Attributing uncertainty in streamflow simulations due to variable inputs via the Quantile Flow Deviation metric. (June 2018)

• Record Type: Journal Article
• Authors: Shoaib, Syed Abu; Marshall, Lucy; Sharma, Ashish
• Is Part Of: Advances in water resources. Volume 116(2018)
• Page Start: 40
• Page End: 55
• Publication Date: 2018-06
• Subjects: Input uncertainty -- Quantile Flow Deviation (QFD) -- Multi-site rainfall -- Model structure; Hydrology -- Periodicals; Hydrodynamics -- Periodicals; Hydraulic engineering -- Periodicals; 551.48
• DOI: 10.1016/j.advwatres.2018.01.022
• Languages: English
• ISSNs: 0309-1708

Abstract

Highlights: We present a method to quantify the impact of variable inputs on streamflow simulations in a typical hydrologic modelling exercise. The study demonstrates the functional relationship between input variability and the availability of multiple rain gauges and spatial observations of potential evapotranspiration in simulating streamflow via hydrologic models. The relative contribution of input uncertainty compared to model structural and parameter uncertainties is evaluated. Abstract: Every model to characterise a real world process is affected by uncertainty. Selecting a suitable model is a vital aspect of engineering planning and design. Observation or input errors make the prediction of modelled responses more uncertain. By way of a recently developed attribution metric, this study is aimed at developing a method for analysing variability in model inputs together with model structure variability to quantify their relative contributions in typical hydrological modelling applications. The Quantile Flow Deviation (QFD) metric is used to assess these alternate sources of uncertainty. The Australian Water Availability Project (AWAP) precipitation data for four different Australian catchments is used to analyse the impact of spatial rainfall variability on simulated streamflow variability via the QFD. The QFD metric attributes the variability in flow ensembles to uncertainty associated with the selection of a model structure and input time series. For the case study catchments, the relative contribution of input uncertainty due to rainfall is higher than that due to potential evapotranspiration, and overall input uncertainty is significant compared to model structure and parameter uncertainty. Overall, this study investigates the propagation of input uncertainty in a daily streamflow modelling scenario and demonstrates how input errors manifest across different streamflow magnitudes.