Some data reduction problems in meteorology solved by means of graph theoretical methods

It is observed that some data reduction problems in meteorology may
be solved by graph theoretical methods.
The solution of a problem in a graph theoretical context corresponds
to the solution of the original data reduction problem.
Three such problems come under review: a plotting problem, a super-
observation problem and a profile problem.
All these problems being optimization problems admit a solution but
computational algorithms generating these solutions are known for the
plotting and profile problem only. As to the super-observation problem an algorithm is proposed which is believed to generate solutions close to the
best possible solution. Super-observations are studied using the
clustering observations provided by AIDS/ASDAR during three proximate
flights.
With respect to the profile problem some applications are described
using rawinsonde and aircraft meteorological observations.
The method proposed in this paper leads to a 30% reduction in the
number of significant levels compared to the WMO method. Similar results
for the plotting and super-observation problem are obtained.