Challenges to Anticipate and Solve

1. Communicating the idea of assemblage nestedness and how to test it. The idea of assemblage nestedness is not easily explained to those unfamiliar with community ecology. Luckily there are multiple ways to attack the problem that appeal to different types of learners. It is easiest to begin by comparing a small number of assemblages that are perfectly nested to a small number of assemblages that are imperfectly or only slightly nested. I suggest introducing the concept first with the Venn diagrams and presence/absence matrices in Figure 2. If some students are still grappling with the concept of nestedness, try the colorfully painted and familiar Matryoshka dolls (AKA Russian nesting dolls). A set of dolls represents a set of perfectly nested assemblages if you assume the following three rules:
   

• each doll represents an assemblage,
• doll size is positively correlated with species richness,
• physical nesting of a smaller doll within a larger doll indicates that all the species in the species-poor assemblage are found within the species-rich assemblage.
  
  Testing nestedness can be just as opaque as the concept of nestedness itself. It is best to discuss how %PN is calculated using the example dataset. Examine the presence/absence matrix and note deviations from perfect nestedness, how the statistical software quantifies deviations from the field data and simulations, and how both types of deviations contribute to the test statistic. See the first of the Questions for Further Thought below.

2. Selecting field sites. If you have an extra lab or two to spend on this experiment, I suggest you have students select field sites. This is a difficult and time consuming part of such an experiment that teaches some of the difficulties inherent in conservation research (e.g., site access, confounding variables, minimum distance necessary for statistical independence of sites, etc.). Assuming you are selecting site, try to select fragments that reasonably cover the available ranges of fragment area (e.g., 3-300ha) and isolation (e.g., 0.1-3km). Identify more candidate sites than you need, check on access for each site, and visit each site to assess general comparability of sites (e.g., lack of extreme physical disturbance to understory and similarity of surrounding landuse). Once you have identified 9-12 viable sites, construct a spreadsheet with five labeled column for distribution to the students:
   

• Column 1 = Site name
• Column 2 = latitude
• Column 3 = longitude
• Column 4 = area (ha)
• Column 5 = isolation (km from closest fragment >10ha)
  
  Columns one through three should have data filling row two through at least row 10.

3. Keeping students on track. The entire experiment progresses quickly. Students who fall behind rarely catch up, make meaningful contributions to group work, and understand all concepts listed in the learning objectives. The best way to maximize the number of students who successfully complete the experiment is to design and clearly communicate appropriate grading incentives for the completion/submission of assigned work. I suggest using the grading rubric below. Additionally, I suggest that you require each student to evaluate his work and that of the students in his group as part of the final draft of the scientific research poster. I attach my peer evaluation form (DOC) to the assignment sheet handed out during the first lab to inspire a more equitable distribution of work.
4. Sorting bird species for nestedness analyses. Bird species have different levels of affinity/aversion for humans and the results of human activities. Most urban ornithologists refer to species that regularly associate with humans or benefit from them as synanthropic and those that do not regularly associate with humans or benefit from them as non-synanthropic. Patterns of nestedness, if present, will be most apparent if student groups sort bird species into these two groups and analyze them separately. See Donnelly and Marzluff (2004) for analyses of sorted species and Johnston (2001) for species sorting.
5. What if bird assemblages are not nested by fragment area or isolation? It is possible that students will find that bird assemblages are not nested by fragment area or isolation. If this occurs, it is useful to discuss potential explanations for the results and what additional fragment attributes if any should be explored if time allowed. Potential explanations for these results include:
   

• fragment area and isolation are not important to bird species in the landscapes where data were collected,
• one visit per site was not enough to accurately determine species presence or absence,
• and there was not enough variation in fragment area and isolation among the selected fragments.
  
  The last potential explanation is likely too be true of isolation in landscapes with >40% native habitat on the landscape; in this case, all fragments are relatively close to one another and likely to exchange individuals. For more information, see the first Question for further thought.

6. Managing the poster session. While driving among field sites, I spend a fair bit of time explaining that poster sessions are a typical part of annual meetings for major academic societies, are an alternative format to oral presentations with greater interaction between presenter and audience, and may be an important step toward a published manuscript. To communicate poster format and content, I refer students to Purrington (2006). Finally, I try to simulate the environment of a real poster session by providing munchies and an audience of lower level biology students and biology faculty.

