UC Berkeley

This dissertation focuses on the spatial distribution of five climatically sensitive plant taxa in central and northwestern Mexico during three time periods: the present, the mid-Holocene and the Last Glacial Maximum (LGM). The results presented here are rooted in recent methods of bioclimatic envelope modeling that incorporate high-resolution paleoclimate model data, GIS, online herbaria data, and cartography. Although the paleoclimate model data and computer methods used in this work are crucial to its outcome, the validation of the model results for the mid-Holocene and LGM rests largely on paleoecological evidence from Mexican fossil pollen sites. In this dissertation, I have devoted much of the discussion to how well the modeled results agree with the paleoecological records and interpretations.

The majority of fossil pollen sites spanning both the mid-Holocene and Last Glacial Maximum occurs in the northwest and central regions of Mexico. The concentration of these sites led to the narrowing of what spatial distribution to consider in the modeling efforts. In both central and northwestern Mexico, there are plant taxa that appear today whose presence has varied over time as environmental conditions fluctuated. The five plant taxa modeled here are Abies, Artemisia, Liquidambar, Picea, and Taxodium, all significant taxa in the pollen record as indicators of temperature and moisture change. In the cases of Abies, Picea and Artemisia, multiple species occur throughout Mexico, which complicates the construction of a climate envelope. In these instances, only one species was modeled in an effort to minimize conflation in the results, which may have happened if multiple species were considered.

The first segment of the work focused on constructing and validating one model for each taxon in the present. The most realistic models from the effort were those for Abies religiosa, Picea chihuahuana and Liquidambar styraciflua, each having very narrow climatic requirements, and occurring in definable biogeographic regions of Mexico. The success also speaks to the appropriateness of species choice, the accuracy of the climate model data as well as the performance of the bioclimatic envelope model. The two models that performed less desirably were those of Taxodium mucronatum and Artemisia ludoviciana. The broad distribution of Artemisia presented challenges in pinpointing an appropriate set of climate parameters for the envelope construction. T. mucronatum is dependent upon flooded riparian pathways for reproduction, an aspect of the environment that was not included in the model construction. Since only bioclimatic parameters were included, the resulting model reflects a clear lack of agreement between the predicted and actual distribution. These results illustrate the importance of selecting an appropriate species for envelope modeling methods.

The second segment of the work was that of projecting the present-day models into the past. Predictably, the three accurate present-day models also performed well when projected into past climate data while the other two models did not agree well with paleoproxy data. However, Picea's mid-Holocene model did generate surprising results by appearing in the central Mexican region. Pollen evidence of Picea in central Mexico during the mid-Holocene is extremely rare and unlikely. Also, some disagreement did appear between the fossil record and the modeled distribution for Liquidambar. According to the fossil record, Liquidambar did appear in the Mexican Basin during the mid-Holocene, but the model does not reflect that finding. The reasons for the disagreement may lie in the variable selection for the envelope, the climate model data, or more likely, the inability of this broad scale technique to model local conditions that may have influenced Liquidambar's distribution.

Finally, the last portion of the work was an experiment comparing species distribution models produced using different climate models modified using EOF analysis. The goal was to determine how, if at all, the ecological models differed from one another. The results indicate some effect, though not particularly significant ones at the broad scale of this work.

In summary, the main finding of the research is that modern bioclimatic modeling techniques with high-resolution climate data are useful in generating paleoecological distributions, but with the following constraints, 1) an appropriate taxon with an easily-defined climatic envelope is modeled, 2) researchers complete a thorough comparison between the model results and palynological records to confirm or refute a models' validity.