Abstract
Irrigation infrastructure development for smallholder farmers in developing countries increasingly gains attention in the light of domestic food security and poverty alleviation. However, these complex landscapes with small cultivated plots pose a challenge with regard to mapping and monitoring irrigated agriculture. This study presents an object-based approach to map irrigated agriculture in an area in the Central Rift Valley in Ethiopia using SPOT6 imagery. The study is a proof-of-concept that the use of shape, texture, neighbour and location information next to spectral information is beneficial for the classification of irrigated agriculture. The underlying assumption is that the application of irrigation has a positive effect on crop growth throughout the field, following the field's borders, which is detectable in an object-based approach. The type of agricultural system was also mapped, distinguishing smallholder farming and modern large-scale agriculture. Irrigated agriculture was mapped with an overall accuracy of 94% and a kappa coefficient of 0.85. Producer's and user's accuracies were on average 90.6% and 84.2% respectively. The distinction between smallholder farming and large-scale agriculture was identified with an overall accuracy of 95% and a kappa coefficient of 0.88. The classifications were performed at the field level, since the segmentation was able to adequately delineate individual fields. The additional use of object features proved essential for the identification of cropland plots, irrigation period and type of agricultural system. This method is independent of expert knowledge on crop phenology and absolute spectral values. The proposed method is useful for the assessment of spatio-temporal dynamics of irrigated (smallholder) agriculture in complex landscapes and yields a basis for land and water managers on agricultural water use.
NSTL
Elsevier