Integration of remote sensing datasets for local scale assessment and prediction of drought

Abstract

Recent attempts to integrate remote sensing-based drought indices with precipitation data seem promising, and can compensate for potential uncertainties from image-based parameters alone, which may be unrelated to meteorological drought. However most remote sensing-based studies have been at regional or global scale and have not considered differences between different land cover types. This study examines a drought-prone region in Central Yunnan Province of China over a four-year period including a notable severe drought event in 2010. The study investigates the phase relationships between meteorological drought from image-based rainfall estimates from the Tropical Rainfall Measurement Mission (TRMM), and imaged drought from a remote sensing drought index, the Normalised Vegetation Supply Water Index (NVSWI) for different land cover types at local scale. The land cover types derived from MODIS and Landsat images were resampled to 250m to match all datasets used. Significant differences between cover types are observed, with cropland and shrubland most highly correlated with 64days' earlier rainfall and evergreen forest most responsive to rainfall 90days earlier, indicating a need to consider detailed land cover information for accurate integrated drought indices. The finding that concurrent rainfall is only weakly correlated with observed drought, suggests that existing drought indices, which compute lowest weightings for the most distant lag period would be unrepresentative.