Monitoring Of Annual Dynamic Changes Of Cultivated Land In Mountainous Areas

The fast-growing population in the world has increased the cultivated land area due to food demand.On the other hand,affected by factors such as high terrain elevation,large slopes,and low agricultural production efficiency,some areas are facing the risk of abandoned land.The abandoned land is particularly serious in mountainous and hilly areas.As a special and important resource,it is important to grasp the trend of cultivated land area change.However,due to the lack of methods to monitor the geographical and spatial distribution of cultivated land using medium and high-resolution remote sensing images,the spatiotemporal distribution of cultivated land in mountainous regions is not clear,and the conversion rules between cultivated land and forestland,grassland,and barren land are unclear.Therefore,this article takes Chongqing,China as an example,and uses ENVI 5.3 software to perform geometric correction,radiation calibration,atmospheric correction,and cloud and shadow preprocessing on the acquired NASA Landsat 8 OLI image data set.NDVI,NDWI,and EVI vegetation index values of pre-processed remote sensing images,and add the three vegetation indices of NDVI,NDWI,and EVI to the pre-processed image data set for interpretation of subsequent remote sensing images.Py Charm 2017.3.4 software was used as the Python language integrated development environment to build four modeling classification methods: random forest(RF),support vector machine(SVM),perceptron,and classification regression tree(Cart)for remote sensing Image interpretation.At the same time,the confusion matrix,Kappa coefficient,and measured sample data were used to evaluate the interpretation accuracy of different models.A method for extracting the spatial-temporal changes in mountainous land based on annual land-use change monitoring was developed.In the past,when using remote sensing data to study the marginalization of cultivated land,the distinction between fallow land and abandoned land was often ignored.The article aimed at the different characteristics of cultivated land as forestland,grassland,and barren land,three identification rules were established: abandoned cultivated land converted to fallow land,abandoned land and rehabilitated land,and the spatial distribution of forestland and grassland converted to cultivated land was calculated.The research shows that the introduction of three vegetation indices enhanced the discrimination of land cover types in the study area.Throughout the study period,among the four classification models,the confusion matrix,Kappa coefficient,and average accuracy of the measured samples of the RF model verification results were 86.18%,0.82,and 87.02%.The stability,accuracy and applicability of image interpretation for different years.The abandonment rate and abandoned land reclamation rate in Chongqing were 17.76% and 33.33%,respectively,and some of the productive arable land was abandoned.Among the newly cultivated arable land,cultivated land is spreading from high-altitude and high-slope areas to low-altitude and low-slope areas with relatively good topographical conditions at the cost of abandoning the original uncultivated land and sacrificing the original forestland and grassland.