COMPANY CONFIDENTIAL. DO NOT DISCLOSE.
Unearthing the secrets of our Pleistocene past
Statistical analysis helps paleontologists uncover new truths about Ice Age megafauna
_____________________________________________________________________________________
Challenge: Sift through data from a collection of nearly 5 million specimens to learn how climate events influences the microevolution of species on Earth.
Solution: Use statistical models to find outliers in the data set that need to be remeasured and reclassified. Dynamic data visualization in JMP® statistical discovery software helps paleontologists quickly spot anomalies that, if not corrected, could threaten the integrity of the results.
Results: “The ability to be interactive with your data, explore problem areas in a visualization and label them, helps you look back at the specimen and make sure the data is really correct,” says paleontologist Sherri Gust. “All of these features in JMP are incredibly valuable for this stage of research.”
_____________________________________________________________________________________
For tens of thousands of years, naturally occurring underground asphalt deposits have seeped to the surface, creating temporary tar sinks with the power to entrap any creature that wandered too close. These natural sites have since become a graveyard for now-extinct species whose bones sank deep into the deposit where the heavy oil of the pits preserved them from the dawn of the last glacial period to today.
The scientific value of these deposits was widely recognized by the time excavations began in 1901 at the Rancho La Brea Tar Pits in what is today the center of urban Los Angeles. Since then, archeologists at the site have amassed the world’s largest collection of late Pleistocene asphaltic fossils including impressive examples of now extinct megafauna – mammoths, giant sloths and saber-toothed cats among them. With between 3.5 and 5 million specimens spanning the last 50,000 years, this collection is perhaps one of the most complete resources scientists have to help reconstruct a portrait of not only Ice Age flora and fauna but major geologic and climate events of the period.
Searching the fossil record for insight into climate change
Some of the biggest research questions in the field of paleontology today deal with the microevolutionary consequences of climate change. A result of mutation, selection, gene flow or genetic drift, microevolution is any change in trait frequency appearing in a population over a relatively short period. Scientists now hypothesize that climate change may not only interfere with the abiotic and biotic aspects of selection, it could also have an impact on the magnitude of nonselective factors like genetic variance.
By mapping species physiology data from the fossil record to a timeline of known climate events, scientists hope to see how climate may correlate with the microevolutions they observe within a given population. To do so, they look to collections like that of La Brea, where each specimen can be meticulously measured, radiocarbon dated and catalogued according to a long list of characteristics and identifiers.
Jaw bones from the Tar Pits collectionJaw bones from the Tar Pits collection“To really look at climate change, you have to not only look at changes in species through time, but you also have to differentiate climate-induced change from random change,” explains paleontologist Sherri Gust, Research Associate of the Vertebrate Paleontology and Rancho La Brea Departments of the Natural History Museum of Los Angeles County.
Gust, who is also Founder and Principal Paleontologist at the resource management firm, Cogstone, notes that researchers expect physiological variance within any sample due to factors like age and sex; e.g., a newborn is likely to be much smaller than a full-grown adult. But while it may be relatively easy to account for these factors in studies involving living animal populations (where researchers’ observations in the field help them clearly identify sex and age at death), making these determinations becomes much more difficult in specimens collected from La Brea.
Archeological dig site at La Brea Tar PitsArcheological dig site at La Brea Tar Pits
The natural geologic phenomena that preserved the bones excavated from La Brea, Gust explains, stratified skeletons across the deposit and their DNA degraded over time. It can therefore be an onerous task to determine not only an animal’s sex and age at death, but also the species and time period in which the animal lived. While modern technological innovations today enable Gust and her colleagues to more accurately make these identifications, the added challenge of the La Brea collection is that many of its 5 million individual bones were measured and catalogued nearly a century ago before the advent of this technology.
Putting a myth to rest
The most prolific of the late Pleistocene animals discovered at the La Brea Tar Pits is the dire wolf, Canis dirus, a carnivorous ancestor of the modern gray wolf, Canis lupus, that once roamed a vast swath of land in North America. Now extinct, scientists believe the species met its evolutionary end roughly 9,500 years ago. With more than 4,000 dire wolf individuals represented in the La Brea collections, the species has become a key player in global North American Pleistocene research.
La Brea’s Canis collection, however, contains some of the first specimens to be catalogued by archeologists at the site in the early 20th century. The data logged from each specimen at this time was likely impacted not only by a lack of technological precision, but also by the now debunked idea that the dire wolf was much larger than its modern counterpart. This mischaracterization, Gust explains, may have originated from – and even perpetuated – a mythology surrounding dire wolves: that of a menacing and fearsome giant. Though it’s not uncommon to read that dire wolves are between 8% and 25% larger than modern wolves, she says, this isn’t true: scientists now know that adult dire wolves, like their modern counterparts weighed around 150 lbs.
