The moisture content of fine dead surface fuel in forests is a crucial metric for assessing its combustibility and plays a pivotal role in the early warning, occurrence, and spread of forest fires. Accurate prediction of the moisture content of fine dead fuel on the forest surface is a critical challenge in forest fire management. Previous research on fine surface fuel moisture content has been mainly focused on coniferous forests in cold temperate zones, but there has been less attention given to understanding the fuel moisture dynamics in subtropical forests, which limits the development of regional forest fire warning models. Here, we consider the coupled influence of multiple meteorological, terrain, forest stand, and other characteristic factors on the fine dead fuel moisture content within the subtropical evergreen broadleaved forest region of southern China. The ability of five machine learning algorithms to predict the moisture content of fine dead fuel on the forest surface is assessed, and the key factors affecting the model accuracy are identified. Results show that when a single meteorological factor is used as a forecasting model, its forecasting accuracy is less than that of the combined model with multiple characteristic factors. However, the prediction accuracy of the model is improved after the addition of forest stand factors and terrain factors. The model prediction ability is the best for the combination of all feature factors including meteorology, forest stand, and terrain. The overall prediction accuracy of the model is ordered as follows: random forest > extreme gradient boosting > support vector machine > stepwise linear regression > k-nearest neighbor. Canopy density in forest stand factors, slope position and altitude in terrain factors, and average relative air humidity and light intensity in the previous 15 days are the key meteorological factors affecting the prediction accuracy of fuel moisture content. Our results provide scientific guidance and support for understanding the variability of forest surface fuel moisture content and improved regional forest fire warnings.

The understory fine dead fuel moisture content (DFMC) is an important reference indicator for regional forest fire warnings and risk assessments, and determining it on a large scale is a critical goal. It is difficult to estimate understory fine DFMC directly from satellite images due to canopy shading. To address this issue, we used canopy meteorology estimated by Landsat images in combination with explanatory variables to construct random forest models of in-forest meteorology, and then construct random forest models by combining the meteorological factors and explanatory variables with understory fine DFMC obtained from the monitoring device to (1) investigate the feasibility of Landsat images for estimating in-forest meteorology; (2) explore the feasibility of canopy or in-forest meteorology and explanatory variables for estimating understory fine DFMC; and (3) compare the effects of each factor on model accuracy and its effect on understory fine DFMC. The results showed that random forest models improved in-forest meteorology estimation, enhancing in-forest relative humidity, vapor pressure deficit, and temperature by 50%, 34%, and 2.2%, respectively, after adding a topography factor. For estimating understory fine DFMC, models using vapor pressure deficit improved fit by 10.2% over those using relative humidity. Using in-forest meteorology improved fits by 36.2% compared to canopy meteorology. Including topographic factors improved the average fit of understory fine DFMC models by 123.1%. The most accurate model utilized in-forest vapor pressure deficit, temperature, topographic factors, vegetation index, precipitation data, and seasonal factors. Correlations indicated that slope, in-forest vapor pressure deficit, and slope direction were most closely related to understory fine DFMC. The regional understory fine-grained DFMC distribution mapped according to our method can provide important decision support for forest fire risk early warning and fire management.

Based on the Google Earth Engine (GEE) platform, this study constructed a Remote Sensing Ecological Index (RSEI) time series for the Poyang Lake Basin from 2000 to 2024. The analysis integrated Theil-Sen median trend estimation, the Mann-Kendall test, the Hurst exponent, and an interpretable machine-learning framework combining extreme gradient boosting (XGBoost) with Shapley additive explanations (SHAP) to examine the spatiotemporal evolution of ecological quality and the factors associated with its spatial heterogeneity. The basin-average RSEI increased slightly from 0.693 in 2000 to 0.710 in 2024, indicating that the basin generally remained at a good ecological level, although interannual fluctuations were evident. Spatially, the RSEI showed a stable pattern of high values in the surrounding mountainous forest belts and lower values in the central plain, urbanized areas, and some lake and wetland zones. The ecological grade structure improved, with the proportion of areas classified as “excellent” increasing from 14.42% to 22.93%. Theil-Sen and Mann-Kendall results showed that areas with historical ecological improvement were more extensive than areas of degradation. However, the Hurst exponent indicated that many pixels exhibited anti-persistent characteristics (H < 0.5), suggesting that future ecological trajectories may be uncertain rather than a simple continuation of historical improvement. XGBoost-SHAP analysis showed that land surface temperature (LST) and NDVI made the largest contributions to RSEI prediction. Because both variables are components of the RSEI, these results should be interpreted as index sensitivity to heat and vegetation conditions rather than independent causal effects. External variables, including GDP, population density, nighttime light intensity, elevation, and slope, showed additional predictive associations with spatial heterogeneity, conditional on the fitted model. These findings not only support differentiated watershed management but also provide a remote sensing perspective on long-term ecological quality changes associated with climate variability and human activities.