Using modern and presettlement pollen analogs to develop transition matrices for northern Wisconsin forests

Background/Question/Methods

To understand how the dynamics of variation in plant communities have responded to past climate changes paleoecologists must devise ways to translate changes in pollen assemblages to changes in plant communities.  Fossil pollen data can be used to interpret vegetation in both typological and continuous ways.  A typological interpretation of pollen records can be used to test models predicting changes in transition probabilities.  We developed a method for vegetation interpretations of pollen assemblages for northern Wisconsin from a library of 77 modern and 65 pre-European analogs.  Each potential analog was matched with vegetation data: LANDSAT classifications for the modern vegetation, and General Land Office (GLO) Survey data for the pre-European settlement period.  Species relative abundance (modern) or relative basal area (pre-European) were summarized within 5km of each site.  Cluster analysis was used to classify vegetation types.  A decision-analysis framework was then applied to pollen assemblages to match them with vegetation types that produce similar pollen assemblages.  An alternate method based on cluster analysis and nonmetric multidimensional scaling of pollen assemblages gave a second classification to compare with the decision analysis interpretation.  Transition probabilities were calculated from a set of 13 late-Holocene pollen records from the northwestern Wisconsin sand plain

Results/Conclusions

Ten modern and 10 pre-European vegetation types were identified based on cluster analysis of LANDSAT and GLO data and named using the most abundant indicator species.  The decision analysis tree was developed using ratios of the pollen types of the most abundant indicator taxa. A transition probability matrix was then calculated for 13 pollen records with century-scale resolution.  564 transitions were observed, with some vegetation types having a 30-50% probability of remaining the same from sample to sample (aspen-birch-oak, jack pine, pine-oak, jack pine oak/herbs, and red pine-birch), and others more likely to transition to other vegetation types (e.g., jack pine-herbs to jack pine-shrubs or jack pine).  Two major clusters of vegetation types were resilient over time, one consisting of jack pine and red pine vegetation types and the other dominated by oak and birch.  Transitions between these two major vegetation clusters were usually through white pine-oak or aspen-birch vegetation types, suggesting an interaction with fire or moisture availability.  Changes in the probabilities of transitions with changes in past climate and disturbance give insight into how the natural variability of vegetation may respond to future climate changes and provide a framework for incorporation of long-term vegetation observations into models.