Annual snowfall over North America during the last 40 years. The applied scale-correction factors for network elevation are: for Alaska—1.23; for Canada, Zone 55-70°N—1.14, Zone 45-55°N—1.19, and Zone 35-45°N—1.47.

important scale correction that was not performed in the work of Karl et al. (1993a). Bogdanova (1977) showed that in mountains the mean long-term ratio of solid to total annual precipitation rises at a mean rate of 27 percent per 1000 m. We used this to adjust the snowfall time series for network elevation. Without this adjustment, the biases in the regionally averaged snowfall values would be very large. Table 2 provides the mean elevation of the network depicted in Figure 3b, together with true mean surface elevation for large regions of the United States and Canada (the mean elevation of the networks depicted in Figures 3a and 3b scarcely differ). From this table it is clear that with our network it would be possible to estimate occurrence of

TABLE 2 Mean Elevations of Meteorological Stations in the United States (Historical Climatology Network) and Canada Versus Real Surface Elevations*


Elevation, m




Western United States**



Eastern United States






Alaska west of 141°W



* Real surface elevations were estimated from NGDC (1988) data with 5-degree resolution

** "Western" and "Eastern" here refer to west or east of a line drawn from New Orleans to Detroit

traces of snowfall only in the valleys and on the plains of western United States, and completely miss its mountainous portion. So our snowfall adjustment for each region was evaluated in the following way:

  1. The mean long-term ratio of solid to total annual precipitation (rat1) was estimated for a plane with the elevation of the network depicted in Figure 3a.

  2. This ratio was increased using a coefficient of 27 percent km-1 to obtain its value (rat2) for a plane with the elevation corresponding to the true surface elevation of the region.

  3. The value of the estimated area-averaged snowfall was then multiplied by an adjustment scale coefficient equal to rat2 divided by rat1.

These coefficients were used for every graph in Figure 5. They are, however, conservative estimates of the adjustment needed if indeed the total precipitation area-averaged over the whole region depends on true surface elevation rather than on the elevation of the network used ( Karl et al., 1993a).

In the following section, area-averaging by the polygon method (Kagan, 1979; Thiessen, 1911) and scale correction (Groisman et al., 1991) were used in to develop regional, unbiased century-scale annual precipitation time series for the contiguous United States and southern Canada, together with 40-year time series of snowfall and total precipitation over all of North America. So that missing data would not affect the results, we averaged the anomalies from the reference period 1971-1990 and then reconstructed the area-averaged precipitation, adding the corresponding area-

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