical downscaling approach C, while the reanalysis allows us to investigate in detail the mechanisms of year-to-year variability and therefore evaluate which elements of the NWS state are likely to be predictable. For the mean climate, the dynamical downscaling approach C has been shown to add value to the simple approach A of reading off variables from the underlying global climate model (e.g. Mathis et al., 2013).
In this study we use the above building blocks to evaluate the potential of
seasonal predictability for the NWS. We address the following questions in
turn.
How well does the global seasonal prediction system GloSea5 represent the
state and inter-annual variability of the NWS (approach A)? Can a regional reanalysis adequately represent the inter-annual variability
of the NWS? If so, can we use it as a proxy to understand mechanisms of
variability in the real world? Do the predictable climate indices (e.g. NAO index) provide actual
predictive skill for the NWS (approach B)? What are the prospects for improving NWS seasonal forecasts (including
prerequisites for approach C)?
From the answers to 1–4, we conclude by assessing the prospects for
seasonal prediction on the NWS and the likely pros and cons of the direct
approach A and the empirical and dynamical downscaling approaches B and
C, and we suggest near-term priorities for applications and research.
Our aim in this paper is to draw evidence on the prospects of seasonal forecasting from the literature and from our own research. We recognise that some aspects will not come as a surprise to some readers; nonetheless, we believe that an overall perspective has not been presented before.
We use data from two modelling systems (NWS reanalysis (CO5) and GloSea5) and several observation datasets. Firstly, we describe and compare the two modelling systems before introducing the observation datasets and the analysis techniques.
The Met Office provides a reanalysis of the NW European Shelf seas to the Copernicus Marine Environment Monitoring Service: this has been extensively described and validated (O'Dea et al., 2012; Wakelin et al., 2014), but here we give a brief overview.
The NWS reanalysis is based on the NEMO coastal ocean model version 5 (CO5)
implementation (Fig. 1). This is on a
regional 7 km grid extending from 40
The use of CORE bulk formulae will have a few important implications for
this study. For example, the bulk formulae assume that the surface air
temperature has an infinite heat reservoir, so the SST will tend to
follow the surface air temperature rather than the opposite, which occurs in
reality. This will affect the relationship between the SST and surface air
temperature. Furthermore, the CORE bulk formulae receive a prescribed downward
component of shortwave and longwave radiation and calculate the
upward component internally using modelled SST. Therefore, while SST data
assimilation will improve the SST and affect upward radiation,
downward radiation is not affected by the SST assimilation. In Sect. 3.3
and 3.4 of this study, we consider correlations of shelf sea variables with
the prescribed
The GloSea5 seasonal forecast system is described in detail by MacLachlan et
al. (2014). Here we give a brief description and
focus on the difference between its ocean component and that of the CO5 NWS
reanalysis. GloSea5 is based on the Met Office Hadley Centre climate model
HadGEM3-GC2 (Williams et al., 2015).
This is a coupled climate model combining the MetUM atmosphere model
(N216,
To make seasonal forecasts, a set of GloSea5 simulations are run to form a forecast ensemble and re-forecast (hindcast) ensemble. Every day, two ensemble members are initialised and run forward for 216 days. The previous 3 weeks are combined into a 42-member ensemble to make a 6-month forecast, which is updated weekly. A re-forecast ensemble is also run every week (with the same modelling system) to correct bias and drift in the forecasts. This includes four start dates for the relevant month for each of the previous 23 years run forward for 216 days. The forecast ensemble is used to predict how the following 6 months will compare to this climatology. The system is also run as a continuous reanalysis (the GloSea5 ocean and sea ice global reanalysis) from 1990–2015 to provide initial conditions for the ocean component of the hindcast ensemble. The atmospheric initial conditions for the forecast are taken from the Met Office operational weather forecast system, and the hindcast atmosphere is initialised from ERA-Interim.
GloSea5 shows improved year-to-year predictions of the major modes of variability compared to the previous system (GloSea4; Arribas et al., 2011). Predictions of the El Niño–Southern Oscillation are improved with reduced errors in the western Pacific. GloSea5 shows unprecedented levels of forecast skill and reliability for both the North Atlantic Oscillation and the Arctic Oscillation (MacLachlan et al., 2014).
The GloSea5 ocean and sea ice global reanalysis and CO5 NWS reanalysis differ in a number of ways. Both rely on the ocean model NEMO, but run with grids of different horizontal and vertical resolutions. The GloSea5 global reanalysis system is a coupled global model system designed to capture the key components of the global climate system in order to make a seasonal forecast, having used data assimilation over a wide range of variables to constrain the model. Conversely, the CO5 NWS reanalysis is a regional reanalysis in which a higher-resolution ocean model is (one-way) forced from ERA-Interim atmosphere forcing using SST-only data assimilation. Effectively, the CO5 NWS reanalysis has a higher resolution and better representation of the NWS physics, whereas the GloSea5 global reanalysis has global scope and assimilates a wider range of observations. Both have high enough resolution to include an open Dover Strait, allowing a route for Atlantic water into the southern North Sea, which is important for simulating the local seasonal cycle of salinity.
The CO5 NWS reanalysis requires lateral boundary conditions (GloSea5 is a global model system and therefore does not need them) which are taken from FOAM before 1990 and from the GloSea5 reanalysis thereafter. The change from FOAM to the GloSea5 reanalysis leads to a discontinuity in the lateral boundary conditions that is important for variables such as subsurface temperature and salinity in the open ocean. While neither subsurface temperature nor salinity are considered in this study, they may influence the NWS properties (such as sea surface salinity (SSS) and SST) that are considered.
An important difference between the two systems is that the CO5 NWS reanalysis is a shelf seas model that includes all the key shelf sea processes, whereas the ORCA025, being a global model, neglects some regionally important processes, including dynamic tides. Tides are particularly important in this region, as the NWS contributes significantly to the global total tidal energy dissipation (Egbert and Ray, 2001). While tides are modelled directly within the CO5 NWS reanalysis, tidal mixing is parameterised in GloSea5. This parameterisation (Simmons et al., 2004) is based on a climatology of turbulence associated with internal tide wave breaking and steep bathymetry (which is very low on the NWS), and thus tidal mixing is much less than in the CO5 NWS reanalysis.
There are also key differences in the riverine forcings and the treatment
of the Baltic Sea. The CO5 NWS reanalysis uses river forcings (with
inter-annual variability) from the E-HYPE river model (which gives too much
discharge) and treats the Baltic as an open boundary where
Much of the evaluation of the CO5 NWS reanalysis (Wakelin et al., 2014) focused on the mean state of the model. Here we are more interested in the modelled temporal variability and therefore undertake additional evaluation. We compare the model to limited observed time series to assess its performance at replicating several observed events. Here we describe the observed time series.
Ferries are well-established vessels of opportunity for oceanographic measurements taking regular long-term samples of surface water while the ship is on passage between ports (Bean et al., 2017). Observations can be in the form of samples taken by crew for subsequent testing in a laboratory or more sophisticated “ferry boxes” as packages of instruments that semi-autonomously monitor temperature, salinity, and other water properties. We use the monthly salinity data from the ferry on the Harwich to Hook of Holland route (see “S” in Fig. 1a), which took quasi-weekly temperature and salinity samples at nine standard stations between 1971 and 2012 (Joyce, 2006) reported in the ICES Report on Ocean Climate (Larsen et al., 2016) and MCCIP Report Cards (Dye et al., 2013). We use point time series from this dataset to compare to the model.
The Western Channel Observatory (WCO) is an oceanographic time series in the
western English Channel (Smyth et al., 2015). In situ
measurements are undertaken fortnightly at open shelf station E1
(50.03
We use annual mean time series data from three stations used in the ICES
Report on Ocean Climate (González-Pola et al., 2018): Malin Head weather
station, Fair Isle, and Helgoland Roads. Malin Head SST (55.37
The North Atlantic Oscillation (NAO) is a climatic phenomenon in the North Atlantic Ocean of fluctuations in the difference of atmospheric pressure at sea level between the Icelandic low and the Azores high. These fluctuations control the strength and direction of westerly winds and storm tracks across Europe (Hurrell, 1995).
We use the NOAA National Weather Service Climate Prediction Center NAO data
(
When analysing the relationships of shelf salinity we find correlation
patterns which suggest that storm track latitude may be important. We therefore
analysed the mean sea level pressure data to produce a storm track latitude
index, following a method adapted from Lowe et al. (2009). The 3-hourly ERA-Interim mean sea level
pressure data from all modelled latitudes at 2
We use regional mean time series of model output from the reanalysis adapted from the region mask from Wakelin et al. (2012) (Fig. 1a): sea surface temperature and salinity (SST, SSS), near-bed temperature and salinity (NBT, NBS), and the difference between surface and near-bed for temperature and salinity (DFT, DFS). We calculated monthly, seasonal, and annual means from these time series. In addition to model output, we extract regional mean time series for the ERAI surface forcings, the GloSea5 ocean lateral boundary conditions (using the masks in Fig. 1b–d), and the E-HYPE river forcings.
Important relationships between shelf drivers and the shelf response are identified by comparing time series. Noting that a statistically significant correlation does not imply a causal relationship, the spatial patterns of the correlation coefficients are used to help interpret the underlying mechanisms behind the correlations. Some possible mechanisms are described, but it is considered beyond the scope of this study to undertake sensitivity studies to explore any mechanisms in detail.
