A Prediction Method of Land Use Type Evolution Based on Long and Short-Term Memory Networks--Taking Gansu Province as an Example

Abstract

In the context of escalating global climate change and human activities, understanding the driving mechanisms behind land use change and predicting future trends is crucial. This study takes Gansu Province as a case, using land use type data from 1990 to 2020 to construct a Long Short-Term Memory (LSTM) model to predict land use changes over the next decade (2021-2030). The results indicate that land use types in Gansu Province exhibit significant dynamic changes, with forest area continuously increasing, built-up land rapidly expanding, and areas of unused land and grassland significantly decreasing. These changes reflect the combined effects of ecological protection policies, urbanization, and land development. The models predictions suggest that built-up land has absorbed substantial areas of unused land, grassland, and cultivated land, with accelerated urbanization. Forest area growth is attributed to the implementation of ecological restoration policies, while grassland and water areas show fluctuating changes, and the area of unused land continues to decrease. The findings not only provide data support for land resource management and ecological protection, but also offer scientific evidence for the formulation of sustainable land use policies, which can serve as an important reference for the sustainable use and management of land resources in Gansu Province and similar regions.