| With the rapid increase in the costs of rural labour and the adjustment of planting structures,the phenomenon of farmland abandonment has appeared in China.It is of great significance to promptly and accurately grasp the information on dynamic temporal and spatial changes in abandoned farmland to ensure national food security and the sustainable use of cultivated land.Aiming at the problem of low precision of abandoned land extraction caused by complex land cover and land fragmentation,this paper proposes a method of extracting abandoned land information based on multi-temporal collaborative change detection.Taking Luquan District,Shijiazhuang City,Hebei Province as the study area,Using Sentinel 2A and Landsat 7multi-spectral images,supported by field samples,the seasonal changes of Normalized Difference Vegetation Index(NDVI)of various types of cultivated land cover were analyzed.Based on the classification system of seasonal abandononment,perennial abandononment,winter wheat,and perennial garden land,a multi-temporal collaborative change detection model was constructed to monitor the abandoned farmland status in the study area by remote sensing.At the same time,using random forest classification and vegetation index change detection methods,the rules of abandoned land identification were formulated,and the spatial and temporal pattern evolution of abandoned land under different location conditions in the study area in recent ten years was analyzed.According to the extraction results of abandoned land,the rules of reclamation extraction of abandoned land were formulated,and the reclamation situation and frequency of abandoned land in the study area were further analyzed.The results are as follows:(1)A method based on multi-temporal collaborative change detection model was proposed,which provides a fast and effective method for abandoned farmland extraction.The total accuracy of abandoned farmland extraction from sentinel images with 10m resolution was 92.19%,and the Kappa coefficient was 0.88;The total accuracy of the abandoned farmland extraction from Landsat 7 images with 30m resolution was 86.33%,and the Kappa coefficient was 0.80.It has high classification accuracy for images with different resolutions.(2)The overall accuracy of the land use classification map from 2010 to 2020 was between 91.81%-96.92%,and the Kappa coefficient was between 0.89-0.96,which has high accuracy and can be used as the basic data for abandoned farmland identification.(3)The maximum abandoned area was 3906.02 hm~2,the minimum was 1618.74 hm~2,the highest abandoned rate was 14.09%,the lowest was 5.83%.Among them,the abandonment rate was high from 2012 to 2017,both of which were above 10%.In recent years,the abandonment rate has declined,and the abandonment area in the study area increased first and then decreased during the monitoring period.(4)The arable land with slope from 6°to 15°has the highest abandoned rate,with an average of 29.19%.The abandoned rate of cultivated land with slope≤2 was the lowest,with an average of 7.65%.The average annual abandonment rate was 12.55%at 0-200m altitude,29.10%at 200-400m altitude,and 5.72%at 400-600m altitude.The abandonment rate of cultivated land in the study area was most likely to occur at 200-400m altitude.(5)The maximum reclamation area of abandoned farmland was 291.49 hm~2,and the minimum was 34.94 hm~2.The highest reclamation rate was 14.26%,the lowest was 0.95%,and the annual average reclamation rate was 4.67%.The overall reclamation rate was low,with reclamation mainly being concentrated in the plain areas,which are suited to cultivation;in contrast,the abandoned cultivated land in the western mountainous areas showed almost no signs of reclamation.(6)The areas with a high frequency of abandonment were mainly concentrated in the western mountainous area of Luquan,while the eastern plain had a relatively low frequency of abandonment.The area of abandoned farmland gradually decreased as the frequency of abandoned farmland increased. |
PDF Full Text Request