GR5J hydrological model code

Published: 12 October 2020| Version 1 | DOI: 10.17632/m5p5xmrr7k.1
, Francesco Pintori


In a river basin, the mass balance equation links total water storage (TWS), precipitation (P), actual evapotranspiration (E) and river discharge (Q). Though, a focused portion of the catchment, located between two river discharge stations, can be studied. Indeed, TWS in a downstream sub-catchment can be estimated based on the mass balance equation dTWS/dt = P + Qin - E - Qout - Qgw; where Qin, Qout, Qgw are respectively incoming river inflow, outcoming river discharge, and potential groundwater import/export in a surrounding basin. Among the different water fluxes, P, Qin and Qout can be measured, whereas actual evapotranspiration and Qgw should be estimated with a model. E and Qgw can be estimated with the lumped parameter rainfall-runoff hydrological model GR5J (Pushpalatha et al., 2011), which allow to quantify daily TWS at the scale of single hydrological basins. Here we provide a MATLAB code of the GR5J model. This model is forced with precipitation, temperature and potential evapotranspiration and computes actual river discharge. The model is based on two storage compartments - production and routing stores - which mimic the typical response of soils and groundwater to antecedent precipitation and evapotranspiration. Snow is considered and it is estimated following the method described in the HBV (Hydrologiska Byrns Vattenbalansavdelning) model (Lindstrom et al., 1997): mean catchment temperature defines both rainfall/snowfall partitioning and snow melt events. It is worth noting that GR5J does not need the specific knowledge of any intrinsic structure/property of the basin. The model is parsimonious and designed to model river discharge. Although the GR5J model is a simplified conceptual model, where only five mathematical parameters define the dynamics of the two stores and their relations, it has proven skillful in predicting river discharge better than more complex models (de Lavenne et al., 2016) and has been successfully applied to represent groundwater storage changes in Nepal rivers (Andermann et al., 2012). GR5J parameters are calibrated using a Marquard-Levenberg least squares regression analysis using root mean square error on the logarithm of observed river discharge to limit the impact of floods and promote the description of the whole water cycle.


Steps to reproduce

Run the file main_GR5J_snow.m. The path of the input files are at the beginning of the code. The only exception are the extraterrestrial radiation data, which are in the read_CRU_OUDIN.m file.