SHAP-Driven Explainable Artificial Intelligence Framework for Wildfire Susceptibility Mapping Using MODIS Active Fire Pixels: An In-Depth Interpretation of Contributing Factors in Izmir, Türkiye

Remote Sensing, Год журнала: 2024, Номер 16(15), С. 2842 - 2842

Опубликована: Авг. 2, 2024

Wildfire susceptibility maps play a crucial role in preemptively identifying regions at risk of future fires and informing decisions related to wildfire management, thereby aiding mitigating the risks potential damage posed by wildfires. This study employs eXplainable Artificial Intelligence (XAI) techniques, particularly SHapley Additive exPlanations (SHAP), map Izmir Province, Türkiye. Incorporating fifteen conditioning factors spanning topography, climate, anthropogenic influences, vegetation characteristics, machine learning (ML) models (Random Forest, XGBoost, LightGBM) were used predict wildfire-prone areas using freely available active fire pixel data (MODIS Active Fire Collection 6 MCD14ML product). The evaluation trained ML showed that Random Forest (RF) model outperformed XGBoost LightGBM, achieving highest test accuracy (95.6%). All classifiers demonstrated strong predictive performance, but RF excelled sensitivity, specificity, precision, F-1 score, making it preferred for generating conducting SHAP analysis. Unlike prevailing approaches focusing solely on global feature importance, this fills critical gap employing summary dependence plots comprehensively assess each factor’s contribution, enhancing explainability reliability results. analysis reveals clear associations between such as wind speed, temperature, NDVI, slope, distance villages with increased susceptibility, while rainfall streams exhibit nuanced effects. spatial distribution classes highlights areas, flat coastal near settlements agricultural lands, emphasizing need enhanced awareness preventive measures. These insights inform targeted management strategies, highlighting importance tailored interventions like firebreaks management. However, challenges remain, including ensuring selected factors’ adequacy across diverse regions, addressing biases from resampling spatially varied data, refining broader applicability.

Язык: Английский

---

Enhancing PM2.5 Air Pollution Prediction Performance by Optimizing the Echo State Network (ESN) Deep Learning Model Using New Metaheuristic Algorithms

Urban Science, Год журнала: 2025, Номер 9(5), С. 138 - 138

Опубликована: Апрель 23, 2025

Air pollution presents significant risks to both human health and the environment. This study uses air meteorological data develop an effective deep learning model for hourly PM2.5 concentration predictions in Tehran, Iran. evaluates efficient metaheuristic algorithms optimizing hyperparameters improve accuracy of predictions. The optimal feature set was selected using Variance Inflation Factor (VIF) Boruta-XGBoost methods, which indicated elimination NO, NO2, NOx. highlighted PM10 as most important feature. Wavelet transform then applied extract 40 features enhance prediction accuracy. Hyperparameters weights matrices Echo State Network (ESN) were determined algorithms, with Salp Swarm Algorithm (SSA) demonstrating superior performance. evaluation different criteria revealed that ESN-SSA outperformed other hybrids original ESN, LSTM, GRU models.

Язык: Английский

---

Effects of confined distance near floor and wire size on electrical wire flame spread behaviors based on heat transfer

International Journal of Thermal Sciences, Год журнала: 2024, Номер 203, С. 109173 - 109173

Опубликована: Май 29, 2024

Язык: Английский

---

Wildfire Scenarios for Assessing Risk of Cover Loss in a Megadiverse Zone within the Colombian Caribbean

Sustainability, Год журнала: 2024, Номер 16(8), С. 3410 - 3410

Опубликована: Апрель 18, 2024

Rising wildfire incidents in South America, potentially exacerbated by climate change, require an exploration of sustainable approaches for fire risk reduction. This study investigates wildfire-prone meteorological conditions and assesses the susceptibility Colombia’s megadiverse northern region. Utilizing this knowledge, we apply a machine learning model Monte Carlo approach to evaluate sustainability strategies mitigating risk. The findings indicate that substantial number fires occur southern region, especially first two seasons year, northeast last seasons. Both are characterized high temperatures, minimal precipitation, strong winds, dry conditions. developed demonstrates significant predictive accuracy with HIT, FAR, POC 87.9%, 28.3%, 95.7%, respectively, providing insights into probabilistic aspects development. Various scenarios showed decrease soil temperature reduces mostly lower altitudes leaf skin reservoir content highest altitudes, as well north Sustainability strategies, such tree belts, agroforestry mosaics, forest corridors emerge crucial measures. results underscore importance proactive measures impact, offering actionable crafting effective amid escalating risks.

