SPARC

• Return period 25 years correspond to 4%
• Return period 50 years correspond to 2%
• Return period 100 years correspond to 1%
• Return period 200 years correspond to 0.5%
• Return period 500 years correspond to 0.2%
• Return period 1000 years correspond to 0.1%

Due to the global nature of the GAR project, flood maps do not take into account flood defenses that are normally present to preserve the value exposed to floods. This can strongly influence the results of the risk calculations and especially of the economic parameters especially at provincial level.

Floods are usually caused by heavy or prolonged rainfall but can also occur due to landslides triggered by heavy rainfall or earthquakes, failure of dams or hydraulic structures, or high sea levels at river mouths. Here is assumed that the source (precursor event) is rainfall and higher levels of precipitation are very likely to lead to high river discharge and consequently to flood events. This approach, sometimes referred as “correlation with triggers”, has been successfully applied in landslide hazard analysis and is based on the assumption that the frequency of the landslides (or flood) is that of the triggering event.

This method, applied to floods, has a few drawbacks (e.g. different triggers can cause an event for example rainfall or snowmelt) yet seems a viable solution for establishing a simple procedure capable of disaggregating the annual probability into a monthly one. In this early phase, however an attempt will be made to consider not only precipitation by adding additional factors (e.g. evapotranspiration) to build a more realistic scenario able perhaps to provide a more reliable short-term monthly estimates of flood occurrences likelihood.For this purpose, several raster datasets were used, amongst those available at FAO geonetwork, for analysing different aspects of the flood phenomena according to the equation above.The precipitation dataset chosen is a grid with estimated precipitation per month with a spatial resolution of 10 arc minutes. This dataset has been constructed from 27 075 stations with 1961-1990 climatological normals. The dataset consists of 12 ASCII-grids with mean monthly data in mm/day * 10, and one ASCII-grid with yearly data in mm/year. In addition, 12 ASCII-grid with monthly values at 5 arc minutes resolution are made available as input data for a global water balance model (GlobWat).The following image depict three different raster containing information on precipitation in January, June and August in Cameroon.

Using the whole set of twelve raster, the annual distribution of rain is calculated for each province worldwide (province are black lines in the previous image) calculating the mean millimetres registered between 1961 and 1990 in order to generate the distribution of rain in that administrative level as visible in the following pictures.

First column represents months, the second the millimetres of rain normalize between 0 and 1 and the number of accidents also normalized to make the comparison easier with precipitation values. Evaporation is known to be the least of these quantities, particularly over short periods of time, but is the only data available together with precipitation on a monthly basis whereas infiltration and interception were assessed only partially as they should be derived from soil data and could easily require a study per se. Finally, the historical variability of precipitation has been used to see if all months are equally affected by high level of variability in terms of precipitation. Al the meteorological data have been compared with the historical occurrences related with flood, in order to evaluate any noticeable correlation.The result is visible in the next figure.