Kumudu Madhawa Kurugama

Kumudu Madhawa Kurugama

Auto-generated profile, based on publicly available information

Kumudu Madhawa Kurugama

Department of Civil and Environmental Engineering Tohoku University Sendai Japan

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 March 2024 in Journal of Flood Risk Management
Verified
Journal Article
Open access
KKKumudu Madhawa Kurugama
Kumudu Madhawa Kurugama
SKSo Kazama
So Kazama
YHYusuke Hiraga
Yusuke Hiraga
CSChaminda Samarasuriya
Chaminda Samarasuriya
Identifying flood‐prone areas is essential for preventing floods, reducing risks, and making informed decisions. A spatial database with 595 flood inventory and 13 flood predictors were used to implement five boosting algorithms: gradient boosting machine (GBM), extreme gradient boosting, categorical boosting, logit boost, and light gradient boosting machine (LGBM) to map flood susceptibility in Rathnapura while evaluating trained model's generalizing ability and assessing the feature importance in flood susceptibility mapping (FSM). The model performance was evaluated using the F1‐score, kappa index, and area under curve (AUC) method. The findings revealed that all the models were effective in identifying the overall flood susceptibility trends while LightGBM model had superior results (F1‐score = 0.907, Kappa value = 0.813 and AUC = 0.970), securing the top scores across all performance metrics compared to the other models (for testing dataset). Based on kappa evaluation, most of the models had finer performance (AUC min = 0.737) while LightGBM had moderate performance for predictions beyond the training region. According to the results, regions with lower altitudes and topographic roughness values, moderate rainfall, and proximity to rivers are more susceptible to flooding. This framework can be adapted for rapid FSM in data‐deficient regions.

1 January 2023 in Journal of JSCE
Verified
Journal Article
Open access
PDF file
KKKumudu Madhawa Kurugama
Kumudu Madhawa Kurugama
SKSo Kazama
So Kazama
YHYusuke Hiraga
Yusuke Hiraga
CPChaminda Samarasuriya Patabendige
Chaminda Samarasuriya Patabendige
Frequent severe flood occurrences in Rathnapura city, Sri Lanka cause damages to both human lives and infrastructures. The existing flood mapping models have certain shortcomings that could be enhanced by utilizing more advanced and combined approaches. In this research, we suggested a new technique to enhance the prediction accuracy of flood susceptibility mapping by integrating the bivariate index of entropy (IoE) and support vector machine (SVM) as an ensemble method. The suggested approach was developed using four SVM kernels; polynomial, linear, sigmoid and radial basis function to examine the robustness of predictability of SVM technique. First a flood inventory map with 445 flood locations was created using satellite images, field survey and documentary sources. A spatial database was created with eight flood conditioning factors(FCFs) including altitude, slope, aspect, topographic roughness index(TRI), soil, Land use, distance from river and rainfall. Initially, IoE was utilized to assess the correlation between the various flood conditioning factors and flood occurrence. Afterwards, the results of the first step were utilized to perform SVM model. Model validation was carried out using seed cell area index(SCAI) and area under curve(AUC) methods. The highest success and prediction rates of 94.82% and 95.81% and lowest SCAI values of 0.18, 0.878 for very high and high susceptibility classes were achieved by ensemble IoE-SVM(RBF) model. Out of all the methods used, lowest accuracies were obtained by the individual IoE and SVM(RBF) method. Ensemble methods, enhance flood prediction capabilities and conditioning factor evaluation resulting in a more accurate final map.

28 August 2023
Verified
Conference Paper
Closed access
KKK.A.K.M. Kurugama
K.A.K.M. Kurugama
SKS. Kazama
S. Kazama
SCS.P. Chaminda
S.P. Chaminda
Flooding is one of the most frequently encountered natural disasters globally. Frequent severe flood occurrences in Rathnapura city, Sri Lanka caused damages to both human lives and infrastructures. Data-driven models have been showing their ability of flood susceptibility mapping (FSM) in data-scare regions as an alternative to traditional hydrological models, but they are not widely used by stakeholders due to their black-box nature. This research suggests utilising the shapley additive explanation (SHAP) method to interpret the results generated by the CatBoost machine learning model and to assess the influence of different variables on flood susceptibility mapping. A flood inventory (445 flooded locations) and thirteen flood conditioning factors were used to implement the model and results were validated using the area under curve (AUC) method, which showed a success rate and prediction rate of 93.1% and 92.5%, respectively. SHAP plots indicated that the regions with lower elevations and topographic roughness values, gentler slopes, closer proximity to rivers, and moderate rainfall are more susceptible to flooding. According to the results obtained, we suggest incorporating SHAP-based datadriven models in forthcoming studies on FSM to enhance the interpretations of model outcomes.