Coherent intra-annual ensemble forecasts of surface and groundwater availability

In many parts of the world, surface water and groundwater are used complementarily to supply agricultural production and to meet urban water demands. Conjunctive management of these water resources requires balancing of the different characteristics of surface water and groundwater with respect to availability, quality and cost of supply. Ensemble forecasts of surface water and groundwater availability can inform management decisions but require explicit representation of the complex processes controlling surface and groundwater interactions. While many methods and operational services exist that provide independent forecasts for surface and groundwater availability, to our knowledge no approaches for coupled forecasting have been developed yet.

In this presentation we introduce an approach that generates coupled forecasts of surface water and groundwater availability. It extends the Forecast Guided Stochastic Scenarios (FoGSS) (Bennett et al., 2016) approach to forecast groundwater level at specified locations, in addition to streamflow totals, to lead times of 12 months at monthly time steps. We adapt a conceptual hydrological model to improve predictions of streamflow and, as a by-product, groundwater level. We then apply independent error models to streamflow and groundwater level to reduce bias, update predictions using recent observations and quantify residual uncertainty. Ensemble streamflow and groundwater forecasts are generated by forcing the hydrological and error models with ensemble rainfall forecasts generated by post-processing ECMWF System 5 outputs. The skill, bias and reliability of the rainfall, streamflow and groundwater level forecasts were assessed for a case-study catchment in South-East Queensland, Australia. We find that skill of forecasts is dependent on the forecast issue month and lead time, with groundwater level forecasts displaying significant skill to lead times of 12 months, while streamflow forecast skill rarely persists beyond 3 months.  We conclude by describing opportunities to improve forecast skill and some of the challenges that may be faced in the operational delivery of water resource forecasts in real-time.

Reference

Bennett, J. C., Wang, Q. J., Li, M., Robertson, D. E., and Schepen, A.: Reliable long-range ensemble streamflow forecasts: Combining calibrated climate forecasts with a conceptual runoff model and a staged error model, Water Resources Research, 52, 8238-8259, 10.1002/2016WR019193, 2016.