Effects of Model Spatial Structure and Basin Characteristics on the Performance of Three Hydrologic Models

Hydrologic model selection plays a pivotal role in ensuring the reliability and accuracy of hydrologic modeling studies. This decision is mostly based on availability or researcher’s prior experience rather on a systematic evaluation of model structure and basin characteristics. This study aims to address this problem by comparing the effect of basin characteristics and hydrologic model structures on the streamflow prediction performance of lumped and distributed hydrologic models. For that, two lumped hydrologic models, Genie Rural à 4 Paramètres Journalier (GR4J) and Technische Universität Wien (TUW), and a distributed hydrologic model, mesoscale Hydrologic Model (mHM), are calibrated and validated in Dandalas subbasin, where Karacasu dam construction has taken place from 1998 to 2012, and in the karstic Çakıtsuyu subbasin. The model calibration was performed using the daily Nash-Sutcliffe Efficiency (NSE) as objective function whereas the validation performance was assessed with both daily and monthly NSE and Kling-Gupta Efficiency (KGE) metrics. Our results revealed the mHM model outperformed the other two models in both the calibration and validation periods in the karstic Çakıtsuyu subbasin, achieving the highest daily NSE during calibration (0.72) and validation (0.61). Notably, GR4J outperformed in Dandalas validation (KGE: 0.53 daily, 0.66 monthly). The superior performance of mHM can be attributed to its multi-parameter regionalization approach, detailed soil structure up to six horizons, elaborated flow routing scheme and representation of geological features using multiple karstic domains.