Improving the estimation of noise from NOAA AVHRR NDVI for Africa using geostatistics

The accuracy of NOAA AVHRR NDVI data can be poor because of interference fr om several sources, including cloud cover. A parameter of the variogram mod el can be used to estimate the contribution of noise from the total variati on in an image. However, remotely sensed information over large areas incor porates non-stationary (regional) trend and directional effects, resulting in violation of the assumptions for noise estimation. These assumptions wer e investigated at five sites across Africa for a range of ecological enviro nments over several seasons. An unsupervised spectral classification of mul ti-temporal NDVI data partially resolved the problem of non-stationarity. Q uadratic polynomials removed the remaining regional trend and directional e ffects. Isotropic variograms were used to estimate the noise contributing v ariation to the image. Standardized estimates of noise ranged from a minimu m of 18.5% in west Zambia to 68.2% in northern Congo. Cloud cover and atmos pheric particulates (e.g. dust) explained some of the regional and seasonal variations in noise levels. Image artifacts were also thought to contribut e noise to image variation. The magnitude of the noise levels and its tempo ral variation appears to seriously constrain the use of AVHRR NDVI data.