In pursuit of reliable flood prediction

Abstract

Flood forecasting necessarily relies on a number of technologies, namely, precipitation measurement at high resolution for large areas; numerical modelling of the hydraulics and hydrology of runoff and routing; and innovative information technology for synthesizing critical information needed for emergency response and risk management. Data sources for precipitation inputs can come from weather radar and automated rain gauges. Regional and urban hydrologic modelling is increasingly based on distributed physics-based hydrologic models that can leverage the high defnition rainfall derived from radar and rain gauge measurements. Statistical quality control in real-time is necessary to transform radar-based precipitation estimates into accurate hydrologic model input, which serves as an input to a physics-based distributed runoff model. Geospatial data are necessary for setup and parameterization of such distributed models. Real-time precipitation estimates from radar and rain gauge monitoring are coupled with the gridded representation of the watershed to produce warning notifications at distributed locations throughout urban, peri-urban, and rural/natural watersheds. Quantification of uncertainties and sources of error are examined within the pursuit of successful food forecasting. The following case studies illustrate distributed hydrologic forecasting across a range of space-time scales. This paper presents forecasting performance, an evaluation of accuracy, and discussion of factors that contribute to uncertainty in food forecasts and risk reduction.