Michael Pohlman

Michael Pohlman
mpohlman@email.arizona.edu

University of Arizona
CALS
SWES

My data entails a detailed look at chemistry and carbon in soils from a single catchment in New Mexico after a major disturbance- in this case wildfire. We have 22 locations about the catchment and depth increments down to 40 cm below the surface. These same locations were sampled just after the fire, and then for two subsequent summers- for a total of 3 years of data (with some scant pre-fire chemistry for six locations ped1-6). 
My hypothesis is focused on the redistribution of chemistry, soil carbon and pyrogenic carbon (fire-derived or black carbon) in the near term following the fire.  I’ve completed some very basic correlellograms that that show some chemistry and carbon do in fact trend together, and that these correlations vary from year to year, in surface and deeper soil. An example of one interpretation is that there are more correlations in the surface just after the fire (2013) pointing to deposition of burned materials associated with that fire. 
QUESTIONS:
~ But I would still like to see if change is driven by some variable such as SLOPE
or LANDSCAPE POSITION
~I’d like to possibly link deposition patterns to pre-fire vegetation distribution (still obtaining the pre-fire veg data, but we could at least discuss this)
~ I could use some thoughts on how to tie together year to year change for a given location or location group (e.g. steep western slope, group ‘x')
 
DATA DETAILS:
Some pre-fire 2012 chemistry results and 2013-2015 chemistry, carbon and black carbon results.
Locations and physical data like GPS coordinates and textural analysis, GWC gravimetric water content. 
Aqueous Extracts (aq ext) are essentially soil water chemistry data. Included are cations, anions, carbon quantification and carbon characterization indices (e.g. HIX) 
Solid sample module (ssm) is carbon and nitrogen totals analyzed on bulk soil samples
BPCA  (black carbon) results are the pyrogenic carbon (fire-derived carbon) results and are found in the last several columns of each sheet
Attached is a spreadsheet with minimal headers ready for Rstudio where I’ve done some initial work. 
See also  informal photo to better visualize sample locations and the catchment (shapes shown are groupings or designations x, y, z from left to right: ‘desig’ in the spreadsheet). The designations are a steep western slope, a central convergent area and a more planar/flat region on the easter slope. 

 


Initial Meeting

I.  Who:

Client: Michael Pohlman (University of Arizona – College of Agriculture and Life Sciences, Department of Environmental Science)

Consultants: Emira, Haozhe, Samir, Amy (author)

 II.  When:

18 September 2019, 3-4pm

III.  What:

A.  Summary of Client’s Problem

Michael is focusing on the redistribution of soil carbon and black carbon (fire-derived carbon) after a fire that occurred in the Valles Caldera National Park and the dependencies of slope at a location on the year to year changes in soil carbon and black carbon concentrations. He has made some visualizations of year to year changes in concentrations of cations of his interest and those of black carbon, but he needs some statistical inferences to validate these visualizations. The main questions, in summary, were the following:

1) Is there a correlation between the slope of a location and outcome variables?

2) Is there a way to tie together year to year change for a given location?

3) How can the data be grouped for the two questions above?

B.  Discussion

1) Data details:

- Sampling process: The data was sampled each summer and collected from soils at different depth and at specific locations. 2012 data is unique because some soils were collected from a deeper depth, but the data does not have much organic carbon and black carbon data because soil analysis (to get the total carbon or the total nitrogen) and BPCA analysis (to get the black carbon concentration) were not performed.

- Location grouping: For 2012, the sample locations are not exactly the same as those for 2013-2015.

2) Background information:

- BPCA carbon: It is a mixture of organic carbon and black carbon. It requires a large number of processes to extract black carbon from organic carbon. When the soil completely burns, then the structure of BPCA is condensed and become more complicated, but when the soil only partially burns, then the structure of BPCA is simpler. Therefore, the severity of the fire can be indicated by the structure of BPCA, and it is important to understand BPCA distribution and concentration.

3) Outcome variables of interest:

- Ratios that Michael wants to look at the correlations with the slope: grams of BPCA/Soil (column CU), grams of TOC/Soil (column CV).

IV.  Next Steps

1) Out suggestions:

- Correlations between vs. outcomes variables

                  - First, Michael needs to specify all outcome variables of interest (he will email us)

                  - Not categorizing the locations, but look at all the locations as a starter

- Tie together year to year changes for a given location

                  - Group locations for this question

2) Things to be discussed (in class)

- Metris for summarizing locations – should Michael just use the average of each location?

 

#####################

Follow-up email from Michael


Attached you’ll find some updated versions of two of the data spreadsheets previously shared. For both the 6 depth-interval chemistry/carbon/black carbon and depth weighted mean (DWM) chemistry/carbon/black carbon, today's date marks the suffix for these files. 

Updates include:
2013 pits 1-6 are now called 131-136 (since 2013 pits are unique locations)
2013 fixed errors in the elevation values for 131-136
2013 changed designation (desig) value for pit 131 to ‘x’ (‘steep' grouping)
2015 fixed data entry error in longitude value (typo: -105 should be -106)
All years: changed order of Physical data columns to begin with ’slope’ and end with ‘GWC’
All years: changed designation (desig) value for pit B to ‘y’ (‘convergent’ grouping)
Overall these changes were minor, but were made to prevent confusion with the 2013 pit locations, to more appropriately designate two of the proximal pits (131 and B), and to possibly improve any correlations associated with elevation. 

Pohlman Sampling Locations

Pohlman 12_15 aqext ssm bpca ENVS

Pohlman 12_15 aqext ssm bpca DWM ENVS

 

#####################

Analysis proposal (Response to Michael's follow-up email)

 

1. Soil Biochemistry Compositional Analysis

- We suggest Michael to start looking at soil composition pre-/post- fire using a statistical test called the "Kolmogorov-Smirnoff Test" (KS for short) which compares how similar two data distributions are and provides a p-value that tests how likely these two distributions come from the same distribution.
- We will do a variety of temporal analyses regarding the soil composition using this KS test, such as comparing year to year as well as combining all pre- and post-fire data and seeing if the data in aggregate show the same behavior as the longitudinal cross-sections. 
- This will allow us to study black carbon composition along the different layers of soil (1-6), and evaluate whether fires cause a change in soil biochemistry composition across the different layers. 
- The temporal component of the analysis will allow us to evaluate whether these differences occur year to year, whether a particular year was an outlier because of different pits or depth measurements, or other interesting patterns that may arise that we may miss if we analyze all the data in aggregate. 
 
2. Aggregate Black Carbon Analysis
- Since Michael is also very much interested in measuring changes in black carbon aggregate amounts year-to-year and across different regions/fire statuses, time permitting we will look at statistical tests that measure whether there is a difference in measurement levels. 
- Because of the spatial challenges, changes in pit locations, and changes in measurement (i.e., different depths), we may have to omit some day or present these results with caveats as they may confound the results. 
- If we are unable to reach this part of the analysis, we can schedule a follow up, or I would recommend contacting Juli again and see if she can add you back into the consulting scheduler that way either us or another group can continue working with you.