LARGE- AND SMALL-SCALE CROPLAND CLASSIFICATION ON THE FOOTHILLS OF MOUNT KENYA BASED ON SPOT-5 TAKE 5 DATA TIME SERIES

摘要

The SPOT-5 Take 5 campaign provided SPOT timernseries data of an unprecedented spatial and temporalrnresolution. We analysed 29 scenes acquired betweenrnMay and September 2015 of a semi-arid region in thernfoothills of Mount Kenya, with two aims: first, torndistinguish rainfed from irrigated cropland andrncropland from natural vegetation covers, which showrnsimilar reflectance patterns; and second, to identifyrnindividual crop types. We tested several input datarnsets in different combinations: the spectral bands andrnthe normalized difference vegetation index (NDVI)rntime series, principal components of NDVI timernseries, and selected NDVI time series statistics. Forrnthe classification we used random forests (RF). Inrnthe test differentiating rainfed cropland, irrigatedrncropland, and natural vegetation covers, the bestrnclassification accuracies were achieved using spectralrnbands. For the differentiation of crop types, wernanalysed the phenology of selected crop types basedrnon NDVI time series. First results are promising.