Accurately assessing forest fire susceptibility (FFS) in the Similipal Tiger Reserve (STR) is essential for biodiversity conservation, climate change mitigation, and community safety. Most existing studies have primarily focused on climatic and topographical factors, while this research expands the scope by employing a synergistic approach that integrates geographical information systems (GIS), remote sensing (RS), and machine learning (ML) methodologies for identifying and assessing forest fire-prone areas in the STR and their vulnerability to climate change. To achieve this, the study employed a comprehensive dataset of forty-four influencing factors, including topographic, climate-hydrologic, forest health, vegetation indices, radar features, and anthropogenic interference, into ten ML models: neural net (nnet), AdaBag, Extreme Gradient Boosting (XGBTree), Gradient Boosting Machine (GBM), Random Forest (RF), and its hybrid variants with differential evolution algorithm (RF-DEA), Gravitational Based Search (RF-GBS), Grey Wolf Optimization (RF-GWO), Particle Swarm Optimization (RF-PSO), and genetic algorithm (RF-GA). The study revealed high FFS in both the northern and southern portions of the study area, with the nnet and RF-PSO models demonstrating susceptibility percentages of 12.44% and 12.89%, respectively. Conversely, very low FFS zones consistently displayed susceptibility scores of approximately 23.41% and 18.57% for the nnet and RF-PSO models. The robust mapping methodology was validated by impressive AUROC (>0.88) and kappa coefficient (>0.62) scores across all ML validation metrics. Future climate models (ssp245 and ssp585, 2022–2100) indicated high FFS zones along the northern and southern edges of the STR, with the central zone categorized from low to very low susceptibility. Boruta analysis identified actual evapotranspiration (AET) and relative humidity as key factors influencing forest fire ignition. SHAP evaluation reinforced the influence of these factors on FFS, while also highlighting the significant role of distance to road, distance to settlement, dNBR, slope, and humidity in prediction accuracy. These results emphasize the critical importance of the proposed approach for forest fire mapping and provide invaluable insights for firefighting teams, forest management, planning, and qualification strategies to address future fire sustainability.

@en

Flood has long been known as one of the most catastrophic natural hazards worldwide. Mapping flood-prone areas is an important part of flood disaster management. In this study, a flood susceptibility mapping framework was developed based on a novel integration of nature-inspired algorithms into support vector regression (SVR). To this end, various remote sensing (RS) and geographic information system (GIS) datasets were applied to the hybridized SVR models to map flood susceptibility in Ahwaz township, Iran. The proposed framework has two main steps: 1) updating the flood inventory (historical flooded locations) using the proposed RS-based flood detection method developed within the google earth engine (GEE) platform. The mosaicked images of multi-temporal Sentinel-1 synthetic aperture radar (SAR) data have been used in this step; 2) producing flood susceptibility map using the standalone SVR and hybridized model of SVR. The hybridized methods were derived from a novel integration of SVR with meta-heuristic algorithms, hence forming the SVR-bat algorithm (SVR-BA), SVR-invasive weed optimization (SVR-IWO), and SVR-firefly algorithm (SVR-FA). A spatial database of flood locations and 11 conditioning factors (altitude, slope angle, aspect, topographic wetness index, stream power index, normalized difference vegetation index (NDVI), distance to stream, curvature, rainfall, soil type, and land use/cover) were built for the susceptibility modelling. The accuracy of the proposed model was evaluated using the statistical and sensitivity indices, such as root mean square error (RMSE), receiver operating characteristic (ROC) and area under the ROC curve (AUROC) index. The results indicated that all hybridized models outperformed the standalone SVR. According to AUROC values, the predictive power of the SVR-FA was the highest with the value of 0.81, followed by SVR-IWO, SVR-BA, and SVR with values of 0.80, 0.79, and 0.77, respectively.

@en

Progress toward habitat protection goals can effectively be performed using satellite imagery and machine-learning (ML) models at various spatial and temporal scales. In this regard, habitat types and landscape structures can be discriminated against using remote-sensing (RS) datasets. However, most existing research in three-dimensional (3D) habitat mapping primarily relies on same/cross-sensor features like features derived from multibeam Light Detection And Ranging (LiDAR), hydrographic LiDAR, and aerial images, often overlooking the potential benefits of considering multi-sensor data integration. To address this gap, this study introduced a novel approach to creating 3D habitat maps by using high-resolution multispectral images and a LiDAR-derived Digital Surface Model (DSM) coupled with an object-based Random Forest (RF) algorithm. LiDAR-derived products were also used to improve the accuracy of the habitat classification, especially for the habitat classes with similar spectral characteristics but different heights. Two study areas in the United Kingdom (UK) were chosen to explore the accuracy of the developed models. The overall accuracies for the two mentioned study areas were high (91% and 82%), which is indicative of the high potential of the developed RS method for 3D habitat mapping. Overall, it was observed that a combination of high-resolution multispectral imagery and LiDAR data could help the separation of different habitat types and provide reliable 3D information.@en

Remote sensing datasets are great resources to map habitat types. In this study, 3D habitat maps were generated using high-resolution multispectral imagery and a LiDAR-derived digital surface model (DSM). Two study areas in the United Kingdom (UK) were selected to investigate the potential of the developed models in habitat classification. The overall classification accuracies for the two study areas were high (91% and 82%), indicating the satisfactory performance of the developed approach for habitat mapping in the study areas. Overall, it was observed that a synergy of high-resolution multi-spectral imagery and LiDAR data could provide reliable 3D information on habitat types.@en