Oral Presentation at SVP 2019:
Comparison of ‘Big Data’ uses in paleoecological multi-proxy models for North American mammalian paleoecological interpretations
The use of ‘big data’ has rapidly increased in recent years across paleoecology and paleobiology. ‘Big data’ can provide robust statistical analyses and a potential basis for answering broad scale questions, but may also suffer from the homogenization of important detailed information. The purpose of this research is to determine the viability of using the FaunMap (FM) and the PaleoBiology (PDBD) databases for paleoecological reconstructions using multi-proxy models. Holocene North American mammalian data were downloaded from FM and PBDB. FM data consist only of locality information, while the PBDB also has associated trait data. We test the viability of these datasets by evaluating their ability to differentiate among modern and Holocene biomes. Data from PBDB and FM were assigned to biomes based on their geographic occurrences within established ecoregions. Diet, locomotion, and body mass were assigned to taxa as traits. Principle component (PCA) analyses were run with and without small-bodied (<500g) mammals.
Across all analyses, we observed no clear association of traits with biomes. FM analyses show forest and semi-desert significantly separating from each other, yet no clear trait associations within biomes are observed. Combined AMNH historical and FM data show slight biome separation among forest, grassland, and semi-desert. However, separation is not significant since 95% confidence ellipses overlap biomes. Combined occurrence data cover a greater area of North America, as AMNH and FM data has different geographic distributions. While AMNH data show an even distribution, semi-desert and desert biomes appear better sampled in the FM data. Combining these distributions better represents regional biomes and as such, more distinct results are not surprising. PBDB analyses also show significant separation between forest and semi-desert. Without small-bodied mammal, grassland appears separate from forest and semi-desert, though grassland does not have enough points for a confidence ellipse. In the PBDB data, grassland and semi-desert are severely underrepresented, and any results may be products of sampling resolution. The PBDB does not appear to have the resolving power or geographic distribution needed to model modern regional biomes using our methods. Both FM and PBD cover the full Holocene, including some extinct megafauna and therefore characterize pre-human mammalian communities. To conclude, FM data do a better job at separating biomes, yet the lack of trait association is intriguing and must be further explored.
Master’s Thesis; Oral Presentation at SVP 2018:
A taxon-free, multi-proxy model for making paleontological interpretations of Neogene North American mammalian faunas
Proxies used for interpreting the paleoecology of extinct vertebrate communities are usually based on modern ecosystems, with many developed from Old World ecosystems. However, because no model is completely taxon-free and phylogenetic influences cannot be entirely discounted, these proxies may not be appropriate for paleoecological interpretations of North American ecosystems. Additionally, many proxies based on modern vertebrate communities exclude small-bodied mammals. Here I explore several new paleoecological models based on the frequency of mammalian traits within three ecological categories: locomotion, diet, and body mass. Since these models are intended for interpreting paleoenvironments occupied by Neogene North American mammals, the data used to develop the models are from historical North American faunas. Pre-existing datasets were augmented with locomotion, diet, and body mass information from a variety of sources. Mammalian geographic occurrences were assigned to digital maps of Bailey’s Ecoregions of North America in ESRI ArcMap and ecoregions were combined into broader biomes in an iterative process using preliminary Principle Component Analysis (PCA). Taxa were sorted by biome and two datasets were created, one where the number of individual occurrences were used to weight traits, and one where only a single taxonomic occurrence was used for each biome. Taxonomic analyses were conducted on unweighted taxa both with and without rodents and lagomorphs. PCA was conducted using frequencies of trait classifications per biome for all datasets. Stacked area charts were created to visualize changing trait frequencies among biomes. PCA analyses using unweighted data without the smallest mammals (<500 g) provides the strongest separation of biomes. High frequencies of grazer, cursorial, and size class G traits (<10500 g) are correlated traits in the grassland biome. Size classes C (500-1000 g) and D (1000 – 1500 g) are the second group of correlated traits, plotting in the opposite direction in grassland. High frequencies of arboreal/scansorial, omnivore, and granivore traits make up key indicators for the forest biome. Weighted datasets without small-bodied mammals (<500 g) work well to distinguish among biomes. I conclude that unweighted analyses excluding small-bodied mammals should provide the best separation of biomes and be most appropriate for certain paleoecological applications in North America.
