Author(s): John Saretzky
Mentor(s): Edward Medeiros, College of Science (Geology, Atmospheric, and Earth Sciences)
AbstractTraditionally, certain morphological features have led researchers to assume different groups of sauropod dinosaurs shared living space in roughly the same geographic area and niche partitioning occurred via the mechanism of dietary preference. This study used fossil occurrence data from the four major sauropod dinosaur taxa (Diplodocids, Dicraeosaurids, Haplocanthosaurids, and Macronarians) of the Late Jurassic Morrison Formation (Western America). Data was binned and put through a series of statistical methods to determine the geographic correlation between the groups, both overall and in a series of more specific regions. Based on results, it is not possible to say which mechanism of niche partitioning is more correct due to low instances of statistical significance. It was determined that Diplodocid and Macronarian fossils were more highly correlated with each other, as were those of Dicraeosaurids and Haplocanthosaurids.
Audio Transcript“Hey, everybody. My name is John Saretzky, my research mentor is Edward Medeiros, and this is my research project for the OSCAR/USRP Fall 2024 semester at George Mason University.
“The title of this project is Paleogeography Database Survey to Examine Niche Partitioning in Sauropod Dinosaurs (Morrison Formation). Let’s break that down. I’ll start with the Morrison Formation itself. A formation is the basic unit of geologic time, a collection of sediments and other rocks and fossils deposited within a certain limited time and limited geographic area. In this case, what is recorded is now in the Western United States between 156.3 – 146.8 mya (during the late Jurassic period). Paleogeography is easy. That’s just what it sounds like: looking at the orientation of geography that existed a long time ago. As you can see from the two globes, continental drift over geologic time has dramatically moved the position of North America. The formation this study focused on was located in the Western US, which moved just around the side of the globe we can’t see here. This is why every occurrence of a taxa that I pulled off the PaleoBiology database (PBDB) had two sets of coordinates to choose from. The modern and the paleocoordinates which if you put into mapping software tells you that you must have just found a dinosaur that lived at the bottom of the ocean.
“Niche partitioning is a highly interesting topic and is my favorite topic in ecology. To show you what it is, just take a look at the image on the right. At one point, there was likely one single species of finch on the Galapagos Islands. What it ate, I have no idea, but at some point, variations started appearing in the size and shape of the beaks. These variants allowed access to more food sources. This happened over and over until eventually the single species with one food source had split into many species that were eating many different foods. That’s evolution at work and it’s actually one of the first steps in Darwin’s development of his theory.
“At this point, some of you are likely wondering what on Earth a sauropod even is. Well, you actually already know what a sauropod is. It is a lineage of long-necked, long-tailed dinosaurs officially arising in the Jurassic period and becoming extinct with the rest of the dinosaurs in the End-Cretaceous mass extinction. If you look on the right, you see there is some debate as to where it fits in the overall phylogeny of Dinosauria. In my opinion, we have the most evidence that the original phylogeny in the upper right is correct and sauropods were more closely related to the group containing T. rex and Velociraptor than to the one containing Triceratops and Stegosaurus.
“Alright. Back to niche partitioning. We’ve got five different species of warblers all living in different areas of the same tree. Why is that? Why doesn’t the biggest species just push all the others out of the tree and have it all to itself? As I said earlier, the processes of natural selection will make it so that some of the birds will change sufficiently to carve out their own little part of the ecosystem. That’s their niche. In the finches a few slides ago, this meant eating different foods. That is what we see on the right. Two sauropods eating different layers of the local plant growth. It’s very similar to the finches, which isnt surprising, given that they are distant cousins. Niche partitioning has been studied (albeit sparingly because of certain limitations and data biases) in dinosaurs. Most of the focus has been on looking at a particular region and seeing if the organisms of a group partition food sources. In sauropods, this has meant saying things like Taxon A eats leaves from the top of the tree, while Taxon B ate low shrubs.” And now we finally get to it: the point of this study was to look at a finite geographic era (in this case, the Morrison Formation) and determine if sauropod taxa partitioned along geographic lines by living in different areas, similar to the warblers on the left, rather than by eating different food sources. This is significant because, if the sauropods were partitioning geographically, rather than by food source, it supports the idea that niche partitioning in birds may have evolved in theropod dinosaurs (the direct precursors of modern birds) rather than earlier in their lineage with the common ancestors of sauropods and theropods.
“The first thing I did was determine what sauropod taxa were living in the Morrison Formation. That I did by Googling “sauropods of Morrison Formation” and checking out Wikipedia. This led to the construction of a list of 19 genera from 5 different taxonomic groups. All nineteen species were entered into the data download feature on PBDB. One of these taxonomic groups (Rebbiachisaurids, represented by the genus Maraapunisaurus) was removed from the list due to the fact that only one occurrence data point was generated. Remaining Paleolatitude and Paleolongitude values were entered into separate Excel tables. From there, two things happened. The first was the binning of data and creation of histograms showing at which paleolatitude and paleolongitude fossil finds of each group were most likely to have occurred. Histograms were made of both the entire data set for each taxa and several zoomed in” parts of the data set. The second thing was uploading those same data sets (including Zooms) into the Past4 software for a more scientifically valid and substantial correlation. A Kolmogorov-Smirnov test was run to determine the significant of correlation between different taxa.
“Before we get into results, here are some images of the taxa examined in this study. They are not to scale, but a quick glance shows some big differences in morphology between groups. For example, Diplodocids have the longest neck proportional to their body, while it would appear Dicraeosaurs have the shortest. The Diplodocid and Dicraeosaur head morphology is similar (at least from this zoomed out view), but look at the much more robust Macronarian head. These are the exact sort of differences, along with the always-raging debate about neck posture, that lead scientists to posit the hypothesis of niche partitioning via diet, rather than geography.
“Before getting into actually analyzing, I had Excel make me a chart that put all occurrence data for the four taxa under analysis (plus that single occurrence of a Rebbiachisaurid species) in one place, charted as paleolatitude on the x-axis vs paleolongitude on the y. What I got is right here. We see three major zones of clustering here, which I have circled in red. As most of the occurrence data was in the Late Jurassic’s Eastern Hemisphere, that is where we will be doing the rest of our analysis. Those ones over to the West are considered outliers for this study and were likely found in Europe, rather than the Morrison.
This same data was made into several histograms showing the relative frequency of occurrence related to both paleolatitude and paleolongitude values. The binned values were inserted into the program Past4 and run through a Kolmogorov-Smirnov Goodness of Fit Test, which essentially tells you to what degree two different data sets overlap and consequently how likely it is that one data set could occur within the other. A perfect correlation is represented by D=0, or 0% variance between data sets (meaning they are identical). No correlation at all is represented by D=1, or 100% variance between the data sets. For a test result to be statistically significant, the D value must be equal to or below .05. Any value above that can provide data about relatedness but is not considered to be a statistically significant correlation between the two data sets.
“So here we are in the East hemisphere, looking at histograms of occurrence data binned every 10 degrees of latitude. This shows us how likely it is that a sauropod fossil of the four taxa is to be found within each 10 degree increment. Again, Rebbiachisaurid data is one data point and does not play a part in data analysis. However, it is interesting to note that all four taxa have almost the exact same distribution, and given this chart, it is not unreasonable to assume that Rebbiachisaurids have a similar distribution. Another thing to notice is that there is no occurrence between degrees latitude 0 and 9.99 and comparatively little between 10 to 19.99.
With that in mind, I zoomed in” on the data, adjusting to search increments of degrees paleolatitude 2.5 (at most approximately 170 miles, close to the projected sauropod migration range of 200 miles), rather than 10. The resulting chart is right here.
“The same process was repeated for paleolongitude, resulting in these histograms. Once again, there appears to be a high degree of visual correlation between fossil occurrence and all taxa. This time, three Zoomed-in views were prepared, one for each of the three “pulses” in relative abundance. These are circled in red and are the same data as the clusters I circled earlier.
This time, three Zooms were necessary, but Zooms 1 and 2 were only made for the Diplodocid and Macronarian groups, as the others exhibited little to no occurrence data within the first two clusters.
“Here are the collected Kolmogorov-Smirnov test results for all taxa and at all resolutions. There are very few tests that resulted in the D is equal to or less than .05 value we need for statistical significance, but what we have can still tell us about relative abundance within each data bin and also relative correlation between taxa. In the case of paleolatitude, we see that we have two sets of correlation.
Diplodocids and Macronarians are most closely related to each other, and this correlation is equal to that between Dicraeosaurids and Haplocanthosaurids. This relationship holds true when “zoomed in” to only include the Eastern Paleohemisphere. For the full data set of paleolongitude, we see the same relationships play out, only this time, Dicraeosaurid and Haplocanthosaurid fossils are even more closely associated, the D value of .0667 coming very close to statistical significance. For Zoom 1, Diplodocids are actually more closely associated with Dicraeosaurids and Haplocanthosaurids than with Macronarians and Dicraeosaurs/Haplocanthosaurs are in perfect association.
For Zoom 2, we see the highest correlation between Diplodocids and Macronarians (D=.125) and weaker correlations between Diplodocids and Dicraeosaurs/Haplocanthosaurs. Dicraeosaurs are perfectly correlated with Macronarians within this cluster.
For Zoom 3, the association between Diplodocids/Macronarians (D=.15) is similar to that of Zoom 2. This is the cluster in which we have our first statistically significant non-perfect correlations. Dicraeosaurids/Haplocanthosaurids have a D value of .05, as do Macronarians and Dicraeosaurs.
“Due to the low correlation values developed from the data gathered for this project, we cannot say definitively whether or not niche partitioning in the sauropod dinosaurs of the late Jurassic Morrison Formation occurred along the lines of geographic separation or differential diet within the same geographic area. What we can say is that, with the fossil record cited in this study, there is not statistically significant evidence pointing to these taxa living in the exact same geographic area. However, it is important to mention once again that the projected sauropod migration range is 200 miles, and the 2.5 degrees paleolatitude bins are at most 170 miles (at the equator). Herds of these animals would likely have been moving around almost constantly, given the amount of browse they ate on a daily basis. It is possible that their migratory patterns, at what time they died, predation or scavenging, or geologic events such as floods could have affected this analysis.
What we can say definitively is that, given the data analyzed in this study, Diplodocids and Macronarians tend to be more closely associated with each other than either group is associated to other taxa. Similarly, Dicraeosaurids and Haplocanthosaurids are more closely associated with each other than either group is associated to other taxa. This is illustrated here in a chart where linear distance represents the degree of uncorrelation. The farther away two circles are, the more uncorrelated they are. This holds true both for the complete paleolatitude/paleolongitude data sets and the zoomed in” data sets.
To continue the work of this study, a few things could be done. First, add more data points. Second, one could try altering the resolution” of the data. You could repeat the same analysis by genera, rather than by sauropod group. It is conceivable that different genera within the same group had different behaviors that could affect the results. One could also look at different-sized data bins (for example, look at 1 degree changes in paleolatitude, rather than 2.5 degree changes).
“And here are my cited sources and acknowledgements of funding and assistance. Thank you for listening!
“The title of this project is Paleogeography Database Survey to Examine Niche Partitioning in Sauropod Dinosaurs (Morrison Formation). Let’s break that down. I’ll start with the Morrison Formation itself. A formation is the basic unit of geologic time, a collection of sediments and other rocks and fossils deposited within a certain limited time and limited geographic area. In this case, what is recorded is now in the Western United States between 156.3 – 146.8 mya (during the late Jurassic period). Paleogeography is easy. That’s just what it sounds like: looking at the orientation of geography that existed a long time ago. As you can see from the two globes, continental drift over geologic time has dramatically moved the position of North America. The formation this study focused on was located in the Western US, which moved just around the side of the globe we can’t see here. This is why every occurrence of a taxa that I pulled off the PaleoBiology database (PBDB) had two sets of coordinates to choose from. The modern and the paleocoordinates which if you put into mapping software tells you that you must have just found a dinosaur that lived at the bottom of the ocean.
“Niche partitioning is a highly interesting topic and is my favorite topic in ecology. To show you what it is, just take a look at the image on the right. At one point, there was likely one single species of finch on the Galapagos Islands. What it ate, I have no idea, but at some point, variations started appearing in the size and shape of the beaks. These variants allowed access to more food sources. This happened over and over until eventually the single species with one food source had split into many species that were eating many different foods. That’s evolution at work and it’s actually one of the first steps in Darwin’s development of his theory.
“At this point, some of you are likely wondering what on Earth a sauropod even is. Well, you actually already know what a sauropod is. It is a lineage of long-necked, long-tailed dinosaurs officially arising in the Jurassic period and becoming extinct with the rest of the dinosaurs in the End-Cretaceous mass extinction. If you look on the right, you see there is some debate as to where it fits in the overall phylogeny of Dinosauria. In my opinion, we have the most evidence that the original phylogeny in the upper right is correct and sauropods were more closely related to the group containing T. rex and Velociraptor than to the one containing Triceratops and Stegosaurus.
