Coupled Data Analytics–SWAT Approach for Flow Generation and Analysis in Ungauged Tropical Watershed.

At a regional scale, decision-making in water resources management is based on the analysis of long-term, continuous observations of hydrometeorological variables including streamflow. Quite often, constraints such as paucity of observed flow data hinder scientific planning in ungauged watersheds of developing countries. In this context, we present an improved framework for predicting streamflow data in an ungauged tropical watershed (Jambhira) in eastern India utilizing coupled data analytics and the Soil and Water Assessment Tool (SWAT) for flow regionalization. In this study, the choice of different dimensionality reduction methods such as principal components analysis (PCA) and Isomap during the clustering stage, as well as the choice of inverse distance weighting (IDW) and physical similarity (PS) approaches for regionalization of SWAT model parameters for the ungauged watershed, are investigated. As a novel aspect, in this case study, we included certain seasonal climate characteristics and changed land-use composition of the study watersheds as pertinent features for clustering of homogeneous watersheds. The proposed framework is then evaluated for its possible limitations, namely, whether the integrated approaches yield the best estimates of hydrological model parameters for ungauged watersheds, and if they guarantee good performance across different flow regimes. The regionalization results for the pseudoungauged watershed (Govindpur, India) indicate that coupled PS-SWAT approach is the most efficient with R2 and Nash–Sutcliffe efficiency (NSE) values ranging from 0.75 to 0.83 and from 0.74 to 0.77, respectively. The hydrological model of the ungauged watershed (Jambhira) is developed using the regionalization framework, and using the simulated flows, flow duration curve (FDC) analysis is also performed. Analysis of regime-based flow characteristics reveals a declining trend in the magnitude of low and high flows by 4% and 22.2%, respectively, in Jambhira. [ABSTRACT FROM AUTHOR]