China’s Forest Light Use Efficiency Rises Driven Strongly by Biological Factors

Between 2000 and 2023, the light use efficiency of forests across China averaged 0.303 ± 0.019 grams of carbon per megajoule and exhibited a statistically significant upward trajectory of 0.003 grams of carbon per megajoule per year, according to a comprehensive national assessment published in Ecological Processes. Light use efficiency, which quantifies the capacity of vegetation to convert absorbed solar radiation into organic matter via photosynthesis, increased across 94.2% of the country’s forested territory, with 72% exhibiting statistically significant gains under non-parametric trend evaluations. The spatial distribution of efficiency values displayed a marked macro-geographic gradient, characterized by substantially higher values in southern and southwestern zones—most notably across the Qinling-Daba Mountains, southeastern Tibet, southern Yunnan, and northern Guangxi—and comparatively suppressed values across northeastern, northern, and northwestern China. To establish these patterns, an investigative team compiled a harmonized repository of 1,710 site-level net primary productivity ground observations gathered across natural forest ecosystems between 2000 and 2023, coupling ground measurements with multi-source remote sensing, gridded atmospheric datasets, and machine learning architectures.

The investigation directly addresses critical knowledge gaps in global carbon cycle science regarding whether terrestrial vegetation productivity is governed primarily by the phenological duration of the growing season or by the physiological rate of photosynthetic carbon assimilation. While classic ecological assumptions posited that warming-induced extensions of growing seasons served as the primary engine of enhanced carbon uptake, contemporary findings indicate that photosynthetic assimilation rate accounts for roughly 65% of the observed temporal changes in gross primary productivity. Despite its foundational importance in governing ecosystem carbon uptake, approximately 60% of the spatiotemporal variability in light use efficiency had remained unexplained in prior scientific literature. Earlier assessments focused almost exclusively on the independent, direct influences of land-surface environmental factors such as ambient temperature and vapor pressure deficit, while largely omitting the regulatory contributions of internal biological characteristics—including forest type, stand age, and canopy architectural traits—as well as the interactive couplings between structural biology and ambient environmental stress.

To overcome the limitations of isolated localized observations, the researchers integrated ground-based field records with extensive spatial rasters at high geographical resolution. Site records were screened to exclude non-natural experimental treatments such as artificial warming, nitrogen addition, or thinning, as well as non-forest vegetation types including bamboo, shrublands, urban woodlands, and mangroves. Spatial predictors included nationwide forest age and forest type data at 1-kilometer resolution spanning 5-year intervals from 2000 to 2015, advanced annually between benchmark nodes and carried forward through 2023, along with gridded forest origin records from 2000 to 2020. Gridded climate variables were compiled from ERA5-Land at 10-kilometer resolution, including mean monthly temperature, total precipitation, solar radiation, and volumetric soil moisture, complemented by 5-kilometer vapor pressure deficit metrics from TerraClimate, standardized precipitation evapotranspiration index records, 250-meter SoilGrids physical clay and sand contents, and annual background atmospheric carbon dioxide measurements from the NOAA Global Monitoring Laboratory.

Site-level annual actual light use efficiency was determined by dividing observed net primary productivity by annual absorbed photosynthetically active radiation, which was calculated from incident solar radiation and canopy light absorption. Canopy absorbed fraction was derived as the arithmetic mean of normalized difference vegetation index and simple ratio formulations to suppress background reflectance distortions and minimize estimation error against ground observations. The authors subsequently trained a non-parametric Random Forest model to link remote spatial predictors to field-derived light use efficiency, utilizing 75% of observations for training and withholding 25% for independent validation. The resulting full model achieved a coefficient of determination of 0.91 with a root mean square error of 0.07 grams of carbon per megajoule on training data, and an external coefficient of determination of 0.80 with an error of 0.09 grams of carbon per megajoule on test records. Stringent leave-one-year-out cross-validation demonstrated robust temporal transferability, maintaining positive predictive correlations in every held-out calendar year, while a longitudinal within-series benchmark confirmed a 23.3% reduction in mean squared prediction error.

To disentangle environmental constraints from structural biological factors, the researchers deployed an ensemble of six conventional light use efficiency models: CASA, EC-LUE, GLO-PEM, VPM, MOD17, and TEC. This six-model framework estimated land-surface environmental constraints by calculating a mean environmental scaling factor representing temperature, moisture, and vapor pressure limitations scaled by an idealized maximum theoretical efficiency of 0.389 grams of carbon per megajoule. Because the six conventional baseline models did not explicitly account for atmospheric carbon dioxide fertilisation, subtracting the ensemble environmental estimate from the full Random Forest prediction yielded a combined residual term comprising biological factors and atmospheric carbon dioxide. Out-of-bag evaluation verified that the Random Forest model exhibited substantially lower absolute prediction error than the environmental ensemble, while an observation-derived residual maintained strong, statistically significant correlations with stand age and forest type after controlling for geographic and temporal gradients, demonstrating an alignment of r = 0.826 with the model-derived residual.

Temporal and spatial patterns evaluated through pixel-level Theil-Sen slope estimations and Mann-Kendall significance tests revealed widespread greening efficiency across China’s forest biomes. Beyond the 72% of forested areas demonstrating significant efficiency increases and the 22.2% displaying non-significant increases, declining efficiency was detected across only 5.8% of the country’s forests, with 5.2% classified as non-significant decreases and a minimal 0.6% exhibiting significant reductions, predominantly concentrated in intensively developed urban fringes and tourism corridors. Land-surface environmental constraints showed pronounced north-south divergence: northern regions experienced severe climatic limitations resulting in depressed environmental efficiency scores, whereas southern regions exhibited higher baseline environmental efficiency.

Quantitative attribution analysis revealed that synergistic interactions between environmental factors and biological processes governed the long-term trend, with areas driven jointly by both components constituting 83.1% of China’s forest expanse. In regions where efficiency expanded over time, model-derived attribution allocated 15% of the relative contribution to land-surface environmental factors and 85% to the combined biological and carbon dioxide component. To resolve the internal structure of this 85% combined share, the authors implemented a counterfactual sensitivity simulation using the revised EC-LUE carbon dioxide response formulation, comparing observed historical carbon dioxide concentrations against a fixed baseline of 368.96 parts per million from the year 2000. This module-specific partition indicated that rising atmospheric carbon dioxide elevated response-module efficiency by 3.82% by 2023, accounting for roughly 16% of the total attribution in expanding zones, whereas internal biological characteristics accounted for the dominant 69% majority.

Climatic drivers exhibited distinct regional mechanisms when examined through partial and multiple correlation analyses focused on mean annual temperature, soil moisture, and atmospheric carbon dioxide. Joint variations in carbon dioxide and soil moisture accounted for 38.3% of the total forest area, while carbon dioxide alone explained 25.8%, exhibiting positive correlations across the Qinling-Daba Mountains, southeastern Tibet, and northeastern China. Soil moisture was positively associated with efficiency across most temperate and subtropical forests, underscoring widespread water limitation. However, in the deciduous needleleaf forests of northeastern China, soil moisture exhibited a notable negative correlation with efficiency. This localized divergence reflects permafrost hydrology and seasonal freeze-thaw cycles: excessive moisture in shallow active thaw layers can impede soil aeration, trigger root hypoxia, and restrict nutrient transport, whereas deciduous needleleaf species possess lower transpiration demands than broadleaf competitors and suffer physiological constraints under saturated, poorly drained conditions.

Stand development stage strongly modulated forest responses to climatic fluctuations, as evidenced by multi-variable stratifications and structural equation modeling. Partial least squares structural equation modeling fitted to all 1,710 ground sites established that climatic conditions were positively linked to forest type distribution, which in turn exhibited a direct positive association with site efficiency, yielding a significant indirect climatic pathway of 0.22. Although all age categories exhibited positive efficiency responses to rising mean annual temperatures—with middle-aged stands showing the steepest slope—and displayed uniform sensitivity to atmospheric carbon dioxide, their moisture responses diverged sharply. Only mature, old-growth forests exhibited a statistically significant gain in efficiency in response to elevated soil moisture. Continuous-age modeling across the common age support interval of 12.5 to 60.3 years demonstrated that both forest-type differentiation and the positive coupling between soil moisture and efficiency strengthened systematically as stands matured.

The authors contextualized these results within clear analytical and structural uncertainties documented in the empirical data. Forest age maps were anchored to 5-year intervals, meaning annual linear progression between benchmark nodes could overlook interim local canopy disturbances; however, independent screening against satellite disturbance layers revealed that only 3.19% of pixels experienced severe disturbance, and excluding disturbed cells shifted the old-forest moisture slope by only 0.79%. Furthermore, while arithmetic averaging of six conventional models reduced reliance on single water-stress functions, unweighted averaging across vapor pressure deficit, soil moisture, and plant indicators cannot fully capture complex hydrological heterogeneity. Old-growth forests in China remain heavily concentrated in northern biomes dominated by deciduous broadleaf and deciduous needleleaf taxa, meaning future demographic shifts may alter ecosystem responses.

Subject of Research: Spatiotemporal dynamics and driving mechanisms of forest light use efficiency in China

Article Title: Light use efficiency in China’s forests: identifying spatiotemporal dynamics and disentangling relative contributions of driving factors

Article References: Qu, Y., Zhou, T., Zeng, J., Tan, E., Zhang, J., Zhang, Y., Wu, X., Zhang, Q., Liu, J., Liu, P., Yu, Y., & Cao, L. (2026). Light use efficiency in China’s forests: identifying spatiotemporal dynamics and disentangling relative contributions of driving factors. Ecological Processes, 15(1), Article 103. https://doi.org/10.1186/s13717-026-00761-x

Image Credits: AI Generated

DOI: 10.1186/s13717-026-00761-x

Keywords: forest ecology, light use efficiency, carbon cycle, gross primary productivity, random forest modeling, soil moisture, stand age, climate change, Light, efficiency, China, forests