We conclude with several remarks about the problem of agriculture risk when a
considered region contains local sub-regions with different local climates. As
a consequence, local crop yields will be different. On the other hand, a
regional market is a typical averaging operator, which averages local
variations of crop production. One of the fundamental (macroscopic) variables
(for the market) is the total amount of crop production. If 
is the area
of the i-th locality (
) and 
is the specific crop yield
(for instance, in tons per hectare) then
is the total amount of regional crop production. Let
, where 
is a stochastic component for
crop yield. Then, if 
, in accordance with the Central Limit Theorem
[17] the total crop production y can be considered as a normally
distributed value with the mean 
and the variance
, where
is the covariance matrix for the local crop yields.
The crucial assumption is that there is a minimal critical value of
. Note that for the market, overproduction as well as
underproduction is dangerous, therefore also the upper critical limit for y
may exist. Here we restrict ourselves to the case of the lower limit, so that
only the event 
is considered as an agricultural
disaster.
In the way sketched above we can scale down our problem to that of local crop
production. It is obviously that the matrix depends on the
covariance matrices for climatic parameters, that is, on the statistical traits
of the regional climate. Any weakening of the correlation between local
climates (as a consequence of the general unsteadiness of the mesoclimate; this
is one of the probable consequences of climate change) would reduce the
regional risk (?!). Let us recall the well-known probability paradox: the
reliability of a system decreases with the reduction of its
diversity[17]. Certainly, our conclusion is correct if the market
scale is close to the scale of the mesoclimate. Scaling up (for markets) tends
to further decrease the regional risk, while scaling down results in the rise
of risk. Formally, the problem described is similar to the problem of river
navigation (the main problem in USA in the course of the ``hot summer'' of
1988), that is, the problem of critical water levels for large rivers. The
water level x is an additive function (functional) of multiple localities
which make up the watershed. On the one hand, the river is an averaging
operator for the local dynamic elements; on the other hand, the mesoclimate
combines all these elements in a statistical way.