“Alright. Back to niche partitioning. We’ve got five different species of warblers all living in different areas of the same tree. Why is that? Why doesn’t the biggest species just push all the others out of the tree and have it all to itself? As I said earlier, the processes of natural selection will make it so that some of the birds will change sufficiently to carve out their own little part of the ecosystem. That’s their niche. In the finches a few slides ago, this meant eating different foods. That is what we see on the right. Two sauropods eating different layers of the local plant growth. It’s very similar to the finches, which isnt surprising, given that they are distant cousins. Niche partitioning has been studied (albeit sparingly because of certain limitations and data biases) in dinosaurs. Most of the focus has been on looking at a particular region and seeing if the organisms of a group partition food sources. In sauropods, this has meant saying things like Taxon A eats leaves from the top of the tree, while Taxon B ate low shrubs.” And now we finally get to it: the point of this study was to look at a finite geographic era (in this case, the Morrison Formation) and determine if sauropod taxa partitioned along geographic lines by living in different areas, similar to the warblers on the left, rather than by eating different food sources. This is significant because, if the sauropods were partitioning geographically, rather than by food source, it supports the idea that niche partitioning in birds may have evolved in theropod dinosaurs (the direct precursors of modern birds) rather than earlier in their lineage with the common ancestors of sauropods and theropods.
“The first thing I did was determine what sauropod taxa were living in the Morrison Formation. That I did by Googling “sauropods of Morrison Formation” and checking out Wikipedia. This led to the construction of a list of 19 genera from 5 different taxonomic groups. All nineteen species were entered into the data download feature on PBDB. One of these taxonomic groups (Rebbiachisaurids, represented by the genus Maraapunisaurus) was removed from the list due to the fact that only one occurrence data point was generated. Remaining Paleolatitude and Paleolongitude values were entered into separate Excel tables. From there, two things happened. The first was the binning of data and creation of histograms showing at which paleolatitude and paleolongitude fossil finds of each group were most likely to have occurred. Histograms were made of both the entire data set for each taxa and several zoomed in” parts of the data set. The second thing was uploading those same data sets (including Zooms) into the Past4 software for a more scientifically valid and substantial correlation. A Kolmogorov-Smirnov test was run to determine the significant of correlation between different taxa.
“Before we get into results, here are some images of the taxa examined in this study. They are not to scale, but a quick glance shows some big differences in morphology between groups. For example, Diplodocids have the longest neck proportional to their body, while it would appear Dicraeosaurs have the shortest. The Diplodocid and Dicraeosaur head morphology is similar (at least from this zoomed out view), but look at the much more robust Macronarian head. These are the exact sort of differences, along with the always-raging debate about neck posture, that lead scientists to posit the hypothesis of niche partitioning via diet, rather than geography.
“Before getting into actually analyzing, I had Excel make me a chart that put all occurrence data for the four taxa under analysis (plus that single occurrence of a Rebbiachisaurid species) in one place, charted as paleolatitude on the x-axis vs paleolongitude on the y. What I got is right here. We see three major zones of clustering here, which I have circled in red. As most of the occurrence data was in the Late Jurassic’s Eastern Hemisphere, that is where we will be doing the rest of our analysis. Those ones over to the West are considered outliers for this study and were likely found in Europe, rather than the Morrison.
This same data was made into several histograms showing the relative frequency of occurrence related to both paleolatitude and paleolongitude values. The binned values were inserted into the program Past4 and run through a Kolmogorov-Smirnov Goodness of Fit Test, which essentially tells you to what degree two different data sets overlap and consequently how likely it is that one data set could occur within the other. A perfect correlation is represented by D=0, or 0% variance between data sets (meaning they are identical). No correlation at all is represented by D=1, or 100% variance between the data sets. For a test result to be statistically significant, the D value must be equal to or below .05. Any value above that can provide data about relatedness but is not considered to be a statistically significant correlation between the two data sets.
