This term I am very lucky to be on sabbatical in France, working with the Ecosys group at INRA in Thiverval-Grignon (http://www6.versailles-grignon.inra.fr/ecosys). My hope is to spend this time catching up on a number of unfinished projects and papers, as well as get involved in some new projects, with collaborators here in France.
The National Science Foundation has just awarded a Macrosystems Biology grant to advance our understanding of the role of forest management in influencing the long-term health of forests and their ecosystems services. I am excited to continue my collaboration with this fantastic group of researchers across the US, and will be looking for a grad student for the final two years of the 5 year project (starting 2020!).
Most forests of the world are managed to provide goods and services, such as wood products, biodiversity protection, and water purification and play an important role in regulating regional and local weather. Forest management is one of the most extensive and continual drivers of forest change affecting the success of associated communities of plants and animals, as well as the patterns of water capture, movement, and storage for future human consumption. This project seeks to understand how forest management decisions influence the long-term health of forests and the services they provide across regional to continental scales across the United States. Using a mapping and computer modeling approaches, the project will evaluate national and regional forest management policies, environmental disturbances, and resulting ecological, social, and economic consequences.
Using satellite technology, maps of forest management types will be developed for forests across the entire US at scales relevant for national and regional policy, and for understanding interactions with the environment. These maps will inform computer models to generate estimates of carbon sequestration, water, and forest characteristics under various management and environmental scenarios. Earth systems models (ESMs), which operate at these very broad scales, will be improved by incorporating forest management and disturbances, allowing better predictions of the effects of changing management policy, disturbance regimes, and environment on forests across the continental US and Alaska. This computer modeling framework will be used to test the relative importance of forest management across environmental conditions and assess the strength of these relationships by region in determining change in forest characteristics (e.g., species composition and size structure), and the services they provide across scales from stands to the continent. A key component of this proposal is evaluating the impacts of changing forest policies and their relative importance in comparison to direct changes in forest condition from the environment, especially in regards to species adaptation and change mitigation strategies. The socioeconomic understanding developed with this project will permit the integration of findings into relevant policy and management. This increased knowledge will assist future decision makers in evaluating potential changes that are economically and socially important.
Our final meeting will be on June 28, and the topic is Occupancy Models. I know this is of great interest to students (and faculty) who do research with animals. However, can it also be of interest to us plant people?
Imagine you are interested in characterizing seedlings densities of Bertholletia excelsa (Brazil nut ). You lay out four 9 ha plots, and sub-sample in smaller, 25m x 25m plots, enumerating the seedlings you find. However, you are deep in the Amazon, in thick rainforest, with occasional dense patches of bamboo. The forest floor is a mosaic of vegetation types, and you often are faced with a sea of plants. The seedlings you are looking for are <1 m tall, their leaves are alternate, simple, entire (or crenate), oblong, 20–35 cm (when mature). And... they look nearly identical to many other species (e.g., Couratari tauari , a species in the same family as Brazil nut and one which is far more common in this region). This is the situation I found myself in a few years ago. I was lucky to be accompanied by some of the best técnicos in the state of Acre. But even so, was the data 'perfect'? When we found zero seedlings, can we be certain there was not one layered under other vegetation? And when we came back the next year to a plot where we recorded 3 seedlings in the previous year, but we only found 2, could we be certain that there was mortality? As the surveys were repeated for 5 consecutive years, we realized our data was not perfect, especially as we noted B. excelsa's uncanny ability to re-sprout after being seemingly broken beyond repair.
Occupancy models were developed to solve the problems created by imperfect detectability. This week, we will get a short introduction and learn about some techniques to use R for these kinds of analyses. Notes are here.
Our summer R gatherings are entering their 3rd week. We meet Tuesdays, 9-11:00am in SEC 2438. To recap, here is what we did so far:
1) Map making: Shapefiles - points/polygons (different symbol/coler levels), Rasters, Inset Maps, Map necessities (Legends, north arrow, scale, text identifiers)
2) Data Management: Cropping, Masking, Filtering, Convert Files (e.g., points to polygons, Minimum convex polys), Random points in a grid (sampling design)
and we might have time for...
3) Analyses: Buffer points, Distance between points/polys (nearest distance), Interpolations, Overlay/Extract, Spatial autocorrelation, Mean center/weighted
Notes for week three here.
Recently, we had a paper revision come back with the suggestion that we use RMA versus ordinary least squares regression. My first thought was: what the heck is RMA? RMA (reduced major axis) regression is a re-branding of GMR (geometric mean regression). Whereas in a "traditional" OLS regression (y=a+bx), our goal is to minimize errors in the y -direction, in GMA, we minimize the product of the distances in the y- and x-directions. I have since gone on to find it called least products regression, diagonal regression, line of organic correlation, least areas line, Deming regression, Model II regression, and an errors in variables model.
(... Seriously? Can we just call a cod fish a cod fish? It does not taste any better if you call it a Sable Fish.)
In any event, I had not seen a GMR since my consulting days back in Vancouver in 2000-2001. I remember my very smart and talented supervisor using it to make adjustments to site index assessments made by experts using field data of top height. These two variables are both subject to a lot of error, so GMR makes sense. GMR is appropriate when the dependent and independent variables in the regression equation are "random"., i.e. not controlled by the researcher. OLS regression can underestimate the slope when the independent variable also contains error. Pierre Legendre gives a great table outlining rules for using GMR, OLS, and some variants here. In particular, OLS is a valid method when the error in the dependent variable is >> error in the independent variable:
"If the magnitude of the random variation (i.e. the error variance) on the response variable y is much larger (i.e. more than three times) than that on the explanatory variable x, use OLS. Otherwise, proceed as follows...."
So... was it appropriate for this paper? Well, we developed predictive equations for tree height and crown width based on diameter as a predictor. Even the most experienced forester would be hard-pressed to reduce the errors in height measurements to <3 times that of diameter. So, my opinion? no way. Let's hope I can convince the reviewers...
Summer seems like a good time to read books, go to the pool, learn Portuguese, or.... get better with R?! My graduate student, Nick, had a great idea. Why not have an informal gathering weekly this summer and improve our R skills? Well, Nick does not really need to improve his R skills from what I can tell, but he has agreed to lead this effort and personally, I am always in need of refreshment in the R department. (If you know me, you already know that I am a die-hard SAS user for almost everything... but I do use R when I want to be Bayesian or I am slaving over a hot spatially weighted regression.) In any event, we will use these meetings to go over some basic R skills, but also to get better at graphing and learn some basic analyses. This is a GREAT opportunity for graduate students who are taking my BSC 695-Experimental Design class in the fall, to get better at programming and get the aggravation of learning R out of the way!
Our first meeting will be Tuesday, May 31 9-11:00am in SEC 2438. During our first meeting, we will do a very quick overview of the basics. We will then cover:
Nick has prepared a group of files with the first week's material, which you can download here.
The second week's (June 7) files are here.