Half-hourly NEE simulation based on flux observations across China's grassland ecosystems: A machine learning fusion approach with interpretability analysis

Accurate simulation of grassland carbon fluxes is critical for regional carbon budget assessment and for supporting China's carbon neutrality goals under climate change. However, high-temporal-resolution estimates of carbon fluxes across China's heterogeneous grassland ecosystems remain largely limited, and single machine-learning models often show inconsistent predictive performance among ecosystem types. Here, we developed a Random Forest–eXtreme Gradient Boosting (RF-XGB) fusion model to simulate half-hourly net ecosystem CO₂ exchange (NEE) based on eddy covariance observations and concurrent environmental variables from 15 grassland flux sites across China. The fusion model achieved an overall R 2 of 0.6425 across all sites and outperformed standalone base models at 80% of monitoring sites with higher R 2 and lower RMSE. Model performance varied significantly across different grassland types, with alpine steppe (AS, R 2 = 0.7394) and temperate typical steppe (TTS, R 2 = 0.7354) showing relatively high prediction accuracy, alpine meadow (AM, R 2 = 0.6145) exhibiting slightly lower accuracy, and temperate desert steppe (TDS, R 2 = 0.5558) showing the lowest prediction accuracy. SHapley Additive exPlanations (SHAP) analysis revealed a hierarchical driver structure for half-hourly NEE, where radiation acts as the primary driver, temperature serves as the secondary regulator, and water-related factors shift from atmospheric dryness (VPD) in mesic grasslands to soil moisture (SWC) in arid systems. This study provides a systematic, type-differentiated evaluation of half-hourly NEE simulation across China's grasslands, demonstrating that multi-algorithm fusion combined with ecosystem-specific parameterization can improve high-resolution carbon flux modeling in heterogeneous landscapes.

Authors

Institutions

Publication Details

Journal
Ecological Indicators
Published
2026-11-14
DOI
https://doi.org/10.1016/j.ecolind.2026.115854
Primary Topic
Plant Water Relations and Carbon Dynamics
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Half-hourly NEE simulation based on flux observations across China's grassland ecosystems: A machine learning fusion approach with interpretability analysis

Li Zhou, Jiaxin Song, Yu Wang, Xudong Fang et al.
Ecological Indicators
Plant Water Relations and Carbon Dynamics
article

Half-hourly NEE simulation based on flux observations across China's grassland ecosystems: A machine learning fusion approach with interpretability analysis

Li Zhou, Jiaxin Song, Yu Wang, Xudong Fang, Chongyang Dong
article en

Abstract

Accurate simulation of grassland carbon fluxes is critical for regional carbon budget assessment and for supporting China's carbon neutrality goals under climate change. However, high-temporal-resolution estimates of carbon fluxes across China's heterogeneous grassland ecosystems remain largely limited, and single machine-learning models often show inconsistent predictive performance among ecosystem types. Here, we developed a Random Forest–eXtreme Gradient Boosting (RF-XGB) fusion model to simulate half-hourly net ecosystem CO₂ exchange (NEE) based on eddy covariance observations and concurrent environmental variables from 15 grassland flux sites across China. The fusion model achieved an overall R 2 of 0.6425 across all sites and outperformed standalone base models at 80% of monitoring sites with higher R 2 and lower RMSE. Model performance varied significantly across different grassland types, with alpine steppe (AS, R 2 = 0.7394) and temperate typical steppe (TTS, R 2 = 0.7354) showing relatively high prediction accuracy, alpine meadow (AM, R 2 = 0.6145) exhibiting slightly lower accuracy, and temperate desert steppe (TDS, R 2 = 0.5558) showing the lowest prediction accuracy. SHapley Additive exPlanations (SHAP) analysis revealed a hierarchical driver structure for half-hourly NEE, where radiation acts as the primary driver, temperature serves as the secondary regulator, and water-related factors shift from atmospheric dryness (VPD) in mesic grasslands to soil moisture (SWC) in arid systems. This study provides a systematic, type-differentiated evaluation of half-hourly NEE simulation across China's grasslands, demonstrating that multi-algorithm fusion combined with ecosystem-specific parameterization can improve high-resolution carbon flux modeling in heterogeneous landscapes.

Ecological IndicatorsVol. 193
Henan University of Science and Technology (CN), Zhengzhou University (CN), Chinese Academy of Meteorological Sciences (CN), State Key Laboratory of Severe Weather (CN)
Ministry of Science and Technology of the People's Republic of China, China Meteorological Administration, Chinese Academy of Meteorological Sciences, National Key Research and Development Program of China
Life on land
Plant Water Relations and Carbon Dynamics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon