90 THE FUTURE OF ARID LANDS 
However, the variability of the annual rainfall is still very 
large, even for a country of the size of Tunisia. It is easy to be- 
come aware of this in comparing the distribution of rainfall during 
two four-year periods, one very dry (1944-47), the other very 
humid (1931-34) (Figure 2). If we assume that desert begins 
where there is less than 200 millimeters annual rainfall, we see 
that the limits of the desert moved some 200 kilometers in the 
Sahel between these two periods. The four-year period was chosen 
because it marks the influence of rainfall not only on annual 
harvests but also on deep-rooted plants such as olive trees. In 
fact, in the Sfax area, famous for its orchards, the olive trees in 
1947 had lost all their leaves and if the drought had lasted longer, 
great numbers would have died. Sand dunes were developing with 
great rapidity all the way into the center of Tunisia. Harvests 
of dry culture were of no account in the center and in the south. 
The drought of 1944-1947, moreover, was felt outside Tunisia, 
in Italy for instance (5). Hence the necessity, when studying 
the variability of climate, to cover territories of different areas to 
determine the extent of regions within which there can be an 
economic compensation for the variability of the climate. 
Forecasts on Variability of Rainfall 
If the rainfall varies much from one year to another, it varies 
also in the course of consecutive groups of years. Statistics give 
some idea of this. For example, the Sfax station gives the data 
in Table 1. The importance of the difference in the average five- 
year precipitation illustrates what has been said earlier on the 
subject of equipment plans. But, the statistics cannot be extra- 
polated with certainty beyond the period of observations. Be- 
yond that period one must have recourse to other methods of 
forecasting, which leads one to interrogate witnesses of a more- 
distant past. 
The only methods employed in Tunisia are the examination of 
tree growth and archaeological studies. The first method bore only 
on a humid locality but, however, one neighboring to the arid 
zone. The Meteorological Service of Tunisia has applied it to the 
study of the climate of Ain Draham since 1736, in examining the 
growth rings of an old oak knocked down in 1905, and in com- 
