Scholarlink

<p>This study investigates the spatiotemporal dynamics of drought and wildfire vulnerability in Nova Scotia from 2001 to 2024 using remote sensing and machine learning approaches. Land Surface Temperature (LST), vegetation and moisture indices (NDVI, NDMI, NDWI, SAVI, and NVSWI), and the Standardized Precipitation Index (SPI) were integrated to assess ecological drought conditions and wildfire susceptibility. Results demonstrated that substantial temporal variability in drought intensity and thermal stress were observed in 2001, 2008, 2023, and 2024. Vegetation and moisture indices indicated higher ecological resilience in agriculturally productive regions, while coastal and upland areas remained highly vulnerable to vegetation degradation and moisture loss. Fire severity analysis using differenced Normalized Burn Ratio (dNBR) identified a widespread moderate-to-high burn intensity across major wildfire years. The ANN–GWR modelling framework effectively measured non-linear and spatially heterogeneous relationships between thermal conditions, vegetation stress, and wildfire occurrence (R<sup>2</sup> &gt; 0.85). Future projections for 2030–2040 suggested an increasing summer LST, expanding drought-prone zones, and elevated wildfire risk under continued climatic stress. The findings provided valuable insights for climate-resilient planning, wildfire mitigation, and sustainable ecosystem management in Nova Scotia.</p> Graphical Abstract <p></p> <p>The graphical abstract represents an integrated assessment of potential drought, forest fire hazard, ecological responses, and predictive modelling in Nova Scotia, Canada. The study integrates satellite remote sensing and machine learning approaches to assess drought stresses and forest fire impacts. LST and biophysical indices including the NDVI, NDMI, NDWI, and SAVI demonstrate the temporal changes of drought and fire hazard potentials for the years 2001, 2008, 2019, 2022, 2023 and 2024. Generated heatmap shows the critical relationships among the drought parameters, LST, and Vegetation Indices. A negative correlation is found among LST, vegetation, and moisture indices. A positive correlation exists between vegetation and moisture indices. Consequently, to address the non-linear relationship, an ANN model is employed using LST, biophysical indices, and spatial variables. Furthermore, the temporal changes for the average summer precipitation showing notable interannual variability and recurring low precipitation periods. Finally, the GWR model acknowledges variations among the considered coefficients over the study area. The approach in this study provides valuable evidence of depicting the risks of critical ecological zones, drought vulnerability and forest fire potential areas.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Spatiotemporal Mapping of Wildfire Zonation and Drought Analysis in Nova Scotia Using Machine Learning and Remote Sensing Techniques

  • 期刊论文

  • 发表日期：

    2026年7月7日

查看全文
  • Md. Tanvir Miah,
  • Md. Rakibul Hasan,
  • Jannatun Nahar Fariha,
  • Rukaya Islam,
  • Niloy Biswas,
  • Raiyan Raiyan,
  • Khan Rubayet Rahaman

摘要

This study investigates the spatiotemporal dynamics of drought and wildfire vulnerability in Nova Scotia from 2001 to 2024 using remote sensing and machine learning approaches. Land Surface Temperature (LST), vegetation and moisture indices (NDVI, NDMI, NDWI, SAVI, and NVSWI), and the Standardized Precipitation Index (SPI) were integrated to assess ecological drought conditions and wildfire susceptibility. Results demonstrated that substantial temporal variability in drought intensity and thermal stress were observed in 2001, 2008, 2023, and 2024. Vegetation and moisture indices indicated higher ecological resilience in agriculturally productive regions, while coastal and upland areas remained highly vulnerable to vegetation degradation and moisture loss. Fire severity analysis using differenced Normalized Burn Ratio (dNBR) identified a widespread moderate-to-high burn intensity across major wildfire years. The ANN–GWR modelling framework effectively measured non-linear and spatially heterogeneous relationships between thermal conditions, vegetation stress, and wildfire occurrence (R2 > 0.85). Future projections for 2030–2040 suggested an increasing summer LST, expanding drought-prone zones, and elevated wildfire risk under continued climatic stress. The findings provided valuable insights for climate-resilient planning, wildfire mitigation, and sustainable ecosystem management in Nova Scotia.

Graphical Abstract

The graphical abstract represents an integrated assessment of potential drought, forest fire hazard, ecological responses, and predictive modelling in Nova Scotia, Canada. The study integrates satellite remote sensing and machine learning approaches to assess drought stresses and forest fire impacts. LST and biophysical indices including the NDVI, NDMI, NDWI, and SAVI demonstrate the temporal changes of drought and fire hazard potentials for the years 2001, 2008, 2019, 2022, 2023 and 2024. Generated heatmap shows the critical relationships among the drought parameters, LST, and Vegetation Indices. A negative correlation is found among LST, vegetation, and moisture indices. A positive correlation exists between vegetation and moisture indices. Consequently, to address the non-linear relationship, an ANN model is employed using LST, biophysical indices, and spatial variables. Furthermore, the temporal changes for the average summer precipitation showing notable interannual variability and recurring low precipitation periods. Finally, the GWR model acknowledges variations among the considered coefficients over the study area. The approach in this study provides valuable evidence of depicting the risks of critical ecological zones, drought vulnerability and forest fire potential areas.