Streamflow forecasting in a climate change perspective using E-FUSE

Abstract

The present work aims to identify the best hydrological model structure suitable for the Lower Godavari River Basin, India, that forecasts streamflows. An extended version of the Framework for Understanding Structural Errors (FUSE), termed E-FUSE, is developed for this purpose. It consists of 1248 model structures. K means cluster analysis (KCA), and Davies Bouldin Cluster Validation Index (DBCVI) are used for identifying optimal clusters, whereas Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) is employed for the best model structure. Correlation coefficient (r), normalized root mean square error (NRMSE), and mean bias error (MBE) are employed as evaluation criteria. The best model structure obtained exhibits r, NRMSE and MBE of 0.734, 0.74 and -0.09 respectively during calibration and 0.69, 0.802 and -0.28 respectively during validation. The best model structure is then used to forecast discharges for a global climate model, EC-Earth3, and four Shared Socioeconomic Pathways, SSP126, SSP245, SSP370, and SSP585 scenarios. Analysis was made for three time horizons, namely, the near-future scenario (2021–2046), mid-future scenario (2047–2072), and far future scenario (2073–2099). It is observed that the July–September months contribute greatly to total runoff for four SSPs and three time horizons. HIGHLIGHTS FUSE is extended by combining with K-means cluster analysis and multi-criteria decision-making technique, TOPSIS, to identify the best hydrologic model structure and applied to Lower Godavari River Basin, India.; Runoff is forecasted using EC-Earth3 and four SSPs, namely SSP126, SSP245, SSP370, and SSP585 for near (2021–2046), mid (2047–2072), and far future (2073–2099) time horizons.;

Keywords

• climate change
• e-fuse
• kca
• runoff
• ssp
• topsis