The research, published in Frontiers in Forests and Global Change, centers on the Canadian Forest Fire Weather Index (FWI) system, a physics-based framework used across the globe to translate routine weather observations into fire danger ratings. The system requires only four inputs—daily temperature, relative humidity, wind speed, and precipitation—but its inner workings are layered. Three fuel moisture codes track the wetness of dead organic material at increasing depths: the Fine Fuel Moisture Code (FFMC) responds to weather within two to three days, the Duff Moisture Code (DMC) integrates conditions over roughly twelve days, and the Drought Code (DC) reflects deep seasonal drought over about fifty-two days. From these codes, the system derives the Initial Spread Index, the Build-up Index, and finally the FWI itself, a composite measure of potential fire-line intensity.

Operational FWI calculations work well where stations are dense, but Shanxi poses a difficult test. More than eighty percent of the province is mountainous or hilly, with the Taihang Mountains to the east, the Lüliang Mountains to the west, and a chain of basins running down the center. Its 111 national automatic weather stations are mostly sited in accessible, open locations, leaving high-elevation slopes and forest interiors unobserved. Traditional interpolation methods such as Kriging tend to smooth away the very spatial heterogeneity that matters most for fire danger. Meanwhile, ERA5-Land, the high-resolution European reanalysis product, provides continuous gridded meteorology but carries known systematic biases in precipitation and near-surface winds over complex topography—biases that the FWI’s nonlinear equations can amplify during daily iteration.

The research team, based at the State Grid Shanxi Electric Power Company Electric Power Research Institute, built a station-supervised machine learning framework to correct these errors. They assembled 25,798 quality-controlled station-day records from March to November 2025, split chronologically so that the model was trained on spring and summer data and validated on autumn conditions it had never seen. Using XGBoost, a gradient-boosted tree algorithm prized for its interpretability, the team predicted the three moisture codes directly, then derived the remaining indices through the standard FWI equations—ensuring the machine learning respected the physics-defined structure of the system.

Crucially, the researchers ran a set of ablation experiments to isolate the value of each data source. The baseline scheme, M1, drove the FWI equations directly with ERA5-Land meteorology and struggled, explaining only about twenty-three percent of the variance in the final FWI. The second scheme, M2, used XGBoost to learn nonlinear corrections of the reanalysis biases, dramatically tightening the fit—for example, raising the coefficient of determination for the Drought Code from 0.524 to 0.761. The third scheme, M3, added fused passive-microwave soil moisture from the SMAP, AMSR2, and Chinese FY-3 satellite sensors, processed by the National Satellite Meteorological Center into a daily national product.

The addition of satellite soil moisture pushed performance higher across the board. The ERA5 plus soil moisture plus XGBoost configuration achieved coefficients of determination of 0.69 for FFMC, 0.63 for DMC, 0.77 for DC, and 0.43 for the overall FWI, while cutting the root-mean-square error for the ignition-sensitive FFMC from 18.74 to 15.15. Time-series analysis showed the fused model tracked sharp FWI spikes during autumn dry-wet swings that the raw reanalysis missed entirely, including an extreme event in early November that the baseline completely failed to capture. The soil-moisture features also suppressed the systematic underestimation of extreme high values that plagued both simpler schemes.

As a physical explanation, the microwave measurements make sense. Passive microwave sensors penetrate clouds and retrieve the dielectric constant of the topsoil, which is strongly coupled through energy and water exchange with the moisture of dead fuels lying on the forest floor. The team engineered multi-scale features from these retrievals—rolling 7-, 14-, and 30-day means capturing short-, medium-, and long-term moisture backgrounds, plus a high-temperature drought stress index combining temperature and soil water content—allowing the model to represent the soil-fuel system’s gradual drying and rewetting rhythms. They also added lagged predictors, cumulative precipitation sums, and sine-cosine encodings of the day of year to capture seasonal transitions, while taking explicit precautions against target leakage in the lagged variables.

Beyond model development, the study mapped where and when Shanxi’s fire danger peaks. The FFMC fluctuated violently with weather, regularly exceeding 80—the extreme risk range—during the low-humidity, windy spring, while the deeper codes accumulated drought steadily toward mid-August. The final FWI showed a single-peak annual structure, with late spring to early summer, from April through June, constituting the province’s absolute high-risk period. Statistically, FFMC and DC were nearly decoupled, with a correlation of just 0.134, demonstrating that surface fine-fuel flammability and deep soil drought are governed by different processes and must be modeled independently.