The region mask of Wakelin et al. (2012)
(Fig. 1a) is used to create regional mean
time series of the results from the NWS reanalysis, including its
atmospheric (downward radiative fluxes, surface air temperature and relative
humidity, total precipitation, mean sea level pressure, wind magnitude) and
riverine forcings. The oceanic
Model and observed time series are compared to one another with Pearson's correlations, and their significance is noted at the 95 % confidence level. Typically, we compare the annual mean time series, but we also compare at the seasonal timescale. We also investigate lag correlations between the shelf response, possible drivers, and climate indices. For example, the time series of DJF NAO will be correlated with the DJF SST across the shelf (at 0-month lag). The DJF will then be compared to the JFM SST (January–March; 1-month lag) and FMA (February–April; 2-month lag; e.g. Fig. 6).
For consistency we have used the same region mask (e.g. Wakelin et al., 2012) for the river forcings as for the shelf sea variables and surface atmospheric forcings. However, this mask was not designed for rivers and several regions must therefore be treated with care or excluded. For example, the northern North Sea region combines the river flow from small sections of the Scottish and Danish coasts, which does not make sense; other regions to be excluded are the central North Sea, Shetland shelf region, and the North Atlantic regions. Other regions combine river flow from different coasts, but in a more sensible manner; for example, the English Channel and Irish Sea regions combine river input from two coasts, but due to the smaller enclosed nature of these regions, this is sensible in terms of local salinity. In the modelling system, the rivers do not have a specified temperature and therefore assume the local temperature when they reach the sea. Therefore, rivers predominantly affect salinity, with only secondary temperature effects (associated with changes in density-driven circulation and stability). As increased river flow reduces the local salinity, most correlations are expected to be negative.
Both GloSea5 and the CO5 NWS reanalysis assimilate similar SST observations,
so their simulated patterns of SST are relatively similar and in
agreement with observations. When looking at the near-bed temperatures (NBT)
and the difference between the surface and bed temperatures (DFT), there are
important differences between GloSea5 and the NWS reanalysis. DFT is an
important diagnostic of stratification; when DFT
A time series showing the seasonal cycle of the areal extent of the
stratification (from a 20-year mean) for both GloSea5 and the CO5 NWS
reanalysis is presented in Fig. 2a. The CO5 NWS
reanalysis has been shown to have a generally good representation of the
seasonal stratification for which independent observations are available
(O'Dea et al., 2012). While both the CO5 NWS
reanalysis and GloSea5 show that the NWS is fully mixed in the winter
(effectively no grid boxes are stratified), in GloSea5 more grid boxes are
stratified in the summer than in the NWS reanalysis; from April to
September, the stratified area of the shelf is
Further evidence is shown under question 2 below that using GloSea5 NWS fields directly can be problematic. We therefore conclude that this approach (approach A) may not be appropriate under some conditions, in which case some form of (dynamic or empirical) downscaling of the GloSea5 fields may be needed to generate reliable NWS forecasts.
While the ability of the CO5 NWS reanalysis to simulate the mean state of the NWS is thoroughly evaluated (O'Dea et al., 2012; Wakelin et al., 2014), its ability to simulate inter-annual variability has received less attention. Evaluation requires long observed time series, preferably of variables that are not assimilated into the reanalysis. Here we focus on five locations across the NWS: the Harwich to Hook of Holland ferry route in the southern North Sea; the Western Channel Observatory (WCO) in the English Channel; the Malin Head weather station north of Ireland; the Fair Isle time series (north-east of Scotland); and the Helgoland Roads time series in the southern North Sea. All these datasets are described in Sect. 2.4.
First, we compare the observed time series of surface salinity in the
southern North Sea (from the ferry route) to that from GloSea5 and the CO5 NWS reanalysis (Fig. 3, both from
the nearest model grid box). The time series exhibits multi-year oscillations
and these are well simulated by the NWS reanalysis (
Time series of observed (ferry samples) southern North Sea salinity (red) compared to CO5 NWS reanalysis (black) and GloSea5 (blue) data as a 2-year running mean. Both modelled time series are from the nearest model grid box to the observations.
WCO E1 observed (black) temperature (surface: SST; 30 m:
Secondly, we compare the observed WCO temperature and salinity profiles to
the daily mean of the nearest CO5 NWS reanalysis grid box
(Fig. 4). The WCO observations are not
assimilated into the NWS reanalysis, but the SST from complementary
satellite products is. Unsurprisingly, the CO5 NWS reanalysis SST is in
close agreement with the WCO observations for both the seasonal cycle and
the year-to-year variations at the surface and at depth (30 m). The
inter-annual variability is well captured in SST for all seasons (
The ICES dataset includes annual mean data from three very different sites (Fig. 5). As we are most interested in the modelled variability, we evaluate against the anomalies (time series minus mean) and consider the standard deviations and correlations.
