Evaluation of Gridded Precipitation Data Products for Hydrological Applications in Complex Topography

David Gampe, Ralf Ludwig
2017 Hydrology  
Accurate spatial and temporal representation of precipitation is of utmost importance for hydrological applications. Uncertainties in available data sets increase with spatial resolution due to small-scale processes over complex terrain. As previous studies revealed high regional differences in the performance of gridded precipitation data sets, it is important to assess the related uncertainties at the catchment scale, where these data sets are typically applied, e.g., for hydrological
more » ... . In this study, the uncertainty of eight gridded precipitation data sets from various sources is investigated over an alpine catchment. A high resolution reference data set is constructed from station data and applied to quantify the contribution of spatial resolution to the overall uncertainty. While the results demonstrate that the data sets reasonably capture inter-annual variability, they show large seasonal differences. These increase for daily indicators assessing dry and wet spells as well as heavy precipitation. Although the higher resolution data sets, independent of their source, show a better agreement, the coarser data sets showed great potential especially in the representation of the overall climatology. To bridge the gaps in data scarce areas and to overcome the issues with observational data sets (e.g., undercatch and station density) it is important to include a variety of data sets and select an ensemble for a robust representation of catchment precipitation. However, the study highlights the importance of a thorough assessment and a careful selection of the data sets, which should be tailored to the desired application. Hydrology 2017, 4, 53 2 of 21 resolution and domain size must be made. While global and continental data sets, provided e.g., through the National Oceanic & Atmospheric Administration (NOAA) [10], the European Centre for Medium-Range Weather Forecasts (ECMWF) [11] and the German Weather Service (DWD) [12] , provide information on precipitation over a large domain and cover a large time period, they lack the high spatial resolution required for regional or catchment scale studies. High-resolution data sets are usually only available on a country level [13] or cover a specific geographical region [14] . These data sets are either reanalysis products [10, 11, [15] [16] [17] , derived through remote sensing [18, 19] or interpolated station observations [12, 14, 20] . Most of these data sets use station data either directly, or assimilate observed precipitation at some stage. The density of the included stations differs greatly between, but also within the data sets, restricting the effective resolution and spatial consistency [21] . Recent efforts were made to merge different data sets to achieve a better representation and combining the advantages of each data set [22] , as well as collecting sub-daily station observations [23] to increase the temporal resolutions. However, observations are prone to severe undercatch of precipitation, which is amplified in case of solid precipitation and over mountainous areas [21, 24, 25] . Therefore, there is merit in including data sets from a variety of sources for a robust estimation of reference precipitation. Various studies to assess the performance of these data sets were carried out in recent years, revealing considerable differences between the products. Most of these studies focus either on global, continental, or regional performance assessment often aiming in the evaluation of Regional Climate Model (RCM) performance [2,26]. Regional studies assess the benefit of higher resolution and consequently a better topographic representation, the issues of station density and the representation of regional weather phenomena [21] . Detailed source specific assessment showed limitations in the representation of several weather patterns for reanalysis data sets [27] and the restriction of satellite information under specific cloud cover conditions [15, 28] . Gridded precipitation data products are very often applied as meteorological input for hydrological modeling exercises in data scarce areas to bridge the data gap at the catchment or smaller regional scale [29, 30] . It is therefore important to also assess the performance of the data set at smaller scales over hydrological basins. Some studies detect the deviations on the catchment scale; however they focus on reanalysis [27] or selected regional data sets [31] . Additional efforts were carried out in the past focusing on 3-hourly and daily error analysis for satellite precipitation products over alpine catchments. They revealed large errors especially in the summer months and showed the difficulties to adequately address errors in satellite products with respect to the station density of in situ measurements [32] . Other studies identified the impact of different satellite precipitation products on the reproduction of flood events in Northern Italy. These studies showed both the potential and advantages of these data products, but also their limitations [33, 34] . However, there is still a need to evaluate long-term uncertainties and differences in existing precipitation products on the catchment scale, and exploiting as many of the available products as possible from different sources. It is important to assess not only the climatology, but also the daily extremes. This study contributes to recent findings by including an ensemble of eight available gridded precipitation data sets derived from observations and through reanalysis or remote sensing. These are evaluated against a high-resolution reference grid derived from local observations over a river basin for the period 1989-2008, with a focus on hydrological implications. Therefore, the comprehensive assessment does not only focus on the general climatology, but also on dry and wet spells, as well as extreme precipitation on the catchment scale. The study area is the catchment of the Adige in located in the southern part of the Alps in Northern Italy, covering an area of 12,100 km 2 . This catchment is selected due to a relatively high network of observation stations, and the challenging orography for the comparison with the ensemble of precipitation data sets. Due to the complex topography, the uncertainty introduced by the spatial resolution of the data sets was of special interest in this study. The overall purpose of this study was motivated from the application point of view, rather than atmospheric assessment of regional patterns. It is clear, that the outcomes of this study are highly
doi:10.3390/hydrology4040053 fatcat:3ld4hibsqrbung6yky6youoyvq