averaged value of the mean precipitation for the same reference period. (While processing the data of the Canadian network that was operating in the 1920s, we used the reference period 1921-1940.) These time series provide our first approximation of contemporary changes of North American precipitation. The first experiments using individual gauge corrections (Groisman, 1991) showed that this approximation does not change substantially on a continental scale after the intra-annual variability in gauge errors has been taken into account.

CURRENT CHANGES IN PRECIPITATION AND SNOWFALL
Snowfall Changes During the Last Four Decades

Figure 5 shows the zonally averaged snowfall over Alaska, Canada north of 55°N, and the zones 45°N to 55°N and 35°N to 45°N for the last four decades. The basic statistics of these time series appear in Table 3.

When the spatial correlation function is known, the accuracy of area averaging over these regions using the networks depicted in Figure 3 can be estimated. The method of estimation, which was developed by R.L. Kagan (1979), can be applied to every isotropic two-dimensional field. In North America we used this method only for northern Canada. Since even today the meteorological network in northern Canada is relatively sparse, we had to estimate the representativeness of the network beforehand in order to analyze the changes over this area.

The spatial correlation function for northern Canadian snowfall can be approximated by the relationship r(R) =, where R is the distance; R0 = 710 km, the radius of correlation; and C = 0.84. The value of (1-C)/C, which equals 0.19, is a measure of the error of measurements combined with microscale snowfall variability near the stations (Kagan, 1979). For the network selected, the mean-square error of spatial averaging of annual snowfall over northern Canada was estimated to be 17 percent of the theoretical variance of the area-averaged snowfall time series in the first half of 1950s; during the period 1956-1990 the error was about 10 percent. The number and spatial distribution of snowfall-measuring stations for southern Canada and the United States during the last four decades were adequate, so there was no need to verify the representativeness of the area-mean values.

The accuracy of area averaging of the data in northern Canada is high enough to confirm the strong positive linear trend in annual snowfall revealed in the zone (19 percent over four decades). This trend, in conjunction with the trend in rainfall data (not shown), confirms the systematic increase in annual totals of precipitation during the entire period of instrumental observations.

Snowfall in the zone 45°N to 55°N is closely connected to temperature changes on continental and hemispheric

TABLE 3 Statistical Characteristics of Time Series (1950-1990) of Annual Snowfall and Total Precipitation Zonally Averaged over North America (see Figures 5 and 6)

Annual Snowfall

Mean and Standard Deviation, in cm

Region

Mean

Std. Deviation

Linear Trend ± its Standard Error, in % per Decade

Alaska

190

35

2.7 ± 2.5

Canada, zone 55-70°N

195

17

5.1 ± 0.8

Canada/U.S., zone 45-55°N

260

27

-0.7 ± 1.4

U.S./Canada, zone 35-45°N

140

18

1.8 ± 1.7

Total Precipitation

Mean and Standard Deviation, in mm

Region

Mean

Std. Deviation

Linear Trend ± its Standard Error, in % per Decade

Alaska

510

59

2.7 ± 1.5

Canada, zone 55-70°N

495

33

4.2 ± 0.6

Canada/U.S., zone 45-55°N

940

41

0.7 ± 0.6

Canada/U.S., zone 35-45°N

905

66

1.9 ± 0.9

United States, zone 28-35°N

1175

137

3.4 ± 1.5



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement