Mine ecological assessment and prediction based on RSEI and a multi-neural-network framework
Abstract
The Remote Sensing Ecological Index (RSEI) maps ecological quality by integrating greenness, wetness, dryness, and heat. However, because it is calculated from observations acquired at the target date, the RSEI is retrospective: it identifies ecological change after it occurs but does not itself forecast the next period. Moreover, the linear construction of RSEI through principal component analysis does not explicitly capture nonlinear temporal interactions between ecological indicators and external environmental drivers. To address these limitations, this study develops a staged convolutional neural network–long short-term memory–backpropagation neural network (CLB) framework for ecological assessment and one-step-ahead prediction in the Wangpo Mine Field, Shanxi Province, China. The framework assigns the three networks complementary roles: CNN extracts features from multivariable time series, LSTM models temporal dependence and predicts the target-period variables, and BP fuses the predicted remote-sensing indicators with climatic, topographic, and spatial variables to estimate the target-period RSEI one step ahead. Landsat-8 imagery from 2015 to 2020 was used to quantify ecological change; data from 2015 to 2019 were used for model development, and 2020 was reserved for testing. The mean RSEI increased by 12.1%, from 0.6136 to 0.6879, accompanied by an expansion of excellent-quality areas and a contraction of poor-quality areas. For the held-out 2020 data, the predicted RSEI grades were consistent with those calculated from Landsat-8 imagery, with grade-specific accuracies of 87.3–94.1%. By coupling feature extraction, temporal forecasting, and nonlinear multivariable fusion, the CLB framework may provide a methodological reference for one-step-ahead ecological-quality prediction in mining areas.
Related
- Associations between social network characteristics and daily social experiences in later life: evidence from ecological momentary assessment
- Matching laboratory and real-world listening environments with ecological momentary assessment improves prediction of speech perception.
- Cognitive Control Neural Correlates of Negative Affective in Psychosis Spectrum Disorders: A Laboratory and Ecological Momentary Assessment Study
- Varietal classification and morphological traits prediction of lettuce cultivar (Lactuca sativa L.) based on RGB-images and convolutional neural network
- PATHNet: A Thermodynamic Path-Integral-Inspired Neural Network for Protein-Ligand Interaction Prediction
BibTeX
@article{li2026mine,
title = {{Mine ecological assessment and prediction based on RSEI and a multi-neural-network framework}},
author = {Peng Li and Bo Wang and Yangyang Guo and Yuan Liang and Junsheng Yan},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-73044-2},
url = {https://www.nature.com/articles/s41598-026-73044-2},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Mine ecological assessment and prediction based on RSEI and a multi-neural-network framework
AU - Peng Li
AU - Bo Wang
AU - Yangyang Guo
AU - Yuan Liang
AU - Junsheng Yan
PY - 2026
DA - 2026/10/19
JO - Scientific Reports
DO - 10.1038/s41598-026-73044-2
UR - https://www.nature.com/articles/s41598-026-73044-2
AB - The Remote Sensing Ecological Index (RSEI) maps ecological quality by integrating greenness, wetness, dryness, and heat. However, because it is calculated from observations acquired at the target date, the RSEI is retrospective: it identifies ecological change after it occurs but does not itself forecast the next period. Moreover, the linear construction of RSEI through principal component analysis does not explicitly capture nonlinear temporal interactions between ecological indicators and external environmental drivers. To address these limitations, this study develops a staged convolutional neural network–long short-term memory–backpropagation neural network (CLB) framework for ecological assessment and one-step-ahead prediction in the Wangpo Mine Field, Shanxi Province, China. The framework assigns the three networks complementary roles: CNN extracts features from multivariable time series, LSTM models temporal dependence and predicts the target-period variables, and BP fuses the predicted remote-sensing indicators with climatic, topographic, and spatial variables to estimate the target-period RSEI one step ahead. Landsat-8 imagery from 2015 to 2020 was used to quantify ecological change; data from 2015 to 2019 were used for model development, and 2020 was reserved for testing. The mean RSEI increased by 12.1%, from 0.6136 to 0.6879, accompanied by an expansion of excellent-quality areas and a contraction of poor-quality areas. For the held-out 2020 data, the predicted RSEI grades were consistent with those calculated from Landsat-8 imagery, with grade-specific accuracies of 87.3–94.1%. By coupling feature extraction, temporal forecasting, and nonlinear multivariable fusion, the CLB framework may provide a methodological reference for one-step-ahead ecological-quality prediction in mining areas.
SN - 2045-2322
ER -