Assessment of landscape diversity in Inner Mongolia and risk prediction using CNN-LSTM model

Abstract

Biodiversity is the central focus of ecological studies, and landscape ecology offers new insights into its preservation. This study employed a landscape-scale and multi-tiered framework to conduct a comprehensive evaluation of the current state of biodiversity in the Inner Mongolia Autonomous Region from 2019 to 2021. A Potential-Connectedness-Resilience framework was used to assess landscape diversity risks from 2010 to 2020, with a Convolutional Neural Network combined with a Long Short-Term Memory (CNN-LSTM) model predicting future risks for 2025. Our findings indicate that landscape diversity in Inner Mongolia was favourable and stable condition during 2019–2021. However, an upward trend in patch dispersion was observed, with landscape types increasingly shaped by human activity and the surrounding environment. Vegetation coverage showed a positive trend, suggesting a reduction in ecological sensitivity. However, the improvement in vegetation quality was not significant. From 2010 to 2020, the potential risks to landscape diversity escalated, although connectivity risks diminished, fragmentation lessened, and landscape structures became more continuous and concentrated, indicating heightened regional anthropogenic influences. The projected landscape diversity risk warning for 2025 mirrors the historical spatial data, with a notable reduction in local disparities and an overall decrease in the average value by 2.73%. This study offers a comprehensive biodiversity risk assessment and leverages the CNN-LSTM model to anticipate future landscape changes, providing a robust tool for a timely understanding of the status, trends, and threats to biodiversity within the Inner Mongolian region.

Keywords

• Landscape diversity
• Risk assessment and prediction
• PCR framework
• CNN-LSTM model
• Inner Mongolia