| Cultivated land has always been a necessary means of production to ensure human survival.However,with the continuous advancement of our country’s industrialization and urbanization,the phenomenon of cultivated land abandonment has emerged in various places and intensified,seriously affecting our country’s food supply and security.At present,remote sensing technology is often used as a means of monitoring abandoned cultivated land at home and abroad.However,in some hilly and mountainous areas with variable climate and complex terrain conditions in the south,most of the remote sensing data obtained are limited by the unavailability of many natural factors.Existing research is still difficult to carry out effective long-term monitoring of such areas.[Objective] In order to explore a set of abandoned cultivated land identification and monitoring technology system suitable for such areas,this paper takes Minqing County,Fujian Province,a representative province in the southeastern hilly area of our country,as an example,based on the Google Earth Engine(GEE)platform,using the second The random forest model was nested and run,and then the LandTrendr algorithm and the TW/TC-DTW algorithm were used to identify and compare the abandoned land,and then draw the time-series distribution map of the abandoned land in Minqing County to study its temporal and spatial distribution characteristics.[Content] This article mainly focuses on Landsat series satellites(4,5,7 and 8)and Sentinel-2 binary star images,and screens the available images from 2000 to 2020 through many remote sensing preprocessing methods;NDVI,BSI,NBR and The time-series point diagram of brightness,humidity and greenness of tasseled cap transformation selects 5 types(farmland,woody,herbaceous,building & bare land and water body)land cover stable calibration samples as the judgment benchmark;it is generated by random forest pre-classification" "Permanent soil cover mask",the original sampling data set is generated in the form of semi-automatic sampling for formal random forest classification;after evaluating the classification accuracy,the LandTrendr algorithm and TW/TC-DTW algorithm are used to realize the identification and monitoring of abandoned land and fallow land.[Method] Random forest algorithm was used for land use classification,and LandTrendr algorithm was used for monitoring and identification of abandoned land/fallow land,supplemented by TW/TC-DTW algorithm.[Results] The interannual land use classification map and the spatio-temporal distribution map of abandoned cultivated land within the research period of Minqing County were generated.[Conclusion] Based on scientific analysis methods and rigorous research attitude,this paper mainly draws the following conclusions:(1)The classification results of the quadratic nested random forest model show that the overall accuracy and the recognition accuracy of cultivated land are both above 90%,which has a significant(2)The GEE-based dynamic monitoring technology system for abandoned cultivated land has a good identification,monitoring and overall planning effect in county-level administrative divisions;(3)LandTrendr algorithm is very effective in detecting changes in land use categories;(4)combined with The TW/TC-DTW algorithm,the complementarity and mutual calibration of two different algorithms can effectively prevent the misjudgment and misjudgment of abandoned land and fallow land;(5)The total area of cultivated land in Minqing County from 2000 to 2020 is roughly doubled with 2010 as the boundary.Gradient transition decreases,and the main part of cultivated land remains stable throughout the year;(6)The average abandoned area in Minqing County from 2000 to 2020 is5301.13 mu,and the average abandoned rate is 2.06%.The abandoned phenomenon mainly occurs in Baizhong Town and Dongqiao Town and Meixi Town. |
PDF Full Text Request