Documentation below is provided as is. The dataset the documentation below relates to is no longer supported and will be removed from the Climate Data Store (CDS) at a later date.
Contributors: Susan Kay (PML)
1. Known issues in model outputs
1.1. Mediterranean phosphate values are too high
The model initial conditions were based on the World Ocean Atlas 2013 (Locarni et al., 2013), which appears to have values higher than observations. This has affected phosphate values throughout the run and means that the modelled ecosystem is limited by nitrogen instead of phosphorus.
However, tests suggest that the resulting productivity is similar.
1.2. Nutrients and pH near river mouths are inaccurate
River inputs of nutrients and dissolved inorganic carbon were inaccurate because of a coding error, and this has affected biogeochemical model outputs around river mouths. River discharge volume and physical model outputs such as temperature and salinity are not affected. pH in the western Mediterranean, influenced by the Rhone output, is particularly badly affected and should not be used. Other biogeochemical variables should be treated with caution near the mouths of larger rivers. The same issue caused excess nitrate input from the Baltic Sea, which may have led to over- high production in the Norwegian Trench.
2. Model validation
Surface chlorophyll and sea surface temperature output by the POLCOMS-ERSEM model (Butenschön et al. 2016) for the North East Atlantic and Mediterranean have been compared to satellite values, using monthly values for years 1998-2015. Satellite chlorophyll data come from the European Space Agency Climate Change Initiative Ocean Colour project (https://climate.esa.int/en/projects/ocean-colour/, v3.1). Sea surface temperature is from the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) dataset (Donlon et al. 2012), downloaded from the Copernicus Marine Environment Monitoring Service (CMEMS) (http://marine.copernicus.eu/, SST_GLO_SST_L4_NRT_OBSERVATIONS_010_001 and SST_GLO_SST_L4_REP_OBSERVATIONS_010_011). Two model runs have been compared to the satellite values: the climate run as delivered to the Copernicus Climate Change Service - Monitoring and Forecasting Centre (C3S-MFC), which was driven by a downscaled general circulation model (GCM), and a separate run driven by reanalysis data (referred to below as the validation run).
Although the GCM-driven run cannot be expected to closely track the observations at time scales of days or months, the two should be comparable at longer time scales (e.g. the 18-year mean shown here).
Figure 1 Seasonal median surface chlorophyll-a for 1998 to 2015 (a) satellite ocean colour (b) model outputs (validation run) (c) model outputs(GCM-driven run) (d,e) model-satellite difference.
The model broadly reproduces the temporal and spatial patterns of chlorophyll concentration across the region – a northward-spreading spring bloom, relatively low chlorophyll in most of the Mediterranean Sea, highest levels in the northwestern European shelf in the April-June season (Figures 1, 2 and Table 1). However, model estimates tend to be higher than satellite, especially in regions of high chlorophyll. The two model runs give similar outputs, with the GCM-driven run tending to give greater overproduction compared to satellite. Model outputs are lower than satellite in the northern winter, however satellite data tends to be sparse for this period and region because of high cloud coverage, so has higher than usual uncertainty. Outputs from the GCM-driven model generally have a higher bias and root mean square difference (RMSD) than the reanalysis- driven model. Additionally, the correlation to satellite data is weaker, which is expected since the climate model cannot reproduce the month-by-month conditions as closely as the reanalysis. The model-satellite correlation is weakest in shallow coastal areas (Figure 2), however satellite chlorophyll estimates are less reliable in these areas than for open seas.
