Flood forecasting is an essential tool for providing people still exposed to risk with advance notice of flooding, in an effort to save life and property.

Different flood forecasting service models exist based on the needs of end users: a system may be developed for the public or strictly dedicated to the authorities. There is no single consistent approach worldwide but the basic principles of a good warning system are shared by all. These comprise:

  • Better detection in times of need well before the actual event occurs
  • Interpretation of the detected phenomena and forecasting this to the areas likely to be affected
  • Dissemination of the warning message to the relevant authorities and public via the media and other communication systems.

The fourth and final aspect is to encourage the appropriate response by the recipients by preparing for the upcoming event. This can be improved through flood response planning by people at risk and their support groups.

Uncertainty in flood forecasting

Models, by definition, are approximations of reality. As described earlier, all models suffer from a certain level of approximation or uncertainty in spite of powerful computing systems, data storage and high level technologies. Decision makers have to consider the effects of uncertainties in their decision-making process. Errors in forecasting of an event, for example stage or time of arrival, may lead to under-preparation (at the cost of otherwise avoidable damage) or over-preparation (resulting in unnecessary anxiety). The balance between failure to warn adequately in advance and the corrosive effects of too many false alarms must be carefully managed.

The reliability of flood forecasting models relies on the quantification of uncertainty. All natural hazards are uncertain. The various sources that give rise to uncertainty in forecasting and early warning can be classified (Maskey. 2004) as:

  • Model Uncertainty
  • Parameter Uncertainty
  • Input Uncertainty
  • Natural and Operational Uncertainty.

It is necessary to gain a better understanding of the options available to deal with the uncertainties within the system arising from these different sources.

In order to produce a forecast, the initial conditions are typically determined by means of observations from rain gauges; these may, however, be unevenly spaced throughout the catchment, leading to uncertainty as to the total volume of rainfall. Where hydrologically important areas (such as steep slopes) are unrepresented, the model may utilize an interpolation method (introducing another element of uncertainty) in order to estimate run-off volume and peak flows. More sophisticated modeling can address these issues, but this in turn may demand high processing speeds and lengthy run-times.

To offset some of this uncertainty, operational flood forecasting systems are moving towards Hydrological Ensemble Prediction Systems (HEPS), which are now the ‘state of the art’ in forecasting science (Schaake et al. 2006; Thielen et al. 2008). This method formed part of initiatives such as HEPEX (Hydrological Ensemble Prediction EXperiment) which investigated how best to produce, communicate and use hydrologic ensemble forecasts for short, medium and long-term predictions. Despite its demonstrated advantages the use of this system is still limited: it has been installed on an experimental basis in France, Germany, Czech Republic and Hungary.

To deal with the uncertainty in spatio-temporal distribution and prediction of rainfall for extreme events, especially through radar derived data, a promising approach has been to combine stochastic simulation and detailed knowledge of radar error structure (Germann et al. 2006a, 2006b, 2009; Rossa et al. 2010). Radar ensembles have the potential benefits of increasing the time for warning especially for flash floods (Zappa et al. 2008). Advanced techniques, such as disdrometer networks (equipment capable of measuring the drop size, distribution and velocity of different kinds of precipitation) and LIDARs are being used to capture small scale rainfall phenomenon, whilst satellite remote sensing is more appropriate for regional and global level applications. A combination of all these methods and blending information is considered to be the most promising way forward.

There are a several useful examples of such systems:

  • DELFT-FEWS: one of the state of the art hydrological forecasting and warning systems developed by Deltares. This system is an integration of a number of sophisticated modules specialized in their individual capacities and the system is highly configurable and versatile. The system can be used as a standalone environment, or it can be used as a compliant client server application. Through its advanced modular system FEWS has managed to reduce the challenges like handling and integration of large datasets to a considerable extent.
  • Automated Local Evaluation in Real Time (ALERT) is the method used within the AUG member states to transmit data and information using remote sensors for warning against flash floods.
  • Central America Flash Flood Guidance is an example of regional flash flood warning. The national Hydrologic Warning Council (NHWC) has member countries across North America and many parts around the world; it is also a major organization in data dissemination for early warning for flood events.
  • The Mekong River Commission flood forecasting system, discussed above, has been operating since 1970. It is an integrated system which provides timely forecasting to its member countries. It consists of three main systems of data collection and transmission, forecast operation and information dissemination at both national and regional level.
  • The Southern African regional model for flood forecasting Stream Flow Model (SFM) has been applied after the Mozambique flood in 2000. The USGS along with Earth Resource Observation System (EROS) supports monitoring and modeling capacities of Southern African Countries.
  • Regional Water Authority of Mozambique (ARA-Sul) is responsible for issuing flood warning and real time forecasting. The system is operational in Southern Africa with a mean area of 3,500 square kilometers. A simplified flood warning system, the Mozambique Flood Warning Project, is specially tailored to the needs of the local population. It also involves the local people and trains them to install, monitor and maintain the structures.
  • Hydro Met Emergency Flood Recovery Project is used in Poland.
  • Bhutan’s Glacial Lake Outburst Flood (GLOFs) Iridium Satellite Communications is used as the telemetry back-bone for Bhutan’s GLOF Early Warning Project.
  • In the Toronto region of Canada, the Toronto and Region Conservation Authority (TRCA) flood forecasting and warning system is used; this is a scalable flood warning system including web-based data and video for nine watersheds.
  • The Automatic Dam Data acquisition and alarm reporting system, is the Puerto Rican System to obtain, monitor and analyze, in real- time, critical safety parameters such as inflows, outflows, gate openings and lake elevations for 29 principal reservoirs
  • Central Water Commission (CWC) in India provides the Turnkey Flood forecasting system across 14 states having 168 remote sites in six river basins.
Literature sources
Maskey, S., Guinot, V. and Price, R.K. 2004. “Treatment of precipitation uncertainty in rainfall-runoff modeling: a fuzzy set approach.” Advances in Water Resources 27 (9): 889-98.
Schaake, J., Franz, K., Bradley, A., and Buizza, R. 2006. “The Hydrological Ensemble Prediction Experiment (HEPEX).” Hydrological and Earth System Sciences Discussions 3: 3321–32.
Thielen, J., Schaake, J., Hartman, R. and Buizza, R. 2008. “Aims, challenges and progress of the hydrological ensemble prediction experiment (HEPEX) following the third HEPEX workshop held in Stres 27-29 June 2007.” Atmospheric Science Letters 9: 29-35.
Germann, U., Berenguer, M., Sempere-Torres, D., and Salvadè, G. 2006a. “Ensemble radar precipitation estimation — a new topic on the radar horizon.” Proceedings of the 4th European Conference on Radar in Meteorology and Hydrology (ERAD). Barcelona. September 18–22, 2006. 559–62.
Germann U., Galli, G., Boscacci, M, and Bolliger M. 2006b. “Radar precipitation measurement in a mountainous region.” Quarterly Journal Royal Meteorological Society 132: 1669–92.
Germann, U., Berenguer, M., Sempere-Torres, D., and Zappa, M. 2009. “REAL — Ensemble radar precipitation estimation for hydrology in a mountainous region.” Quarterly Journal Royal Meteorological Society 135: 445–56.
Rossa, A. M., Cenzon, G. and Monai, M. 2010. “Quantitative comparison of radar QPE to rain gauges for the 26 September 2007 Venice Mestre fl ood.” Natural Hazards and Earth System Science 10 (2): 371–7.
Zappa, M., Rotach, M.W., Arpagaus, M., Dorninger, M., Hegg, C., Montani, A., Ranzi, R., Ament, F., Germann, U., Grossi, G., Jaun, S., Rossa, A., Vogt, S., Walser, A., Wehrhan, J., and Wunram, C. 2008. “MAP D-PHASE: Real-time demonstration of hydrological ensemble prediction systems.” Atmospheric Science Letters 2: 80–7.
Measure category