Monitoring water stress using time series of observed to unstressed surface temperature difference

Abstract : Remote sensing data in the thermal infra red (TIR) part of the spectrum provides indirect estimates of water stress – defined as a function of the ratio between actual and potential evaporation rates – at the earth surface. During the first stage of evaporation (‘‘energy limited'' evaporation), this ratio is close to one. During the second stage of evaporation (‘‘soil controlled'' evaporation) water stress occurs and as a result this ratio drops below one. Recently, methods using TIR data to monitor stress have shifted from establishing empirical relationships between combined vegetation cover/temperature indices and soil moisture status to data assimilation of surface temperature into complex soil–vegetation–atmosphere transfer models. However, data and expertise are often lacking to widely apply those methods. In this paper we investigate the proof-of-concept of using solely the difference between actual and unstressed surface temperature as a baseline to monitor water stress. The unstressed temperature is the equilibrium temperature of a given surface expressed in potential conditions, computed with an energy balance model. Theoretical, modelling, and experimental documentation of the proof-of-concept are shown for datasets acquired within the frame of two international experiments in semi-arid region. We show that the difference between the observed and the unstressed surface temperatures is almost linearly related to water stress. A sensitivity study is carried out to test the impact of modelling errors on the evaluation of the unstressed temperature. We found that even with inaccurate but realistic values of the surface parameters used to solve the energy balance and compute the unstressed temperature, the observed to unstressed surface temperature difference is still more relevant to detect second-stage processes than the difference between the observed surface temperature and the air temperature. The perspective of using an empirical index based on this difference is also investigated. These results are especially attractive for application based on TIR satellite imagery at a regional scale.