Objective 2 will involve modelling hemlock occurrence as a function of landscape attributes and will not, at least initially, involve any additional field sampling. Landscape and hemlock occurrence data will be derived from existing GIS maps, and crown health data that has been collected at Shenandoah park will be used to evaluate stand mortalities. Addressing these issues will be reasonably straight forward and methods will not be discussed here.
Common-site Selection
The following narrative describes options for common site selection. Throughout this narrative we will use the term sampling design to refer to the process of site selection and the term common sampling unit to refer to the minimum unit of land for which all cooperators will make inference. This means that parameters of interest (e.g., diversity) will be estimated for each selected Common Sampling Unit. In turn, the set of selected Units will be used to estimate park-wide diversity. The Common Sampling Units will likely be too large to be inventoried completely so sub-sampling will likely be necessary. For example, if the Common Sampling Unit is a 30x 30 m parcel of land and species richness is a parameter of interest then all cooperators must subsample the parcel so that species richness can be estimated reliably. Each cooperator will subsample differently depending on the taxonomic group being studied.
The options for defining Common Sampling Units discussed later share a basic sampling design. That design is to select units to allow comparisons across vegetation type (most notably hemlock versus non-hemlock forests) while controlling for confounding variables. We recommend 2 methods to control for confounding variables: 1) blocking (or stratification) by topographic variables, and 2) covariate modeling of site specific variables. For example, aspect, slope, and elevation are topographical variables which may influence hemlock occurrence and species richness. Apparent relationships between richness and hemlock occurrence may be proximate to ultimate causes due to topography. Thus, we propose to control for these potentially confounding variables by using combinations of aspect, slope, and elevation (or other topographical variables) to form blocks or strata within which comparisons of richness can be made across vegetation type.
A major concern is limiting the number of strata to allow sufficient replication. Thus, some confounding will need to be controlled at the analysis stage by use of covariates. Covariates are variables that can be measured in the field which may interact with design variables (i.e., topographic strata or vegetation type) to influence the response of interest (e.g., species richness). In our opinion, a major advantage of a common site strategy is that measurements by one cooperator can be used as covariates by others to help explain their response of interest.
The remainder of this narrative describes four options for defining Common Sampling Units and our preliminary evaluation of each. In our view the ideal common site should meet the following criteria: 1) be consistent with the major study objective (i.e., comparing diversity among hemlock versus non-hemlock vegetation types), 2) permit straightforward selection to facilitate straightforward data analysis and modeling, and 3) be compatible among all taxonomic groups being studied.
Reality Check
As stated previously, the general approach agreed upon is that Common Sampling Units will be selected to represent i) landscape diversity of the parks, and ii) a gradient in the abundance of hemlock. Assuming sufficient statistical power, this will allow us to derive park-wide estimates of diversity, and compare diversity patterns among different landscape types and hemlock abundances. We are currently attempting to estimate statistical power for the pertinent comparisons and hope to have those estimates by our meeting. However, we think it is imperative that we ensure a minimum of replication in our study design. The aquatic group has indicated our logistical potential in terms of the number of sampling units we could effectively sample is between 30 and 40. We have not as of yet evaluated the potential of the terrestrial groups. Assuming this number is reasonable for everyone, we would have to limit the number of topographic unit types to about 5, and the number of hemlock abundance classes to about 3 for a total of 15 different types. This would yield between 2 and 3 replicates/type).
Another concern that warrants discussion is the difficulty in locating selected common sites on the ground.
Specific Options for Common Sampling Units
After considerable discussion, we have selected four options to delineate Common Sampling Units. All four will begin with a base map that includes information on topography and hemlock abundance. That map will be constructed using GIS as follows:
| Introduction | Option 1 | Option 2 | Option 3 | Option 4 |
|---|
A Forum for Discusion
We encourage discussion on these options. We do not consider these to be exhaustive nor final, but rather starting points. However, we do feel that the design underlying these options provides a general framework to guide our discussion and help us converge on a satisfactory site selection strategy.
Towards this end, an email discussion list (called HEMLOCK-L) has been set up to distribute comments regarding the Collaborative Hemlock Study. The discussion list approach will allow all concerned to comment not only on the original proposal but on others' comments, as well. In this way we can make progress prior to our physical meeting on the 19 Sep.
To subscribe to the list, send an email message to listserv@aegis.lsc.nbs.gov with the message SUBSCRIBE HEMLOCK-L YOUR NAME. A successful subscription to the list will be followed by a message with instructions on how to send messages to the list for distribution and on how to unsubscribe when the time comes.
After you have successfully subscribed, send another message to listserv@aegis.lsc.nbs.gov with the message GET HEMLOCK-L DIGEST and you will receive one file which contains all the messages posted to the list to date.