Subsurface Initiation of Deep Convection near Maud Rise

Abstract. In 2016 and 2017, an open-ocean polynya appeared over Maud Rise. The formation of these polynyas has been attributed to the occurrence of intense winter storms. However, the evolution and lifetime of the two polynyas was quite different. Here, we use model output of a century long high-resolution climate model simulation to explain the differences between the 2016 and 2017 Maud Rise polynyas. Analysis of the results, using convective available potential energy to measure subsurface convection, leads us to the interpretation that the first polynya event is (partly) initiated by subsurface static instabilities, 5 leading to subsurface convection. Subsurface convection associated with the formation of the 2016 polynya preconditioned the Maud Rise region, resulting in a weakly stable surface layer and eventually leading to the 2017 polynya event. Based on this, we argue that, apart from atmospheric variability, subsurface convection is important to initiate a Maud Rise polynya.

August 2016, but these storms did not lead to any MRP formation. Surprisingly, when the 2017 MRP opened, even a longer quiescent period (1.5 months) in storm activity occurred and the MRP did not close during that year. The differences in the evolution and lifetime between the 2016 and 2017 MRP were attributed to the stronger vertical overturning in 2017 (Campbell et al., 2019). 5 In this study, we aim to provide an alternative explanation for the differences between the evolution and lifetime of the 2016 and 2017 MRPs. In a simulation with a high-resolution version of the Community Earth System Model (CESM, Hurrell et al. (2013)), van Westen and Dijkstra (2020) found a multidecadal time scale preconditioning of the Maud Rise region by subsurface processes. Consequently, by deep convection, MRP formation occurred at the same multidecadal time scale, but the formation mechanism was not discussed in van Westen and Dijkstra (2020). 10 Here, we analyse the formation of one of these MRPs and investigate the role of subsurface initiation of (deep) convection on polynya formation. Section 2 provides a brief overview of the climate model simulation used and information on the observational data and reanalysis products analysed. In Section 3, the formation of one multiyear MRP in CESM is analysed, together with a comparison to observations and reanalysis for the 2016 -2017 period. A summary and discussion of the results with the main conclusions are given in the final Section 4. We analysed the last 101 years (model years 150 -250) of the CESM simulation. Four (multiyear) polynyas are identified over Maud Rise in this simulation: model years 158 -159, 178 -182, 205 -209 and 231 -237 (van Westen and Dijkstra, 2020). 25 Here, the formation of the last multiyear MRP (model years 231 -237) is analysed in detail, where we use both monthly-mean and daily-mean quantities. Most quantities are spatially averaged over the Polynya region (2 • E -11 • E × 63.5 • S -66.5 • S, see The CESM results are compared to observations and reanalysis data, in particular for the period 2016 -2017. We used sea-ice measurements by the Scanning Multichannel Microwave Radiometer (SMMR) and Special Sensor Microwave Imager 30 (SSM/I) (http://nsidc.org/data/G02202, Peng et al. (2013); Meier et al. (2017)). Also vertical profiles of an Argo float (5904468, http://www.coriolis.eu.org), which was deployed in 2015 and remained active till 2018, and was located near Maud Rise during the 2016 -2017 MRP, were analysed. In addition, we retained model output from the Operational Mercator global ocean analysis and forecast (http://marine.copernicus.eu/services-portfolio/access-to-products/) at 1/12 • horizontal resolution.
Mercator assimilates various observational data sets and the model is 'steered' towards observations. The Mercator output is stored as daily averages and the output is available from 2016 to present.

Results
In Subsection 3.1 below, we first present a detailed analysis of the last multiyear MRP event (model years 231 -237) in CESM, 5 which most clearly shares phenomena as observed during the 2016 -2017 MRP. The details of the formation of this MRP are described in Subsection 3.2 and a comparison with observations is made in Subsection 3.3.

The Maud Rise Polynyas in Model Years 231 -237
The monthly averaged sea-ice fields ( Figure 1a) show 7 consecutive years where an MRP appears, starting in September of model year 231 and ending in November of model year 237. There is no overall decline in sea-ice fraction, f ice , between model 10 years 227 -230 (except from seasonality) and all September sea-ice fractions over this period are above the climatology mean (f ice,Sep = 89.1 %), which has been determined using all non-MRP years between model years 150 -250.
Prior to the MRP in model year 231, the subsurface (below 100 m depths) is relatively warm and salty (compared to the time mean) over Maud Rise (colour plots in Figures 1a, b). To obtain the anomalies, we averaged the temperature and salinity (T k and S k , respectively) over the Polynya region (at depth z k ), the time-mean temperature (model years 150 -250) was 15 subtracted from the time series and the result was quadratically detrended. Accumulation of subsurface heat and salt leads to preconditioning of the density in the Maud Rise region (van Westen and Dijkstra, 2020).
The MRP forms in model year 231 when the September sea-ice fraction strongly deviates from climatology (∆f ice = −26.1%), as well as for the other months during that year ( Figure 1c). The mixed layer depth (black curve in Figure 1a) also increases in model year 231, mixing positive subsurface temperature and salinity (colours in Figure 1a     4 https://doi.org/10.5194/os-2020-33 Preprint. Discussion started: 4 May 2020 c Author(s) 2020. CC BY 4.0 License.

