不同尺度转换方式对土地利用格局模拟的影响效应研究

Exploring Effect of Different Scaling Methods on Simulation of Land Use Patterns

以张家界市永定区为研究对象,根据2005年土地利用现状图和数字高程模型数据源得到的土地利用、地形、河流以及道路等空间数据,在不同的尺度聚合方法下构建了多种尺度上的土地利用格局模拟模型,以试图探讨不同尺度转换方式对土地利用格局模拟的影响。研究结果显示:①在空间尺度逐渐增大过程中,中心值原则(RCC)下,模型的拟合优度(ROC值)变化相对较大,表现出很强的随机性,在400 m×400 m和700 m×700 m的尺度时,ROC值出现明显的波动;②面积最大原则(RMA)和优化组合方法(OIM)下,随着聚合尺度的升高,模型的拟合优度值变化相对较为平缓,且ROC值逐渐增大的趋势比较明显;③比较RCC、RMA和OIM三种方法,优化组合方法(OIM)具有更高的拟合优度,在各种聚合尺度上,土地利用类型的ROC值均得到了明显的提高(耕地、林地和建设用地的平均ROC值分别提高到0.767,0.785,0.895);④各种尺度转换方法下,在1 700 m×1 700 m尺度时各种地类格局的拟合均出现ROC的峰值,可以认为1 700 m是永定区土地利用格局的最佳模拟尺度。研究结果对区域进行土地利用的规划、制定相应的土地开发与利用政策等也具有一定的参考意义。

Land uses in Yongding county,which is one of the typical karst mountain area in northwestern Hu＇nan province,were investigated by means of modeling the land use types with different scaling methods at multiple scales with the purpose of studying on the scale effect of scaling with regional land use saptial data,and using the model to simulate and analyze the spatial land use patterns in Yongding county.The following results were obtained：（1） The ROC values of different land use types from raster data with the rule of centric cell（RCC） varied widely when spatial scale increases and appeared strong randomicity.The ROC values showed obvious fluctuation at 400 m×400 m and 700 m×700 m scales.（2） The ROC values appeared less fluctuation under the rule of maximum area（RMA） and the optimized integration method（OIM） with the increase of the aggregation levels and the increasing trend of ROC values were obvious.（3） Compared with RCC and OIM,the results indicate that the land use simulation model showed better goodness of fitting and higher accuracy of fitting under the scaling method of OIM.The area under ROC curves（AUC） of cultivated land,forest land and construction land were improved to 0.767,0.785 and 0.895 respectively when using OIM.（4） The simulation of land use patterns all showed maximum values of ROC at 1 700 m×1 700 m scale in different scaling methods,so it is argued that the 1 700 m×1 700 m is an significant level for simulating the distribution of land use types in the study area.The results will also provide some reference for the land use planning and ploicy-making of sustainable utilization of land resources.