Assemblages of fossil plant remains have been widely used to reconstruct past climatic conditions, usually through the application of methods that involve either finding vegetation analogues on the modern landscape (and using the modern associated climatic values as the basis for an estimate) or using the modern climatic ranges of individual taxa in an assemblage to determine the range of a given climate variable that would allow these plant taxa to live together. Although these approaches are relatively straightforward, it is difficult to assess the uncertainties associated with each approach, particularly in regard to their application to plant macrofossil assemblages. To explore the uncertainty that may arise from inaccuracy and imprecision in climate reconstructions and from ecological considerations we used variants of both approaches to estimate climate from two data sets of modern vegetation assemblages from the southwestern United States: (1) 1752 gridded “virtual plant assemblages” based on plant range maps that provide uniform spatial coverage of the presence or absence of major woody plant taxa across the study area; and (2) 43 modern packrat (Neotoma spp.) midden presence-absence assemblages that are similar to fossil midden assemblages. By comparing observed and estimated climate values, we evaluated the quality of the climate estimates, identified sources of uncertainty, and characterized the nature and magnitude of the effects of these uncertainties on the climate estimates.

Uncertainties in estimating climate from vegetation assemblages arise because any given plant taxon (or assemblage) must have the resiliency to survive a range of climatic variability, and because of the strong intercorrelations among climatic variables in the modern climate data. Additional sources of uncertainty in climate estimates from plant assemblages include: (1) the modern climate and plant distribution data that are selected as the basis for estimation; (2) the particular quantitative approach that is used to estimate climate; (3) the sufficiency of the number of taxa in the analysis for providing an unbiased representation of the vegetation community as it existed for each time period in the analysis; and, (4) the location of the assemblage on the climatic and environmental gradients in the calibration data set for each climate variable under consideration.

We conclude that vegetation assemblages can provide valid and reproducible estimates of climatic variables and that the primary trends and mapped patterns in the observed climate data can be reconstructed from such estimates. However, many factors may affect the quality of an estimate from a given plant assemblage, including aspects of data selection, data adequacy, methodologies, and the location of the assemblage site relative to gradients in the base climate data. It is particularly difficult to accurately estimate extreme values in the observed climate data, because estimated values from either end of an observed climate gradient necessarily “move toward the middle” of the gradient. In addition, the interval chosen to represent modern climate (here we used 1961 to 1990) may have a large impact on the size of the estimated difference between modern and past climate at a given site.