The relationship between isotopic input and isotopic uptake in human tissues is mediated by a plethora of cultural and environmental factors, among them gender, political organization, social status, climate, ecological and geological context, and resource accessibility (Tung & Knudson, 2018; Britton, 2017; Bogaard & Outram, 2013). Despite its complexity, investigation of broad, regional patterning of dietary isotopic ratios spanning multiple cultural and temporal contexts is still infrequent, though this type of isotopic analysis has the potential to illuminate trends that are environmentally influenced and therefore geospatially contingent. This poster utilizes geospatial and statistical analyses to identify clusters and outliers within the δ15N dataset while also testing the validity long-established assumptions surrounding environmental δ15N value distribution in relation to coastal access and elevation. It also serves as a first attempt at mapping δ15N value distribution onto patterns of various ecological zones and attributes in Peru to better illustrate the impact of environment on nitrogen uptake. Overall, the findings support the hypothesis that environmental patterns do play a significant role in nitrogen uptake in human populations, not just through elevation and coastal access, but through ecological attributes such as ground coverage type, bioclimate, and phenological type. Future researchers should strive to expand and balance existing datasets as this is a topic that is worth exploring further in Peru, and possibly in other global regions with robust and well-distributed isotopic datasets.
12 thoughts on “Spatial Visualization of Environmental Factors Contributing to Delta 15N Isotope Uptake Patterns in Pre-Columbian Peruvian Populations”
Comments are closed.
Hi All!
Welcome to this poster! I wanted to explore ways that we could map larger-scale patterns of isotope uptake and diet n bioarchaeological contexts especially those that were environmentally and spatially contingent as opposed to socially produced or motivated. I’m interested in the ways that environmental attributes of a region might be predictors for nitrogen uptake levels. I believe this is critical to developing a more balanced and nuanced understanding of isotopic analysis, which will also help us to better answer questions about diet and nutrition that are socially and culturally derived. If you have any questions/comments/feedback, pleas leave them in the comments below!
Nicely done, Sylvia! It’s good to see those high stable nitrogen isotope values along the coast– that makes sense, and it’s good to see that pattern so clearly (though the range is remarkable). I like the Figure 1, showing the high-low outliers– do you think this might be related to chronological affiliation (same place, but high-low range because of samples coming from really distinct time periods)? How might temporal affiliation help to decipher some of these patterns in the stable nitrogen isotope values? I enoyed seeing the data pulled together and visually displayed and analyzed!
Hi Tiffiny!
I absolutely think that the outliers are related to chronological/cultural affiliation. Huari is MH-LIP and Vinchos is Inka. Vinchos also includes samples from other tissues besides bone, which likely impacted the average delta 15N value. In future analyses I intend to more aggressively weed out data from other tissues, as you recommended! What I loved about doing this project is the cluster analysis illuminated BOTH environmentally contingent clustering (for example, high isotopic Nitrogen values on the coast), while also illustrating outliers, which could be investigated in greater detail to better understand what cultural/temporal influences may have caused these sites to stand out. If a dataset can be built that allows for us to conduct temporally-specific cluster analyses, I would love to attempt that, but in order for that to work we would need a minimum of 40-50 sites with Nitrogen data per time period, which is a goal the community of researchers conducting stable isotope analyses in Peru will need to build towards!
Hi Sylvia. It is really cool to see stable isotope research presented spatially like this, and to see all of the “rules” we learn about stable isotopes displayed in such a way. I had a question regarding one of your conclusions, in which you found (surprisingly) that some of the coastal sites were not represented as high-high clusters, despite the likely access to marine foods, which boost N values, and also the high N outliers in the highlands. Do you think if you added data to this analysis such as known trade routes (which could reference trading of different food resources), that these would account for some of the results, or do you suspect other reasons? Thank you for sharing this!
Hi Alyssa!
I do think that adding in trade routes might provide additional lines of consideration to explain the lack of clustering in some areas, BUT I think in the case of the data set I was working with, a lot of the lack of clustering was produced by unequal representation of certain geographic vicinities in the dataset as a whole. A lot of stable isotope research is conducted at a local/small-regional level, and large, regional metadata-type analyses are not at the forefront of people’s minds as they conduct their individual research at the site/project scale. As the stable isotope research community continues to grow and build a body of research in Peru, I think its important to keep in mind that in order for metadata analyses to work most efficiently, the distribution of data needs to be as even and comprehensive as possible. There are definitely places in need of additional isotopic data points more than others and this need is something that could help to determine where future research is conducted!
Hi Sylvia, this is such a great poster! I love all the little details, especially the QR code to save space and the multi-colored background to keep the eye engaged. Why do you suppose that there has been so few regional investigations of dietary isotope patterning? Is it due to a lack of samples or is there a methodological component as well? Also, how are you results similar or different from Dr. Tung and Dr. Dillehay’s recent paper on Huaca Prieta and Paredones?