Introducing the Experiment to Your Students

I run this experiment during the mid- and latter-stages of my upper-level General Ecology course, after covering abiotic and biotic factors limiting species distributions and equilibrium theories of species richness. Much of this material is heavily steeped in theory and can put the students to sleep even if you try to illustrate it with real world examples. One can regain the attention of the theory-resistant students by introducing this experiment as a chance to apply theories already studied to conservation planning. Who can resist saving imperiled species?

Data Collection and Analysis Methods Used in the Experiment

The most challenging portions of this experiment for the students are bird identification and statistical analysis. It is much easier to identify birds to the species level than individuals of many other taxa. However, students may need training on binocular operation, bird morphology relevant to visual identification, and identification of species by sound. Have the students practice the act of quickly locating and focusing on stationary birds or fixed objects. If conducting the lab in North America, use the Golden Guide to Birds of North America (Robbins et al. 2001), Thayers Guide to Bird Identifcation (Thayer 1998), and/or the Patuxent Wildlife Research Centers website (Gough et al. 1998) to practice bird identification. The Golden Guide has a nice description and figure of external bird morphology near the front of the book and printed sonograms for most species. The latter is very rare in field guides. The CD has pictures, audio clips, sonograms, and mnemonic devices for song identification. It is best if students use the book and the CD in the lab to learn a reasonable number of the species most likely to be encountered during the experiment. Even if students have some experience with bird identification, I highly recommend looking over their shoulders during bird surveys to ensure that they detect species that are present and do not detect species that are absent. Students typically find some species easier to identify than others. As long as these abilities/limitations do not overlap excessively, a group of three students will be able complete a relatively accurate survey with minimal input from the instructor.

Statistical analysis runs smoothly if you use an example dataset to illustrate formatting of data for input, running of the software, and production of graphics associated with the statistics. All necessary files including results of example dataset analysis dataset analysis are available under the Downloads page.

There are many ways for students to collect the data necessary for this experiment. In order to save time in the lab, I suggest instructing each group to estimate area and isolation for a subset of the forest fragments and to enter their results into a common excel file that will be disseminated to all groups. This system is effective as long as individual groups do not produce consistently biased estimates of area and or isolation. I suggest a slightly different approach to surveying birds. Have each group record all birds that it detected on a transect. Then, pool the species lists from all groups for that transect to create an aggregate species list for that transect, or fragment. This approach will compensate for considerable variation among students in the ability to identify bird species. The wise instructor will still as much as possible monitor bird identifications in the field.

Questions for Further Thought

• Unfortunately, there is no simple answer. Theoretical, meta-analytical, and empirical studies (Andren 1994, Fahrig 1997, Bunnell 1999, Harrison and Bruna 1999, Lichstein et al. 2002) suggest that habitat loss is more detrimental to retention of native species than habitat fragmentation until native habitat becomes fairly rare on the landscape; once the landscape falls below this threshold, habitat fragmentation is more important than habitat loss. Andren determined this threshold to be 30% native habitat on the landscape for birds and mammals, but the exact value is likely to vary among taxa and to be higher for less vagile taxa. To demonstrate thresholds in fragment connectivity, use a simple applet available from the American Museum of Natural History (2007; exercise 9)
• See Faaborg (2002) for many acceptable answers. Most bird conservationists would put the following near the top of the list: smaller fragments have a larger edge to core ratio and higher nest predation rates, smaller fragments have a larger edge to core ratio and higher brood parasitism rates, and more isolated fragments have smaller chances of population rescue.
• If presence of species in a fragment is largely dictated by dispersal ability and species vary in their ability to reach (i.e., colonize) more distant islands, then assemblages in fragments will be nested. Species with poor dispersal ability will be absent from distant fragments, while species with good dispersal ability will be ubiquitous. If presence in fragments is largely dictated by extinction and species vary in their ability to persist in smaller fragments, then assemblages in fragments will be nested. Species that are sensitive to area, edge, or habitat degradation associated with edge will be absent from smaller fragments, while the others will be ubiquitous. Given their powers of flight, dispersal ability probably explains nestedness of fewer bird assemblages than assemblages of less vagile taxa (e.g., amphibians and small mammals). Nonetheless, there are studies showing that some bird species do not like to cross even small gaps in fragmented landscapes and that such behavior can limit fragment occupancy (e.g., chickadees, tree creepers). Nested habitats or resources and unequal sampling effort can also cause assemblages to be nested or to appear nested, respectively .
• Hypotheses regarding species-specific extinction assume that the entire regional species pool was present throughout that habitat prior to fragmentation. For this lab, that means that all species found in the largest contemporary fragment were found in all fragments prior to their separation from neighboring habitat.
• The primary weakness of any presence/absence analysis is false absence (i.e., failure to detect a species when that species is present). In general, we assume that surveyors are skilled enough to avoid detecting species that are absent. It is possible, however, for even skilled surveyors to miss species that are present, but have low detectability (e.g., they are rare or otherwise inconspicuous) or range over large areas. One can sufficiently cope with these problems by increasing survey effort, estimating detection probabilities for each species, or removing species with low detectability from the analysis. Remember, no analysis is perfect and that presence/absence analyses may be the only option for natural resource managers mandated to conserve native species without sufficient resources. See recent literature by Vojta (2005) and Fleishman et al. (2007).
• If nestedness of assemblages in habitat fragments is highly correlated with fragment area, a small fragment will contain only a small sample of the species pool and that sample will vary little among all small fragments. On the other hand, a large fragment will support most or all of the species pool. To protect as many species as possible in this scenario and with limited resources, one should protect a single large fragment rather than several small fragments.
• Possibly. Nestedness analyses, as presented here, determine whether assemblages are nested by fragment attributes. If assemblages are nested by a particular attribute, the attribute could directly affect a species fitness (e.g., some species apparently select only large fragments for strictly behavioral reasons) or indirectly affect a species fitness via a correlated aspect of the habitat (e.g., size dictates edge to core ratio and the rate of nest predation). If the latter is true and one can identify the aspect of the habitat that is most important to assemblage structure, management outcomes will improve with more thorough study.