The Malin Head time series shows a good agreement between the reanalysis and
observed SST, with a significant Pearson's correlation of
The Fair Isle time series monitors the properties of one of the main
pathways of Atlantic water into the North Sea, so evaluation is important to
lend credibility to the results of this study. Here the observations are for
the upper 100 m, so the data assimilation is less dominant in this
comparison. As with the Malin Head data, there is a very good agreement with
the observations, with a significant correlation of
Reanalysis evaluation with ICES temperature and salinity data. Data
are presented as anomalies (the time series mean is removed) to highlight the
variability (reanalysis red; observations black); however, the mean and
standard deviation (along with Pearson's correlation coefficient,
The Helgoland Roads time series shows an excellent temperature agreement (
There are two other ICES time series in the vicinity of the North Sea, Utsira B, within the Norwegian Trench, and the Norwegian site Svinøy, off the shelf to the north of the Norwegian Trench, both in regions which are known to be difficult to model, and outside the main focus of the reanalysis. Both showed much lower modelled salinity variability, and low correlations.
Further evaluation of the CO5 NWS reanalysis against other long time series is planned through the Copernicus NWS regional Marine Forecasting Centre. These initial results suggest that the reanalysis can provide valuable information on inter-annual variability in the NWS, where there is a strong signal.
We now investigate empirical forecasts based on the response of the CMEMS reanalysis to the observed NAO that are then applied to the GloSea5 forecast NAO.
Correlation maps between winter (DJF) NAO and atmospheric surface
fields associated with shelf temperature (surface air temperature, SAT:
In the literature, there are many empirical relationships between climate indices and various physical and biological responses. The CMEMS reanalysis (through data assimilation) combines observations with models to give the best possible state estimate of the NWS and therefore provides a powerful tool to develop such relationships. We focus on the winter NAO, as it is an important source of year-to-year variability in the NWS, and GloSea5 has predictive skill for it. By investigating relationships between the CMEMS reanalysis fields and the observed (NOAA) NAO and then considering how these relationships change when we use the GloSea5 forecast NAO, we can explore the empirical approach to NWS seasonal forecasting. We note that many of the relationships we find between the NAO and the NWS are not new and are underlain by published relationships (e.g. Becker and Pauly, 1996; Dippner, 1997; Hurrell and Deser, 2009; Hurrell and Van Loon, 1997).
As Fig. 6, but for atmospheric surface fields associated with SSS
variability: mean sea level pressure (MSLP;
First we focus on shelf temperature and restrict our analysis to the surface forcing that we consider important for shelf temperatures. We investigate the correlations (and lagged correlations) between the winter (DJF) NAO and this subset of surface forcing (Fig. 6). We find a positive correlation of the NAO with the DJF surface air temperature (consistent with Hurrell and Van Loon, 1997) (Fig. 6a) and humidity (not shown), and this persists for 1 month (to January–March) in most regions (SAT in Fig. 6b). This reflects one of the main characteristics of a NAO positive winter: warmer, wetter winters over northern Europe, with cooler, drier winters over southern Europe. There is a significant negative correlation between DJF NAO and the incoming solar radiation for shelf regions west of the UK (SSRD in Fig. 6g), although this only persists to a significant level in the Irish shelf (Fig. 6h). DJF NAO is strongly positively correlated to the downward component of thermal (longwave) radiation for most shelf regions west of the UK and for the Norwegian Trench (NT), and this persists for 3 months in some regions (English Channel) (STRD in Fig. 6d–f). Under NAO negative winters, there are more cold, clear days with greater downward solar radiation compared to NAO positive years; hence the negative correlations (between NAO and downward solar radiation). During positive NAO winters, the greater cloud cover reduces the downward solar radiation and increases the downward thermal radiation from clouds and water vapour, thus supporting the positive correlations between the NAO and downward thermal radiation. Note that the SST assimilation increments will affect the upward radiation fluxes (which are calculated from the modelled SST), but, as the reanalysis system is an ocean-only uncoupled system, the downward component of the radiation (which we consider here) is prescribed (by the ERA-Interim data), so these correlations are not affected by this assimilation.
We now consider the surface forcings we think are likely to be important for shelf salinity (Fig. 7). The DJF NAO is strongly correlated with the DJF 10 m wind magnitude (UV10; defined as the magnitude, i.e. wind speed irrespective of direction, of the 10 m wind) across the domain (consistent with Hurrell and Deser, 2009), and this persists into the third month (March–May) for the southern and central North Sea (UV10 in Fig. 7d–f). The correlation between winter (DJF) NAO acts in opposite ways for winter (DJF) mean sea level pressure (MSLP) and for total precipitation (not shown). The DJF mean sea level pressure (total precipitation) is negatively (positively) correlated with DJF NAO in the northern regions and positively (negatively) correlated in the southern regions. These correlations persist for a few months in some regions (Fig. 7a–c) and are consistent with the correlations of Hurrell and Deser (2009). The observed north–south gradient in correlations between MSLP and NAO reflects the pressure gradient nature of the NAO. The stronger winds associated with the NAO positive phase lead to positive correlations with wind across the domain. The north–south dipole in wetter and/or drier weather (between northern and southern Europe) leads to the north–south gradient in correlation between NAO and total precipitation.
