Comparison of different rainfall inputs in a continuous rainfall-runoff model – a case study for Argentina
Precipitation data is the main input parameter in order to simulate rainfall-runoff processes, since it is strongly dependent on the accuracy of the spatial and temporal representation of the precipitation. In regions where rainfall stations are scarce, additional data sources may be considered necessary. In this manner, remote sensing from satellite platforms has provided a satisfactory alternative due to its global coverage. Although a wide range of satellite-based estimations of precipitation is available, not all the satellite products are suitable for all regions. Most of the studies performed with the purpose of evaluating their accuracy are focused in particular areas of the world. In this fashion, particular models have to be conducted in order to evaluate their performances, specially in regions with complex geography as high mountains.
Additionally, to perform an appropriate spatial representation of the rainfall and consequently to improve the available data, interpolation techniques are used, e.g. simple techniques as Nearest Neighbour or Inverse Distance methods, and some more complex as geostatistical (Kriging) methods. This last one offers the advantage of adding relevant additional information in the interpolation, providing a chance to compensate a low network density. Moreover, in data scarce regions in which interpolation schemes are applied, it becomes difficult to have an accurate performance assessment; in this manner, other comparison tool is required as rainfall-runoff models.
In this manner, the aim of this study is to perform a comparison between different types of available rainfall data by means of a hydrological model. The work is focused in Neuquén catchment, a mountainous region of Argentina where several rainfall stations and flow gauges are available. In this fashion, a satellite-based estimated precipitation already validated in mountainous areas and southern latitudes, CMORPH, is used as well as the available rainfall stations as input. Moreover, to improve the rainfall stations measurements, CMORPH data and topography are used during the interpolation as additional variables. Consequently, five precipitation input cases are generated and compared. To accomplish the main objective, at first several interpolation techniques are tested and assessed by means of cross-validation for each precipitation input. Subsequently, a hydrological model HEC-HMS is set up for every case and thus its outcomes are compared using indices of reliability.
Regarding the cases that consider the rainfall stations data, assess from the interpolation technique showed that the best performance is obtained with the case without external drift. Conversely, the hydrological model showed the most accurate precision when topography was used as additional information. Input cases with the satellite-based estimations as external drift improved considerably the results in comparison to the case in which rainfall stations are considered alone. However, results showed that the case with CMORPH data as only input, the estimation of the observed discharge was not able to be reproduced precisely. Finally it could be concluded that, in those cases in which the rainfall stations networks are not dense enough and do not represent the spatial variability of the area correctly additional information is extremely useful to simulate more accurately the observed discharge in the area.
Notes: Supervisor: Ana Claudia Callaú Poduje
Format: Digital (Free)