农业工程学报2025，Vol.41Issue(2)：271-280,10.DOI:10.11975/j.issn.1002-6819.202408131
结合RF和LandTrendr算法的黄河流域下游撂荒耕地提取与时空变化分析
Extraction and spatiotemporal analysis of abandoned cultivated land in the Lower Reaches of Yellow River Basin using RF and LandTrendr algorithms
摘要
Abstract
Cultivated land is one of the most basic natural resources in agricultural production for human survival.However,the large-scale cultivated land has been abandoned in China,due to the rural migration to the non-agricultural employment in the urban areas during rapid urbanization in recent years.Taking the vital grain production base in the Yellow River basin as the study area,this study aims to accurately and efficiently obtain the spatial distribution of abandoned cultivated land,particularly for national agricultural production and food security.The downstream regions of the Yellow River were specifically selected as Lankao,Changyuan,Fengqiu,and Dongming County.Landsat imagery was utilized to collect the image data from March to October,covering the years 2003 to 2023.The images were then preprocessed in this period.According to the differences among cultivated land and the rest land types,the vegetation indices,spectral,texture,topographic,and Kauth-Thomas features were extracted to construct the datasets for each year.Stratified sampling and visual interpretation were combined to obtain the sample datasets.The random forest and LandTrendr models were employed to extract the abandoned and fallow lands,excluding urban areas.The accuracy of the improved model was then validated on the abandoned cultivated land using F1-Score.The influencing factors were finally determined on the abandoned cultivated land.The results indicate that the EVI_p80,NDVI_p80EVI,and nir variables shared high importance and discriminability in the output using random forest.Statistical analysis was also performed on the area of abandoned and fallow land within the study area from 2004 to 2022.The results revealed that the area of abandoned cultivated land was found as a trend of initial increase followed by fluctuating decline,with an average area of abandoned cultivated land of 28.24 km².The maximum area of abandoned cultivated land was reached in 2007,amounting to 49.35 km².The abandonment area was effectively controlled after 2020.The area of fallow land was found to be relatively small and scattered.Furthermore,the F1-score values of 0.87 and 0.89 were achieved in the abandoned and non-abandoned cultivated land,respectively,indicating higher accuracy,compared with the land use/cover change extraction.A multivariate regression analysis was conducted to investigate the influencing factors on the abandoned cultivated land.A systematic investigation was implemented to incorporate the total agricultural production value,total mechanical power,grain production,and the number of employed individuals in the primary industry,particularly in relation to the area of abandoned cultivated land.The increasing level of agricultural mechanization can reduce the abandonment of cultivated land,while grain production can raise the risk of abandonment.Some regulations are required to prevent land abandonment.The cultivated land can be effectively protected to reduce the area of abandoned cultivated land,in order to ensure the national food security and the sustainable development of agriculture.This finding can provide an effective technical approach to monitoring the abandoned cultivated land in the Yellow River basin.The spatiotemporal patterns and driving factors of cultivated land abandonment can also offer valuable references to formulate the decision-making on agricultural production in China.关键词
黄河流域/耕地撂荒/粮食安全/时空变化Key words
Yellow River Basin/ccultivated land abandonment/food security/spatiotemporal changes分类
农业科技引用本文复制引用
郭益梦,何珊,邵怀勇..结合RF和LandTrendr算法的黄河流域下游撂荒耕地提取与时空变化分析[J].农业工程学报,2025,41(2):271-280,10.基金项目
四川省青年科学基金项目(24NSFSC4165) （24NSFSC4165）
