Study On The Spatial Distribution Characteristics Of Multi-Scale Farmland And Its Prediction By Spatial Regression Model

The Land-Use and Land-Cover system has strong spatial dependency and spatial non-stationary, also the formation and changes of farmland spatial pattern was influenced by natural environment factors and socio-economic factors, so these driving factors have different influence to farmland in different time and geographical location. It has important significance to simulate and forecast farmland area throng choosing the suitable models and driving factors, and this can provide theory to protect farmland. The suitable farmland models are not only influenced by base models and driving factors, but also influenced by the study region, especially the minimal research unit. This paper want to construct the suitable farmland spatial models by discussing the influence of base models/driving factors and study scale.This paper takes spatial auto-regression models and geographical weighted regression model as the base model, to explorer the difference between global spatial model and local spatial model. In the aspect of driving factors, this paper chosen natural environment factors and socio-economic factors as the influence factors, and given different weighted coefficients when constructed the farmland spatial models. In the aspect of study region, this paper chosen different scales as the suitable scale to construct the farmland model. This paper got the following conclusions:(1) Constructing farmland spatial model through spatial regression model in different scales, and got the correlativity between farmland and other factors. The result shows that slope and human activities has great influence to the formation and changes of farmland. Through comparing the difference of goodness-of-fit and residual sum of squares(SSE) between spatial auto-regression model and GWR model, both of the models have the same goodness-of-fit, but have great difference in the SSE and GWR model has good simulation and smaller residual error.(2) This paper constructed farmland empirical model, farmland prediction model and fusion farmland prediction model to simulate and forecast the spatial distribution pattern of farmland from1999to2009, and checking the prediction accuracy of farmland models. Besides this paper forecasted the farmland area of different villages and towns of Gucheng by fusion farmland prediction model in2014, the results shows although the farmland area has downward trend and the speed was gradually slow down, and the model analyzed the influence of driving factors to the changes of farmland. This provide theory policy and technical support.