Beyond the Remote Sensing Ecological Index: A Comprehensive Ecological Quality Evaluation Using a Deep-Learning-Based Remote Sensing Ecological Index

• Published in: Remote sensing (Basel, Switzerland) Vol. 17; no. 3; p. 558
• Main Authors: Gong, Xi, Li, Tianqi, Wang, Run, Hu, Sheng, Yuan, Shuai
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.02.2025
• Subjects: Biosphere; China; Correlation coefficient; Correlation coefficients; Datasets; Deep learning; Depth indicators; Earth resources technology satellites; Ecological conditions; ecological index; ecological quality evaluation; Economic development; Ecosystem integrity; Environmental economics; Environmental quality; Geospatial data; Geosphere; Image retrieval; Land use; Landsat; Long short-term memory; Machine learning; Methods; Natural resources; Neural networks; Nonlinear systems; Population growth; Quality assessment; Remote sensing; Risk management; Rivers; Socioeconomic factors; Socioeconomics; Sustainable development; Terrestrial ecosystems; Urban areas; Urbanization; China; Hubei China; Yangtze River
• DOI: 10.3390/rs17030558
• ISSN: 2072-4292

Ecological integrity is fundamental to human survival and development. However, rapid urbanization and population growth have significantly disrupted ecosystems. Despite the focus of the International Geosphere-Biosphere Programme (IGBP) on terrestrial ecosystems and land use/cover changes, existing ecological indices, such as the Remote Sensing Ecological Index (RSEI), have limitations, including an overreliance on single indicators and inability to fully encapsulate the ecological conditions of urban areas. This study addresses these gaps by proposing a Deep-learning-based Remote Sensing Ecological Index (DRSEI) that integrates human economic activities and leverages an autoencoder neural network with long short-term memory (LSTM) modules to account for nonlinearity in ecological quality assessments. The DRSEI model utilizes multi-temporal remote sensing data from the Landsat series, WorldPop, and NPP-VIIRS and was applied to evaluate the ecological conditions of Hubei Province, China, over the past two decades. The key findings indicate that ecological environmental quality gradually improved, particularly from 2000 to 2010, with the rate of improvement subsequently slowing. The DRSEI outperformed the traditional RSEI and had a significantly higher Pearson correlation coefficient than the Ecological Index (EI), thus demonstrating enhanced accuracy and predictive performance. This study presents an innovative approach to ecological assessment that offers a more comprehensive, accurate, and nuanced understanding of ecological changes over time. Integrating socioeconomic factors with deep learning techniques contributes significantly to the field and has implications for ecological risk control and sustainable development.