Early drought detection by spectral analysis of satellite time series of precipitation and Normalized Difference Vegetation Index (NDVI)

Abstract

The time lag between anomalies in precipitation and vegetation activity plays a critical role in early drought detection as agricultural droughts are caused by precipitation shortages. The aim of this study is to explore a new approach to estimate the time lag between a forcing (precipitation) and a response (NDVI) signal in the frequency domain by applying cross-spectral analysis. We prepared anomaly time series of image data on TRMM3B42 precipitation (accumulated over antecedent durations of 10, 60, and 150 days) and NDVI, reconstructed and interpolated MOD13A2 and MYD13A2 to daily interval using a Fourier series method to model time series affected by gaps and outliers (iHANTS) for a dry and a wet year in a drought-prone area in the northeast region of China. Then, the cross-spectral analysis was applied pixel-wise and only the phase lag of the annual component of the forcing and response signal was extracted. The 10-day antecedent precipitation was retained as the best representation of forcing. The estimated phase lag was interpreted using maps of land cover and of available soil water-holding capacity and applied to investigate the difference in phenology responses between a wet and dry year. In both the wet and dry year, we measured consistent phase lags across land cover types. In the wet year with above-average precipitation, the phase lag was rather similar for all land cover types, i.e., 7.6 days for closed to open grassland and 14.5 days for open needle-leaved deciduous or evergreen forest. In the dry year, the phase lag increased by 7.0 days on average, but with specific response signals for the different land cover types. Interpreting the phase lag against the soil water-holding capacity, we observed a slightly higher phase lag in the dry year for soils with a higher water-holding capacity. The accuracy of the estimated phase lag was assessed through Monte Carlo simulations and presented reliable estimates for the annual component.