Scholarlink

Background <p>Wildfires in mountainous regions pose significant ecological and socio-economic threats, with their frequency and severity expected to increase due to climate change. Therefore, accurately and promptly characterizing wildfires is crucial for understanding their behaviour and potential damage. Remote sensing and geostatistical methods have become increasingly relevant for providing this timely and precise characterization. This study analyzed the wildfire event that occurred in November 2022 in the Golden Gate Highlands National Park (GGHNP), South Africa, using the spectral-rich imagery from Sentinel-2. The fire severity in the region was quantified using the difference Normalized Burn Ratio (dNBR). This study employs a Generalized Linear Model (GLM) and Random Forests (RF) to identify and quantify the relationships between fire severity and key environmental factors. These environmental factors include topographical variables, climatic factors, and the Normalized Difference Vegetation Index (NDVI), along with various vegetation communities.</p> Results <p>The results revealed that over 50% of the GGHNP was affected by wildfires, with 0.4% of the area experiencing severe burns and less than half of the area remaining unburned. The GLM explained 52.1% of the variation in fire severity (R<sup>2</sup> = 0.521) and identified NDVI, aspect, solar radiation, terrain ruggedness, evaporation, and temperature as the most influential factors. RF identified NDVI, aspect, and solar radiation as the most influential factors based on variable importance (%IncMSE and IncNodePurity) values.</p> Conclusions <p>Overall, these findings enhance wildfire prediction, management, and mitigation strategies in the ecosystem. Future research should focus on investigating long-term fire trends and the effects of climate change to inform conservation efforts in fire-prone mountainous regions. This study demonstrates the novelty of using GLM and RF to integrate Sentinel-2 dNBR, vegetation communities, climate and topography interaction in the South African grassland ecosystem. Additionally, the study analyzes fire severity by management compartments and validates severity maps with SANPark control points, providing insight for fire management.</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

Potential drivers of fire severity in Golden Gate Highlands National Park: a case study of the November 2022 big fire event

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月1日

查看全文
  • Morena Mapuru,
  • Sifiso Xulu,
  • Irvin D. Shandu,
  • Ernest Daemane,
  • Lufuno Munyai

摘要

Background

Wildfires in mountainous regions pose significant ecological and socio-economic threats, with their frequency and severity expected to increase due to climate change. Therefore, accurately and promptly characterizing wildfires is crucial for understanding their behaviour and potential damage. Remote sensing and geostatistical methods have become increasingly relevant for providing this timely and precise characterization. This study analyzed the wildfire event that occurred in November 2022 in the Golden Gate Highlands National Park (GGHNP), South Africa, using the spectral-rich imagery from Sentinel-2. The fire severity in the region was quantified using the difference Normalized Burn Ratio (dNBR). This study employs a Generalized Linear Model (GLM) and Random Forests (RF) to identify and quantify the relationships between fire severity and key environmental factors. These environmental factors include topographical variables, climatic factors, and the Normalized Difference Vegetation Index (NDVI), along with various vegetation communities.

Results

The results revealed that over 50% of the GGHNP was affected by wildfires, with 0.4% of the area experiencing severe burns and less than half of the area remaining unburned. The GLM explained 52.1% of the variation in fire severity (R2 = 0.521) and identified NDVI, aspect, solar radiation, terrain ruggedness, evaporation, and temperature as the most influential factors. RF identified NDVI, aspect, and solar radiation as the most influential factors based on variable importance (%IncMSE and IncNodePurity) values.

Conclusions

Overall, these findings enhance wildfire prediction, management, and mitigation strategies in the ecosystem. Future research should focus on investigating long-term fire trends and the effects of climate change to inform conservation efforts in fire-prone mountainous regions. This study demonstrates the novelty of using GLM and RF to integrate Sentinel-2 dNBR, vegetation communities, climate and topography interaction in the South African grassland ecosystem. Additionally, the study analyzes fire severity by management compartments and validates severity maps with SANPark control points, providing insight for fire management.