Geographically, the fire danger showed a clear south-high, north-low structure. The highest mean values of all four indices clustered in the central-southern and southwestern mountainous regions, around the edges of the Jinan Basin and the southern Lüliang range, roughly between latitudes 35 and 37 degrees north. The researchers attribute this to the southern region’s earlier spring warming, more intense solar radiation, and vigorous evaporation, amplified by terrain-driven winds on mountain slopes. Using K-means clustering on the standardized mean values of the four indices, they produced a five-level fire-weather zoning in which high-risk areas formed contiguous blocks in the south while the cooler, wetter north remained largely low risk—a pattern consistent with the province’s natural geographic conditions.

The authors are candid about limitations. No spatially independent validation or high-resolution dynamical baseline was performed, so extrapolation to unsampled mountain locations remains uncertain. The FWI system’s empirical constants were originally calibrated in Canadian boreal forests, and direct transfer to Shanxi’s warm-temperate deciduous and mixed forests may introduce seasonal biases. Passive microwave soil moisture represents only the near-surface layer, which limited the gain for the long-memory Drought Code and produced none for the DMC. Stations also under-sample the very microclimates where fires ignite, and the 0.01-degree working grid does not reflect the true resolution of the sensors. Still, the study demonstrates a practical path forward: correcting reanalysis products against ground truth, enriching them with satellite observations of the land surface, and quantifying exactly which data sources earn their place in the model. Future work will benchmark the approach against mesoscale atmospheric simulations and deep spatiotemporal architectures, and localize the FWI parameters for Chinese fuel types—steps toward fire danger maps precise enough to guide prevention at the scale of a single ridge or valley.

Subject of Research: Machine learning and satellite soil moisture fusion for improving fire weather index estimation in mountainous terrain

Article Title: Spatiotemporal variation characteristics and multi-combination modeling assessment of fire weather index in Shanxi Province

Article References: Lu, Z., Liang, J., Li, J., Yuan, H., Wang, S., & Duan, Z. (2026). Spatiotemporal variation characteristics and multi-combination modeling assessment of fire weather index in Shanxi Province. Frontiers in Forests and Global Change, 9, Article 1892417. https://doi.org/10.3389/ffgc.2026.1892417

Image Credits: AI Generated

DOI: 10.3389/ffgc.2026.1892417

Keywords: fire weather index, XGBoost, ERA5-Land, soil moisture, satellite remote sensing, Shanxi Province, forest fire risk, machine learning, Canadian FWI system, mountainous terrain, reanalysis bias correction, microwave remote sensing

Carbon storage in terrestrial ecosystems is largely a function of land cover. Forests hold vast quantities of carbon in trunks, roots and soils; grasslands store smaller but still significant amounts; croplands cycle carbon quickly through seasonal growth and harvest; and built-up areas store comparatively little. When farmland is converted to housing estates or forests are cleared for mining, the carbon accounting of an entire region can shift dramatically. Tracking those shifts requires maps of land use that are both accurate and finely resolved in time, which is precisely what the research team set out to produce for the Fen River Basin.

The first pillar of the methodological chain is PLUS, the Patch-generating Land Use Simulation model. PLUS is designed to mine the rules of land-use change from historical maps and then project future landscapes under alternative scenarios. It works by analyzing how individual land patches expand or contract in response to terrain, accessibility, population and policy constraints, and it uses a random-forest-based rule discovery engine to capture the nonlinear relationships between those drivers and observed land transitions. In the Fen River Basin, PLUS allowed the researchers to generate internally consistent land-use maps for multiple time points and to simulate plausible futures under different development pathways.

The second pillar is the Integrated Valuation of Ecosystem Services and Tradeoffs, or InVEST, model developed with support from the Natural Capital Project. Its carbon storage module estimates the carbon held in four pools—above-ground biomass, below-ground biomass, dead organic matter and soil organic carbon—by assigning each land-cover class a set of carbon densities and multiplying those densities by mapped areas. The result is a spatially explicit picture of where carbon is concentrated and how the regional total moves as land cover changes. By coupling InVEST to the PLUS-generated land maps, the researchers could translate every hectare of forest loss or gain into a quantified change in stored carbon.

