Land cover change detection in Chinese Zhejiang Province based on object-oriented approach

Many factors such as geomorphic features, economic development, expansion of cities, the implementation of new policy, etc. are changing land cover. Therefore, it is necessary to monitor Land-Use and Land-Cover Change (LUCC) by using new technologies and methods at high tempo-spatial resolution. Based on one supervised classification approach combining object-oriented method and binary decision tree, this study mapped land cover of Zhejiang Province, China, at the scale of 1:250000 in 2000, 2005, and 2010. After image segmentation, object features such as normalized difference vegetation index (NDVI), modified normalized difference water index (MNDWI), area, the ratio of length to width, density, etc. were applied to assign each object to specific class. Being verified through confusion matrix, the mapping results were satisfactory. Taking land cover map in 2010 as an example, the lowest user accuracy was 84.14%, with the average of 92.15%; The lowest production accuracy was 62.00%, with the average was 86.69%; The overall accuracy was 0.8928; The Kappa coefficient was 0.8752. Under the influence of geomorphic features and economic development, changes of land cover in Zhejiang were mainly distributed in the areas with lower elevation and higher GDP from 2000 to 2010. Under the influence of natural factors and human activities, many croplands and wetlands were lost from 2000 to 2010. For croplands, there were 2106.608 km² croplands changed into other types from 2000 to 2005, and 1897.809 km² from 2005 to 2010. Most of croplands were changed into artificial lands, with 1520.601km² from 2000 to 2005 and 1446.826 km² from 2005 to 2010. For wetland, there were 209.085 km² wetlands changed into other types from 2000 to 2005, and 292.975 km² from 2005 to 2010. Most of wetlands were changed into croplands, with 134.652 km² from 2000 to 2005 and 122.979 km² from 2005 to 2010.