Submitted as: model description paper | 20 Oct 2016
Review status: this preprint was under review for the journal GMD but the revision was not accepted.
eWaterCycle: a hyper-resolution global hydrological model for river
discharge forecasts made from open source pre-existing components
Rolf Hut1,Niels Drost2,Maarten van Meersbergen2,Edwin Sutanudjaja3,Marc Bierkens3,4,and Nick van de Giesen1Rolf Hut et al.Rolf Hut1,Niels Drost2,Maarten van Meersbergen2,Edwin Sutanudjaja3,Marc Bierkens3,4,and Nick van de Giesen1
Received: 25 Aug 2016 – Accepted for review: 18 Oct 2016 – Discussion started: 20 Oct 2016
Abstract. eWaterCycle is an open source hyperresolution (10 km × 10 km) global hydrological forecasting framework that runs an ensemble of hydrological models. Forced with a weather forecast ensemble, it predicts river discharge and river discharge uncertainty nine days ahead. Daily satellite soil moisture observations are assimilated into the state of the model ensemble using an Ensemble Kalman Filter. We demonstrate that it is feasible to build such a system using pre-exisiting, open source, components that communicate through standard interfaces. The PCRGLOBWB2.0 (van Beek et al., 2011; Sutanudjaja et al., 2014) model is used to model hydrology globally, forced with GFS (Kanamitsu, 1989; Kanamitsu et al., 1991; Moorthi et al., 2001) weather forecast. The operational soil moisture product from the HSAF (Drusch et al., 2009; De Rosnay et al., 2011) service is assimilated into the model ensemble using OpenDA (Velzen et al., 2016), a data assimilation framework. Output of the model ensemble is presented in a Cesium (Analytical Graphics, 2011) based visualization. All communication between framework components is through standard file types (NetCDF)(Rew and Davis, 1990) and services (Web Map Service) (de La Beaujardiere, 2006). Communication between model and data assimilation framework is through the Basic Model Interface (BMI) (Peckham et al., 2013). The forecasts is available at forecast.ewatercycle.org. By using standard open interfaces, the different components of the model can be replaced with relative ease, facilitating future model comparison studies without the need of extensive Computer Science support. This makes eWaterCycle, in addition to an operational forecasting model, a testbed environment where the impact of different model structures, input sources and/or data assimilation schemes can easily be studied. Setup instructions to run the eWaterCycle project on local hardware are provided, allowing the hydrological community to build on this open source framework.
How to cite. Hut, R., Drost, N., van Meersbergen, M., Sutanudjaja, E., Bierkens, M., and van de Giesen, N.: eWaterCycle: a hyper-resolution global hydrological model for river
discharge forecasts made from open source pre-existing components, Geosci. Model Dev. Discuss. [preprint], https://doi.org/10.5194/gmd-2016-225, 2016.
A system that predicts the amount of water flowing in each river on earth, 9 days ahead, is build using existing parts of open source computer code build by different researchers in other projects.
The glue between all pre-existing parts are all open interfaces which means that the pieces system click together like a house of LEGOs. It is easy to remove a piece (a brick) and replace it with another, improved, piece.
The resulting predictions are available online at forecast.ewatercycle.org
A system that predicts the amount of water flowing in each river on earth, 9 days ahead, is...