GEOG 566

         Advanced spatial statistics and GIScience

June 11, 2017

My Spatial Problem: Revised

Filed under: Final Project @ 11:48 pm

I am interested in the understanding how mammals have responded to changes in their habitat through time and how the distribution of resources on the landscape structures small mammal communities, historically and today. I predominantly utilize stable isotope analysis of carbon and nitrogen to evaluate niche dynamics of rodents along elevation gradients and through time, approximately the last 8,000 years, in the Great Basin of the intermountain west of north America. In this regard, I want to know, how do rodents in the Great Basin respond to changes in the resource base driven by human land-use and climate change?

Original spatial problem.

When I outlined my original spatial problem, I was interested in replicating a geospatial analysis performed by Elsen and Tingley in their study, “Global mountain topography and the fate of montane species under climate change,” published in the journal Nature – climate change, in 2015. In their analysis, Elsen and Tingley calculated the proportional distribution of horizontal surface area along the slope of mountain ranges, with the objective of evaluating the assumption that available area (habitat) decreases with increasing elevation. This is an important assumption to test as it has direct implications for the way ecologists think about biodiversity in mountains. Specifically, it has been assumed that in general species richness decreases with increasing elevation due to decreasing area, as would be predicted by the species area relationship. I wanted to perform this analysis on three mountain ranges in the Great Basin that are of particular interest for my dissertation research. However, the data sets which Elsen and Tingley uses to perform their analysis were far too large for me to handle on my own, and my knowledge of the requisite skills far too limited to perform the analysis. Consequently, my spatial project began to morph, and followed a circuitous trajectory for the remainder of the term.

Exercise 1: Hot spot analysis of the North American deer mouse in the Toiyabe mountains.

Upon choosing to ditch the area-by-elevation analysis I thought I would dig into a dataset that I had explored previously with other methods. This dataset contained two matrices of data collected in the Toiyabe mountain range, NV in both 1930 and 2011. The first matrix was a species by site matrix which included the trap-effort standardized abundance of ~20 small mammal species, predominantly rodents; the second matrix contained site characteristic data including elevation, latitude, longitude, mean summer high temperature, mean winter low temperature, and mean monthly precipitation. In the past I had explored this data using principal components analysis and non-metric multidimensional scaling. I wanted to know which factors explained changes in elevational distribution of species between these two time-periods. I found that climate factors described most of the variation. Notably, all climate factors were collinear with elevation, either increasing or decreasing in proportion with elevation.

Given the spatially explicit nature of this data set, I thought it would be interesting to perform hotspot analysis on the most abundant species in the data set. In performing hotspot analysis on the distribution of this species I hoped to gain insight into the influence in might have on the spatial structuring of the rodent community in the Toiyabe mountain range, at large. However, what this analysis revealed was that transect data is inadequate for evaluating this type of question. Spatial analysis of this nature requires a distinctly different sampling design, specifically, a grid across some larger portion of the mountain range would have been ideal for the question I was hoping to answer. Instead, I discovered that hotspot analysis on transect data reveals more about the behavior of those doing the trapping, than it does of the mammals being trapped.

Exercise 2: Energy flow in the Great Basin rodent community and a test of space for time substitutions.

Following this I chose once again to shift gears and leverage the elevational information in the Toiyabe data set. Again, I borrowed inspiration from novel research, this time more directly related to my own work. In 2015, Terry and Rowe published “Energy flow and functional compensation in Great Basin small mammals under natural and anthropogenic environmental change,” in PNAS. They demonstrated that daily energy flow (kJ/day) through Great Basin rodent community had responded differentially through the Holocene to the modern to the changing climate. Furthermore, they demonstrated that functional groups within the rodent community responded differentially through time as well, and that many of the patterns that had held through the Holocene changed dramatically at the onset of the Anthropocene. I wanted to know if the dynamics demonstrated by these functional groups through time could be approximated along a spatial gradient that captures a similar change in climate. In this regard, I was testing a space for time hypothesis. I found that many of the patterns observed through the Holocene were well approximated by the spatial substitution. However, the analogy between elevation and time did not hold into the Anthropocene. Although interpretation of my analysis was predominantly qualitative, I feel comfortable making the claim that the space for time substitution of the rodent community in the Toiyabe mountain range for the rodent community of the Great Basin approximates patterns from the end-Pleistocene to the end-Holocene, but is decoupled from the dramatic changes in the Anthropocene.