But the prevelance of such misconceptions led 20th century archeologists to characterize larger skeletons Canis dirus and smaller skeletons Canis lupus. As a result, Gust explains, some specimens at La Brea are likely mislabeled. “The original scientists here at La Brea nearly a hundred years ago were working with very small sample sizes,” she says. “They didn’t have the variation in their data that would have let them be much more clear. And they didn’t have access to a data sample of gray wolves [to compare dire wolf samples to].”
“As a researcher, you must have the right criteria and identify the specimens for yourself so that you know what you’re saying in your research findings is truly correct,” Gust says. But with such an enormous collection and limited time and resources, it is impractical for researchers to remeasure and rerecord every single specimen in the dataset.
That’s why Gust is using a statistical approach in JMP® software to identify outliers. “One of the first things most statisticians do is to look at the distribution of their data,” she explains, pointing to a data visualization with a sloping bell curve. Gust is looking at a mandible length data set, noting outliers in the little peaks and bumps in the fit. When she spots an unexpected peak, she looks first to the data to see if there’s an explanation – perhaps this specimen needs to be remeasured.
Taking the time to verify the quality of her underlying data sample, she says, leads to more accurate results – even if those results may lead only to more questions. “With statistics, we really have the opportunity now to put down in print the criteria for being a Canis dirus versus a Canis lupus. Previously, that really didn’t exist,” she explains.
Tooth and nail
Some of the most valuable information collected from any wolf specimen is found in the jawbone. “Because bone is such a responsive tissue, the more pressure that’s placed on it, the thicker the bone should be,” Gust explains. By taking 3 enamel depth measurements – the second molar to the bottom of the mandible, the fourth premolar and length of the mandible itself – she and her team gain more insight into the age of the animal at the time of its death. She can also look at muscle scars, ridges and crenulations for insight; older animals will have more differentiation while younger animals have smoother bones. All these characteristics help Gust’s team to develop an age sequence and define age staging for each sample.
Another option is radiocarbon dating, a technique that measures the extent of radiocarbon decay to ascertain the length of time that has passed since an animal’s death; the older the sample, the less radioactive isotope there is to be detected. Because the method requires destructive analysis, Gust explains, La Brea technicians use 3D imaging and a 3D printer to create a replica so that the real specimen can be drilled into without losing the physical record.
After reclassifying outliers in the Canis dirus data, Gust then looks at known-sex data sets from organizations like the Yellowstone Wolf Project and Parks Canada. “We can now compare the two [species] and the result is very interesting,” she says, explaining that the minute she plotted the two sets of data together in JMP, she noticed something surprising. 
“I pulled in all the data, looked at it and said ‘wow, why is it like that?’ I did not expect that any Yellowstone specimens would be in the middle of the range for dire wolves. But that is what JMP is all about, right? Discovering the unexpected.”
Running down the rabbit holes with data exploration in JMP
Since first learning about JMP a number of years ago, Gust has used the software to create interactive visualizations where outliers become visually apparent. And with the tool’s dynamic interface, she can interact with the data distributions. “I love JMP,” Gust says. “It’s been a wonderful tool to use in looking at these changes and trying to figure out what’s going on in the data. The interactivity is one of my favorite things – you can say, ‘I’d like to explore that more.’ And you can run down the rabbit holes really easily with JMP, which is a wonderful thing to be able to do.”
One of the features that helps with this exploration, Gust says, is the ability to attach specimen photos and specs to each data point in a visualization – or even use the specimen number in place of a dot. Demonstrating a distribution of mandible data, Gust hovers her cursor over one of the outliers as a photo pops up. “Having these photos is really helpful because if, for example, I’d forgotten to measure a tooth, all I would have to do is call this photo up in [our measurement software] and update the measurement. That's been an important part of our data checking process,” she says.
“The ability to be interactive with your data, explore problem areas in a visualization and label them, helps you look back at the specimen and make sure the data is really correct. All of these features in JMP are incredibly valuable for this stage of research – when you’re trying to just figure out what the variation is. And figure out the source of the variation we’re seeing in these graphs.”
As her research progresses, the possibilities are endless and Gust is exploring what more she can do to make use of the comparative data. For example, she says, data from known-sex individuals from Yellowstone (whose sex can be determined through DNA sampling) can be used to classify the unknowns via a discriminant function. The results might transform scientists’ thinking about how size is – or is not – correlated with sex and age.
