.png){kind=logo}
Current issue
Volume 1 Issue 1
Comparative Mapping of Flood-Susceptible Zones Using AHP and Machine Learning Models in a Data Scarce River Basin of Northeast India
Asesh Rudra Paul and Tilottama Chakraborty
Flood-susceptible zone mapping is essential for effective flood risk management, enabling the identification of vulnerable areas and guiding targeted mitigation strategies. However, delineating flood-susceptible zones in data-scarce and topographically complex regions poses significant challenges. This study addresses these limitations by integrating Geographic Information Systems (GIS) with the Analytic Hierarchy Process (AHP) to assess flood susceptibility in the Haora River Basin, located in West Tripura, India. A total of nine flood-influencing parameters, including rainfall, elevation, slope, land use, and hydrological indices, were considered to develop a Flood Susceptibility Index (FSI). The resulting flood susceptibility map categorizes the basin into five classes: very low, low, moderate, high, and very high. The "Very High" zone covers 28.36 km², primarily concentrated in the low-lying urban areas around Agartala. The AHP model's predictive accuracy was validated using Receiver Operating Characteristic (ROC) curve analysis, which yielded an AUC of 0.848, indicating acceptable reliability. To enhance the robustness of the flood assessment, two machine learning models—Random Forest (RF) and Support Vector Machine (SVM)—were also employed. These models achieved AUC values of 0.9483 (RF) and 0.9260 (SVM), and demonstrated superior performance through lower MAE, MSE, and RMSE values compared to AHP. The integration of AHP-GIS with machine learning approaches offers a reliable and scalable framework for flood susceptibility mapping, especially in resource-limited environments. The methodology is generalizable to other vulnerable catchments across Northeast India and provides actionable insights for disaster planners and urban managers to prioritize high-risk zones and improve flood resilience.
Prediction of monthly summer monsoon rainfall of four meteorological subdivisions in India
A. K. Singhania,A. J. Borthakur, G. Lal,N. Khokher and Ganesh D. Kale
Chhattisgarh (CG), Goa and Konkan (G&K), Gangetic West Bengal (GWB) and West Uttar Pradesh (WUP) Meteorological Subdivisions have exhibited significant trends in south-west monsoon rainfall over the period of 100 years. Therefore, monthly summer monsoon rainfall’s (MSMR’s) prediction is necessary for aforesaid meteorological subdivisions. In the domain of water resources management, monthly rainfall values of monsoon are more effective as compared to total of monsoon rainfall. Thus, current study is performed to assess the hydroclimatic teleconnection (HCT) between MSMR of aforesaid meteorological subdivisions and lagged circulation indices, which will be useful to predict the MSMR of aforesaid meteorological subdivisions few months ahead. Four models are prepared for assessment of aforesaid HCTs having periods of model development as 1950-1999, 1950-1994, 1950-1989, and 1950-1984 with a common testing period of 2000-2014 for each meteorological subdivision. For each model of each subdivision following methodology is adopted. Significant lagged circulation indices (SLCIs) impacting MSMR of given subdivision are identified by using significant linear correlation. Then, multi-collinearity existing among these SLCIs is eliminated to derive significant and independent lagged circulation indices (SILCIs). SILCIs are used in formulation of monthly composite indices (MCIs) between MSMR and corresponding SILCIs by employing multivariate linear regression. These MCIs are then used for predicting MSMR of given subdivision over a common testing period. SILCIs derived in the current study have shown effect of other indices on MSMR of aforesaid meteorological subdivisions besides ENSO and EQUINOO. Correlation coefficient values for testing period are found to be higher for G&K and GWB Meteorological Subdivisions as compared to other two subdivisions.