Use of Infrared Satellite Observations for the Surface Temperature Retrieval over Land in a NWP Context

Abstract

In Numerical Weather Prediction (NWP), an accurate description of surface temperature is needed to assimilate satellite observations. These observations produced by infrared and microwave sensors, help retrieving good quality land surface temperature (LST) by using surface sensitive channels and emissivity atlases. This work is a preparatory step in order to assimilate LSTs in Météo-France NWP models surface analysis. We focus on IASI and SEVIRI sensors. The first part of this work aims at comparing the SEVIRI retrieved LST to local observations from two stations included in the meso-scale AROME-France domain over four periods from different seasons. Diurnal cycles of local LST and SEVIRI LST show a good agreement especially for the summer period. Averaged biases show seasonal variability and are smaller during Winter and Autumn with less than 1 K values for both stations. The second part of the study deals with the comparison of LST values retrieved from different infrared sensors in AROME-France model. First results show encouraging agreement between both LSTs. A comparison during Autumn period for clear sky conditions reveals an almost null bias and a standard deviation of about 1.6 K. More detailed comparisons were performed over contrasted seasons with a special attention to diurnal cycles for both sensors. A better agreement is noticed during nighttime. The last step of this inter-comparison was to study the simulation of SEVIRI and IASI brightness temperatures by using a fast radiative transfer model. Thus, several simulations have been run covering various dates from different seasons by daytime and nighttime using SEVIRI LSTs, IASI LSTs and AROME-France model LSTs. Simulated brightness temperatures were then compared to observations. As expected, the best simulations are the ones using the LST retrieved from the sensor for which simulations are performed. However, the LST retrieved from another sensor provides better simulations than the predicted LST from the model especially during nighttime. For IASI simulations, SEVIRI LSTs increase RMSE by 0.2 K to 0.9 K compared to IASI LSTs for nighttime case and by around 1.5 K for daytime.

Keywords

• remote sensing
• satellite observations
• land surface temperature retrieval
• data assimilation
• surface analysis
• infrared
• numerical weather prediction