A Forest Fire Prediction Framework Based on Multiple Machine Learning Models

Forests, Journal Year: 2025, Volume and Issue: 16(2), P. 329 - 329

Published: Feb. 13, 2025

Fire risk prediction is of great importance for fire prevention. maps are an effective tool to quantify regional risk. Most existing studies on forest mainly use a single machine learning model, but different models have varying degrees feature extraction in the same spatial environment, leading inconsistencies accuracy. To address this issue, study proposes novel integrated framework that systematically evaluates multiple and combines their outputs through weighted ensemble approach, thereby enhancing robustness. During selection stage, factors including socio-economic, climate, terrain, remote sensing data, human were considered. Unlike previous eight evaluated compared using performance metrics. Three based Mean Squared Error (MSE) values, cross-validation results showed improvement model performance. The achieved accuracy 0.8602, area under curve (AUC) 0.772, superior sensitivity (0.9234), outperforming individual models. Finally, was applied generate map. Compared with prior studies, multi-model approach not only improves predictive also provides scalable adaptable mapping, valuable insights future sustainability issues.

Language: Английский

---

A 1 km monthly dataset of historical and future climate changes over China

Scientific Data, Journal Year: 2025, Volume and Issue: 12(1)

Published: March 13, 2025

High-resolution climate data are important for understanding the impacts of change on multiple sectors worldwide. In this study, based latest released meteorological records during 1991–2020 and recently updated general circulation models (GCMs), we established a 30-year averaged 0.01° (≈1 km) dataset 5 basic variables 23 bioclimatic variables, using ANUSPLIN software, delta correction (DC) downscaling, cubic spline resampling method. Each variable contained monthly gridded historical bias-corrected future over three periods (2021–2040, 2041–2070, 2071–2100), scenarios (SSP1-2.6, SSP2-4.5, SSP5-8.5) 10 GCMs (including an ensemble model). The interpolations generated by ANUSPLIIN software showed good fit (above 0.91) with observations. DC improved accuracy most GCM original simulations, reducing bias 0.69%–58.63%. This new therefore demonstrates reliable quality, further provides high-resolution long-term across China ecological impact studies.

Language: Английский

---

Forest Wildfire Risk Assessment of Anning River Valley in Sichuan Province Based on Driving Factors with Multi-Source Data

Forests, Journal Year: 2024, Volume and Issue: 15(9), P. 1523 - 1523

Published: Aug. 29, 2024

Forest fires can lead to a decline in ecosystem functions, such as biodiversity, soil quality, and carbon cycling, causing economic losses health threats human societies. Therefore, it is imperative map forest-fire risk mitigate the likelihood of occurrence. In this study, we utilized hierarchical analysis process (AHP), comprehensive weighting method (CWM), random forest Anning River Valley Sichuan Province. We selected non-photosynthetic vegetation (NPV), photosynthetic (PV), normalized difference index (NDVI), plant species, land use, type, temperature, humidity, rainfall, wind speed, elevation, slope, aspect, distance road, residential predisposing factors. derived following conclusions. (1) Overlaying historical fire points with mapped revealed an accuracy that exceeded 86%, indicating reliability results. (2) primarily occur February, March, April, typically months characterized by very low rainfall dry conditions. (3) Areas high medium were mainly distributed Dechang Xide counties, while low-risk areas most prevalent Xichang city Mianning country. (4) Rainfall, NPV emerged main influencing factors, exerting dominant role occurrence fires. Specifically, higher coverage correlates increased fire. conclusion, study represents novel approach incorporating PV key factors triggering By mapping risk, have provided robust scientific foundation decision-making support for effective management strategies. This research significantly contributes advancing ecological civilization fostering sustainable development.

Language: Английский

---

Fuel Load Models for Different Tree Vegetation Types in Sichuan Province Based on Machine Learning

Forests, Journal Year: 2024, Volume and Issue: 16(1), P. 42 - 42

Published: Dec. 29, 2024

(1) Objective: To improve forest fire prevention, this study provides a reference for risk assessment in Sichuan Province. (2) Methods: This research focuses on various vegetation types Given data from 6848 sample plots, five machine learning models—random forest, extreme gradient boosting (XGBoost), k-nearest neighbors, support vector machine, and stacking ensemble (Stacking)—were employed. Bayesian optimization was utilized hyperparameter tuning, resulting models predicting fuel loads (FLs) across different types. (3) Results: The FL model incorporates not only characteristics but also site conditions climate data. Feature importance analysis indicated that structural factors (e.g., canopy closure, diameter at breast height, tree height) dominated cold broadleaf, subtropical mixed forests, while mean annual temperature seasonality) were more influential coniferous forests. Machine learning-based outperform the multiple stepwise regression both fitting ability prediction accuracy. XGBoost performed best coniferous, with coefficient of determination (R2) values 0.79, 0.85, 0.81, 0.83, respectively. Stacking excelled achieving an R2 value 0.82. (4) Conclusions: establishes theoretical foundation capacity It is recommended be applied to predict broadleaf suggested FLs Furthermore, offers management, assessment, prevention control

Language: Английский

---