Impact of Spatio-Temporal Disaggregation of Rainfall on Hydrological Modelling

Uncertainty assessment of rainfall patterns and the accompanying hydrological effects is essential to formulate effective adaptation strategies. Although the problem of equifinality in hydrological modelling has long been debated, its impact on hydrological analysis has not been sufficiently investigated. Traditional calibration techniques assume that input error is minimal, which might add a bias to the parameter estimates and impair the model predictions. Existing methods to overcome this issue are often weak due to both challenges in comprehending sampling errors in rainfall and processing limitations during parameter estimation. Such approaches consider structural and parameter uncertainties, whereas input and calibration data errors are often unaccounted for. This study aims to enhance the computational effectiveness of uncertainty analysis and separate the sources of uncertainty. Also, the implications of model input uncertainty to coupled human-natural-hydrologic systems and environmental changes are evaluated. A regression-based technique is developed to measure the level of uncertainty in the monsoon precipitation patterns for an urban catchment in Bangalore city, India. Sub-hourly rainfall datasets for various stations are estimated using disaggregation techniques such as scale-invariance and k-nearest neighbours-based methods. These datasets are fed into a hydrological model to connect the proposed method with the common framework for hydrological modelling. The findings demonstrate that the performance of a hydrological model is highly dependent on the spatio-temporal scale of the input rainfall in urban catchments where flash flood situations are envisaged.