Object-based locust habitat mapping using high-resolution multispectral satellite data in the southern Aral Sea basin

Three different object-based image classification techniques are applied to high- resolution satellite data for the mapping of the habitats of Asian migratory locust(Locusta migratoria migratoria) in the southern AralSea basin, Uzbekistan. A set of panchromatic and multispectral Systeme Pour l’Observation de la Terre-5 satellite images was spectrally enhanced by normalized difference vegetation indexand tasseledcap transformation and segmented into image objects, which were then classified by three different classification approaches: a rule-based hierarchical fuzzy threshold (HFT) classification method was com- pared to a supervised nearest neighbor classifier and classification tree analysis by the quick, unbiased, efficient statistical trees algorithm. Special emphasis was laid on the discrimination of locustfeeding and breeding habitats due to the significance of this discrimination for practical locustcontrol. Field data on vegetation and land cover, collected at the time of satellite image acquisition, was used to evaluate classification accuracy. The results show that a robust HFT classifier outperformed the two automated procedures by 13% overall accuracy. The classification method allowed a reliable discrimination of locustfeeding and breeding habitats, which is of significant importance for the application of the resulting data for an economically and environmentally sound control of locustpests because exact spatial knowledge on the habitat types allows a more effective surveying and use of pesticides.