96 THE FUTURE OF ARID LANDS 
stations, one in the most humid region, the others in very dry 
areas. 
From our experience, it seems that the water needs thus calcu- 
lated are too weak for Gabes and especially so for Tozeur. Cor- 
rections would be necessary. One is evident: the water at Tozeur 
is saltier than at Tabarca and the rate of irrigation must therefore 
be increased. Other corrections would perhaps be necessary if 
there were at our disposal more observations on the evapotranspi- 
ration in arid regions. 
Moreover, dry cultures are s pase tle in the arid zone by re- 
ducing evaporation surfaces by plant spacing, or by concen- 
trating on only a part of the land the water which falls on more 
extensive impluvium. 
Discontinuity of Procedures for Establishment of Water Resources 
In the first part of this report, we stressed the variability of 
rainfall. We limited our treatment of the topic to annual rainfall. 
It is now necessary to push the analysis farther by showing the 
discontinuity of the rainfall and its distribution over the year. 
Not only is the rainfall discontinuous, but in general it falls on 
non-saturated surfaces. The effects of rain on the runoff and on 
the supply of underground aquifers are felt only after exhaustion 
of a volume necessary to saturate the soils, or when the rate of 
rainfall exceeds the rate of absorption by the soil. 
Insufficiency of Climatological Information 
In order to study the flow of a great river of the humid zone, 
one can be content with several spaced-out measures, for example, 
once a week, and thus obtain a sufficiently accurate evaluation 
of runoff. The discontinuous nature of the flow in arid zones makes 
much more necessary the continuity of observation. It is very 
dificult to get an observer to maintain a continuous watch. 
When he observes a watercourse for months without seeing any 
flow, his attention is likely to relax, and he may absent himself 
precisely during the several hours when there is something to 
observe. This continuity of observation is even more necessary if 
one wishes to study the silt flow of streams. Available observa- 
tions are of insufficient duration precisely in those regions in 
