Abstract: Locust habitats have a close relationship with occurrence of locust plagues. Correct classification of locust habitats into types is a key step to study the relationship between locust and its habitat in prediction and early-warning of locust plagues. Based on the multi-temporal Landsat TM data of the New Bohai Development Zone of the years of 1992 -2010, locust habitats in the three typical historical periods in that region were classified and the modified locust habitat type extraction method and factors affecting occurrence of locust plagues were analyzed, taking into account the meteorological data of the period. Results show that this modified method makes full use of the vegetative phenological information in the multi-temporal data and fairly well solve the problems of the same-object-has-different-spectra-simultaneously and different-objects-have-the-same-spectrum in identifying critical habitat types. The locust habitat type classification maps of the periods reached 89.92%, 90.84% and 92.23% in overall accuracy, with Kappa factor being 0.8695, 0.8809 and 0.9047, respectively. Comparative analysis shows that during the three periods reed fields remained almost unchanged in area, while grasslands expanded to 596.2 km² at the time of a large outbreak. Further analyses reveal that outbreaks of locust plagues during the period were a result of the joint effects of habitat types and climatic factors in the research area. For instance, in 2000, the outbreak of locust plague was mainly attributed to the higher temperature, less rainfall and expansion of grassland in area; while in 2009 a much small outbreak was to the expansion of alfalfa fields, shrinkage of grassland and more rainfall in the study area.

Key words: locust, habitat type, remote sensing, oriental migratory locust