Scholarlink

This study employs advanced geospatial techniques to analyze landslide susceptibility in Uttara Kannada district, Karnataka, from 2025 to 2050. We identified approximately 1852.18 sq km of high-risk zones using the Maxent model. The MGWR analysis revealed significant correlations with landslide occurrences, incorporating the highly influencing environmental parameters (water proximity, TWI, soil type, slope, rainfall, NDVI, LULC, lineament proximity, and geomorphology). The standardized residuals vs. predicted plot indicated a generally well-fitted model, with residuals randomly dispersed around zero. However, some areas showed patterns suggesting potential model refinements. The WP significance map highlighted regions with a strong statistical impact of these parameters, guiding targeted mitigation efforts. Emerging hotspot and cold spot analyses provided insights into spatial and temporal trends in landslide activity. Hotspots, primarily located in the Western Ghats and certain coastal areas, indicate regions of increasing landslide occurrences due to high rainfall, steep slopes, and other contributing factors. Conversely, cold spots in the northern parts of the district reflect areas with decreasing landslide activity, likely due to effective land management and mitigation measures. These findings underscore the importance of focused interventions in hotspot areas and the need to maintain successful practices in cold spot regions, thereby enhancing disaster risk management and promoting sustainable land-use practices in Uttara Kannada. The application of the artificial neural network (ANN) model has further proven the model’s significance for landslide hazard zonation mapping, as evidenced by an R2 value indicating a strong model fit. Using a network architecture of 15-10-10 with 75% training data, the ANN effectively captured the complex relationships between environmental variables and landslide occurrences. The model’s high R2 value signifies its robustness in predicting high-risk zones, complementing the spatial analysis performed by Maxent and MGWR. These findings validate the ANN model’s reliability and underscore its potential for enhancing landslide risk mitigation strategies in geospatial hazard assessments.

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

Geospatial Analysis and Prediction of Landslide Susceptibility in Uttara Kannada District, Karnataka (2025–2050): A Multi-Model Approach Using Maxent, MGWR, and ANN

  • 章节

  • 发表日期：

    2026年

查看全文
  • M. Sasi,
  • Sawant Sushant Anil

摘要

This study employs advanced geospatial techniques to analyze landslide susceptibility in Uttara Kannada district, Karnataka, from 2025 to 2050. We identified approximately 1852.18 sq km of high-risk zones using the Maxent model. The MGWR analysis revealed significant correlations with landslide occurrences, incorporating the highly influencing environmental parameters (water proximity, TWI, soil type, slope, rainfall, NDVI, LULC, lineament proximity, and geomorphology). The standardized residuals vs. predicted plot indicated a generally well-fitted model, with residuals randomly dispersed around zero. However, some areas showed patterns suggesting potential model refinements. The WP significance map highlighted regions with a strong statistical impact of these parameters, guiding targeted mitigation efforts. Emerging hotspot and cold spot analyses provided insights into spatial and temporal trends in landslide activity. Hotspots, primarily located in the Western Ghats and certain coastal areas, indicate regions of increasing landslide occurrences due to high rainfall, steep slopes, and other contributing factors. Conversely, cold spots in the northern parts of the district reflect areas with decreasing landslide activity, likely due to effective land management and mitigation measures. These findings underscore the importance of focused interventions in hotspot areas and the need to maintain successful practices in cold spot regions, thereby enhancing disaster risk management and promoting sustainable land-use practices in Uttara Kannada. The application of the artificial neural network (ANN) model has further proven the model’s significance for landslide hazard zonation mapping, as evidenced by an R2 value indicating a strong model fit. Using a network architecture of 15-10-10 with 75% training data, the ANN effectively captured the complex relationships between environmental variables and landslide occurrences. The model’s high R2 value signifies its robustness in predicting high-risk zones, complementing the spatial analysis performed by Maxent and MGWR. These findings validate the ANN model’s reliability and underscore its potential for enhancing landslide risk mitigation strategies in geospatial hazard assessments.