Hydrological hysteresis and its value for assessing process consistency in catchment conceptual models

O Fovet,M Hrachowitz,M Faucheux,C Gascuel-Odoux,L Ruiz

doi:10.5194/hess-19-105-2015

Abstract

Abstract. While most hydrological models reproduce the general flow dynamics, they frequently fail to adequately mimic system-internal processes. In particular, the relationship between storage and discharge, which often follows annual hysteretic patterns in shallow hard-rock aquifers, is rarely considered in modelling studies. One main reason is that catchment storage is difficult to measure, and another one is that objective functions are usually based on individual variables time series (e.g. the discharge). This reduces the ability of classical procedures to assess the relevance of the conceptual hypotheses associated with models. We analysed the annual hysteric patterns observed between stream flow and water storage both in the saturated and unsaturated zones of the hillslope and the riparian zone of a headwater catchment in French Brittany (Environmental Research Observatory ERO AgrHys (ORE AgrHys)). The saturated-zone storage was estimated using distributed shallow groundwater levels and the unsaturated-zone storage using several moisture profiles. All hysteretic loops were characterized by a hysteresis index. Four conceptual models, previously calibrated and evaluated for the same catchment, were assessed with respect to their ability to reproduce the hysteretic patterns. The observed relationship between stream flow and saturated, and unsaturated storages led us to identify four hydrological periods and emphasized a clearly distinct behaviour between riparian and hillslope groundwaters. Although all the tested models were able to produce an annual hysteresis loop between discharge and both saturated and unsaturated storage, the integration of a riparian component led to overall improved hysteretic signatures, even if some misrepresentation remained. Such a system-like approach is likely to improve model selection.

Highlights

Rainfall-runoff models are tools that mimic the low-pass filter properties of catchments
The 2-dimensional observed relationship between saturated storage in the hillslope (HSS) or in the riparian zone (RSS) and stream discharge (Q) for each year was hysteretic, highlighting the nonuniqueness of the response of discharge to storage depending on the initial conditions and a lag time between both variable dynamics, in particular during the recharge period, as illustrated in Fig. 4 for two contrasting water years
The direction of the hysteretic loop was different depending on the topographic position of the piezometer: loops were always anticlockwise for the piezometer located at the top of the hillslope HSSF5b(Q), mostly anticlockwise for the midslope piezometer hillslope saturated storage (HSS)-F4(Q) and mostly clockwise for the piezometer in the riparian zone riparian saturated storage (RSS)-F1b(Q) (Fig. 5)

Summary

Introduction

Rainfall-runoff models are tools that mimic the low-pass filter properties of catchments. They aim at reproducing observed stream flow time series by routing time series of meteorological drivers through a sequence of mathematically formalized processes that allow a temporal dispersion of the input signals in a way that is consistent with the modeller’s conception of how the system functions. Modelling efforts on the catchment scale typically face the problem that, on that scale, neither integrated internal fluxes nor the integrated storage and the partitioning between different storage components at a given time can be observed within limited uncertainty. While spatial aggregation of storage estimates (e.g. catchment averages) in lumped models may lead to a loss of crucial information

Objectives

Methods

Results

Conclusion