LAUSR

ABSTRACT In this article, correlations to derive minimum (Tmin), mean (Tmean), and maximum (Tmax) air temperatures from land surface temperature (Ts) in Morocco are studied. Land surface temperature was derived from 349 decadal cloud-free images acquired by the Advanced Very High Resolution Radiometer sensor, onboard the National Oceanic and Atmospheric Administration’s satellites. Ts was first compared with Tmin, Tmean, and Tmax, at the level of nine meteorological stations between 1995 and 2012. This first step showed the existence of good linear correlations between Ts and air temperatures. The coefficient of determination, R2, varied between 0.59 and 0.79 for Ts versus Tmin, between 0.60 and 0.76 for Ts versus Tmax, and between 0.67 and 0.79 for Ts versus Tmean. The root mean square error (RMSE) varied between 2.4°C and 3.9°C for Ts versus Tmin, between 2.5°C and 4.6°C for Ts versus Tmax, and between 2.2°C and 3.8°C for Ts versus Tmean. The mean absolute error (MAE) varied between 2°C and 3.1°C for Ts versus Tmin, between 1.9°C and 3.6°C for Ts versus Tmax, and between 1.8°C and 3°C for Ts versus Tmean. Second, Ts was compared with Tmin, Tmax, and Tmean gridded at the level of agricultural areas of 46 provinces in Morocco. For 42 provinces, the mean value of R2 was 0.71 for both Tmin versus Ts and Tmax versus Ts, and 0.76 for Tmean versus Ts. The mean values of the RMSE and the MAE were 3.1°C and 2.4°C for Ts versus Tmin, 3.6°C and 2.8°C for Tmax versus Ts, and 3°C and 2.3°C for Tmean versus Ts, respectively. Finally, the accuracy of the regression models between air and surface temperatures was tested using a k-fold cross-validation method that showed high stability of these models. The relationships obtained in this work could be very useful for further monitoring and modelling agriculture and meteorological parameters.