Effects of Climate on Numbers of Northern Prairie Wetlands

Results: Models

For parkland, the best model explained 65% of the variation in percentage of basins holding water:

% wet basins_p = 17.12 + 0.30(basin1_p) + 0.16(yrpc_p) - 3.47(sprtma_p) - 3.07(falltma_p),

where basin1 is the percentage of wet basins counted the previous May, yrpc is total precipitation for the year, sprtma is spring (March through May) mean maximum temperature, and falltma is fall (September through November) mean maximum temperature. The best model for Canadian grassland explained 63% of the variation in percent wet basins:

% wet basins_gc = 61.34 + 0.19(basin1_gc) + 0.15(yrpc_gc) + 0.15(fallpc1_gc)
- 5.54(daprt_gc) - 1.34(falltma_gc),

where fallpc1 is fall precipitation during the previous year and daprt is the difference between mean minimum and mean maximum April temperatures. The United States grassland model also explained 63% of the variation in percent wet basins and contained the same variables, but with different coefficients:

% wet basins_gus = -4.42 + 0.098(basin1_gus) + 0.042(yrpc_gus) + 0.040(fallpc1_gus)
- 0.0042(daprt_gus) - 0.0021(falltma_gus).

No lags greater than one year entered into any model. Moisture indices derived from combinations of climate variables, although explaining substantial amounts of variation on their own, produced multicollinearity when combined with other variables. Because as much variation could be explained by other, more easily interpretable variables, the moisture indices were excluded from these models.

Previous Section -- Methods
Return to Contents
Next Section -- Results: Testing and Validation