Investigating Parameter Transferability across Models and Events for a Semiarid Mediterranean Catchment

Enrica Perra,Claudio Paniconi,Roberto Deidda,Monica Piras,Giuseppe Mascaro

doi:10.3390/w11112261

Abstract

Physically based distributed hydrologic models (DHMs) simulate watershed processes by applying physical equations with a variety of simplifying assumptions and discretization approaches. These equations depend on parameters that, in most cases, can be measured and, theoretically, transferred across different types of DHMs. The aim of this study is to test the potential of parameter transferability in a real catchment for two contrasting periods among three DHMs of varying complexity. The case study chosen is a small Mediterranean catchment where the TIN-based Real-time Integrated Basin Simulator (tRIBS) model was previously calibrated and tested. The same datasets and parameters are used here to apply two other DHMs—the TOPographic Kinematic Approximation and Integration model (TOPKAPI) and CATchment HYdrology (CATHY) models. Model performance was measured against observed discharge at the basin outlet for a one-year period (1930) corresponding to average wetness conditions for the region, and for a much drier two-year period (1931–1932). The three DHMs performed comparably for the 1930 period but showed more significant differences (the CATHY model in particular for the dry period. In order to improve the performance of CATHY for this latter period, an hypothesis of soil crusting was introduced, assigning a lower saturated hydraulic conductivity to the top soil layer. It is concluded that, while the physical basis for the three models allowed transfer of parameters in a broad sense, transferability can break down when simulation conditions are greatly altered.

Highlights

Based distributed hydrologic models (DHMs) are useful tools for simulating watershed hydrologic response in applications that include water resources estimation and management, flood forecasting, and support of numerical weather prediction models [1,2,3,4]
CATchment HYdrology model (CATHY) and TOPographic Kinematic Approximation and Integration model (TOPKAPI) simulate quite well the recession curves, while tRIBS has a tendency to simulate steeper recession limbs [62], probably due to higher terrain resolution adopted in this model, which better captures steeper terrain slopes, leading to higher water velocities in the routing schemes of hillslopes and channels
This study examined the potential of direct parameter transferability by assessing the comparative runoff response of three physically based distributed hydrologic models for a small Mediterranean catchment during two contrasting periods

Summary

Introduction

Based distributed hydrologic models (DHMs) are useful tools for simulating watershed hydrologic response in applications that include water resources estimation and management, flood forecasting, and support of numerical weather prediction models [1,2,3,4]. To account for the spatial variability of physiographic watershed characteristics, DHMs discretize the spatial domain into smaller elements and, in each of them, solve physics equations of the different processes. These equations depend on parameters that could potentially be measured in the field and whose values fall within often known admissible ranges. As a result, when different DHMs are applied to the same basin or the same DHM is applied to different events or even diverse sites, similar parameter values should, in principle, be adopted for the simulation of the same hydrologic processes. There is a gap between theory and hydrologic models stemming from the pragmatic manner in which models are often developed, limiting the ability to generalize about hydrologic behaviors and resulting in poor parameter transferability [9]

Objectives

Methods

Results

Conclusion