Kretzschmar, Ann and Tych, Wlodek and Beven, Keith and Chappell, Nick (2017) Utilising Reverse Hydrology to quantify and improve the spatio-temporal information content of catchment rainfall estimates for flood modelling. PhD thesis, Lancaster University.
2017KretzschmarPhD.pdf - Published Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.
Download (38MB)
Abstract
Reverse Hydrology is a term describing methods for estimating rainfall from streamflow. The method presented here is based on combining inversion of a causal rainfall-runoff model with regularisation. This novel method, termed RegDer, combines a continuous-time transfer function model with regularised derivative estimates and is compared with an alternative method for direct inversion of a discrete-time transfer function using sub-hourly data from two catchments with contrasting rainfall and catchment storage characteristics. It has been demonstrated to recover the prominent features of the observed rainfall enabling it to generate a streamflow hydrograph indistinguishable from the observed catchment outflow. The loss of temporal resolution of the resultant rainfall series is the price paid for the numerical stability of the RegDer method, however this does not affect its ability to capture the dynamics required for streamflow generation. The inferred rainfall series was initially interpreted as an estimate of catchment rainfall but was later more precisely described as the rainfall necessary for generating streamflow – Discharge Generating Rainfall (DGR). The spatial aspect of the method was investigated using data from a densely gauged catchment. Frequency domain aspects of RegDer dual interpretation as a composite spectral decomposition method are analysed and discussed in the context of catchment data. Potential applications and developments of the approach include in-filling and extending rainfall records, reducing uncertainty in both gauged and ungauged catchments by improving rainfall estimates, assessing and refining rain gauge networks and re-evaluating areal rainfall estimation.