A Patch Area Scaling Model Based On Large Sample

As an important “lens” for geographers to observe and understand the real world,scale is one of the basic issues in geoscience research.Scale effect is an important aspect of the scale issues,which is widespread in the research of geographic information science,remote sensing science,ecology,hydrology and meteorology,and has received more attention.Landscape pattern indicators change with the change of grain is the major form of the scale effect of raster data.As a basic unit of the landscape,patch area changes with the scale.Even the scale span is small,the change is significant.Therefore,investigating the relationship between patch area and scale is crucial to adequately characterize and analyze the scale effect.A study has shown that patch geometry has a decisive influence on the evolution of patch as up scaling.it also proposes an estimation model for interpreting and estimating the evolution of patch and the different type area as scale changes in raster class data aggregation.However,this model is based on a small number of samples and lack of evaluation and analysis.Based on this model,we takes the GlobeLand30 data mode aggregation as an example to build an areascale evolution model based on large samples at the patch level,and introduces an area-tostandard deviation model in order to further calibrate and optimize he original model.Firstly,based on a small number of samples,we build a patch area scaling model and a corresponding accuracy calibration model.The model and its accuracy model was further calibrated at patch and verificated the change rate at single category area.The verification results show that the model can well explain and estimate the patch area evolution and predict the accuracy of the model and explain the uncertainty of the model.The result also shows that though the accuracy of the model verification is high under the condition of a small number of samples,there are still problems such as a large proportion of empty grids and poor samples within the grid.Secondly,we analyzed the main influencing factors of model accuracy,and the impacts of sample capacity and grid size was investigated,respectively.The l results show that by increasing the samples,the statistical distribution of samples within the grid is effectively improved.The step length of logarithmic relative support radius is 0.02 and the fill step is 0.05 is identified an approprite grid size range for modeling.In this range,the overall accuracy of the model is stable as the grid size decreases.Finally,we constructed the patch area scaling model and its accuracy model which based on a large sample size by significantly increasing the sample capacity.the evolution of patch area scale was analyzed from two aspects of the overall model and single factor change,respectively.The model and its accuracy model was further calibrated at patch and verificated the change rate at single category area.Compared with the area-scale evolution model developed on a small number of samples,the large-sample model has a better statistical distribution of samples,and the accuracy of the model based on the large sample has obviously improved in the patch and class synthesis.