Multi-feature fusion for grassland information extraction and temporal change monitoring
- 1Shaoyang Grassland Resources Protection Center, Shaoyang, Hunan, China
- 2College of Civil Engineering, Xiangtan University, Xiangtan, Hunan, China
Abstract
Grasslands play a vital role in maintaining ecological security and supporting sustainable development, yet accurate extraction and dynamic monitoring remain challenging in complex mountainous regions. This study takes Chengbu Miao Autonomous County, Hunan Province, as the study area and develops a multi-feature fusion framework on the Google Earth Engine platform by integrating optical imagery, Sentinel-1 SAR data, vegetation indices, texture features, topographic factors, and time-series characteristics. A Random Forest model was employed for high-accuracy grassland extraction, while multi-temporal remote sensing observations from selected cloud-free years between 2000 and 2025 were used to analyze long-term spatial distribution and temporal dynamics. Fractional vegetation cover (FVC) was introduced to assess grassland ecological conditions, and Nanshan Pasture was selected to investigate the impacts of topography, human activities, and climatic factors on grassland variation. The results show that multi-feature fusion significantly improves classification accuracy, achieving an overall accuracy (OA) of 93.79% and a Kappa coefficient of 0.7165 in the optimal feature combination scheme, with optical and SAR data exhibiting strong complementarity under complex terrain conditions. Grassland area in Chengbu County remained relatively stable from 2000 to 2025, with annual variations ranging from 47.32 to 52.15 km² and no significant long-term change trend. The mean FVC exhibited relatively stable interannual fluctuations, with values generally ranging from 0.59 to 0.72 during most years. However, spatial analysis of FVC trends revealed considerable heterogeneity, with significant decreasing trends observed in localized areas, particularly in the southwestern part of Nanshan Pasture, indicating potential vegetation degradation risks. In Nanshan Pasture, FVC showed a significant negative correlation with DEM (r = −0.2839) and precipitation (r = −0.589), while correlations with human activities and land surface temperature were relatively weak. Overall, grassland dynamics in the study area may reflect the combined effects of environmental conditions and human activities, although the relationships identified in this study mainly represent statistical associations rather than causal mechanisms. The proposed framework provides an effective methodological reference for grassland monitoring, ecological assessment, and sustainable management in subtropical mountainous grassland environments.
Related
- Fine-scale grassland land-cover classification using UAV RGB and multispectral imagery with staged feature selection
- MaSkel and napari-MaSkel: fast morphological skeleton feature extraction from segmentation masks
- A multimodal deep learning for rice crop health monitoring using vegetation index feature selection
- UniWave-2: A Hybrid Model for Nucleic Acid Waveform Feature Extraction Enhanced by Fourier and Wavelet Transforms
- eFEL: electrophysiology feature extraction library
BibTeX
@article{liu2026multifeature,
title = {{Multi-feature fusion for grassland information extraction and temporal change monitoring}},
author = {Xiyuan Liu and Jie Chen and Zijin Liu and Xiaohua Zhu and Shigao Jiang and Wanli Kang},
journal = {PLOS One},
year = {2026},
month = nov,
doi = {10.1371/journal.pone.0347821},
url = {https://doi.org/10.1371/journal.pone.0347821},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Multi-feature fusion for grassland information extraction and temporal change monitoring
AU - Xiyuan Liu
AU - Jie Chen
AU - Zijin Liu
AU - Xiaohua Zhu
AU - Shigao Jiang
AU - Wanli Kang
PY - 2026
DA - 2026/11/06
JO - PLOS One
DO - 10.1371/journal.pone.0347821
UR - https://doi.org/10.1371/journal.pone.0347821
AB - Grasslands play a vital role in maintaining ecological security and supporting sustainable development, yet accurate extraction and dynamic monitoring remain challenging in complex mountainous regions. This study takes Chengbu Miao Autonomous County, Hunan Province, as the study area and develops a multi-feature fusion framework on the Google Earth Engine platform by integrating optical imagery, Sentinel-1 SAR data, vegetation indices, texture features, topographic factors, and time-series characteristics. A Random Forest model was employed for high-accuracy grassland extraction, while multi-temporal remote sensing observations from selected cloud-free years between 2000 and 2025 were used to analyze long-term spatial distribution and temporal dynamics. Fractional vegetation cover (FVC) was introduced to assess grassland ecological conditions, and Nanshan Pasture was selected to investigate the impacts of topography, human activities, and climatic factors on grassland variation. The results show that multi-feature fusion significantly improves classification accuracy, achieving an overall accuracy (OA) of 93.79% and a Kappa coefficient of 0.7165 in the optimal feature combination scheme, with optical and SAR data exhibiting strong complementarity under complex terrain conditions. Grassland area in Chengbu County remained relatively stable from 2000 to 2025, with annual variations ranging from 47.32 to 52.15 km² and no significant long-term change trend. The mean FVC exhibited relatively stable interannual fluctuations, with values generally ranging from 0.59 to 0.72 during most years. However, spatial analysis of FVC trends revealed considerable heterogeneity, with significant decreasing trends observed in localized areas, particularly in the southwestern part of Nanshan Pasture, indicating potential vegetation degradation risks. In Nanshan Pasture, FVC showed a significant negative correlation with DEM (r = −0.2839) and precipitation (r = −0.589), while correlations with human activities and land surface temperature were relatively weak. Overall, grassland dynamics in the study area may reflect the combined effects of environmental conditions and human activities, although the relationships identified in this study mainly represent statistical associations rather than causal mechanisms. The proposed framework provides an effective methodological reference for grassland monitoring, ecological assessment, and sustainable management in subtropical mountainous grassland environments.
SN - 1932-6203
ER -