Язык: Английский

---

Advancing the LightGBM approach with three novel nature-inspired optimizers for predicting wildfire susceptibility in Kauaʻi and Molokaʻi Islands, Hawaii

Expert Systems with Applications, Год журнала: 2024, Номер 258, С. 124963 - 124963

Опубликована: Авг. 5, 2024

Язык: Английский

---

Assessing Pan-Canada wildfire susceptibility by integrating satellite data with novel hybrid deep learning and black widow optimizer algorithms

The Science of The Total Environment, Год журнала: 2025, Номер 977, С. 179369 - 179369

Опубликована: Апрель 15, 2025

In light of the rising frequency severe wildfires and their widespread socio-ecological impacts, it is essential to develop cost-effective reliable methods for accurately predicting mapping wildfire occurrences. This study aimed several novel deep-learning models determine probability occurrence on a national scale in Canada by integrating remote sensing data, deep learning, metaheuristic algorithms. present study, standalone long short-term memory (LSTM), recurrent neural network (RNN), bidirectional LSTM (BiLSTM), RNN (BiRNN) were developed, these hybridized with black widow optimizer (BWO). To train test models, 4240 historical (2014-2023) large locations collected across Canada. Fourteen wildfire-related predictors used map susceptibility, Gini coefficient determining each predictor's importance occurrence. Finally, developed evaluated tested using area under receiver operating characteristic curve (AUC), other statistical error metrics. During testing stage, hybrid BiLSTM-BWO model outperformed (AUC = 0.9686), followed RNN-BWO 0.9683), LSTM-BWO 0.9672), BiRNN-BWO 0.9643), BiLSTM 0.9420), 0.9367), BiRNN 0.9247) 0.8737). Based model, 19.7 %, 42.6 13.4 14.5 9.8 % was classified as having very low, moderate, high, high susceptibility future wildfires, respectively. Saskatchewan, Manitoba, British Columbia Alberta among provinces areas while Prince Edward Island Newfoundland Labrador from Atlantic had lowest According coefficient, windspeed, land use cover, precipitation, specific humidity maximum temperature strongest impact highlights effectiveness prediction potential improve management, prevention, mitigation strategies Canada's future.

Язык: Английский

---

Assessing Forest Road Network Suitability in Relation to the Spatial Occurrence of Wildfires in Mediterranean Forest Ecosystems

Fire, Год журнала: 2024, Номер 7(6), С. 175 - 175

Опубликована: Май 22, 2024

Identifying the relationship between forest roads and wildfires in ecosystems is a crucial priority to integrate suppression prevention within wildfire management. In various investigations, interaction of these elements has been studied by using road density as one anthropogenic dependent variables. This study focused on use broader set metrics associated with networks, such density, number links (edges), access percentage based two effect zones (road buffers 75 m 97 m). These were employed response variables assess network suitability relation wildfires, specifically size fires (2000–2021), Apulia region (Italy) case study. addition, enhance comprehensive understanding networks this considered affecting factors, including land-cover data (forest, maquis, natural grassland), geomorphology (slope, aspect), vegetation (Normalized Difference Vegetation Index (NDVI)), morphometric indexes (Topographic Position (TPI), Terrain Ruggedness (TRI), Topographic Wetness (TWI)). We used geographically weighted regression (GWR) ordinary least squares (OLS) analyze Results showed that GWR models outperformed OLS term statistical results R2 Akaike Information Criterion (AICc). found among metrics, do not effectively demonstrate correlation singular criterion. However, they prove be beneficial supplementary variable when alongside percentage, particularly 75-m buffer zone. Our findings are discuss implications for planning management analysis. one-dimensional static infrastructure; rather, multi-dimensional dynamic structure. Hence, need analyzed from perspectives, accessibility ecological approaches, order obtain an integrated understating their wildfire.

Язык: Английский

---

A Critical Review on Metaheuristic Algorithms based Multi-Criteria Decision-Making Approaches and Applications

Archives of Computational Methods in Engineering, Год журнала: 2024, Номер unknown

Опубликована: Авг. 2, 2024

Язык: Английский

---