Table 1: Model-satellite comparison for surface chlorophyll-a, using monthly data for 1998 to 2015, for the whole model domain and sub-regions. The Atlantic (North) region is from 46°N to 65°N, Atlantic South from 25°N to 46°N. Bias = model mean – satellite mean (mg m-3); RMSD = root mean square difference between model and satellite (mg m-3); Spearman-r = Spearman rank correlation coefficient.
| Reanalysis-driven model | GCM-driven model | ||||
region | Bias | RMSD | Spearman-r | Bias | RMSD | Spearman-r |
whole | 0.36 | 1.27 | 0.68 | 0.67 | 1.61 | 0.60 |
Mediterranean | 0.04 | 0.43 | 0.51 | 0.22 | 0.64 | 0.48 |
North Sea | 0.64 | 2.04 | 0.54 | 1.12 | 2.50 | 0.45 |
Atlantic (North) | 0.70 | 1.55 | 0.62 | 0.82 | 1.74 | 0.58 |
Atlantic (South) | 0.35 | 0.91 | 0.62 | 0.92 | 1.56 | 0.57 |
(a 
(b)
Figure 2: Spearman correlation between monthly mean and satellite values of surface chlorophyll, 1998- 2015. The model runs are (a) validation (reanalysis-driven) (b) GCM-driven. Areas where the correlation is not statistically significant are marked in grey.
There is also a good spatial and temporal match between modelled and satellite-derived sea surface temperature (Figures 3, 4 and Table 2). The reanalysis-driven run of the model tends to overestimate sea surface temperature, especially in the spring and summer. The GCM-driven run has lower bias, but underestimates winter temperatures in much of the region. Model-satellite correlation is high in all regions, though rather lower for the GCM-driven run: as noted above, a climate model is not expected to accurately reproduce conditions month-by-month, even in a hindcast. Correlations are weaker for the Atlantic regions than for the North Sea and Mediterranean.
Figure 3 Seasonal median sea surface temperature for 1998 to 2015 (a) satellite (b) model outputs (validation run) (c) model outputs(GCM-driven run) (d,e) model-satellite difference.
Table 2 Model-satellite comparison for surface temperature, using monthly data for 1998 to 2015, for the whole model domain and sub-regions. The Atlantic (North) region is from 46°N to 65°N, Atlantic South from 25°N to 46°N. Bias = model mean – satellite mean (°C); RMSD = root mean square difference between model and satellite (°C); Spearman-r = Spearman rank correlation coefficient.
| Reanalysis-driven model | GCM-driven model | ||||
region | Bias | RMSD | Spearman-r | Bias | RMSD | Spearman-r |
whole | 0.42 | 0.98 | 0.99 | 0.07 | 1.26 | 0.98 |
Mediterranean | 0.57 | 0.94 | 0.99 | -0.19 | 1.14 | 0.97 |
North Sea | 0.26 | 0.78 | 0.98 | 0.55 | 1.24 | 0.96 |
Atlantic (North) | 0.17 | 0.89 | 0.97 | 0.37 | 1.28 | 0.90 |
Atlantic (South) | 0.75 | 1.06 | 0.97 | -0.46 | 1.06 | 0.94 |
(a 
(b)
Figure 4: Spearman correlation between monthly mean and satellite values of sea surface temperature, 1998-2015. The model runs are (a) validation (reanalysis-driven) (b) GCM-driven.
References
Butenschön, M., Clark, J., Aldridge, J. N., Icarus Allen, J., Artioli, Y., et al. (2016). ERSEM 15.06: A generic model for marine biogeochemistry and the ecosystem dynamics of the lower trophic levels. Geoscientific Model Development, 9(4), 1293–1339.
Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli, P., Kobayashi, S., … Vitart, F. (2011). The ERA-Interim reanalysis: Configuration and performance of the data assimilation system.
Quarterly Journal of the Royal Meteorological Society, 137(656), 553–597. https://doi.org/10.1002/qj.828
Donlon, C. J., Martin, M., Stark, J., Roberts-Jones, J., Fiedler, E., & Wimmer, W. (2012). The Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA) system. Remote Sensing of Environment, 116, 140–158.
Locarnini, R.A., Mishonov, A.V., Antonov, J.I., Boyer, T.P., Garcia, H.E., Baranova, O.K., Zweng, M.M., Paver, C.R., Reagan, J.R., Johnson, D.R. and Hamilton, M., (2013). World ocean atlas 2013
