Related Experiment Video
Updated: Jan 8, 2026
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
2.0K
Optimizing methane flux prediction and key feature identification based on a novel hybrid machine learning model
Xinqin Gu1, Li Yao1,2, Xiang Xiao3
1College of Water Conservancy, Jiangxi University of Water Resources and Electric Power, Nanchang 330099, China.
Iscience
|December 16, 2025
Summary
Accurate methane (CH4) flux prediction is crucial. Machine learning models, especially hybrid XGBoost, improved CH4 emission predictions using key features like soil temperature.
Area of Science:
- Environmental Science
- Atmospheric Chemistry
- Computational Science
Background:
- Natural landscape methane (CH4) emissions constitute 50% of the global total.
- Quantifying these emissions is challenging due to high costs and ecosystem complexity.
Purpose of the Study:
- To assess machine learning models (XGBoost, Random Forest, SVM) for predicting CH4 fluxes.
- To optimize CH4 flux prediction using a hybrid GWOPSO-XGBoost model.
Main Methods:
- Evaluated XGBoost (XGB), Random Forest (RF), and Support Vector Machine (SVM) on 36 FLUXNET-CH4 sites.
- Applied Grey Wolf Optimizer-Particle Swarm Optimization (GWOPSO) to hybridize XGB for feature and hyperparameter optimization.
Main Results:
- Ensemble models (XGB, RF) outperformed SVM in CH4 flux prediction.
- The hybrid GWOPSO-XGBoost model enhanced prediction accuracy, requiring only five features for most wetland types.
- Reduced Root Mean Square Error (RMSE) by 1.1%-10.8% compared to the all-feature XGB model.
- Soil temperature emerged as the most critical predictive variable.
Conclusions:
- The study recommends a reliable CH4 prediction method using optimized machine learning.
- This approach provides a valuable reference for CH4 flux prediction across diverse site types.
