Towards reducing uncertainty in rainfall–streamflow model parameter regionalisation (TRUMPER)
An IAHS-PUB international community project organised by the Top-Down modelling Working Group (TDWG)
For gauged basins the uncertainty in statistical relationships between rainfall–streamflow model parameters and physical catchment properties is typically large, with knock-on consequences for flows estimated from rainfall at ungauged (flow) sites using such relationships. One likely reason for the large uncertainty is that the rainfall–streamflow model parameters are often derived using data at a given time-step (e.g. daily), irrespective of whether or not that data time-step is small enough to capture the dynamics of the flow regime at each of the gauged sites. To date, this has received little attention in parameter regionalisation studies or by commentators on rainfall–streamflow modelling. This lack of attention would be acceptable if the effect of data time-step was always small. However, recent work has shown that for the 10.6 km² Wye at Cefn Brwyn (one of the Plynlimon research basins in Wales) each of the five parameters of an IHACRES model changes by between 50% and 80% as different time-step data (daily to hourly) are employed for model calibration. The analysis also outlined a simple method for normalising IHACRES parameters for the Wye at Cefn Brwyn (i.e. giving parameters essentially independent of data time-step). Further work includes similar analysis of a much larger (298 km2) basin, also in Wales.
Anticipating that statistical relationships between physical basin properties and normalised model parameters (rather than data-time-step-dependent parameters) will be better predictors of model parameters at ungauged sites, the TDWG has initiated TRUMPER whereby analyses on datasets (at 1-hour, 2-hour, 4-hour, … 24-hour, …, time-steps) for many basins will be undertaken using IHACRES and other rainfall–streamflow models. Although the initial focus will be on the data time-step issue many related measurement and modelling issues will arise in this work, so other PUB WGs (e.g. WG2-MOPEX and WG7-uncertainties), the ERB Network and EURO-FRIEND researchers are warmly invited to participate in TRUMPER.
Within the overall PUB objective of reducing predictive uncertainty the initial objectives of TRUMPER are to: