Implementation of an upscaled probabilistic fill-and-spill method to simulate wetland-dominated landscapes

Runoff characteristics of wetland-dominated landscapes are highly influenced by the heterogeneity of wetland properties such as contributing areas, local storage deficits, and degree of connection to the wetland network. Field observations indicate the fill-and-spill process is a common and controlling process in wetland-dominated landscapes in which wetlands receive water until a threshold is satisfied and then release water to the next water bodies in a basin. The upscaled probabilistic fill-and-spill algorithm provided here aims to take the understanding from observation of the fill-and-spill process in individual wetlands at local scales to estimate the response of thousands of wetlands to precipitation or snowmelt events at a larger scale. For this purpose, a novel derived distribution approach is implemented using the probabilistic characteristics of wetlands, i.e., deficit and concentrating factor probability distribution functions. The method proposed here is a generalization of the Probability Distributed Model (PDM) and Xinanjiang probabilistic runoff models with the inclusion of the effects of wetland contributing area and cascading networks. The analytical solution of this upscaled fill-and-spill processes has been compared with a Monte Carlo solution and then implemented in RAVEN, a semi distributed hydrological modelling framework. The accuracy and ability of the model simulation has been tested on ten watersheds in the Qu'Appelle River Basin in Saskatchewan, Canada. The promising simulation results obtained from manual and automated calibration shows strength of the proposed upscaled fill-and-spill algorithm in simulation of low gradient landscapes.