The Spatial Structure Of Agricuitural Land Price Based On GWR Model

The Agricultural land price plays an important role in the social and economic developments, and it has great practical reference value for optimal allocation of land resource and land-use structure. The analysis of agricultural land price has long been focused on the traditional methodologies, which ignore the spatial attributions. Besides, the traditional agricultural land price calculation is always based on the whole structure, which is inappropriate due to the diversities of influence factors and their heterogeneities in spatial region.Based on the input-output data of agricultural land, the Income Approach Investment Method, Geographic Information System(GIS) and Geographically Weighted Regression Model (GWR), this paper discusses different influence factors and analyzes their impacts on the spatial structure of agricultural land price. The main aspects and their results as following:(l)The price calculation and cartography for sample points. On the basis of GUCHENG’s input-output investigation data of agricultural land from2006to2008, the pure income was calculated for all sample points. After deleting the oversize and undersize outliers,262sample points were saved. The reinstate rate which equals to4.5%in GUCHENG was subsequently determined based on the method of Security Interest Rate plus Risk Adjusted Value. Finally, the agricultural land prices of sample points were calculated and analyzed.(2) The non-stationary test for agricultural land price space. By using semi-variogram, the distribution of agricultural land price in four directions including0°,45°,90°and135°, were analyzed. The results showed that the geometric anisotropy existed. In other words, the change of agricultural land price was non-stationary, which was the precondition of applying GWR model to analyze the relationship between agricultural land price and their influence factors.(3)The calculation of GWR model. Based on Delphi method, eight influence factors related to society, nature and location were selected, including the influence degrees of national roads, provincial roads, county roads and towns, as well as guaranteed rate of water resource, the density of road network in farmland, per capita average agrarian area, slope. The GWR model was built after quantizing the influence factors. Then, four comprehensive methods, including Fixed Spatial Kernels combined with Cross-Validation(CV), Fixed Spatial Kernels combined with Akaike Information Criterion(AIC), Adaptive Spatial Kernels combined with Cross-Validation(CV), Adaptive Spatial Kernels combined with Akaike Information Criterion(AIC), were applied to calculate the GWR model. The results showed that the comprehensive method of Adaptive Spatial Kernels combined with Akaike Information Criterion(CV) had the highest fitted value.(4) The spatial structure analysis of agricultural land price. With the support of GWR function in ArcGIS software, the raster maps were created for eight influence factors’regression coefficients in GWR, which could intuitively show the influence degree of each factor to the agricultural land price in location region of GUCHENG.(5) The precision test of GWR model. The comparison of GWR model and Kriging model was conducted. After randomly selecting10%points from the total262sample points, the differences between their real values and the predicted values of agricultural land price generated from two models were calculated, which showed that the GWR was more precious and reliable in practice.