The usability of 250 m resolution data from the UK Meteorological Office Unified Model as input data for a hydrological model

Abstract

This article examines whether very-high resolution (250 m) data provided by the United Kingdom Meteorological Office Unified Model (UM) could be used as an input to a hydrological model for hydrological forecasting purposes. A summer convective event and a winter stratiform event are both considered and it is found that in these two cases, despite errors in both the positioning and timing of storms the predictions were of a sufficient quality to provide beneficial inputs for hydrological modelling. There are encouraging indications that this technique may be valuable in improving flood forecasting generally.

Detecting the effects of spatial variability of rainfall on hydrological modelling within an uncertainty analysis framework

“…Here, we are not considering the quality of the performance measures used in terms of the likely error characteristics, we are only interested in the relative changes of this measure to assess changes in the resultant prediction bounds. An assessment of performance measures defined from observational uncertainties and an appropriate methodology for this assessment can, however, be found in Younger et al (2008).…”

“…Dynamic TOPMODEL has previously been run on the Brue catchment (Younger et al, 2008). In the present study, the model has been calibrated (for each scenario) using a 1-month calibration period at the end of the dataset (not shown).…”

“…First, if any raingauges were situated within the same 2 km gridsquare, then they were considered to be the same raingauge with the total taken to be the mean of the raingauges (again considered to be acceptable because of the low variability of nearby raingauges). Each point in the catchment was associated with its nearest raingauge (the grid was a design specification of a previous experiment involving the same catchment (see Younger et al, 2008)). This gave a total of 28 'rain areas' (Figure 1).…”

“…A study by Bell and Moore (2000) on this catchment into the effects of spatial rainfall variability showed that lower resolution rainfall data gave the best performance when using a simple grid distributed model. Cluckie et al (2006) used ensemble rainfall from a numerical weather prediction (NWP) model on this catchment, whereas Younger et al (2008) used NWP output at 250-m resolution on the Brue and found that NWP rainfall could be used as acceptable input to a distributed hydrological model.…”

Response to Reviewer 1 and 2

“…Water balance is calculated as mean annual rainfall minus mean annual discharge and potential evapotranspiration (as actual evapotranspiration is not available). NSE and RRBias for the best ranked simulation according to the combined score described in Sect ditions (Fowler et al, 2016), for different hydrological extremes (Coron et al, 2012;Veldkamp et al, 2018;Zaherpour et al, 2018) and for multiple objectives simultaneously (Kollat et al, 2012) and incorporate input and flow data uncertainty (Coxon et al, 2014;Kavetski et al, 2006;McMillan et al, 2010;Westerberg et al, 2016).…”