Flood Prediction in data-scarce basins : Maximising the value of limited hydro-meteorological data

Sammanfattning: Floods pose a threat to society that can cause large socio-economic damages and loss of life in many parts of the world. Flood-forecasting models are required to provide simulations at temporal resolutions higher than a day in basins with concentration times smaller than 24 h. However, data at such resolutions are commonly limited or not available, especially in developing or low-income countries. This thesis covers issues related to the scarcity and lack of high temporal-resolution hydro-meteorological data and explores methods where the value of existing data is maximised to improve flood prediction.By varying the starting time of daily records (the day definition), it was shown that this definition had large implications on model calibration and runoff simulation and therefore, should be considered in regionalisation and flood-forecasting applications. A method was developed to treat empirically model-parameter dependencies on the temporal resolution of data. Model parameters seemed to become independent of the temporal resolution of data when the modelling time-step was sufficiently small. Thus, if sub-daily forcing data can be secured, flood forecasting in basins with sub-daily concentration times may be possible using model-parameter values calibrated from time series of daily data. A new calibration method using only a few event hydrographs could improve flood prediction compared to a scenario with no discharge data. Two event hydrographs may be sufficient for calibration, but accuracy and reduction in uncertainty may improve if data on more events can be acquired. Using flood events above a threshold with a high frequency of occurrence for calibration may be as useful for flood prediction as using only extreme events with a low frequency of occurrence. The accuracy of the rainfall forecasts strongly influenced the predictive performance of a flood model calibrated with limited discharge data. Between volume and duration errors of the rainfall forecast, the former had the larger impact on model performance.The methods previously described proved to be useful for predicting floods and are expected to support flood-risk assessment and decision making during the occurrence of floods in data-scarce regions. Further studies using more models and basins are required to test the generality of these results.