Testing temporal mixture analysis for quantification of land cover at sub-pixel level

With the rapid increase of the population, the quantification of the impact from the human activities becomes critical not only for the environmental science but also the urban engineering under the global urbanization and climate change. This paper focused on the estimation of two human activity-related indexes, fractions of impervious surfaces and crop fields with remote sensing technology. Temporal mixture analysis (TMA) was applied to the sorted NDVI profile from MODIS to calculate the fractions of impervious surfaces and crop (plowed and paddy) fields in Japan. The optimal length of the sorted NDVI profile for accurate estimation was also investigated. The accuracy of fractional impervious surfaces was assessed by impervious surface map made by Landsat images; fractional plowed and paddy fields were assessed by local geographic database. The smallest RMSEs of the TMA were 0.096 for impervious surface, 0.177 for plowed field and 0.141 for paddy field, respectively. In summary, TMA has the potential to map impervious surfaces and crop fields over large areas.