Impartially resolving reservoirs of all sizes for seamless hydrological forecasting using multiscale Lake Module (mLM)

Abstract:

In regions with regulated hydrology, real-time water resources planning and management is only possible when reservoirs are well represented as an integrated component of hydrologic forecasting system. Modeled reservoirs, however, appear or disappear depending upon the size of the lake with respect to the model resolution of the hydrological simulation. To address this shortcoming of state-of-the-art hydrological models, we propose and demonstrate how the sub-grid catchment contribution (SCC) preserves lake topology and the drainage area contributing to the lake inflow, at all scales. This in turn sustains forecast reliability across scales by improving precision of reservoir inflow. The implication is showcased by augmenting the mesoscale hydrological model (mHM, git.ufz.de/mhm) with multiscale lake module (mLM).

The experimental setup constitutes the SaWaM project regions encompassing six semi-arid basins across three continents (São Francisco, Jaguaribe, Piranhas in Brazil, Blue Nile, Atbara in Sudan, Karun in Iran). Reservoirs varying in catchment size by four orders of magnitude (<500 km² to > 500,000 km²) are incorporated. The model is setup at six spatial resolutions ranging from 5 km to 50 km. SCC utilises dual spatial discretization i.e. it extracts the lake details from a constant finer "sub-grid" resolution morphology (~ 220 m) while model runs at coarser modelling resolutions. A hindcasting experiment is carried out using state-of-the-art meteorological reanalysis and forecast data. The forecasting performance of three phases of model development, a) mHM without mLM, b) mHM with mLM but without SCC, and c) mHM with mLM and with SCC, are evaluated using the Brier Skill Score.

Preliminary results show that SCC improves the forecast precision across modeling scales in regulated domains. Sub-grid level lake delineation and in-/outflow calculations of mLM lead to scalable reservoir state variables, fluxes and, overall, a quasi-scale independent basin hydrology. Seamless forecasts for soil moisture, streamflow, reservoir inflow and reservoir water level were achieved across scales. The new SCC based spatial discretization approach will help modellers in water resource management and improve forecasting systems in regulated regions, ranging from local to global scale.