Progressive development of vegetation resources and its relationship to animal use in Amboseli ecosystem, Kenya : a remote sensing approach
NDVI and rainfall data covering 1982-90 of Amboseli ecosystem, Kenya were analysed with a view to detecting patterns which could be related to the structure and function of vegetation types. NDVI data were first normalised using a square root transformation and the 1988-89 integrated NDVI was calculated for months from August 1988-July 1989, which represented a biologically active year. Integrated NDVI was used to extract quartiles which were used in a linear mixture model to derive four vegetation classes: grassland, shrubland, bushland and woodland. Vegetation classes were further examined against mean annual and monthly NDVI/rainfall. A lag time analysis of peak NDVI/rainfall was carried out and the probability of any one month's contribution to the peak NDVI/rainfall assessed. Two hypotheses relating animal abundance and distribution to vegetation were developed and tested with Systematic Reconnaissance Flight animal data collected in March 1983, April 1986, March 1987 and May 1990. The animal species selected were Wildebeest, Zebra and livestock. The hypothesis related to animal abundance was tested using a one-way analysis of variance, while the hypothesis related to animal distribution was tested by a Chi-squared analysis. Finally, a correlation was calculated between integrated NDVI of a vegetation class and animal density for each species. The results show that in 1982-90, good rainfall years were 1988-90 with bimodal peaks occurring in April-May (long rains) and November-December (short rains). In general, woodlands showed highest NDVI, followed by bushlands, shrublands and lastly grasslands. Analysis of vegetation classes temporal profiles showed that dry years 1982-83, 1983-84 and 1986-87 had only one peak of rainfall while 1984-85, 1985-86, 1987-88, 1988-89 and 1989-90 had two peaks of rainfall. Years with bimodal peaks showed better vegetation classes NDVI compared to years with only one peak of rainfall. There were very high correlations between vegetation classes cumulative mean monthly NDVI against cumulative mean monthly rainfall. The lag time between peak rainfall and peak NDVI is usually two months, except in dry years, e.g. 1982-83/84 when it was 3 months and one month respectively. The peak rainfall (99%) occurs in April (long rains), November and December (short rains) while January, February and June contributed to the peak NDVI (99%).