Upcoming events and meetings

As a reference, here are dates of interest:

October 29, 2008: Conference call at 9:30am (AZ time)
November 12, 2008: Conference call at 9:30am (Mountain Standard Time, includes AZ)
November 20-21, 2008: Moab meeting with Greg Okin
December 3, 2008: Meeting with 3 CO-PIs and Travis in Flagstaff

Piecewise regression for Semidesert Sandy Loam (Fourwing Saltbush)

I have been working with Mark's data as I try to build the STM and think about the first ecosite (Semidesert SaL). I ran a multiple regression investigating factors that influence soil aggregate stability. "Good" BSC points showed the strongest influence on soil stability. I then ran a piecewise regression, which yielded the following pattern (breakpoint of 22 ± 5). Red points are "grazed", open points are "ungrazed". A similar pattern is shown in the presentation Mark put together that was presented at the SLC workshop (slide 9 of SemidesertSandyLoam.ppt).

Updated List of Datasets

After meeting with Sue Phillips in Moab on Monday, October 20th, 2008, we have acquired new datasets. The list below includes the names given to the folders by Sue Phillips, but I am not yet sure the nature of the datasets (i.e., what was measured, where, when, etc.).

Here is a list of our datasets to date:

(1) data from Mark (CANY_Dugout_Combined_12Jun08_mem.xls)
(2) data from John Spence via Chris (Upland_Data.xls)
(3) data from John Spence via Mark (CARE_Exclosures.xls)
(4) data from Sue Phillips/Belnap lab:
a. ccows
b. Glen Canyon
c. I&M BSNE
d. Kleiner
e. rainfall sim
f. SERDP
g. SEUG
h. thumper
i. Needles (Virginia Park, Chesler Park)

State and Transition Monitoring Workshop Presentations

Some good presentations here for reference:
http://oregonstate.edu/dept/range/events/archivedevents/StateAndTransitionMonitoringPage.php

Datasets to date

As a reference for you to know which datasets I have, I will update this within 24 hours of receiving a new dataset.

As of 11:00AM on 10.06.08, I have:

(1) data from Mark (CANY_Dugout_Combined_12Jun08_mem.xls)
(2) data from Jayne/Sue via Steve (belnap_needles file folder)
(3) data from John Spence via Chris (Upland_Data.xls)
(4) data from John Spence via Mark (CARE_Exclosures.xls)

Travis

Piecewise Regression for Blackbrush (Desert Shallow Sandy Loam and Desert Sandy Loam)

I was "messing around" with piecewise regression and Mark's CANY/BLM data, exploring patterns between vegetation and soil stability for Blackbrush dominated ecological sites. It didn't take long to find an interesting pattern and after running a piecewise regression, I was able to calculate the breakpoint (threshold?) and confidence level around it (standard error). I plotted the data coding the observations by "grazed" or "ungrazed". These are very preliminary, but I thought this was interesting. Someone might argue that because soil stability is measured categorically you would expect this kind of "saturating" relationship. But, I thought I would go ahead and share in case anybody actually looks at this.

Updated framework

"All models are wrong, but some are useful." - G.E.P. Box

"Box and arrows won't get rid of cedar, but a box of matches will" - John Weir, 2006 (cited by Fuhlendorf)

I continue to update the framework using concepts derived from the existing framework, the SLC workshop, and the literature (using STM literature and threshold literature). My hope is to produce an updated framework that coincides nearly perfectly - in terms of concepts - with an updated table.

Here is the current draft framework (9.22.08). Click on the images to enlarge.

And the corresponding draft Threshold Attribute Table.

Piecewise regression & discriminant analysis(?)

I found what looks like helpful references to piecewise regression. The first is an Ecology paper. The second is a USFS report with SAS code and data that looks like it might be very helpful. Sorry to be redundant if you already have these.

Has there been any thought/work on using some kind of multivariate methods of determining thresholds? I am thinking of something like a discriminant function (linear or nonlinear) that distinguished between groups. If data are available where "states" are identifiable using multivariate datasets (like Mark's data), then a discriminant function may provide some way of determining thresholds among groups using several important variables. I'm thinking back to a class from>5 years ago, but thought I would see if you know whether this has been attempted. I've searched some papers and found some interesting methods from folks in Maryland looking at aquatic systems (Bob Hilderbrand et al.). I've been going through papers from the Duluth DVD, but haven't found such an analysis used yet. But, there's a lot on that DVD.

On other fronts: I've set the blog so that it should be easier to comment. Just use the "anonymous" choice for now until you all sign in and get register with the blog. Make sure to sign your comment so we know who said what.

Looks like Mark and I are scheduled to meet with Sue Phillips on October 20th. She mentioned data that she gave you; Are those the data you're planning on sending me this week?

Thanks,
Travis

Revised Framework and Threshold Attribute Table

Hi all. Just as a demonstration for how this blog might work: This might be a way for me to communicate an idea or question to you all.... I've been mulling over the framework, the threshold attribute table, and the Jayne's (Blackbrush) and Craig's (PJ) STMs. I attempted to revise the framework a little by combining boxes and revising terms a little to fit some of the concepts from the meeting. So below you will find four boxes. Box 1 (Reference conditions) represents pristine sites with no measurable disturbance. Box 2 (Manageable conditions) are those sites where a disturbance or alteration (climatic) has shifted the structure and function of a site but that a management action that removes the disturbance would allow "natural" feedbacks to shift the site back to Reference conditions (i.e., through resilience). Box 3 (Restorable conditions) are sites that have been disturbed or altered such that active restoration is necessary to return the system to acceptable structure and function. Box 4 is the degraded condition where negative feedbacks maintain conditions that no longer provide ecosystem services. I was trying to simplify the existing framework while maintaining the overall concepts and updating some based on our discussions at the meeting. I also added some terminology from the literature and that Mark added as an afterthought to Craig’s STM. This is just my attempt to get my head around material and synthesize our conversations from the meeting. I don’t mean to try to change things in a dramatic way. We may not use any of this but it has helped me get my head around some of the material.

Also, I think we might want to revise the table so that it fits with the framework more seamlessly. I will make some changes and present to you all soon.

Talk to you tomorrow.
Travis

A viable forum for communication?

This is just a test to see if a blog like this would be a viable form of communication. We can limit the viewers to only the collaborators. Something to consider on conference call scheduled for Friday, September 19th.