As Fig. 6, but for the shelf response itself: SST
River systems can give additional predictability by continuing to respond after the forcing or can reduce predictability by having such long response times that they act as a low-pass filter. Furthermore, different river catchment areas are located in different climate regimes and can therefore respond in different ways, which can further complicate the response. The river run-off forcings in the Norwegian Trench (Fig. 7g–i) are highly correlated with the DJF NAO, and this persists until the following summer (July–September, not shown). The run-off in regions influenced by northern British and Irish rivers (e.g. Irish shelf) is also positively correlated with the NAO and shows persistence beyond the winter season. The regions which include much of the European coast (Armorican shelf, English Channel, and southern North Sea) show little correlation of run-off with the NAO, perhaps reflecting the larger catchments not having time to respond to the NAO and their more southerly location. The river correlation patterns were consistent with those of Rödel (2006) and Bouwer et al. (2008).
As with Fig. 7, but using the GloSea5 forecast DJF NAO:
SST
Having shown the correlations of the NAO with the important surface forcings, we now look directly at the relationship between the observed DJF NAO and the shelf response (Fig. 8). We find a significant positive correlation between the winter NAO and the winter SST in the most southern and eastern shelf regions (Fig. 8a), consistent with previous studies (Dippner, 1997; Hurrell and Van Loon, 1997). In most of these regions, the significance of these correlation (at the 95 % confidence level) persists for 1 month (Fig. 8b), and in the English Channel and southern North Sea a second month (February–April (FMA) SST; Fig. 8c). The NBT correlations also show significant correlations with the NAO and have memory in some regions. The DJF NAO is generally not significantly correlated with SSS (sea surface salinity) in most regions, although there is a significant correlation in the Skagerrak–Kattegat, which persists until FMA (Fig. 8g–i). There could be a much lower-frequency salinity response to the NAO (e.g. Belkin et al., 1998; Mysak et al., 1996), which would not be captured by these correlations. Such a response may provide predictability on longer timescales.
The above results suggest that knowledge of the NAO index could provide some skill for important variables at modest lead times of 1–2 months, even if the DJF NAO is only determined from observations (at the end of the December–February period). Because the GloSea5 system has skill in predicting the DJF NAO index from the previous November (Scaife et al., 2014), it is possible that the lead time could be increased by using the predicted rather than the observed NAO index. In Fig. 9 we examine correlations for the same predicted NWS variables as in Fig. 8, but this time using the DJF NAO index predicted from the ensemble mean of the GloSea5 forecast run the preceding November. Unsurprisingly the correlations are generally lower than for the observed NAO index, and many are not statistically significant at the 95 % confidence level. However, the correlations are largely of the same sign and pattern as for the observed NAO index. This suggests that a prototype seasonal forecast based on the GloSea5 NAO may be possible. For SST and NBT, there are regions that exhibit persisting significant correlations (e.g. the English Channel), which is promising. The correlation patterns for SSS are, however, quite different to those from the observed NAO. There is a general negative correlation across the shelf that gets stronger with an increasing lag (Fig. 9g–i). The persistence in the NAO–SSS correlation reflects the longer-term nature of salinity anomalies. The difference in the SSS correlation patterns between the observed and GloSea5 NAO perhaps act as an error estimate to this approach, suggesting that caution and further assessment are needed before relying on an empirical seasonal forecast of this form for SSS. Overall, the results with the GloSea5 NAO suggest that real relationships exist between the forecast NAO and the observed NWS fields and that further improvements in the seasonal NAO forecast would deliver higher levels of forecast skill and/or regional detail.
Exemplar regional seasonal forecast of English Channel SST (selected
for the strength of the correlations): the seasonal forecast is based on the
linear relationship between the reanalysis SST and the observed NAO (blue,
left column:
The correlations between the NAO and the NWS fields describe how strong a linear relationship exists between the two. Where there is significant skill (a significant correlation, Fig. 8) this linear relationship can be used to predict the NWS fields from the NAO. A simple approach would be to find the slope and intercept between the observed (NOAA) NAO and the shelf sea variable and then apply this equation to the GloSea5 forecast NAO. This provides a simple empirical forecast giving information about the future state of the NWS (e.g. greater than average, less than average). Other non-linear relationships (e.g. quadratic) may exist between the NAO and NWS fields that could be used as the basis of an empirical forecast; further analysis (and curve fitting) would be required for their identification. The reanalysis provides a coherent dataset in order to explore such relationships.
