In pursuit of Sustainable Development Goals (SDGs), land cover change (LCC) has been utilized to explore different dynamic processes such as farmland abandonment and urban expansion. The study proposed a multi-scale spatiotemporal pattern analysis and simulation (MSPAS) model with driving factors for SDGs. With population information from the census, multi-scale analysis criteria were designed using the combination of administrative and regional divisions, i.e., district, province, nation and ecological region. Contribution and correlation of LCC or population were quantified between multiple scales. Different kinds of driving factors were explored in the pattern analysis and then utilized for the definition of adaptive land suitability rules using the Cellular Automata-Markov (CA-Markov) simulation. As a case study of the MSPAS model, Nepal entered into a new era by the establishment of a Federal Republic in 2015. The model focused on four specific land cover classes of urban, farmland, forest and grassland to explore the pattern of Nepal’s LCC from 2016 to 2019. The result demonstrated the performance of the MSPAS model. The spatiotemporal pattern had consistency, and characteristics between multiple scales and population were related to LCC. Urban area nearly doubled while farmland decreased by 3% in these years. Urban areas expanded at the expense of farmland, especially in Kathmandu and some districts of the Terai region, which tended to occur on flat areas near the existing urban centers or along the roads. Farmland abandonment was relatively intense with scattered abandoned areas widely distributed in the Hill region under conditions of steep topography and sparse population. The MSPAS model can provide references for the development of sustainable urbanization and agriculture in SDGs.

Large-Scale land cover mapping (LLCM) based on deep learning models necessitates a substantial number of high-precision sample datasets. However, the limited availability of such datasets poses challenges in regularly updating land cover products. A commonly referenced method involves utilizing prior products (PPs) as labels to achieve up-to-date land cover mapping. Nonetheless, the accuracy of PPs at the regional level remains uncertain, and the Remote Sensing Image (RSI) corresponding to the product is not publicly accessible. Consequently, the sample dataset constructed through geographic location matching may lack precision. Errors in such datasets are not only due to inherent product discrepancies, and can also arise from temporal and scale disparities between the RSI and PPs. In order to solve the above problems, this paper proposes an LLCM framework for generating labels for use with PPs. The framework consists of three main parts. First, initial generation of labels, in which the collected PPs are integrated based on D-S evidence theory and initial labels are obtained using the generated trust map. Second, for dynamic label correction, a two-stage training method based on initial labels is adopted. The correction model is pretrained in the first stage, then the confidence probability (CP) correction module of the dynamic threshold value and NDVI correction module are introduced in the second stage. The initial labels are iteratively corrected while the model is trained using the joint correction loss, with the corrected labels obtained after training. Finally, the classification model is trained using the corrected labels. Using the proposed land cover mapping framework, this study used PPs to produce a 10 m spatial resolution land cover map of Cambodia in 2020. The overall accuracy of the land cover map was 91.68% and the Kappa value was 0.8808. Based on these results, the proposed mapping framework can effectively use PPs to update medium-resolution large-scale land cover datasets, and provides a powerful solution for label acquisition in LLCM projects.