Introduction
Rosdi et al. (2017) made significant strides in understanding sinkhole susceptibility in Kuala Lumpur and Ampang Jaya by combining Geographic Information Systems (GIS) with the Analytical Hierarchical Process (AHP). Their work laid a solid foundation for assessing sinkhole risk, but there remains an opportunity to refine and enhance these models using more advanced spatial analysis techniques. One promising approach is Geographically Weighted Regression (GWR), which has the potential to improve both the accuracy and granularity of sinkhole susceptibility assessments. This article examines how incorporating GWR, along with other advanced GIS methodologies, could lead to more precise and insightful analyses of sinkhole hazards.
1. Application of Geographically Weighted Regression (GWR)
Geographically Weighted Regression (GWR) represents an evolution from traditional regression models by allowing for spatial variability in the relationships between variables. Unlike global models that assume a uniform relationship across the study area, GWR acknowledges that these relationships can vary from one location to another. This spatial flexibility is crucial for understanding sinkhole formation, as it reveals how different factors influence sinkhole risk in distinct geographical contexts (Fotheringham et al., 2002).
Applying GWR to the analysis of sinkhole susceptibility in Kuala Lumpur and Ampang Jaya could illuminate how key factors such as lithology, groundwater level decline, soil type, land use, and proximity to groundwater wells affect sinkhole risk differently across various regions. For instance, the impact of lithology might be more pronounced in areas with specific geological features, while groundwater decline could play a more significant role in other areas. By capturing these spatial differences, GWR would provide a more nuanced and accurate understanding of sinkhole susceptibility (Brunsdon et al., 1996).
GWR offers several advantages for sinkhole susceptibility analysis. It allows for localized insights by identifying areas where certain factors disproportionately affect sinkhole formation, thereby enabling more targeted and effective mitigation strategies. Additionally, by accounting for spatial heterogeneity, GWR can enhance the accuracy of susceptibility models, leading to improved predictions and risk assessments. The results from GWR can also be visualized as spatially varying coefficients, providing a clear and interpretable representation of how each factor’s influence varies across the study area (Fotheringham et al., 2002).
2. Integration of High-Resolution Remote Sensing Data
The current study’s reliance on existing land use data can be significantly improved by incorporating high-resolution remote sensing imagery from satellites or unmanned aerial vehicles (UAVs). This approach would allow for the development of more accurate and up-to-date land use and land cover maps, which are essential for assessing areas at risk of sinkhole formation (Li et al., 2019).
High-resolution satellite imagery also enables time-series analysis, which can track changes in land use and land cover over time. Such analysis is crucial for identifying trends and patterns that contribute to sinkhole development, including urban expansion, deforestation, and alterations in groundwater extraction practices (Wu et al., 2015).
3. Incorporation of Additional Spatial Variables
In addition to the factors considered in the current study—lithology, groundwater decline, soil type, land use, and proximity to groundwater wells—incorporating topographical factors such as slope, elevation, and aspect could provide additional insights. These topographical variables often influence water drainage and soil stability, both of which are important in sinkhole formation (Gao et al., 2014).
Furthermore, integrating detailed hydrological modeling into the GIS analysis could enhance our understanding of how water movement through the landscape affects sinkhole susceptibility. Simulating scenarios of heavy rainfall or prolonged drought could provide valuable information on their impact on groundwater levels and sinkhole risk (Beven & Kirkby, 1979).
4. Improved Data Integration and Validation Techniques
A more comprehensive GIS framework that integrates diverse datasets—such as geological surveys, hydrological models, and remote sensing data—would facilitate a thorough analysis of sinkhole risk. Utilizing machine learning techniques could further help in identifying complex patterns and interactions among various factors that contribute to sinkhole formation (Hengl et al., 2015).
Expanding the sinkhole inventory and performing rigorous cross-validation of the model would enhance its reliability. Incorporating data from other regions with similar geological and environmental conditions could also test the model’s generalizability and robustness (Chen et al., 2020).
5. Exploring Alternative Multicriteria Decision-Making (MCDM) Techniques
The Fuzzy AHP method could bolster the robustness of the susceptibility model by addressing the uncertainty and vagueness inherent in geological and hydrological data. This technique provides a way to incorporate and manage these uncertainties in decision-making processes (Saaty, 2008).
The Weight of Evidence (WoE) method is another promising approach, particularly for binary classification problems such as identifying areas prone to sinkholes. WoE calculates the probability of sinkhole occurrence based on the presence or absence of certain factors, offering a probabilistic perspective on risk assessment (Bonham-Carter, 1994).
Conclusion
The study by Rosdi et al. (2017) significantly advanced our understanding of sinkhole susceptibility in Kuala Lumpur and Ampang Jaya. However, the integration of advanced GIS methods such as Geographically Weighted Regression (GWR), high-resolution remote sensing data, and additional spatial variables holds the potential to further enhance the accuracy and utility of sinkhole susceptibility models. By exploring these and other advanced techniques, future research could provide more reliable tools for predicting and mitigating sinkhole hazards, contributing to safer and more resilient urban environments.
References
Bonham-Carter, G. F. (1994). Geographic Information Systems for Geoscientists: Modelling with GIS. Pergamon Press.
Beven, K. J., & Kirkby, M. J. (1979). A physically-based variable contributing area model of basin hydrology. Hydrological Sciences Bulletin, 24(1), 43-69.
Brunsdon, C., Fotheringham, A. S., & Charlton, M. (1996). Geographically weighted regression: A method for exploring spatial nonstationarity. Geographical Analysis, 28(4), 281-298.
Chen, C., Wu, J., & Zhang, Y. (2020). Enhancing sinkhole susceptibility mapping with deep learning: A case study in southern China. Environmental Monitoring and Assessment, 192(9), 1-15.
Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2002). Geographically Weighted Regression: The Analysis of Spatially Varying Relationships. Wiley.
Gao, J., Wang, H., & Zhao, J. (2014). A new approach to sinkhole susceptibility mapping using GIS and remote sensing techniques. Environmental Earth Sciences, 71(6), 2721-2734.
Hengl, T., de Jesus, J. M., Heuvelink, G. B. M., & Kempen, B. (2015). SoilGrids250m: Global soil information based on machine learning. PLoS ONE, 10(9), e0134086.
Li, J., Li, X., & Lu, S. (2019). An improved method for land use/cover classification using high-resolution remote sensing imagery. Remote Sensing, 11(11), 1302.
Rosdi, M. A. H. M., Othman, A. N., Zubir, M. A. M., Latif, Z. A., & Yusoff, Z. M. (2017). Sinkhole susceptibility hazard zones using GIS and analytical hierarchical process (AHP): A case study of Kuala Lumpur and Ampang Jaya. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLII-4/W5, 145–151. https://doi.org/10.5194/isprs-archives-XLII-4-W5-145-2017
Saaty, T. L. (2008). Decision Making with the Analytic Hierarchy Process. Springer.