Formation of the Maud Rise Polynya
Convection is a generic aspect of the occurrence of the multiyear polynya event during model years 231 -237, as well as for the other polynya events (van Westen and Dijkstra, 2020). In this subsection, we analyse the initiation of vertical mixing near the Polynya region in more detail by focusing on the polynya event in model year 231.
Convection is caused by a statically unstable water column. The CESM provides a measure of this static stability (part of 5 the standard output) through the vertical gradient of potential density ρ θ at each depth level, i.e., It turned out that the monthly-averaged CESM output is not appropriate to analyse the initiation of convection due to the strong temporal variability of this process. Therefore, we analyse daily-averaged output for the sea-ice fraction, mixed layer depth, potential density and static stability. Due to storage limitations, these daily-averaged quantities are only available for model To investigate the role of the subsurface static instability in the polynya formation, we use the concept of Convective Available Potential Energy (CAPE), which is a measure of the potential energy available for conversion to kinetic energy of a fluid particle to move upwards or downwards in a statically unstable situation (Su et al., 2016). Using ρ θ (z) as the background, CAPE at depth z is defined as where g the gravitational acceleration and ρ θ (z) the potential density of the particle at depth z. Starting at a reference level (z ref ) and the corresponding potential density (ρ θ (z ref )), we integrate upwards until CAPE becomes zero (or the surface is reached); this depth is indicated by z 0 and is referred to the final depth. Fluid elements which are initially located in a statically unstable (S > 0) layer are only considered here, since particles need to deviate from their reference depth in order to initiate   (not shown), outside of the regions where the two polynyas form (cf. Figures 2a, b), but still they enhance upwelling, turbulent mixing and sea-ice divergences and hence can favour MRP formation.
After (deep) convection is initiated which vertically mixes heat and salt upwards, the surface water in the Polynya region remains relatively warm and saline compared to the time mean (Figure 1a, b). The relatively warm surface water slows down and prevents the formation of sea ice near Maud Rise during the following year (model year 232, Figure 1a). The open-5 ocean water area in the Polynya region is strongly cooled during austral winter, which decreases the static stability of the water column near the surface. Deep convection also mixes subsurface salinity anomalies towards the surface which decrease the static stability near the surface. The combination of positive temperature and salinity anomalies result in surface driven convection which is shown in Figure 3d by the relatively large surface instabilities after model year 231. After several years, the subsurface heat reservoir is depleted and convection ceases, leading to a statically stable water column. year, an MRP formed in September 2017 and relatively low sea-ice fractions (with respect to climatology) were observed 20 over the Polynya region between September and December. On 16 September 2017, the sea-ice fraction was f ice = 58% 1990 -2015). Note that the CESM model results (also plotted in Figure 4a) display comparable seaice fractions (averaged over the Polynya region) of about f ice = 60% when the MRP formed in model year 231 (cf. Figure 1c).
Second, the oceanic state between 2016 -2017 is analysed using model output from Mercator, which provides daily averages of oceanic potential temperature, salinity, mixed layer depth and sea-ice fraction. The Mercator sea-ice fractions (averaged 25 over the Polynya region) in 2016 and 2017 are also displayed in Figure 4a and reasonably agree with observations. The vertical profiles of potential temperature and salinity are shown in Figure 4b and compared to those of an Argo float (5904468) that was present near Maud Rise in 2016 -2017. Although these Argo float observations are too sparse (every 10 days) to analyse the oceanic state (and e.g. convection), they can be used to validate the Mercator data which has a similar vertical background profile as observations.

30
Applying the same CAPE analysis as in section 3.2 to the Mercator data, we determined the final depth of (subsurface) convection in the Polynya region. Because the static stability is not provided in the Mercator data, it was assumed that layer dz k is statically unstable (S k > 0) when ρ θ (z k+1 ) < ρ θ (z k−1 ), where index k increases with increasing depth. Next, from the  The final depth of (subsurface) convection for the 5%-percentile level of the Mercator potential density profiles in the Polynya region is shown in Figure 4d for