Abstract
We modeled soil temperatures at 50-cm depth, using 1951–2000 air temperature and precipitation data from 194 National Weather Service stations in Wisconsin and Michigan. The accuracy and bias of the physical model used in this study were validated by comparing its output data to 22,401 actual soil temperature readings taken from sandy soils at thirty-nine forested sites throughout northern Michigan; the model was shown to have almost no temperature bias. Although mean annual air temperatures across the region show no strong spatial or temporal trends over the fifty-year period, at many sites, especially in Wisconsin, wintertime air temperatures have been increasing slightly in recent years. Conversely, mean annual soil temperatures have been decreasing at most sites in the region, some by more than 0.5°C. Likewise, wintertime soil temperatures are also decreasing, especially at sites downwind from the Great Lakes—many of which are in snowbelt locations. Increasing wintertime air temperatures over the past fifty years coincide with (and probably have led to) more variable and thinner snowpacks, lessening their insulating impact and contributing to decreasing wintertime soil temperatures that our model show are occurring in the eastern and northern parts of the region. These findings illustrate the complex response of natural systems to slow atmospheric warming, and draw attention to the potential changes that are occurring in growing season characteristics, phenology, and spring runoff characteristics in the Great Lakes region.
Acknowledgments
We thank Bruce Knapp, John Werlein, and Marty Kroell at the USDA-NRCS for long-standing support of, and help with, the field phase of this project.
Notes
Note: Model output versus actual soil temperatures (at 50-cm depth) was measured at thirty-nine Michigan locations. After Schaetzl, Knapp, and Isard (2005). RMSE=root mean square error; MBE=mean biased error.
aA positive error or bias indicates that the model predicted higher temperatures than actually existed in the field. Negative errors or biases indicate the opposite.