Flood website: Indication of accuracy

The flood forecasts have certain accuracy, of course. No forecast can be exact given the uncertainties in location and in the parameters that are needed as inputs. Furthermore, the hydrological and hydrodynamic models that are used for the MRC flood forecast are by definition a simplified representation of the real world. Given these considerations, and the fact that rainfall predictions are notoriously uncertain, it is unrealistic to expect 100% accuracy. Still, it is very useful to have an idea of the errors.

It can easily be understood that the accuracy is higher if the rainfall forecast and the other daily input parameters (primarily water levels to determine river flow) are more certain. This means that forecasts that look further ahead depend on forecasts that are less certain. Also, the impact of a given river flow at an upstream station becomes less significant the further it travels downstream, because (uncertain) river flows of tributaries join the mainstream. Finally, if we move upstream, we have less upstream river monitoring stations, adding to the uncertainty. Overall measurement and other errors should be considered too.

The conclusion is that the MRC Mekong Flood Forecast is in general more accurate for downstream locations, while it becomes less accurate if we look further ahead.

This discussion of uncertainties may give the impression that the forecast is not very useful. Still, the forecasts are actually rather accurate, considering. However, the final judge of the usefulness is the user. If the user demands 10 cm accuracy with 5 days lead time the RFMMC nor anyone else can help him for most of the stations; if the user needs to know if the water level is going to reach the flood level within three days with an uncertainty of 25 to 50 cm the product is useful in all but 5 of the forecast locations. Typically, a user of flood forecast information will also accept a larger margin of error more upstream because the maximum rise rate of the Mekong is (much) higher.

An indication of the accuracy is given in the figure on this page. The graph uses the mean absolute error of the MRC Mekong Flood Forecasting System, based on the 2008 data and the configuration for the 2009 flood season.

For the actual flood forecast these values are further adjusted by the Forecaster-In-Charge using knowledge of bias and system behavior, and of weather patterns and specifics of the input data. This usually improves the forecast significantly.

In the future the performance of the MRC Mekong FFS will be measured against a set of performance indicators that is established by combining international standards and the specific circumstances in the Mekong River Basin.

Please note that the flood levels used in the flood forecast are determined by the member states. The definition of flood level is thus depending on national standards; usually it indicates the level where the local or national authorities need to take urgent measures to prevent significant damage. Alarm stage is defined by the MRC as the level where flood stage is predicted within the next three days. Alarm level is defined by the member states and is mentioned as reference.