The third pillar, GeoDetector, addresses a question that pure mapping cannot answer: which factors actually explain the observed spatial variation in carbon storage. GeoDetector is a statistical technique rooted in spatial variance analysis. It compares the variance of a variable within spatial strata defined by a potential driving factor with the variance of the whole study area. If carbon storage values are far more homogeneous inside, say, zones of a given land-use type or precipitation band than across the basin as a whole, that factor is judged to have strong explanatory power. The method is particularly valuable because it captures nonlinear interactions without imposing a predefined functional form, and it quantifies how pairs of factors act together.

Applying this three-part chain to the Fen River Basin revealed a landscape under measurable stress. The basin’s carbon stock, the analysis indicates, has been shaped by the expansion of construction land and the pressures of cultivated agriculture on one side, and by afforestation and grassland restoration programs on the other. China’s sweeping ecological initiatives, including the Grain for Green Program that converts sloping cropland back to forest and grassland, have pushed in the direction of greater carbon storage, while urbanization and mining-related land disturbance have pushed against it. The net trajectory reflects the balance of these opposing forces, and the study quantifies that balance in tonnage terms for successive time windows.

The spatial detail matters as much as the totals. The InVEST maps show that carbon storage is concentrated in the basin’s forested uplands and along vegetated margins, while the densely settled valley floor, where Taiyuan and other industrial cities sit, functions as a persistent carbon deficit zone. This uneven geography means that a hectare of forest protected in the loess hills yields far more carbon benefit than a hectare of marginal grassland elsewhere, and it gives provincial planners a concrete basis for prioritizing restoration investment. The maps also expose hotspots where carbon losses have accelerated, often coinciding with mining expansion or the outward growth of urban built-up areas along transport corridors.

The GeoDetector analysis adds explanatory depth by ranking the drivers behind this spatial pattern. Land-use type itself emerges as the dominant single factor, an expected but important confirmation that vegetation class controls the carbon ledger. Beyond land use, climatic variables such as precipitation and temperature, topographic features including elevation and slope, soil properties, and human-intensity measures such as population density and gross domestic product all contribute, with their relative influence varying by sub-period. Crucially, the method shows that interactions between factors—between precipitation and land use, or between slope and human activity—often explain more of the spatial variance than any single factor alone, underscoring that carbon storage in the basin is governed by coupled natural and human systems rather than by any one lever.

Perhaps the most forward-looking element of the study is its use of scenario simulation. By driving the PLUS model with different assumptions about future development—ranging from trajectories that prioritize economic growth and continued urban expansion to ones emphasizing ecological protection and cropland preservation—the researchers generated contrasting projections of the basin’s land system and, through InVEST, of its carbon future. The comparisons suggest that policy choices made in the coming years will have consequences measurable in millions of tons of carbon: under protection-oriented scenarios, continued afforestation and constrained urban sprawl raise projected storage, while growth-oriented scenarios erode it, particularly around existing urban cores. These scenario outputs give decision-makers a quantified trade-off table rather than a vague warning.

The significance of the work extends beyond one river basin. The Fen River drains a portion of the Yellow River system and typifies the ecological fragility of China’s loess plateau region, where soil erosion, water scarcity and industrial land use intersect. A transferable PLUS-InVEST-GeoDetector framework, validated in such a demanding setting, can be redeployed in other basins facing similar tensions between development and carbon goals. As China pursues its commitments to peak carbon emissions and to enhance carbon sinks, tools that connect land-use planning to carbon accounting at landscape scale are likely to become standard instruments of regional policy. This study demonstrates that when land change modeling, ecosystem service valuation and driver detection are chained together, the result is not just a set of maps but an actionable diagnosis of where a landscape’s carbon is going, why it is moving, and what choices could keep more of it in the ground.

Subject of Research: Spatio-temporal evolution and driving forces of terrestrial carbon storage in the Fen River Basin using an integrated PLUS-InVEST-GeoDetector modeling framework

Article Title: Spatio-temporal evolution and driving force analysis of carbon storage coupled with PLUS-InVEST-GeoDetector in the Fen River Basin of China

Article References: Chen, J., Gao, P., & Hou, Y. (2026). Spatio-temporal evolution and driving force analysis of carbon storage coupled with PLUS-InVEST-GeoDetector in the Fen River Basin of China. Scientific Reports. https://doi.org/10.1038/s41598-026-71784-9

Image Credits: AI Generated

DOI: 10.1038/s41598-026-71784-9

Keywords: carbon storage, Fen River Basin, PLUS model, InVEST model, GeoDetector, land use simulation, ecosystem services, Shanxi Province, afforestation, urbanization, carbon sinks, spatial analysis