In response to peer feedback I have set out to determine if the patterns observed in the above analysis would hold in a second mountain range. This analysis is in progress, and I hope to add to it, eventually, a third mountain range. I have in hand the data for the Ruby mountains in the Great Basin, and will soon have access to a similar dataset for the Snake range.

Exercise 3: Creating a map of NDMI for the Toiyabe mountain range in the Great Basin.

Despite the interesting findings of the second part of my analysis, I felt as though my work in this class had become tangential to my own interests and the focus of my dissertation. Although it is unfortunate that it took most of the term to come to this realization it has not been a loss. I have a new target; the Normalized difference moisture index (NDMI) uses landsat-7 data to estimated vegetation density and soil moisture content. I want to use this data to develop a better understanding of the spatial distribution of vegetation in the Toiyabe and Ruby mountain ranges (the focus of my dissertation). I have several questions in mind; 1) how does NDMI vary with elevation and aspect? 2) what is the NDMI value at each of the trap sites in the Toiyabe and Ruby mountain ranges? 3) are there correlations between trap-line species abundances and NDMI upon which to base predictions about range wide distributions of species? And 4) What is the degree of patchiness of NDMI in these two mountain ranges, respectively, and what predictions might we make about how this could limit or promote the movement of species laterally or up and down slope?

The above outlined questions would all be exceptionally useful to address and I hope to do so as I become more comfortable with the tools available in ArcMap and Cran-R for these types of analyses. In the interest of exploring these data I have calculated NDMI for the Toiyabe mountain range using Landsat 7 data for June 2011, a time during which small mammal trapping at this site occurred. While I have not yet been able to address the questions I outlined above, I have generated a map showing the NDMI at a 30m x 30m resolution. The color scale for this image is continuous from blue, to green, to yellow, to orange, to red, and reflects the -1(low moisure vegetation, low density) to 1(high moisture, or high vegetation density) range of values possible from each grid cell. What is most apparent in this map is that peaks in the mountain range, especially those with snow remaining, have particularly high NDMI values, and much of the remaining landscape appears uniform with intermediate to low NDMI values. While the general impression given by this map is not surprising, given the arid environment, the distribution of NDMI values evaluated at different scales could prove interesting.

The most valuable thing I learned through this course was not a software package or a particular statistical method. I learned what type of data is truly necessary to perform spatial analyses, and how to get it. Specifically, I learned that there is a vast wealth of landsat data available for free and that it can be clipped down to manageable sizes for the analyses I am interested it. While I still feel somewhat uncomfortable finding datasets, downloading them, and navigating the endless options of ArcMap, I have also made a new connection with an incredibly knowledgeable graduate student, Jessica Pierson. I hope that with her help and that of a several other peers, more knowledgeable by far than I am in these methods, I can become confident in accessing data and performing the necessary analyses in ArcMap.

Reflection on Julia’s guidance and peer feedback

I have come to this point in my spatial project by trial and error and with the encouragement of Julia Jones. Julia’s advise, as I progressed through this course was to give things a shot, and if they didn’t work that would be fine, we could just try something else. This low pressure, low stakes approach was very helpful. This has been a particularly busy term, especially outside of this class, and outside academia as well. Knowing that this course would be student driven, and what I got out of it would be entirely up to me felt daunting at first, but I have come to accept that the bulk of my learning happened at the tail end, and I’m excited that the product I am walking away with is new questions to ask and new methods to learn. I did not bend my data to just perform new analyses if I did not think they would be informative, nor did I choose to recycle old analyses. Following Julia’s guidance, I gave a few new ideas a shot, and while some of them didn’t work out, I am excited to take what I did learn beyond the context of this class to inform important aspects of my dissertation.

Feedback from my peers on the first tutorial was to suggest that different experimental design (e.g. data collection) would have been more appropriate for hotspot analysis, or to change my approach to one that could leverage the structure of my spatial data. I responded to this by shifting from lat-long coordinates to elevation as my spatial parameter. Peer feedback on my second tutorial was largely limited to acknowledgement of my limited sample size, specifically one mountain range, and that by performing similar analyses in additional mountains, I might be able to perform statistical tests, such as linear regression. As stated above, this is underway, and will be a product outside of the context of this class, as both my graduate major advisor and I have an interest in knowing how energy flow through functional groups will vary in different mountain ranges, and if spatial findings will be concordant with the temporal analysis.

Print Friendly, PDF & Email

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

© 2019 GEOG 566   Powered by WordPress MU    Hosted by