Here we present such an example forecast of the English Channel SST initialised in November for the following winter. Using the reanalysis, we construct a linear empirical relationship between the observed (NOAA) DJF NAO and the English Channel winter (DJF) SST. The correlation in Fig. 8 quantifies the strength of this linear relationship and its significance. We then apply this equation to the GloSea5 forecast NAO (Fig. 10). Due to the persistence of the NAO SST correlation in this region, we are able repeat the process to extend the lead time (February–April forecast initialised in November), beyond which the underlying correlation significance is greater than 0.05. These (normalised) forecasts are illustrated in Fig. 10.
While our example has focused on a region and variable with a relatively high
correlation, the English Channel winter SST may have a direct application.
For example, European sea bass (
Developing a range of empirical forecasts is a possible way of producing NWS seasonal forecasts, especially if based on predictable climatic indices such as the NAO. Investigating the relationships between the climatic drivers and the shelf response within an NWS climate control simulation (i.e. using a multi-century global climate model run with fixed climate forcings to drive a multi-century NWS simulation; Tinker et al., 2018) may allow much subtler relationships to be established than is possible with the relatively short modern observed period. However, due to empirical forecasts relying on past observations, there can be limits to their use in the future when conditions are outside the present day range.
With the maturity of seasonal forecasting systems and shelf sea dynamic downscaling systems, it will not be long before a seasonal forecast system for the NWS based on dynamical downscaling is technically possible. The underlying skill of the global seasonal forecasting system would be the basis for any such NWS downscaling, and its skill at predicting the surface and lateral boundary conditions of the NWS will be the leading-order limit to the subsequent skill of the NWS seasonal forecast. However, the skill from the global seasonal forecasting system must be able to propagate from its ocean and atmosphere and manifest itself on the NWS for an NWS forecast to have any skill. We start to explore whether this is the case with the CMEMS reanalysis. If there is only a weak relationship between the boundary (lateral and surface) and the interior of the shelf with internally generated chaotic variability dominates, any year-to-year variability modelled by GloSea5 will not manifest in the NWS. The NWS is considered to be quasi-isolated from the North Atlantic (Wakelin et al., 2009); however, it is a broad continental shelf sea and interaction with the atmosphere is therefore important. We now investigate the relationships between the state of the NWS in the CMEMS reanalysis and the boundary conditions that forced it to see how much of the NWS variability is driven by large-scale drivers and how much is internally generated. We focus on the surface boundary conditions, as the advective lag between the lateral boundaries and the shelf makes them less relevant for inter-annual variability and seasonal forecasting.
We consider the inter-annual correlations between the CO5 NWS reanalysis surface and open ocean boundary forcings and between the NWS temperature and salinity in the regions defined in Fig. 1 (see Fig. 11). Our interpretation of the correlations must be informed by physical insight, recognising that correlation does not imply causation. We examine processes influencing temperature and salinity separately, since temperature is likely to be strongly influenced by surface heat exchange, whereas advective processes may play a stronger role for salinity due to a lack of direct feedback between surface salinity and the atmosphere.
Correlation maps between time series (annual mean unless stated
otherwise) of model forcings and shelf response (split into
three sub-panels). Insignificant correlations (at the 95 % confidence
level) are not given. The individual panels relate the shelf response to the
forcing:
NWS surface temperatures are significantly correlated with temperatures at
the open ocean boundaries of most of the domain (the northern boundary west
of 10
The salinity on the NWS is primarily a balance between salty water entering the shelf from the North Atlantic and its modification due to water exchanges with the atmosphere (e.g. precipitation) and dilution from rivers (and exchange with the Baltic). This leads to some intuitive relationships: the saltier the Atlantic, the saltier the NWS; the greater the river flow and rainfall into the NWS, the fresher the NWS. These can be considered the direct mechanisms that control salinity. An important secondary mechanism is the rate at which the NWS water is exchanged with the Atlantic.
Salinity in the regions of the CO5 NWS reanalysis domain off the shelf is strongly correlated with the salinity of the boundaries. For example, the salinity in the north-western oceanic part of the domain is strongly correlated with the salinity boundary forcings along the western edge of the domain (e.g. western boundary–surface_central is correlated with SSS in Fig. 11b), and this penetrates onto the shelf in the Shetland shelf region. The salinity of the large south-western oceanic region is positively (though not significantly) correlated with the surface salinity of the southern boundary (southern boundary–surface_east), but there is a suggestion that the influence of boundary forcing is advected onto the shelf through the Armorican shelf, Celtic Sea, and English Channel and into the southern and central North Sea due to significant correlations (Fig. 11e). Note that the exemplar ocean boundary conditions in Fig. 11a–b and d–e are not necessarily the strongest correlations between the shelf and the ocean boundary conditions, but have been chosen to provide a pair of regions to highlight the west–south temperature and salinity differences in correlations. The lack of significance in the correlations between the SW oceanic region and the southern boundary salinity may reflect this large region blurring out a smaller correlated area (i.e. the tightly defined European slope current, which flows northward, following the continental shelf slope from the south of the domain through the Bay of Biscay, around Ireland, and then towards Norway across the top of the North Sea).