References and Links Cited in Comments to Faculty Users

• American Museum of Natural History. 2007. Problem solving in conservation biology and wildlife management. Available at http://cbc.amnh.org/solving/.
• Andren, H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: A review. Oikos 71:355-366.
• Bunnell, F.L. 1999. What habitat is an island? In Forest Fragmentation: Wildlife and Management Implications (J.A. Rochelle, L.A. Lehmann, and J. Wisniewski, eds.), pp. 1-31. Brill, Boston.
• Donnelly, R. and J.M. Marzluff. 2004. Importance of reserve size and landscape context to urban bird conservation. Conservation Biology 18:733-745.
• Faaborg, J. 2002. Saving migrant birds: developing strategies for the future (chapters 4 and 5). University of Texas Press, Austin.
• Fahrig, L. 1997. Relative effects of habitat loss and fragmentation on population extinction. Journal of Wildlife Management 61:603-610.Harrison and Bruna 1999
• Fleishman, E., R. Donnelly, J. Fay, and R. Reeves. 2007. Applications of nestedness analysis to conservation of biodiversity in developing landscapes. Journal of Landscape and Urban Planning 81:271-281.
• Gough, G.A., J.R. Sauer, and M. Iliff. 1998. Patuxent Bird Identification Infocenter v.97.1. Patuxent Wildlife Research Center. Available at http://www.mbr-pwrc.usgs.gov/id/framlst/infocenter.html.
• Harrison, S. and E. Bruna. 1999. Habitat fragmentation and large-scale conservation: what do we know for sure? Ecography 22:225-232.
• Johnston, R.F. 2001. Synanthropic birds of North America. In Avian Ecology and Conservation in an Urbanizing World (J. Marzluff, R. Bowman, and R. Donnelly, eds), pp.49-68. Kluwer Academic, Norwell.
• Lichstein, J.W., T.R. Simons, and K.E. Franzreb. 2002. Landscape effects on breeding songbird abundance in managed forests. Ecological Applications 12:836-857.
• Lomolino, M. L. 1996. Investigating causality of nestedness of insular communities: Selective immigrations or extinctions. Journal of Biogeography 23:699-703.
• Purrington, C.B. 2006. Advice on designing scientific posters. Available at http://www.swarthmore.edu/NatSci/cpurrin1/posteradvice.htm.
• Robbins, C.S., B. Bruun, H.S. Zim, and A. Singer. 2001. Birds of North America: A Guide to Identification. Golden Guide from St. Martins Press
• Thayer, P. 1998. Birds of North America v. 2.5 (CD). Thayer Birding Software.
  Vojta, C. 2005. Old dog, new tricks: innovations with presence/absence information. Journal of Wildlife Management 69:845-848.

