Satellite-Derived Bias-Corrected Air Temperature for Understanding Crop-Climate Interactions in Tropical Agricultural Systems

Abstract:

The magnitude of local/regional temperature variability and crop responses to such changes must be well understood to accurately assess the impacts of climate variability and change on agriculture. Challenges arise when meteorological stations are sparsely distributed such as in much of the tropics including cerrado Brazil—one of the largest agriculturally-important areas in the tropics. Currently available gridded climate datasets are coarse (i.e. 2.5º to 0.5º), heavily-interpolated and hence not adequate for regional/local scale assessments of agricultural impacts from climate. In this context, we aim to develop a new method for gridded standard retrieval of a number of agriculturally-relevant near-surface air temperature indicators. We employ Moderate Resolution Imaging Spectroradiometer (MODIS)-derived land surface temperature, vertical temperature profile, meteorological station data, and other biophysical data. We will contrast the accuracy of different air temperature derivation approaches including multiple linear regression, machine learning, the use of vegetation indices to proxy for air temperature variation, and physical approaches based on surface energy balance parametrizations. Our method aims to control for potential biases in surface temperatures due to differences in land cover types, changes in vegetation cover, and variation in plant growth. Satellite-derived near-surface air temperature datasets can support enhanced analysis of climate impacts and climate change adaptation in tropical agriculture.