There can be considerable advective lags between the ocean–lateral ocean
boundaries and the NWS. The analysis of Holt et al. (2012) suggests
that modelled conservative tracers in the open ocean took
Run-off from large rivers tends to be (negatively) correlated with salinity across the shelf (e.g. the European rivers that flow into the southern North Sea region; Fig. 11f), while run-off into regions that only include smaller (e.g. UK and Irish) rivers tends to have only local effects (such as the rivers flowing into the Irish Sea; Fig. 11c).
Correlations between winter (DJF) storm track latitude and annual
mean winter and spring (left to right) mean sea level pressure (MSLP;
The salinity across the shelf is significantly negatively correlated with total precipitation in the south of the domain (Fig. 11p), but not with precipitation in the north of the domain (Fig. 11m), which is insignificantly positively correlated. Conversely, strong wind magnitudes in the north of the domain are correlated with salinity across the domain (Fig. 11o), but wind magnitude in the south of the domain is not (Fig. 11r). Mean sea level pressure (MSLP) in the south of the domain is positively correlated with salinity across the shelf (Fig. 11q), but northern mean sea level pressure is not (Fig. 11n). These patterns suggest that the NWS salinity is responding to a large-scale driver (rather than simply local rainfall and river flow) such as varying exchange with the Atlantic. The meteorological relationship between these surface forcings (low mean sea level pressure, high wind and rain) suggests stormy conditions, and the north–south correlation patterns of mean sea level pressure, wind magnitude, and total precipitation suggest that storm track latitude may be important.
When we compare these variables against the winter storm track latitude (Fig. 12), we find that a more northern storm track location does correlate negatively (positively) with mean sea level pressure (wind magnitude and total precipitation) in the north of the domain, and the opposite in the south. We find that these correlations tend to reduce in strength through the seasons, being highest in winter (DJF) and reducing through to the spring (MAM). This suggests that the index of storm track latitude is important for these variables.
We find that shelf salinity is (insignificantly) positively correlated with a more northerly storm track, particularly in the Irish and Shetland shelf and the northern and central North Sea, and there is a suggestion of a lag in the correlation towards the North Sea. One possible mechanism that could explain these correlations is the rate of exchange of water with the Atlantic.
The exchange of water between the NWS and the Atlantic tends to dominate the mean salinity of the North Sea (Sündermann and Pohlmann, 2011). One of the main pathways of water into the North Sea is via the Fair Isle current (Sheehan et al., 2017), fed by the European slope current. Both the Fair Isle current and the European slope current have been shown to correlate with the NAO (Marsh et al., 2017; Sheehan et al., 2017; Winther and Johannessen, 2006). This positive correlation means more (relatively) high salinity is advected onto the NWS under NAO positive condition, which would tend to increase the NWS salinity. We do not explicitly look at the European slope current in this study, but understanding its predictability could provide an important mechanism for NWS salinity predictability.
Because the strength of the inflow of North Atlantic water onto the shelf (the indirect mechanism affecting the year-to-year shelf salinity) and the direct mechanism (dilution from rivers and precipitation) are both correlated with NAO, but in an opposing manner, it is possible that more information (than is contained within the simple NAO index) is needed to predict NWS salinity variability. Such a balance of opposing mechanisms may explain the relatively low NAO SSS correlations (Fig. 8g–i) despite the main drivers of salinity being correlated with NAO. In principle, dynamic downscaling may provide additional skill by modelling these opposing mechanisms directly.
The spatial patterns in the relationships in Fig. 11 show that (some of) the variability on the NWS is strongly coupled to the large-scale boundary conditions; this is consistent with the established view that the NWS is a boundary-driven system (with the caveat that we are using a non-eddy-permitting model for the NWS). Hence the overall concept of (empirical or dynamic) downscaling based on large-scale boundary drivers that may be predictable by a global seasonal forecast system is plausible. This suggests that a key area for scientific effort is to evaluate and improve the predictability of the NWS boundary drivers as produced by GloSea5.
