Identification of Hydrological Models for Enhanced Ensemble Reservoir Inflow Forecasting in a Large Complex Prairie Watershed

Abstract

Accurate and reliable flow forecasting in complex Canadian prairie watersheds has been one of the major challenges faced by hydrologists. In an attempt to improve the accuracy and reliability of a reservoir inflow forecast, this study investigates structurally different hydrological models along with ensemble precipitation forecasts to identify the most skillful and reliable model. The key goal is to assess whether short-and medium-range ensemble flood forecasting in large complex basins can be accurately achieved by simple conceptual lumped models (e.g., SACSMA with SNOW17 and MACHBV with SNOW17) or it requires a medium level distributed model (e.g., WATFLOOD) or an advanced macroscale land-surface based model (VIC coupled with routing module (RVIC)). Eleven (11)-member precipitation forecasts from second-generation Global Ensemble Forecast System reforecast (GEFSv2) were used as inputs. Each of the ensemble members was bias-corrected by Empirical Quantile Mapping method using the Canadian Precipitation Analysis (CaPA) as a training/verification dataset. Forecast evaluation is performed for 1-day up to 8-days forecast lead times in a 6-month hindcast period. Results indicate that bias-correcting precipitation forecasts using verifying datasets (such as CaPA) for a training period of at least two years before the forecast time, produces skillful ensemble hydrological forecasts. A comparison of models in forecast mode shows that the two lumped models (SACSMA and MACHBV) can provide better overall forecast performance than the benchmark WATFLOOD and the macroscale Variable Infiltration Capacity (VIC) model. However, for shorter lead-times, particularly up to day 3, the benchmark distributed model provides competitive reliability, as compared to the lumped models. In general, the SACSMA model provided better forecast quality, reliability and differentiation skill than other considered models at all lead times.2 of 27 soil moisture, and evaporation from summer to fall [1]. Relevant methodologies were proposed to assess several aspects of the hydrological cycle such as snowpack, spring melt, soil moisture, rainfall frequency, and evaporation, in the Canadian Prairie regions [2][3][4][5]. The effect of climate, land use, and ecosystem change on the hydrological processes of cold and wetland regions were also studied [6][7][8]. Even though some efforts were made to formulate the realistic representation of wetland processes in hydrological models [9][10][11][12][13], challenges of hydrological forecasting and flood predictions in such complex watersheds remain at large.Several important works have already been performed for enhancing flood prediction in several watersheds: for example, using single or multiple hydrological models [14-22], or feeding ensemble numerical weather products to models [23][24][25][26][27][28][29]. Velázquez et al. [18], for example, analyzed 16 lumped hydrological models with 50-member ensemble weather inputs. They detected that the multi-model approach of a grand member ensemble provided more...

Evaluation of Radar-Gauge Merging Techniques to Be Used in Operational Flood Forecasting in Urban Watersheds

“…A well-developed flood forecasting system that can deliver accurate and reliable forecasts with proper lead time is a vital part of nonstructural flood management. At present, flood forecasting and warning utilizing hydrological models to forecast river flow have a widespread application in disaster management [3,4]. In urban watersheds, the frequency and magnitude of flooding are mostly influenced by precipitation [5,6].…”

Evaluation of Snowmelt Estimation Techniques for Enhanced Spring Peak Flow Prediction

“…Their results demonstrated poor performance when the HSPF and HBV-EC conceptual models were assessed for the spring season. Awol et al [60] performed a comparative analysis of lumped conceptual models (SAC-SMA and MAC-HBV), a distributed WATFLOOD model, and a macroscale land surface model (VIC) for shortand medium-range flood forecasting at UASR. Their results also suggested that SAC-SMA provided better performance than other models investigated at the study site.…”

Understanding Uncertainty in Probabilistic Floodplain Mapping in the Time of Climate Change

“…A PFC of zero shows a perfect simulation [89]. VE varies between 0 and ∞ and values close to 0 show better performance [90]. NSE values range from −∞ to 1 where 1 indicates the highest simulation performance, and so higher values are preferable.…”