AVHRR NDVI for monitoring and mapping of vegetation and drought in East African environments

Abstract: In this thesis, an assessment is made of the performance of the Normalized Difference Vegetation Index (NDVI) derived from coarse resolution data from the NOAA AVHRR sensor, with particular reference to mapping and monitoring of drought and vegetation in East African vegetation and climatic conditions. Methods for analysing time series of satellite data are investigated. By studying the empirical relationship between long term averages of rainfall and NDVI, a strong spatial relationship is established. The influence of additional environmental factors on the relationship is studied, and prediction models are established. The temporal relationships between NDVI and rainfall are investigated for individual rainfall stations, by formulating distributed lag models. These relationships are found to be very weak, on 10-day, monthly and annual basis. Noise, or random variation in the NDVI data, is estimated using a geostatistical method. Noise levels are found to be very high for individual scenes, but low for a long term average NDVI image. The studies suggest that the capability of AVHRR NDVI for vegetation monitoring is very limited, although the data can be used for mapping of spatial vegetation distribution. Fourier Series are used to extract the seasonality of vegetation, timing and length of the growing seasons, minima, maxima and rates of increase and decrease in the NDVI data. Applying this method, very important parameters for vegetation mapping and for global and regional ecosystem modelling, can be estimated. It is concluded that, although NDVI data from NOAA AVHRR have serious limitations for monitoring, they are an important source of information on the spatial distribution of global and regional vegetation.