Developing a satellite-based method of landscape drought assessment.

摘要

This dissertation describes the major research components involved in the implementation, modification, and testing of a method for assessing surface moisture status with NOAA-AVHRR data, known as the Surface Moisture Index (SMI). Chapter two presents the initial landscape-scale development of the SMI model which is predicated on previous large-scale studies reviewed in chapter one. The initial landscape model relies on a moving neighborhood analysis that exploits the relationship between radiant surface temperature (T s) and a spectral vegetation index, namely the Normalized Difference Vegetation Index (NDVI), to elucidate surface moisture status.; Chapter three describes the Water Deficit Index (WDI), another large-scale technique, the logic of which was adapted and used to modify the original SMI model. The WDI logic dictates the incorporation of near surface air temperature (Ta) as an input for the model. Since Ta is a critical component, a method for estimating this variable was necessary. Chapter four discusses the implementation and testing of the Ta technique used in subsequent model development. Three methods of estimating Ta were tested. These methods differed only in the manner that earth-sun-sensor geometry was used to impose constraints in the model calculations. The most rigorous method, which had the most sophisticated geometry constraints, resulted in the fewest actual predictions, but also showed the best results (R = 0.742, MAE = 6.09°C).; The fifth chapter treats the final version of the SMI model. That model incorporates four significant revisions from the original model. Those are: (1) adaptation of the WDI logic for landscape-scale implementation, (2) inclusion of geometric constraints, (3) near-surface air temperature estimates, and (4) specification of a new landscape reporting unit. Validation exercises relied on a ten year comparison of SMI to the Palmer Z index, a popular index for tracking drought that relies on data from meteorological stations. The ability to predict Palmer Z from SMI values was assessed using the Autoregressive Integrated Moving Average (ARIMA) technique and resulted in excellent predictability with all coefficients being significant at the alpha = 0.05 level.