Hydrological assessment of meteorological network density through data assimilation simulation

A stochastic hydro-geostatistical assessment approach is proposed to evaluate the predictability of streamflow simulation as a function of precipitation observation network density. This approach is based on the paradigm of a controlled data assimilation simulation algorithm, known as Observing System Simulation Experiment (OSSE). The OSSE-based algorithm is designed to simulate a data assimilation technique known as Statistical Interpolation (SI). In SI, observations are combined with short-term precipitation forecasts through variographic analysis. The assessment approach was designed in two stages. At the first stage of the assessment, SI precipitation analysis is stochastically simulated to construct a hypothetical true reference precipitation field, while at the second stage, the variographic by-products of the SI Gaussian process are used to perturb the reference field in order to produce the pseudo counterparts of the SI input components. The pseudo terms are then combined in SI to compute the pseudo precipitation analysis while various network scenarios are considered. Distributed hydrologic modeling is then performed to convert reference and pseudo analysis datasets into streamflow. The proposed assessment approach can be used to assess observation networks of various densities; however, it is particularly effective in sparsely gauged basins. To demonstrate the application of the approach to a real-world case, the sparse observation network in the Churchill River basin (CRB), which covers an area of roughly 296,000 km2 in the Canadian boreal ecoclimatic region, was assessed. Regional Deterministic Precipitation Analysis from 2002 to 2012 was extracted from the Canadian Precipitation Analysis (CaPA) data assimilation system and used while generating the reference field. Gridded precipitation data from both stages of the assessment were used as input to WATFLOOD™ distributed hydrologic model. Quantitative assessment of streamflow suggested that an observation network with a density in the order of 3 stations/10,000 km2 in the CRB can satisfy hydrological criteria while maintaining the network as an integral part of the CaPA’s SI data assimilation technique.