Sampling And Precision Of Land Use Based On Spatial Autocorrelation And Assistant Variable

In order to accessing the status of land resources and taking effective prevention measures according to local conditions,land use survey must be carried out.It is of great significance to develop and use land use resources reasonably,maintain the stability of the ecological environment,and drive for economic growth quality.Affected by natural factors and socio-economic factors,land use has spatial autocorrelation and variability in spatial distribution.However,traditional land use sampling survey methods do not consider its spatial characteristics,and there are defects in sample design and overall estimation.With the development of remote sensing technology,spatial sampling methods that combine remote sensing images with spatial autocorrelation theory have been widely used in land use surveys in many countries,effectively improving the accuracy of sampling surveys.In view of the above problems,this paper selects Mengyin County in Yimeng mountain asresearch area.Using GF-1 and GF-6 satellite remote sensing images to extract land use information,and using spatial autocorrelation theory to analyze the spatial autocorrelation characteristics of different sampling unit sizes.And designing the best sampling unit size combined with the coefficient of variation.Four spatial sampling methods including spatial simple random sampling,spatial systematic sampling,spatial stratified sampling,and sampling based on representative samples are used for land use sampling.And use sample representativeness and sampling cost to determine the best sampling plan.Different spatial interpolation methods are used to carry out spatial prediction research on the degree of land use,so as to realize the rapid investigation and spatial prediction of land use status.The main conclusions are as follows:(1)This study designed five sampling unit size plans of 200 m×200 m,300 m×300 m,400m×400 m,500 m×500 m and 600 m×600 m.The soil erosion modulus was used as auxiliary variable.Global Moran's I and Local Moran's I were used to analyze the spatial autocorrelation and variability of soil erosion modulus.It was found that the soil erosion modulus showed strong spatial autocorrelation characteristics(Global Moran's I>0.5),and the local autocorrelation types were all HH types.Under different sampling unit sizes,the coefficient of variation decreases as the sampling unit size increases,and the change of it was stable gradually(less than 7%),and the average number of spots increases as the sampling unit size increases.A comprehensive comparison of the two spatial autocorrelation indexes,the coefficient of variation and the average number of patches in the unit,and finally select a 400 m×400 m grid as the optimal size of the sampling unit.(2)Four sampling plans(spatial simple random sampling,spatial systematic sampling,spatial stratified sampling and sampling based on representative samples)were used to sample the land use in the study area based on the 400 m×400 m sampling unit.In order to evaluate the applicability of these four methods,Sample representativeness and sampling costwere used as evaluation indicators.The sampling method based on representative samples is the most suitable method.Because this method comprehensively considers natural factors and social factors.The sampling accuracy of this method was 96.64%.The absolute error of the land use area of this method was less than 3.0%,and the coefficient of variation was 91%.The sampling cost was lower.(3)Further research was taken on the basis of sampling based on representative samples.Four interpolation methods including inverse distance weighting method,spline function method,kriging method and co-kriging method(slope is an auxiliary variable)with different mutation models are used to interpolate the degree of land use in the study area.The cross-validation results show that when the same method is used for interpolation,the selection of parameters or models has a greater impact on the accuracy of interpolation.In general,the accuracy of ordinary kriging and co-kriging is higher.In terms of spatial distribution,with the aid of slope,the spatial distribution of co-kriging is more detailed.The inverse distance weighting method is susceptible to extreme values and the "bull's eye" phenomenon appears.The surface generated by the spline function method is relatively rough.(4)Analyzing the relationship between natural factors,socio-economic factors.and land use in the study area.It is showed that there is a correlation between land use degree and altitude,slope,landform type,soil type and other natural factors.In the study area,the degree of land use decreases with the increase of slope.When the altitude is less than 700 m,the intensity of land use decreases with the increase of altitude.And in areas with an altitude of higher than 700 m,the degree of land use is low and continues to be stable.There are differences in the degree of land use between different landforms or altitudes.There are also correlations between the population density and per capita disposable income of each township and village,and the degree of land use,but the correlation coefficients are low,0.32 and 0.42,respectively.