Overall, we conclude that much of the year-to-year NWS variability is relatively tightly linked to the variability in the boundary conditions, which is a prerequisite for dynamic downscaling. We note that in some cases, there may be a balance of opposing mechanisms or a response to a sequence of NAO events that requires more information than is encapsulated in the simple NAO index; this may provide a pathway for additional predictability from dynamic downscaling when compared to empirical downscaling. Furthermore, much of the temperature and salinity variability on the NWS is linked to large-scale climate variations (including river outflow which integrates rainfall over a large area) rather than to more local effects (such as the direct effect of rainfall on the synoptic scale). This increases the prospect of useful seasonal predictions since global seasonal forecast systems are beginning to show significant skill in predicting large-scale climate indices.
Our exploratory investigation shows that despite the useful skill that GloSea5 has in predicting certain large-scale climate indices, its output may limit direct application for shelf sea seasonal forecasts because of limited resolution, missing shelf sea processes, and simplified treatment of river run-off. However, we have shown evidence that many aspects of inter-annual variability on the NWS are driven by large-scale variations in elements of atmospheric, oceanic, and riverine forcings that are closely linked to the winter NAO index, for which there is considerable predictive skill at a lead time of several months. Indeed, Fig. 10 shows that a simple empirical downscaling approach driven by the forecast NAO index can provide significant skill in some variables and regions at a lead time of several months. Further improvements to the reliability of the GloSea5 NAO forecast may enhance the forecasts for the NWS region. While being skillful, GloSea5 has been shown to be under-confident (meaning there is too high a proportion of unpredictable noise in the forecast) for the NAO, the wider Atlantic region (Eade et al., 2014), and for inter-annual predictions (Dunstone et al., 2016). This problem is common to most skillful seasonal forecasting systems (Baker et al., 2018), but as there are exceptions, it is not inherent to such systems. Solving the under-confidence issue is an active research area.
Based on our results we can make an assessment of the three possible
forecasting methods identified in the Introduction.
assess user value for cases in which we have already demonstrated some skill
and establish what skill and/or reliability would be needed to provide actionable
forecast information in various sectors; carry out in-depth assessment of regional reanalyses as a tool to develop empirical
downscaling relationships; assess if, when, and where the GloSea5 seasonal forecast system can be used
directly for NWS forecasts; assess potential added value of dynamical downscaling, initially through
case studies; identify the largest sources of uncertainty in downscaled predictions (e.g.
seasonal forecast fields for specific drivers, downscaling model) to inform
where to focus development effort; and assess predictability in seasons other than winter. To date seasonal
forecasting systems have shown less skill in the summer, but this is an
active research area (e.g. Hall et al., 2017). It may also
be possible to demonstrate some degree of memory in the shelf seas
themselves, which would add to forecast lead times.
Even the limited level of predictive skill we have shown here for some
regions of the NWS may be useful for certain applications, e.g. SST forecasts
for February–April may give early indications of increased risk of harmful
algal blooms, and predictions of near-bottom temperature and its impacts on
the gas supply network may inform more resilient energy planning. Further
developments will be needed to deliver a seasonal prediction system for the
NWS with sufficient skill and reliability to inform user planning decisions
over a wide range of applications. Specific research priorities are the following:
Many challenges remain before we can derive seasonal forecasts for the NWS
that are accurate and reliable for a wide range of regions and variables. It
is likely that some variables will prove to be inherently unpredictable to
any useful degree. But the early results presented here show that current
seasonal forecast systems can already provide meaningful information with
the potential for applications in marine operations and planning. As our
understanding and capability develops, a close interaction between climate
scientists, marine scientists, and end users will be needed to bring the
added value of seasonal forecast information into decision-making in marine
policy, planning, and operations.
The data used in this study are available from the Copernicus Marine Environment Monitoring Service (CMEMS) North-west European Shelf reanalysis (NORTHWESTSHELF_REANALYSIS_PHYS_004_009; Wakelin et al., 2014) available from their online catalogue.
JT designed the analysis, which was undertaken by JK and JT. JT and RW prepared the paper with contributions from all co-authors. RB and SRD provided advice on observations datasets, policy, and user relevance.
The authors declare that they have no conflict of interest.
The authors thank Leon Hermanson, Craig MacLachlan, and Drew Peterson (MOHC)
for advice and support with working with GloSea5, John Pinnegar (CEFAS)
for his comments and proofreading, the reviewers for questions and comments
that strengthened the paper (particularly relating to the reanalysis
evaluation), and Enda O'Dea, Dave Storkey, and Robert King (MOHC) for NEMO
advice. This work was supported by the Joint BEIS/Defra Met Office Hadley
Centre Climate Programme (GA01101) and by Defra ME5317 ForeDec – developing
the capability to use marine forecasts over seasons to decades. The southern
North Sea ferry time series work has been supported by Defra ME Cefas-SLA15/25 and is a contribution to the ICES Report on Ocean Climate
(