Abstract:

Given that the ordinary least square estimate is not adaptable to analyzing the driving mechanism of land use pattern represented by the area percentage data, this paper develops an estimation approach according to the partial least squares regression algorithm capable of analyzing the driving mechanism of land use pattern identified at the pixel specific resolution. The approach can be used to obtain the robust estimates for the determinant model of land use pattern and guarantee that the sum of predicted area of all kinds of land use categories equals the total land area of each grid pixel in addition to deducting the estimation bias resulting from the multi-collinearities between explanatory variables. This paper elaborately deduces the estimation algorithm for the explanatory model for the land use pattern according to the partial least-squares regression algorithm and based on the area percentage data. By using this kind of estimation approach, this paper analyzes the driving mechanism of land use pattern for cultivated land, built-up area and other kinds of land and obtains the estimates with high goodness of fit in Huang-Huai-Hai Plain of China. The research result shows that the partial least squares regression analysis can be used to estimate the driving mechanism of land use pattern represented by the area percentage data. The estimation approach proven applicable in this study is with unbiased, efficient and robust characteristics and worthy of being promoted to use in relevant case studies.

Key words: area percentage data, Huang-Huai-Hai Plain, land use pattern; driving mechanism, partial least squares regression analysis