Hi Sam! Thank you so much! In the past few years, I’ve fully converted to Adobe InDesign for poster construction purposes and I haven’t once regretted the decision! Would highly recommend to anyone frustrated with PPTs poster capabilities. As for the lack of larger-scale regional investigations, I think its a combination of researchers being focused in on answering more local questions (for good reason – there’s a lot that needs answering!) with stable isotope analysis, combined with the fact that it takes time to build up a larger regional database with a solid distribution that can be used for metadata analysis. Even this dataset, which is impressive, was not ideally distributed. As for the Dillehay and Tung paper, I’m not sure which one you’re referring to, but if you email it to me, or give me the year so that I can find it, I’d love to chat about it!
Very well-presented–good going! I think this is a good start to a journal article, and have lots of ideas for development. One thing that jumps out at me is the scatter plot. I think there are two distinct linear correlations from 0-1500 m and from 1500 and up. There is a good rationale to run separate OLS regressions for those two groups, both in the structure of the data, and in the structure of the regional geography, as 1500 m is roughly a separator between more coastal and more highland ecozones and lifeways. There is just not as much human occupation in that mid-altitude belt on the western slopes, as valleys (which are like desert oases) tend to be steeply incised at those elevations, while the open up below (coastal valleys) and above (highland valleys). There would likely be strong and distinct betas for those two regressions. That would be guts of the punchline of the paper, together with the Local Anselin Moran’s I. The ecozone data looks inconclusive, but it is good that you explored that (and logistical regressions could be run on those to quantify the weak/non-existent correlations). I wonder if you have any thoughts on the high and low extremes among the coastal sites? It’s also clear that there is some heteroskedasticity in the coastal data (the lower the altitude, the more variance in the data). What do you think accounts for that? Good going.
That was me, Steve, not sure how/why I’m logged in as James!
Ha! How did that happen???
Actual James here… though I don’t disagree with anything the above Steve-James said.
Excellent work! I hope this turns into something cool for you in the future! It’s great to see the expected trends in action. The variation of δ15N with environmental variables, I think, is especially interesting and adds nuance to the standard coastal/land resource model we would generally think of. I’m super intrigued by your map A where you seem to have some Low-Lows in the highlands (to be expected) but then some High outliers there too. Any guesses what that’s about?
I do think you’re limited somewhat (as archaeologists always are) by:
a) the small sample size available to you and,
b) the opportunistic sampling strategy that leads to very small regions with an extremely high-density sampling and large gaps of nothing in between.
For example, it looks like near Trujillo, or Pisco, you only have samples from one site. This means it is impossible for Local Moran’s I to find a cluster or outlier because it really only has one data point to work with. There’s nothing for it to compare with anywhere even remotely close by, so there’s no way it could find a significant local cluster or outlier. This is one reason Luc Anselin says not to use Local Moran’s I on small datasets, as it becomes unreliable. (Totally fine for this project, but something to be addressed should you want to publish)
As a corollary, I also suspect that your Local Moran’s I values are going to be highly dependent on how you model spatial relationships. So if you use a fixed distance band vs an inverse weighted distance, or vary the distance/weights on those models, you will l likely end up with substantially different results. And choosing a reasonable model for spatial relationships is *hard,* especially given the data distribution.
James,
I think the high-low outliers inland are a combination of cultural/temporal conditions that are irregular – Huari, for example was a focal point of Wari cultural activity. It would be interesting to look into cultural specificities that may have produced the unusually high N values there. At the other site, we may be looking at an outlier caused by difference in tissue-type of the samples. The article that the Vinchos data came from utilized hair and soft tissue samples as well as bone samples. I discussed this with Tiffiny and it would probably be good to redo the analysis taking that into consideration and see if Vinchos remains an outlier.
As for your critiques of dataset size and distribution, I could not agree more. One goal if I continue working with this is to keep on expanding this dataset so that it is more representative and also so that maybe I can better control for distribution (in the current conditions I had to use every site I had because the dataset was relatively small overall). If I had more sites, I could have omitted some (randomly, to avoid any bias) in areas that were overrepresented. Basically, I am well aware of the limitations of this, but am optimistic that as more research is done, this issue can be improved upon. I would even argue that stable isotope researchers in Peru as a whole could use this information on areas that are unequally represented to inform location/region of future research, hopefully creating conditions for metadata-type analysis that is more fruitful and accurate in the long run.
Dr. Wernke,
I agree with the 1500m separation and would like to try to do separate regressions for those two groupings. I would also love to build this dataset so that low-altitude and high-altitude sets were substantial enough that they could be considered separately with respect to spatial analysis. I think overall, I’d want to look into expanding the dataset and balancing the distribution of data points better across the region (as James mentioned in his comment below). I agree that the ecozone data is inconclusive, but I do think it is interesting to consider the eco-types of coastal regions when looking into answering questions of heteroskedasticity, like you raise here. The low-altitude coastal areas are hyperdesertic and relatively barren, with only annual grassland coverage. Is it possible that the more hostile ecological conditions in these regions necessitated greater versatility and resourcefulness with respect to subsistence that produced the increased variability in N values? As Alyssa raised in her comment, perhaps trade and import of food played a role, which may have been easier along waterways and the coast as opposed to further inland. Hypotheses like these would likely need to be tested at a more local level (perhaps just for the Moquegua valley since we have a relative abundance of data points there), and more specific, local ecological data would need to be obtained. It’s an interesting question to explore.