基于栅格面积成分数据的土地利用格局解释模型稳健估计

Robust Estimation for the Determinant Model of Land use Pattern Based on the Area Percentage Dataset of Land Uses

针对最小二乘估计不能应用于栅格尺度以面积成分表征的土地利用格局驱动机理分析的难题,本文提出了一种利用偏最小二乘回归法稳健估计该类土地利用格局解释模型的方法。利用该方法能在解释变量间存在多重共线性的情况下,获得基于栅格面积成分数据的土地利用格局解释模型的稳健估计。本文推导了应用偏最小二乘回归分析的数据处理和建模估计过程,并运用该方法开展了针对黄淮海地区耕地、建设用地分布格局及其驱动因子的建模分析,得到了拟合优度高的估计结果。研究表明,偏最小二乘回归分析方法在开展栅格尺度以面积成分表征的土地利用格局驱动机理分析时具备高效与稳健的特征,适宜在类似研究中推广应用。

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.