obvious sources of inhomogeneities. The data are therefore free of the large-scale inhomogeneities shown in Figure 2, but other inhomogeneities may still exist.
The Canadian network, which seems quite large in Figure 3b, declined rapidly in size during the first half of the century (see Table 1). In southern Canada it contained only 190 stations by 1940, and 26 stations in 1900. As a consequence, we selected a special network of Canadian stations that were operating in the 1920s (Figure 3c) to use in to studying Canadian precipitation prior to World War II. The area-averaged precipitation time series based on this network were combined with the corresponding time series developed from the main network shown in Figure 3b.
The precipitation data for the 593 contiguous U.S. stations and the snowfall data were used without adjustments.