Abstract
Rain gauge data suffers from spatial errors because of precipitation variability within short distances and due to sparse or irregular network. Use of interpolation is often unreliable to evaluate due to the aforementioned irregular sparse networks. This study is carried out in the Nette River catchment of Lower Saxony to alleviate the problem of using gauge data to measure the performance of interpolation. Radar precipitation data was extracted in the positions of 53 rain gauge stations, which are distributed throughout the range of the weather surveillance radar (WSR). Since radar data traditionally suffers from temporal errors, it was corrected using the Mean Field Bias (MFB) method by utilizing the rain gauge data and then further used as the reference precipitation in the study. The performances of Inverse Distance Weighting (IDW) and Ordinary Kriging (OK) interpolation methods by means of cross validation were assessed. Evaluation of the effect of the gauge densities on HBV-IWW hydrological model was achieved by comparing the simulated discharges for the two interpolation methods and corresponding densities against the simulated discharge of the reference precipitation data. Interpolation performance in winter was much better than summer for both interpolation methods. Furthermore, Ordinary Kriging performed marginally better than Inverse Distance Weighting in both seasons. In case of areal precipitation, progressive improvement in performance with increase in gauge density for both interpolation methods was observed, but Inverse Distance Weighting was found more consistent up to higher densities. Comparison showed that Ordinary Kriging outperformed Inverse Distance Weighting only up to 70% density, beyond which the performance is equal. The hydrological modelling results are similar to that of areal precipitation except that for both methods, there was no improvement in performance beyond the 50% gauge density.
Highlights
Hydrological modelling to simulate historical discharges and to forecast future runoff is extensively used as a tool to understand catchment processes and to optimize water allocation and management
Radar precipitation data was extracted in the positions of 53 rain gauge stations, which are distributed throughout the range of the weather surveillance radar (WSR)
The common univariate methods are Nearest Neighbour, Thiessen Polygon and Inverse Distance methods, Ordinary Kriging (OK) and Indicator Kriging (IK) while the multivariate methods are Kriging with external Drift (KED) [9]
Summary
Hydrological modelling to simulate historical discharges and to forecast future runoff is extensively used as a tool to understand catchment processes and to optimize water allocation and management. Rain gauges provide rainfall measurements at individual points but different uncertainties are associated with the use of gauge data to estimate rainfall with appropriate temporal and spatial scale variation at basin scale [4]. Even if rain gauges are equipped with real-time rainfall information at very fine temporal resolution under the help of automatic rainfall recording equipment, it is still a challenge to characterize the spatial variation of rainfall without a network of well-defined rain gauges in the catchment [5]. The choice of the location of rain gauges is not planned nationally but rather is an ad-hoc localized process. This leads to irregular and inefficient allocation of gauges. Radar provides precipitation information at longer ranges, higher temporal resolution and continuous spatial information suffers measurement errors among which include clutter, bright band effect (a dark region in range height indication (RHI) scans due to melting of precipitation as it descends) and volume errors [8]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
