Real-time river flow forecasting web application

Back to list


Public Service of Wallonia, waterways management


Real-time web application for river flow forecasting and flood alarms.


The Walloon region in Belgium enjoys beautiful rivers but also suffers from recurrent flooding. River flow rates and water level forecasting therefore play an important role as part of the region’s rain hazard management strategy.

Together with the Centre for Systems Engineering and Applied Mechanics (CESAME) from the Catholic University of Louvain (UCL), the Walloon region initiated a project to design a mathematical model to predict river flow rate and water level at various critical locations. These predictions are based on precipitation measurements and forecasts.

In 2005 this model was given to N-SIDE with a view to developing and maintaining a professional real-time river flow forecasting web application for everyday use by Public Service of Wallonia staff to anticipate flooding.


The mathematical model combines four components (see [1] for more details):

  1. An optimal minimum variance interpolator that computes the mean areal rainfall on the watershed, using on on-line field measurements of both rainfall depths in rain gauges and water levels in rivers
  2. A non-linear conceptual production function that describes the water storage in the watershed and computes the effective rainfall from the mean areal rainfall.
  3. A linear ARX predictor that describes the superficial runoff of the net rainfall towards the watershed outlet and computes the short term river flow forecast.
  4. A simulation model that produces long-term river flow forecasts from either meteorological data or other rainfall scenarios.

The parameters of the model components are estimated by regression techniques from historical data. This estimation is performed on a regular basis as more data is collected, using automated procedures. This ensures continuous improvement of the model and alignment with the natural evolution of the system (maintenance of the rivers, new constructions that impact river flows, additional field measurements …).

The model is embedded in a user-friendly web application that multiple users can run simultaneously to consult predictions, compare past predictions with real data, perform what-if analysis (worst-cast vs. best-case rainfall scenarios), verify and correct input data.

The figure below illustrates flow rate predictions with the warning (black) and alarm (red) thresholds, for one specific location in Belgium, during a period of intense precipitation. Several prediction scenarios are compared, based on different rainfall forecasts, from the worst case (red) to the most optimistic case of no further rainfall (black).

Flow rate predictions with the warning and alarm thresholds


Thanks to this forecasting software, the flood prevention department is able to anticipate flooding hazards during high intensity rainfall periods with a lead-time from a few hours to two-days. Based on the alarm thresholds, the department can communicate flood warnings to the general public sufficiently well in advance for damage avoidance action to be taken. In case of flooding, the prediction model also helps to estimate the time needed for a return to normal, which is also important information to communicate.

The system has now run for several years and, thanks to the continuous recalibration of the underlying model parameters, it has consistently produced reliable predictions.


[1] “HYDROMAX : A Real-Time Application for River Flow Forecasting”, G. Bastin, L. Moens, Proceedings International Symposium on Flood Defence, Vol. 1, pp. D-33-40, Kassel Reports of Hydraulic Engineering, N° 9/2000, Hercules Verlag Kassel, Germany, September 2000.

Print Friendly
Back to list