During the internship at Royal HaskoningDHV, the additional value of using rain radar for rainfall-runoff models with respect to rain gauges is investigated. The area of waterboard Roer en Overmaas as well as Hupselse Beek are taken as pilot areas.
A study has been performed to investigate the additional value of rain radar in rainfall-runoff-models in relation to more conventional methods using raingauges. With rain radar, it is possible to have a spatial distributed overview of rainfall (1 x 1 km). Rain gauges give point measurements at one location, which can be spatially interpolated using different spatial interpolation techniques. In this study the conventional method of using Thiessen polygons was used. The additional value is investigated for two different components: the precipitation variability and the rainfall-runoff-process. The main study area is the Geul catchment (343 km2) in Limburg (NL) and Germany/Belgium. This catchment is divided in multiple smaller subcatchments. Additionally a small study was performed for the Hupsel catchment (6 km2).
Firstly, the rainfall variability was determined for different months by investigating radar data from 2009-2012. As expected, the variability in the Geul catchment is the highest in summer months and lowest in winter months. So, the additional value of radar measurements is the highest during summer months. Afterwards, a theoretical study was performed to test how many rain gauges are needed to approach the spatial variability of radar.. In the Geul catchment, different amounts of random selected rain gauges got the value of the belonging radar pixel. With Thiessen polygons, they were interpolated to a catchment average precipitation. These catchment average values were compared to the radar average. Based on the threshold that a deviation of 10% is allowed this leads to the following main conclusions. During winter, 3 rain gauges on 343 km2 give reliable results for 80% of the days. To reach the same reliability in summer months, around 10 rain gauges should be installed.
Rain radar data were also used as input for the HBV-model, in which the smaller subcatchments of the Geul can be modelled. These simulated discharges are considered as the reference situation. Again, randomly rain gauges were selected and interpolated to a subcatchment average using Thiesenpolygons, which is the input for the HBV- model. We see that the range of the peak discharges becomes smaller when the amount of rain gauges increases and the reference run is approached better. Also the timing with respect to the reference run can be different when only a few rain gauges are used. So, the additional value of radar is a better area average as well as a better timing.
In the Hupsel catchment, the input of rain radar and rain gauge data were compared for the extreme event in August 2010 with a more physical model. This model is developed by Royal HaskoningDHV. We see that the computed discharge for this event with radar data is slightly higher than with rain gauge data.