Browsing by Author "Mechenich, Michael"

Ecometric analysis allows one to identify relationships between traits measured in organisms and conditions measured in the organisms' local environment. In developing an ecometric model, one selects phenotypic features of organisms which potentially are adaptations to local environment, aggregates measures of these features by organismal community, and quantitatively relates these to measures of the environment via statistical modeling, such that organismal traits may be used to predict environmental conditions. Once established, these models may be applied to fossil assemblages, to reconstruct local paleoenvironment using traits preserved in the fossil record. In this ecometrics case study, we addressed a number of related research questions: what bioclimatic and threshold measures of temperature and/or precipitation are most closely correlated with large, herbivorous mammal communities' mean hypsodonty (HYP) in sub-Saharan Africa? Do these correlations differ at differing spatial scales; specifically, do mammal communities in Kenya's national parks and reserves together relate differently to local environment than do communities in the Afrotropics ecoregions? Finally, what do results obtained by ecometric analysis suggest concerning organismal evolution and dispersal in sub-Saharan Africa? In pursuing answers to these questions, we also pursued the case study's primary objective: we developed and implemented a set of best practices for ecometric analysis, based on an assessment of the sensitivity of ecometric models to model-building decisions, and to uncertainty in source datasets. Ecometric analyses frequently make use of raster datasets - continuous coverages of temperature, precipitation, productivity, and other environmental properties - in characterizing local environment at study localities. Thus we primarily asked: are ecometric results dependent on raster resolution or resampling algorithm? Results of these sensitivity analyses are encouraging. Using the recommended mean resampling algorithm, change in linear regression equations with raster resolution is predictable, and not significant until very low resolutions, in which the average raster cell area is an order of magnitude or more greater than the average locality area. Moreover, using the recommended area-weighted averaging in calculating environmental observations at localities ameliorates this predicable trend in derived regressions. In working 'behind the scenes' addressing these methodological questions, we enable more informed interpretation of ecometric results, and allow future researchers to proceed to the real business of ecometrics with a vetted set of analytical methods. Moreover, in promoting better understanding of ecometric results, we promote better understanding of the complex relationships between organisms and environment, essential to understanding the biosphere's past and present, and to protecting it in the future.