Parameter Uncertainties in a Conceptual Rainfall-runoff Model and Implications on Surface Water Management and Planning Decisions

Abstract

This paper investigates the uncertainties arising from parameter identification in a conceptual rainfall-runoff model and implications on surface water management and planning decisions. A conceptual rainfall-runoff model, the Probability Distributed Rainfall-Runoff Model (PDM), is applied within the Dove River catchment (UK) using 1km2 resolution radar rainfall as inputs and 15minutes resolution gauged flow data for calibration and validation. In most conceptual/lumped models, some parameters lack physical basis and cannot be inferred from direct measurements. The DiffeRential Evolution Adaptive Metropolis (DREAM) algorithm developed by Vrugt et al. [1] is employed for Bayesian inference to automatically calibrate the model against observed flow data. Probabilistic flow predictions are determined based on the resulting posterior parameter distributions, which reflect the residual model parameter uncertainty. These uncertainties associated with model parameters are propagated through a reservoir management model to assess its impacts on reservoir performance in maintaining adequate supply demand balance. The impact of using various reservoir operational rules on the characteristics of uncertainty propagation and associated impact on predicted supply demand balance are investigated. The results in this study suggest that adaptive management of reservoir operational rules can be used to achieve optimum balance between environmental impact, drought reliable supply and operation cost.