Historical (1800s) vegetation in Wisconsin: Mapping forests and the controlling variables

Background/Question/Methods

Understanding how environmental factors and disturbance regimes control vegetation and the scales at which they operate is a perennial goal of ecology.  Historical tree records before significant human activities provide valuable vegetation information. The US Public Land Survey (PLS) data from the US general Land Office contain systematically collected tree data prior to large scale Euro-American settlement in the state of Wisconsin. This data set provides a unique opportunity to examine relationships between tree vegetation and environmental variables before large scale human disturbances. In this study, we first classify the tree vegetation using hierarchical clustering of PLS tree records in Wisconsin. The classification was then assessed by a data-dividing, cross validation technique. Classification and Regression Trees (CART) was then used to examine the relationship between vegetation classes and environmental variables. The variables include soil variables extracted from SSURGO database and climate variables from the PRISM climate dataset.

Results/Conclusions

A total of 19 vegetation classes were obtained for pre-European tree vegetation in Wisconsin. The North was mostly covered by closed canopy forests while the South was mainly savanna, prairie and open forest. Pine forests mainly occurred in the sand plains. The northern hardwood forests are extensively distributed in the central portion of the North with mesic conditions. Oak species are prevalent in the South either as savannas or closed canopy forests in the driftless area. CART model indicate that potential evapo-transpiration (water balance variable) separate the vegetation classes into the southern classes (with oak dominated) and the northern classes (with pine and hardwood species dominated). The southern classes were then further separated by precipitation and temperature variables and then by soil variables into finer oak classes. The northern classes were first separated by silt into pine and hardwood species and then further by soil and precipitation variables. Some classes were not well separated in the CART model, probably due to their subtle differences with other classes both in species composition and environmental conditions. Results of this study provide some references of how environmental variables influence spatial distribution of tree species under minimum human disturbances.