Abstract
A multi-temporal analysis of Landsat panchromatic data was applied to the Usanga wetlands in Tanzania, where there are concerns that irrigated agricultural development is reducing downstream water availability. As there are many small fields, and even smaller areas of continuous vegetable cultivation throughout the year, 15-m resolution multi-temporal Landsat Enhanced Thematic Mapper Plus (ETM+) panchromatic data, acquired in the months of March, May and October, were used to extract information on the seasonal land use/cover in order to investigate agricultural water use. Object-oriented analysis was used to interpret these high-resolution data through definition of image segments, which were then classified using a pixel-based unsupervised classification method. This hybrid approach helps to process high-resolution data quickly and efficiently, so as to economically and effectively understand and classify land use/cover in highly heterogeneous and seasonally dynamic tropical conditions, where land units are typically small.