Comparison of MODIS and AATSR Data With n Situ Temperature Measurements During Upwelling Events in the Baltic Sea
Uiboupin, Rivo; Laanemets, Jaan
Marine Systems Institute at Tallinn University of Technology, ESTONIA

Besides the field measurements the satellite SST data carry substantial additional information about the spatial extent and structure of SST field. The high variability in SST field that is observed during upwelling events provides a good basis for studying the performance of different remote sensing sensors, in particular when in situ temperature measurements are available. The remote sensing imagery provides valuable information for characterizing upwelling events in the coastal sea.
The two infrared sensors that are used in current study - MODIS (MODerate Resolution Imaging Spectroradiometer) and AATSR (Advanced Along-Track Scanning Radiometer) - provide up to 3 daytime images per day over the Gulf of Finland, Baltic Sea. The SST data from different sensors have different characteristics due to different measuring principles. Previous studies have characterized the accuracy and bias of AATSR sensor globally and regionally. Reynolds et al. (2010) showed that AATSR SST retrievals have variable biases during summer day time retrievals in regions higher than 50° N. The bias between different sensors needs to be understood to be able to use the data from multiple sensors for characterizing upwelling event in coastal sea.
The objectives of current study, which is devoted to upwellings in Gulf of Finland, are (1) to quantify the differences between in situ temperature and MODIS and AATSR SST products, (2) to quantify the biases between MODIS and AATSR SST products, (3) to develop a method to derive bias corrected daily SST maps in cases of highly variable SST fields using MODIS and AATSR data and (4) to characterize upwellings in the Gulf of Finland using SST data.
In current study the data from MODIS and AATSR were used for estimation of sea surface temperature (SST) in the Gulf of Finland (Baltic Sea) during upwelling events. The set of MODIS SST images include 5 upwelling events along the Finnish coast and 3 events along Estonian coasts. In total 24 daytime SST images acquired by MODIS onboard of Terra and Aqua satellites were used. Also 6 AATSR SST images were used in the study - showing 4 upwelling events along the northern coast and 1 event along the southern coast of the Gulf. In addition to daily remote sensing SST imagery that provide information over the Gulf of Finland there is another source of operational upper layer temperature (SST4m) data - the temperature measurements conducted regularly by ship of opportunity at a fixed depth. The high spatial resolution (~150m) surface layer flow through (FT) temperature data were collected along the transect between Tallinn and Helsinki unattended on passenger ferries. Also mooring buoy temperature data at the central part of Gulf of Finland were used. For comparison of the MODIS and AATSR SST products with the corresponding measured data the root mean square temperature differences (RMSD), the correlation coefficient (R) and bias were calculated. Prior to the comparison, a wind speed threshold criterion was set to define the conditions when in situ measurements at 4m depth can be used for satellite SST validation. The SST data obtained with MODIS and AATSR were also compared. Also bias corrected RMSD along the transect between the two products were calculated. Three types of composite maps were calculated in cases when data from both satellite sensors was available: (1) uncorrected composite maps, (2) composite maps where bias correction was applied to AATSR data and (3) maps where the bias correction was applied for both images (MODIS and AATSR) in order to get composite SST map that best represents in situ measurements.
The analysis showed that in the most of the occasions the correlation between MODIS SST and in situ temperature was high - exceeding 0.9. This was expected as the study concentrates on images that were collected during upwelling events i.e. the variability of temperature distribution is well expressed. The mean RMSD of all MODIS SST data was 0.46 °C. It was shown that FT temperature measurements from a depth of 4 m could be used for validation of SST data in cases when the wind speed was over 5 m s-1 as the bias is reduced significantly compared to the low wind conditions. Comparison between MODIS SST retrieval and FT temperature at the time of moderate-high winds showed a good correlation (R>0.97), with an RMS difference of 0.28°C. A warm bias of 0.11°C was observed between the remote sensing SST and the FT temperature. All of the AATSR images coincided with one or two MODIS images within 3 h time window. The correlation between AATSR SST and in situ temperature was in the range of 0.87- 0.99. The comparison of the MODIS and AATSR SST data along the transect showed that mean correlation between the two products was 0.95. The correlation between the data from two remote sensing sensors is higher, compared to the satellite data correlation with FT measurements. This is due to the fact that both satellite sensors provide measurements from the skin layer temperature (SSTskin) while FT measurements provide information about subsurface temperature (SSTbulk). The mean bias was 0.78 °C, showing that AATSR nadir-view over estimates the SST values compared to MODIS SST.
The coastal upwellings in the Gulf of Finland area characterized based on the bias corrected multi-sensor composite SST imagery.