Assessment of Student Learning Outcomes

If you stick to the staged approach to the experiment outlined in Assessment of Student Learning Outcomes, you should be able to identify students who fall behind early and help keep them motivated by continually mentioning the peer evaluation portion (DOC) of the assessment process. I suggest using the results of this peer evaluation and your observation of group dynamics to "adjust" individual scores. Start with the group score (does not include bird ID quiz or report on Faaborg reading) and add or subtract points equivalent to 1-1.5 letter grades. Of course, this grading procedure must be announced at the front end. Here is a suggested grading rubric for the entire exercise

Figure 5. Suggested grading rubric.

Evaluation of the Lab Activity

I have taken two approaches to formative evaluation of this experiment. First, I have used the questions listed under Questions for Further Thought to evaluate the degree to which students understand the concepts fundamental to their hypotheses (Q1-3), the methods they will use to test their hypotheses (Q4-5), and the applications of their results to conservation (Q6-7). I assign sets of questions at strategic intervals throughout the experiment so that I can correct any misunderstandings with written feedback and discussion before we move to the next step in the experiment. I have found that this approach corrects misconceptions before they snowball and prevents students from abandoning all hope of understanding the experiment because they did not understand fundamental concepts presented in early steps. Second, I periodically assign and review responses to minute papers on important and confusing concepts. These papers help me identify barriers to learning that I did not anticipate and how to adjust my list of questions for further thought and discussion.

Translating the Activity to Other Institutional Scales or Locations

In general, this experiment readily translates to many different settings. It translates to medium-sized classes (10-18 students) by increasing the number of groups. With more groups, the class can cover more hypotheses (e.g., nestedness by habitat degradation) and multiple taxa. While the later may require the help of a lab coordinator or teaching assistant to collect field data, it is instructive to compare how well fragment area and isolation explain nestedness of taxa varying in their general dispersal ability (e.g., most frogs and salamanders do not disperse as well as birds). In fact, the instructor may not wish to use birds at all if the academic schedule does not align with bird breeding season, if he does not feel comfortable surveying birds in an afternoon lab, or if he is particularly comfortable with another taxon. Many assemblages are nested, regardless of taxonomic status and ecosystem identity. I firmly believe that this experiment does translate to classes populated by students with little exposure to organism identification, if you (the instructor) are comfortable with bird identification and provide students with teaching tools mentioned in the section titled Comments on Data Collection and Analysis Methods Used in the Experiment.

This experiment is not appropriate for students with physical disabilities or for pre-college courses.

Instructions for using NestedSim.exe

The Basics
Nested uses a simple simulation program designed by Mark Lomolino to test whether biotic assemblages are nested by a variable of the users choice. The program requires a current PC operating platform (Windows 2000, Windows XP, or Windows emulator on a Macintosh) and a carefully formatted input file.

Formatting the Data for Input
The file must be comma delimited. This can be achieved by saving an excel spreadsheet with a .csv extension. Rows must represent sites. In order from left to right, columns must represent fragment area, fragment isolation, and species.

Figure 6. Example of a formatted input file in .xls format. For the program to read the file, it must be converted to .csv format.

Values for area and isolation must be entered with at least one decimal place. Column 3 through column n+2 (where n = number of species detected across all fragments) must represent species. Use a one to denote presence and a zero to denote absence. Now, sort rows in your input file using fragment area (the highest value must be in the top row) and add a final column (column n+3) that contains the row number.

NOTE: After testing for nestedness with respect to area, the program will test for nestedness with respect to isolation. For this second test, it will sort the rows so that the value of the variable in column two decreases from the top of the matrix to the bottom of the matrix (i.e., the program assumes a negative correlation between the variable in column two and species richness).

Performing the Test
Once you have formatted the input file, open program Nested and type in the complete path to that file (e.g., c:/temp/NestedSimIn.txt). Click the checkfile box and compare the numbers of sites and species read by the program to known values. If these numbers appear correct, set the number of random simulations you prefer (1000 is standard) and click the run box. The large simulation window will show the progress of the simulations and confirm the number of simulations selected. When the simulations are complete, values will appear in the bottom five windows. Calculate the final test statistic:

Percent Perfect Nestedness = 100 * ((R-D)/R).

Note that the output includes a D-value for both area and isolation, so be careful which D-value you select when calculating percent perfect nestedness.

Example Data
You have been provided with a text file of formatted example data named NestedSimIn.csv. It is composed of 29 sites and 20 species. Darea equals 87. Disolation equals 110. R, P-values, and percent perfect nestedness will vary with each test, but they should be near the following: R = 107, P-valuearea < 0.01, P-valueisolation > 0.5, percent perfect nestedness with respect to area ~ 19.