Abstract: Taking the junction of the three provinces of Hebei, Shandong and Henan under the influence of "marginalization law" as a study area, based on the 30 m resolution Landsat data, the visible spectrum, near-infrared, mid-infrared band and multiple normalization indices are combined in multiple bands. Using CART decision tree classification method to extract the annual change of abandoned area and abandonment rate in each district and county of the study area, the spatial pattern of abandoned farmland is analyzed to explore the driving factors of abandoned farmland. The results show that the average accuracy of classification is 89.15%. Between 2000 to 2018, the area of abandoned land in the study area fluctuated slightly but the overall trend was declining. In 2002, the area of abandoned land was the largest, and the area of abandoned land after 2015 was maintained at a low level. Most of the abandoned farmland area for the duration of abandonment in the study area only existed for 1~2 years, there were fewer years of long duration of abandoned farmland. Abandoned farmland reclamation rate in study area is relatively good, and the average annual abandoned farmland reclamation rate is 32%. Abandoned farmland in study area is related to the proportion of agricultural population, agricultural production efficiency and government policies.

Key words: abandoned farmland, CART, decision tree, Hebei, Shandong and Henan provinces