“So here we are in the East hemisphere, looking at histograms of occurrence data binned every 10 degrees of latitude. This shows us how likely it is that a sauropod fossil of the four taxa is to be found within each 10 degree increment. Again, Rebbiachisaurid data is one data point and does not play a part in data analysis. However, it is interesting to note that all four taxa have almost the exact same distribution, and given this chart, it is not unreasonable to assume that Rebbiachisaurids have a similar distribution. Another thing to notice is that there is no occurrence between degrees latitude 0 and 9.99 and comparatively little between 10 to 19.99.
With that in mind, I zoomed in” on the data, adjusting to search increments of degrees paleolatitude 2.5 (at most approximately 170 miles, close to the projected sauropod migration range of 200 miles), rather than 10. The resulting chart is right here.
“The same process was repeated for paleolongitude, resulting in these histograms. Once again, there appears to be a high degree of visual correlation between fossil occurrence and all taxa. This time, three Zoomed-in views were prepared, one for each of the three “pulses” in relative abundance. These are circled in red and are the same data as the clusters I circled earlier.
This time, three Zooms were necessary, but Zooms 1 and 2 were only made for the Diplodocid and Macronarian groups, as the others exhibited little to no occurrence data within the first two clusters.
“Here are the collected Kolmogorov-Smirnov test results for all taxa and at all resolutions. There are very few tests that resulted in the D is equal to or less than .05 value we need for statistical significance, but what we have can still tell us about relative abundance within each data bin and also relative correlation between taxa. In the case of paleolatitude, we see that we have two sets of correlation.
Diplodocids and Macronarians are most closely related to each other, and this correlation is equal to that between Dicraeosaurids and Haplocanthosaurids. This relationship holds true when “zoomed in” to only include the Eastern Paleohemisphere. For the full data set of paleolongitude, we see the same relationships play out, only this time, Dicraeosaurid and Haplocanthosaurid fossils are even more closely associated, the D value of .0667 coming very close to statistical significance. For Zoom 1, Diplodocids are actually more closely associated with Dicraeosaurids and Haplocanthosaurids than with Macronarians and Dicraeosaurs/Haplocanthosaurs are in perfect association.
For Zoom 2, we see the highest correlation between Diplodocids and Macronarians (D=.125) and weaker correlations between Diplodocids and Dicraeosaurs/Haplocanthosaurs. Dicraeosaurs are perfectly correlated with Macronarians within this cluster.
For Zoom 3, the association between Diplodocids/Macronarians (D=.15) is similar to that of Zoom 2. This is the cluster in which we have our first statistically significant non-perfect correlations. Dicraeosaurids/Haplocanthosaurids have a D value of .05, as do Macronarians and Dicraeosaurs.
“Due to the low correlation values developed from the data gathered for this project, we cannot say definitively whether or not niche partitioning in the sauropod dinosaurs of the late Jurassic Morrison Formation occurred along the lines of geographic separation or differential diet within the same geographic area. What we can say is that, with the fossil record cited in this study, there is not statistically significant evidence pointing to these taxa living in the exact same geographic area. However, it is important to mention once again that the projected sauropod migration range is 200 miles, and the 2.5 degrees paleolatitude bins are at most 170 miles (at the equator). Herds of these animals would likely have been moving around almost constantly, given the amount of browse they ate on a daily basis. It is possible that their migratory patterns, at what time they died, predation or scavenging, or geologic events such as floods could have affected this analysis.
What we can say definitively is that, given the data analyzed in this study, Diplodocids and Macronarians tend to be more closely associated with each other than either group is associated to other taxa. Similarly, Dicraeosaurids and Haplocanthosaurids are more closely associated with each other than either group is associated to other taxa. This is illustrated here in a chart where linear distance represents the degree of uncorrelation. The farther away two circles are, the more uncorrelated they are. This holds true both for the complete paleolatitude/paleolongitude data sets and the zoomed in” data sets.
To continue the work of this study, a few things could be done. First, add more data points. Second, one could try altering the resolution” of the data. You could repeat the same analysis by genera, rather than by sauropod group. It is conceivable that different genera within the same group had different behaviors that could affect the results. One could also look at different-sized data bins (for example, look at 1 degree changes in paleolatitude, rather than 2.5 degree changes).
“And here are my cited sources and acknowledgements of funding and assistance. Thank you for listening!
One reply on “Paleogeography Database Survey to Examine Niche Partitioning in Sauropod Dinosaurs (Morrison Formation)”
Nice work. Would more detailed anatomical examination of the heads give more information about diet? Would you look in finer detail at the ones that are closely correlated?