Area-average estimates of monthly P-E (mm) were compiled for
the Ob, Lena, Yenesei and Mackenzie watersheds using the
"aerological" approach.  Daily rawinsonde data for the period
1973-1995 were processed into monthly-mean vertically-integrated
vapor fluxes and precipitable water.  These data, interpolated to
the basin boundaries, were then used to compute the net influx
(convergence) of water vapor into each watershed. Figure 1 shows
a schematic diagram of an air mass over a drainage basin. Net
influx acts in the horizontal direction. P-E acts in the vertical
direction. P-E was determined by adjusting the monthly convergences
by basin-averaged time changes in precipitable water.


Evaluation of the mean seasonal cycles (Figure 2) reveals
that P-E for all watersheds is highest  during late autumn and
early winter.  Except for the Yenesei,  P-E (and the flux
convergence) exhibits a summer minimum.  Summer values are
negative for the Ob, indicating mean vapor flux divergence. 
Since precipitation in these watersheds has a mean summer
maximum,  this illustrates that a large part of summer
precipitation results from within-basin evaporation and
recycling.  Studies by W. Gutowski and Z. Otles (Iowa State
Univ.) using analyzed wind and moisture fields from the 
NCEP/NCAR reanalysis capture this same pattern of flux
convergence.


Over a sufficiently long time period,  P-E should equal
runoff.  Table 1 shows the mean annual P-E and runoff (mm)
averaged over 1973-1995 for the water year October through
September.  For the Mackenzie,  mean P-E and runoff match
closely.  The two values are also within 20% of each other for
the Yenesei.   However, large discrepancies  are found for the Ob
and Lena Basins.  


Figure 3 shows the time series of annual (October-September)
P-E and runoff.  The match between the time series is very poor.
While in part,  this may reflect storage effects, it more
fundamentally illustrates limitations of the aerological
approach.  The P-E estimates are sensitive to the size, shape and
topographic characteristics of the domains relative to the
distribution of the rawinsonde data.  The large interannual
variability in P-E is primarily driven by the summer months, when
water vapor is relatively abundant but winds are light.  Small
errors in winds on the original data and those introduced during
interpolation have pronounced effects on the estimated flux
convergence.  It is possible that better results may be obtained
by estimating flux convergences using analyzed wind and moisture
data from NCEP/NCAR or ERA reanalysis fields.