Evaluation of accuracy and consistency in a multisource cultivated land dataset: a case study of the Pengqu basin at the northern foot of Mount Qomolangma

Land cover data is an important data source for evaluating cultivated land change. The Pengqu Basin at the northern foot of Mount Qomolangma is a key area for the construction of the South Asia Corridor in my country. Through the study of the intensity of human activities such as the change of cultivated land in this area, it has important strategic significance for the development of the South Asia Corridor, the stabilization of borders, and the maintenance of homeland security. In this paper, 11 sets of land cover data sets are selected to compare and analyze the cultivated land in this area from the three aspects of cultivated land quantity, precision evaluation and spatial consistency. The results showed that: (1) The consistency of 11 sets of land cover data in Pengqu Basin was poor, and the high consistency was only 0.64%. The number of cultivated land in each dataset is also quite different, and the closest to the statistical data are CRLC, Globeland30 and TP datasets, and the OA and Kappa coefficients of the TP dataset are the highest, indicating that all datasets have poor consistency in such high altitude area as the Qinghai-Tibet Plateau are poor. (2) Compared with other data sets, TP data set is the most suitable data set for cultivated land classification in this study area. Its OA and Kappa coefficients are the highest, and there is a relatively accurate cultivated land area. It shows that the higher the spatial resolution of the original data, the higher the accuracy of the data set, and although manual visual interpretation is more time-consuming, it can identify the fragmented and difficult-to-identify cultivated land. (3) The findings of this study can serve as a valuable reference for reconstructing the distribution pattern of historical cultivated land, providing insight into human activities in the region, and providing feedback for data producers. In the future, targeted improvements in classification techniques and methodologies for regional cultivated land datasets are needed to enhance data accuracy.