The scale of the underlying data compilation is one of the study’s most striking features. The researchers gathered 2,311 in-situ measurements of woody fine root biomass and 3,580 measurements of herbaceous fine root biomass from peer-reviewed studies spanning the country. These field observations, collected from tropical forests in the south to boreal woodlands in the north, and from alpine meadows on the Qinghai–Tibet Plateau to arid grasslands in the northwest, form the empirical backbone of the new dataset. For decades, such measurements have sat scattered across hundreds of individual papers, each describing a single site at a single moment in time. By pooling them, the team converted a fragmented literature into a coherent national picture—one that can be compared, modeled, and validated at scales far beyond any individual field campaign.

Turning thousands of point measurements into a continuous map requires a modeling bridge, and the researchers built theirs from machine learning. They trained random forest models—a technique that combines hundreds of decision trees, each trained on random subsets of the data and environmental predictors, to produce robust estimates even when relationships between variables are nonlinear—to predict fine root biomass in the surface soil layer, defined as the top 20 centimeters. Crucially, they built separate models for woody ecosystems and herbaceous ecosystems, recognizing that trees and grasses deploy their roots under fundamentally different constraints. Environmental covariates, including climate, soil properties, and vegetation characteristics, provided the spatial context that allowed the models to extrapolate from the measurement sites to every one-kilometer grid cell across China.

But surface biomass alone does not tell the full story. Fine roots are distributed through the soil profile in patterns that vary enormously with species, climate, and water availability, and this vertical structure matters enormously for carbon accounting. To capture it, the team modeled a vertical attenuation coefficient—a parameter, conventionally denoted k, that describes how quickly root biomass declines with depth. By pairing the predicted surface-layer biomass with these depth-distribution functions, the researchers generated estimates of fine root biomass in successive 20-centimeter layers reaching down to two meters, for every grid cell, for all woody and herbaceous ecosystems in China, and for each year from 2005 through 2025. The result is a four-dimensional picture of the country’s underground biomass: across longitude, latitude, depth, and time.

The models were rigorously evaluated against independent observations, and the performance figures are impressive for a variable as noisy as root biomass. The random forest models explained approximately 86 percent of the observed variation in surface fine root biomass in woody ecosystems and roughly 81 percent in herbaceous ecosystems. The team also reported that their predictions reasonably captured interannual dynamics—meaning the models did not merely reproduce average conditions but tracked year-to-year fluctuations in the observations. For a quantity that is notoriously difficult and labor-intensive to measure, involving careful excavation, sieving, and separation of living fine roots from soil, this level of predictive skill represents a substantial methodological achievement.

The mapped patterns reveal a continent-scale geography of root life. Woody fine root biomass density is generally higher in southern China than in central and northern regions, a gradient consistent with the greater productivity and denser vegetation of the country’s humid subtropical forests. Herbaceous fine root biomass, by contrast, reaches its highest densities on the Qinghai–Tibet Plateau, where cold-adapted grasslands concentrate a remarkable share of their biomass below ground—an evolutionary strategy for surviving grazing, frost, and short growing seasons. These contrasting regional hotspots underline a key message of the study: belowground biomass cannot be inferred from aboveground greenness alone, and the two storages obey different rules in different biomes.

Perhaps the most scientifically provocative finding concerns depth. The analysis showed that woody plants send proportionally more of their fine root biomass deeper into the soil in arid regions, a pattern consistent with the physiological imperative to reach reliable water sources when precipitation is scarce. Herbaceous species, meanwhile, maintained consistently surface-concentrated rooting patterns across all climate zones, keeping their fine roots packed into the uppermost soil layers regardless of how dry their environment became. This divergence means that the same meter of soil beneath a desert shrubland and beneath a steppe grassland holds very different amounts of carbon-bearing root tissue—and that global models assuming a universal root-depth profile are likely misplacing carbon in both cases.

The environmental drivers of fine root biomass also split along vegetation lines in ways that matter for future projections. Woody fine root biomass was more strongly influenced by root traits and nutrient availability, suggesting that the chemical and architectural properties of tree species, together with soil fertility, set the ceiling on how much fine root mass a forest maintains. Herbaceous fine root biomass, in contrast, tracked climatic conditions more closely, implying that grassland roots may respond more directly and rapidly to warming and shifting precipitation regimes. As climate change alters both temperature belts and nutrient cycles, these differing sensitivities could cause woody and herbaceous ecosystems to diverge in their belowground carbon dynamics over the coming decades.

Why does this matter beyond China’s borders? Earth system models that simulate the global carbon cycle must partition photosynthesized carbon among leaves, wood, and roots, and must decide how quickly root carbon enters the soil through turnover. Poorly constrained root distributions propagate uncertainty into estimates of soil carbon stocks and their feedbacks to climate. An observation-constrained, spatially explicit benchmark of the kind produced by Zheng and colleagues gives modelers something they have rarely possessed: a gridded ground truth against which simulated root biomass and its vertical arrangement can be tested, corrected, and improved. The dataset’s yearly coverage from 2005 to 2025 adds a temporal dimension that can support studies of how belowground carbon storage is already shifting as climate and land use change.

The dataset itself has been made openly available, archived on Figshare, and the underlying preprint is undergoing open review at Earth System Science Data, a journal dedicated to publishing data products of lasting value to the geosciences. The authors suggest that this resource will serve as a benchmark for quantifying belowground carbon stocks and their dynamics, but its applications extend further—to nutrient cycling studies, hydrological modeling, land restoration planning, and the verification of carbon offset schemes that depend on honest accounting of what lies beneath. For a country that spans tropical reefs, karst forests, alpine meadows, and deserts, a unified, vertically resolved map of fine roots is more than a technical curiosity. It is a detailed portrait of the hidden half of China’s ecosystems, and an invitation to look more carefully at the roots that quietly run the planet’s most important carbon exchange.

Subject of Research: Vertically resolved fine root biomass mapping across China's ecosystems

Article Title: A High-Resolution Gridded Dataset of Vertically Resolved Fine Root Biomass Across China

Article References: Zheng, W., Zhao, M., Li, W., Zhang, D., Tang, W., Luo, J., Jiao, S., Li, F., & Zhang, H. (2026). A High-Resolution Gridded Dataset of Vertically Resolved Fine Root Biomass Across China. https://doi.org/10.5194/essd-2026-777

Image Credits: AI Generated

DOI: 10.5194/essd-2026-777

Keywords: fine root biomass, soil carbon, random forest, machine learning, Qinghai-Tibet Plateau, root depth distribution, Earth System Science Data, gridded dataset, carbon cycle, China ecosystems, belowground biomass, ecosystem modeling