EGU22-6155
https://doi.org/10.5194/egusphere-egu22-6155
EGU General Assembly 2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

How long do initial conditions prevail over boundary conditions in streamflow forecasting in South America?

Ingrid Petry1, Fernando Fan2, and Louise Crochemore3
Ingrid Petry et al.
  • 1Water Research Institute, Federal University of Rio Grande do Sul, Porto Alegre, Brazil (ingrid.petry@ufrgs.br)
  • 2Water Research Institute, Federal University of Rio Grande do Sul, Porto Alegre, Brazil (fernando.fan@ufrgs.br)
  • 3Institut Geosciences Environnement, Université Grenoble Alpes, Grenoble, France (louise.crochemore@univ-grenoble-alpes.fr)

In the model-based streamflow forecasting context, initial conditions (ICs) and meteorological forcings (boundary conditions) are two important drivers of predictability. While the meteorological forcings increasingly influence forecasts at long time horizons, ICs’ influence tends to decrease, being nonetheless the predominant source of predictability at short time horizons. Quantifying the period of time when the ICs contribute to streamflow predictability can help researchers and water managers to choose the best forecasting method for each study area and horizon, balancing computational effort and streamflow forecast accuracy. In this work we quantify IC's prevalence time over boundary conditions in natural river flow forecasts in large basins (>1000km²) in South America (SA) from the MGB-SA model, a continental and hydrodynamic version of the MGB conceptual semi-distributed hydrological model. The methodology consisted of forcing MGB-SA with null precipitation as meteorological forcing, so that all the predictability obtained from the experiments was due to ICs. Streamflow experiments had a 215-day horizon, monthly initialization, daily timestep and comprehended the period of 1990 to 2010. The results were compared to the MGB-SA streamflow simulations with MSWEP as observed rainfall. Errors from the hydrological model were thus not considered in this analysis. The prevalence (T50) was estimated by the horizon (from 1 to 215 days) when streamflow predictability was degraded by 50%, i.e. when meteorological forcings start prevailing over ICs. Predictability was estimated by the performance indicator KGE, and the T50 for each of the river reaches of MGB-SA was presented in a map. The T50 map shows that the shortest IC’s prevalence on streamflow predictability is observed on riverheads, ranging from 1 to 3 days. IC's influence increases near the main reaches of the great rivers of SA, reaching up to 10 days on the Iguaçu River and up to 20 days on the Oricono River and Araguaia River. T50 is up to 40 and 90 days on reaches of the Amazon River, Atlantic coast of North Argentina and Pantanal plains. In general, IC’s influence is higher in the main river reaches of basins with flat relief, due to their greater drainage area and the slow response time of the basin.

How to cite: Petry, I., Fan, F., and Crochemore, L.: How long do initial conditions prevail over boundary conditions in streamflow forecasting in South America?, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6155, https://doi.org/10.5194/egusphere-egu22-6155, 2022.