CREST v2.1 Refined by a Distributed Linear Reservoir Routing Scheme

Wednesday, 17 December 2014
Xinyi Shen1, Yang Hong2, Ke Zhang1, Zengchao Hao1 and Dacheng Wang3, (1)University of Oklahoma Norman Campus, Norman, OK, United States, (2)University of Oklahoma, Norman, OK, United States, (3)Institute of Remote Sensing and Digital Earth Chinese Academy of Science, Beijing, China
Hydrologic modeling is important in water resources management, and flooding disaster warning and management. Routing scheme is among the most important components of a hydrologic model. In this study, we replace the lumped LRR (linear reservoir routing) scheme used in previous versions of the distributed hydrological model, CREST (coupled routing and excess storage) by a newly proposed distributed LRR method, which is theoretically more suitable for distributed hydrological models. Consequently, we have effectively solved the problems of: 1) low values of channel flow in daily simulation, 2) discontinuous flow value along the river network during flood events and 3) irrational model parameters. The CREST model equipped with both the routing schemes have been tested in the Gan River basin. The distributed LRR scheme has been confirmed to outperform the lumped counterpart by two comparisons, hydrograph validation and visual speculation of the continuity of stream flow along the river: 1) The CREST v2.1 (version 2.1) with the implementation of the distributed LRR achieved excellent result of [NSCE(Nash coefficient), CC (correlation coefficient), bias] =[0.897, 0.947 -1.57%] while the original CREST v2.0 produced only negative NSCE, close to zero CC and large bias. 2) CREST v2.1 produced more naturally smooth river flow pattern along the river network while v2.0 simulated bumping and discontinuous discharge along the mainstream. Moreover, we further observe that by using the distributed LRR method, 1) all model parameters fell within their reasonable region after an automatic optimization; 2) CREST forced by satellite-based precipitation and PET products produces a reasonably well result, i.e., (NSCE, CC, bias) = (0.756, 0.871, -0.669%) in the case study, although there is still room to improve regarding their low spatial resolution and underestimation of the heavy rainfall events in the satellite products.