Study on the effect of rainfall spatial variability on runoff modelling

Abstract Rainfall spatial variability was assessed to explore its influence on runoff modelling. Image size, coefficient of variation (Cv) and Moran's I were chosen to assess for rainfall spatial variability. The smaller the image size after compression, the less complex is the rainfall spatial variability. The results showed that due to the drawing procedure and varied compression methods, a large uncertainty exists for using image size to describe rainfall spatial variability. Cv quantifies the variability between different rainfall values without considering rainfall spatial distribution and Moran's I describes the spatial autocorrelation between gauges rather than the values. As both rainfall values and spatial distribution have an influence on runoff modelling, the combination of Cv and Moran's I was further explored. The results showed that the combination of Cv and Moran's I is reliable to describe rainfall spatial variability. Furthermore, with the increase of rainfall spatial variability, the hydrological model performance decreases. Moreover, it is difficult for a lumped model to cope with rainfall events assigned with complex rainfall spatial variability since spatial information is not taken into consideration (i.e. the VIC model used in this study). Therefore, it is recommended to apply distributed models that can deal with more spatial input information.