Poster Presentation at SVP 2017:
Vertebrate species richness change from the late Miocene to early Pliocene of Lothagam, Turkana Basin, Kenya
Turkana Basin in Kenya, Africa is home to many discoveries of both hominin and non-hominin fossils. To date, a myriad of isotopic analyses has been conducted to interpret the paleoenvironment of the region. These include stable carbon isotope analysis in fossil tooth enamel and fossil eggshell to determine diet, stable carbon isotope analyses in paleosols to determine the amount of C4 biomass in the paleoecosystem, and oxygen stable isotope analyses in fossil enamel and paleosols to determine precipitation patterns. The purpose of the present study is to ascertain if paleoenvironmental shifts from the late Miocene to early Pliocene are associated with changes in vertebrate (mammals, turtles, and crocodiles) diversity (taxonomic richness) at Lothagam site in the Turkana Basin. The upper Miocene is represented by the Lower and Upper members of the Nawata Formation, and the lower Pliocene is represented by the Apak and Kaiyumung members of the Nachukui Formation. Both formations consist of alternating sandstone and mudstone, representing a perennial fluvial system. The Lower and Upper Nawata members also show repeated volcanic activity. The Apak Member is separated from the Kaiyumung Member by a basalt layer and lacustrine strata that have been excluded due to lack of vertebrate fossils, except for fish. The fossils utilized in this study were all collected from fluvial deposits, but further details about collecting methods and deposits are not available. Thus, possible taphonomic differences among the faunas cannot be ruled out and could conceivably be affecting the analyses of species richness. To determine richness changes, I compiled specimen counts for terrestrial, semi-aquatic and aquatic fossil species for each member, excluding fish and birds. Rarefaction analysis from the Lower and Upper Nawata, Apak, and Kaiyumung Members shows a significant (p<0.05) decrease in species richness from the Apak to Kaiyumung Member. There is also a decrease from the Lower to Upper Nawata Member, but it is not statistically significant (p>0.05). Paleoenvironmental interpretations show a shift from C3 to C4 vegetation and a transition from browsing to grazing ungulates at the Miocene-Pliocene boundary. The general change of vegetation and a shift of ungulate abundance are plausible drivers for the decline in species richness. Further broad scale richness analyses in the Turkana Basin would be required to determine if regional climatic changes were driving the decrease of diversity observed in the early Pliocene or if the pattern was localized to Lothagam.
Click for my 2017 SVP Poster: Hock SVP Poster 2017
Poster Presentation at SVP and Oral Presentation at GSA 2015:
A comprehensive study of key paleoenvironmental changes using major faunal turnovers focusing in the Turkana Basin, Kenya: A development of a new model to determine environmental change.
Lake Turkana in Kenya, Africa has been home to many discoveries that are critical for understanding human evolution. These include a Paranthropus bosei cranium, Homo ergaster type specimen, cranium and full skeleton, Homo rudolfensis cranium, Homo habilis cranium, Paranthropus aethiopicus cranium, Austrolopithecus anamensis mandible, Kenyanthropus platyops cranium, and hominin footprints. However, we have limited understanding of the factors that drove adaptations observed in hominins. To date, efforts to understand the environmental underpinning of these adaptations have been based mainly on isotopic analysis of paleosols, using carbon and strontium isotopes from paleosols and comparing carbon dioxide ratios taken from paleosols to modern day carbon dioxide ratios taken from soil. The environmental information that is extracted from these isotopic analyses is limited. The purpose of this study was to diagnose significant environmental transitions based directly on faunal turnover of aquatic/amphibious and fully terrestrial biotas in the middle-late Miocene to the Recent. By compiling and creating a comprehensive synthesis of previous research in the Turkana Basin, I was able to document faunal turnover and then determine environmental changes. Based on analysis of hippopotamids, equids, suids, elephantids, rhinocerotids, proboscideans, antelopes and primates, I was able to diagnose significant environmental changes at the late Miocene transitioning to early Pliocene. At this time, there was a change from lowland wooded tropical forests with alluvial grassland to a savannah grassland with riparian tropical rainforests. This compilation of environments from modern and recent faunal habitats is supported with previous isotope research, supporting this method of determining environmental change. By comparing the results gathered through this research against environmental changes gathered previously through isotope data, this research would begin to establish a new model of diagnosing environmental changes through fossil records alone.
Click for my 2015 SVP poster! poster_48x721