Assimilation of sea surface temperature by MET Norway's coastal ocean forecasting model Norkyst
The Norwegian Meteorological Institute (MET Norway) and the Norwegian Institute of Marine Research (IMR) have developed an ocean forecasting model specifically for the Norwegian coastal and shelf seas, called Norkyst. This model enables MET Norway to provide forecasts such as Sea Surface Temperature (SST) and ocean currents, which are crucial for oil spill response, search and rescue operations, and plankton dispersion tracking. Norkyst assimilates near real-time SST products from OSI SAF catalogue to enhance its forecasting capabilities.
Figure 1: Rendering of the Norkyst model data (bathymetry and temperature). made with ParaView visualization engine.
To produce a reliable forecast, models need to assimilate observation data. Norkyst assimilate both in situ data from a wide range of sources, and satellite data. While in situ data have less bias, and are giving information closer to the coast line and below the surface, the satellite data allow a wider coverage of the measurements, even from remote areas. The two are then complementary (fig 2)
As the model is able to handle observation data at their correct time and location, the MET Norway team decided to use satellite data with a few processing, provided on the native satellite swath grid (level 2P products), rather than more processed ones where the data has been spatiotemporally resampled (for instance, the data has been sampled to lower the effect of day/time variability) and the gaps mainly due to the cloud cover have been virtually filled (Level 3 and level 4 satellites products). The choice of L2P data allows as well to keep a high spatiotemporal resolution for each data.
A specific setup of Norkyst (Norkyst_DA), with reduced horizontal resolution is used for data assimilation. It is based on the Regional Ocean Modeling System (ROMS) with a 4D-variational data assimilation scheme. The 2.4 km horizontal resolution was selected to match the scale of available observations, including OSI SAF products, as well as to reduce computational costs
assimilated SST
Figure 2: Assimilated surface temperature data for the 1/06/2024. | a. map displaying the 1 199 866 assimilated satellite sea surface temperature measurements. | b. map displaying the 1 740 assimilated in situ surface temperature measurements.
(daily data available online)
L2P data come however with less bias corrections than the more processed ones. To overcome this, Norkyst uses data from a wide range of different satellites:
OSI SAF products made from the Advanced Very-High-Resolution Radiometer (AVHRR) aboard Metop-B and Metop-C (OSI-204-b and OSI-204-c)
EUMETSAT (headquarter) produced Copernicus products from the Sea and Land Surface Temperature Radiometer (SLSTR) aboard Sentinel-3A and sentinel-3B.
NOAA products from the Visible Infrared Imaging Radiometer Suite (VIIRS) aboard NOAA-20 and NPP
JAXA products from the Advanced Microwave Scanning Radiometer 2 (ASMR2) aboard GCOM-W1
Being a passive microwave sensor, ASMR2 gives valuable observation data on cloudy conditions. However, compared to the other sensors used here, which are infrared sensors, its spatial resolution is much coarser. This has to be taken into account by the model during the assimilation process. A specific mode, called supermod operator, has been introduced into the ROM 4D-var System algorithm to handle that issue.
Additionally, SST products from different satellites and sensors are subject to regional and temporal biases, originating from both retrieval process and algorithms. To handle these, AVHRR products, together with the NOAA products are bias corrected prior to assimilation. The SLSTR products are used as reference for the SST bias correction scheme, and, together with data from drifting buoys, used for validation of the Norkyst forecasts.
The whole assimilation procedure has been described in a dedicated article by S.C. Iversen, A.K. Sperrevik and O. Goux.
