The Pearl River Delta is experiencing a fast urban growth in recent years which is responsible for rapid loss of the valuable agricultural land in the region. There is a great need to monitor the urban expansion using remote sensing for urban planning and management purposes. However

it has been well recognized that there is significant over-estimation of land use change inusing multitemporal images in change detection. The problem is due to inadequate creation of classification signatures in the classification of remote sensing images. This paper presents an improved method using principal components analysis of stacked multi-temporal images. It is found that this method can reduce the errors in land use change detection and provide a very useful method in monitoring land use changes in the Pearl River Dalta.