88 THE FUTURE OF ARID LANDS 
One might ask whether, in desert areas, the mean annual 
rainfall is of any interest. If we take a station such as Adrar, 
where the average annual rainfall is 17.4 millimeters, there is 
evidently little chance of that amount of rainfall in any one year, 
the standard deviation being 16.7 millimeters. On the other 
hand, there is much more chance that 174.0 millimeters will fall 
in ten years. The annual average value has consequently little 
bearing on the most probable value, but it gives an idea of the 
amount of precipitation that is more realistic in less arid regions. 
Such is the conclusion drawn by Dubief who has specially studied 
the noalnancan@)): 
The element which may most easily be considered is the annual 
rainfall. The yield of crops depends not only on the total annual 
rainfall, but also on the distribution of the rain by season and 
even by days. 
However, to simplify matters, the discussion in this section 
will be limited to: 
1. What is the probability of having, at a given point, a total 
annual rainfall greater than a given amount? 
2. What is the probability of having a given annual total rain- 
fall over the entire country? 
3. Can one forecast variations in the annual rainfall over a 
period of time? 
Rainfall at Given Point 
Suppose that it is possible to make observations over a long 
period of time, say more than 50 years. The variability of the 
rainfall at a given point can then be represented during the 
length of observation by the law of Galton Gibrat. If H is the 
annual amount of rainfall in millimeters, HM the medium amount, 
and HO the corrective amount, normal distribution will be the 
logarithm of (H + HO). On a Gausso logarithmic diagram, we 
obtain a straight line equation: 
log (HW + HO) = ax + log (HM + HO) 
Figure 1 gives the frequencies of the annual rainfall for Tunis, 
and for the average of five stations scattered over all Tunisia 
