Ecological Archives A024-046-A2

Ralph Grundel, Krystalynn J. Frohnapple, David N. Zaya, Gary A. Glowacki, Chelsea J. Weiskerger, Tamatha A. Patterson, Noel B. Pavlovic. 2014. Geographic coincidence of richness, mass, conservation value, and response to climate of U.S. land birds. Ecological Applications 24:790–811. http://dx.doi.org/10.1890/12-0823.1

Appendix B. Possible effects of differences in species detection on results.

The Breeding Bird Survey (BBS) protocol requires the observer to watch and listen for birds for three minutes at each of fifty points separated by 0.8 km along a route (https://www.pwrc.usgs.gov/BBS/participate/instructions.html). In such a survey, as with many surveys, not all birds present in an area will be detected. Detection probability can be affected by habitat structure, observer ability, species differences and probably other factors. Differences in detection among routes can affect interpretation of status and trends in bird populations determined from BBS data (Boulinier et al. 1998, Sauer and Link 2011, Rittenhouse et al. 2012). In our study of spatial congruence among five measures of conservation value, each of the measures is potentially affected by detectability and if detectability differentially affects the measures, the relationships among the measures can be different if detectability is, or is not accounted for.

If undetected birds are not accounted for by modeling, we can expect that MASS will be lower since it represents the product of the number of birds observed along and the mean weight of each bird. RICH will be lower if undetected species are not estimated. The Conservation Index, CI, represents the portion of the global population of a bird species present along a hectare of the BBS route, summed across all bird species observed on that route. If undetected birds are not accounted for on that route, CI will be lowered. The Combined Score, CS, represents the average Combined Score for the five species along a BBS route with the highest Combined Scores (higher Combined Scores indicate a higher risk of extirpation for a species). The rationale for averaging five scores was that managers will often pay increased attention to highly threatened species when creating management plans so having species with high CS in an area will tend to focus management on helping those highly threatened species. We averaged the five highest CS scores to represent a handful of threatened species at a location driving management focus. Detectability could affect CS if, for example, the most threatened species were also the rarest at a site and would be less often encountered or detected. Finally, we examined the one-dimensional principal curve ordination score, PC. PC ordination is based on some measure of abundance, or presence, of species across sites so differences in number of birds counted in one part of the country or one type of habitat compared to another, due to differences in detectability, can affect the ordination score of those regions or habitats relative to other regions or habitats.

To evaluate how detectability might affect the relationships among the five responses – MASS, RICH, CI, CS, and PC – we examined how correlations among these responses and between these responses and latitude and longitude were affected by calculating these responses while taking detectability into account and when ignoring possible detectability influences, i.e., using unadjusted BBS counts as measures of species abundances for a route. For MASS, CI, and PC, we compared results based on raw counts to results based on counts that were adjusted using the methodology adopted by Partners in Flight to estimate global population sizes of landbird species (Rosenberg and Blancher 2005, Thogmartin et al. 2006) (http://www.rmbo.org/pif_db/laped/download/Guide%20to%20PIF%20Population%20Estimates%20Database%202.pdf). The methodology we describe here was finalized in 2007, using BBS data from 1990–1999 to make global population estimates. To make these global estimates, species counts from individual routes were modified in three ways: (1) a pair adjustment, (2) a time-of-day adjustment, and (3) a detection distance adjustment. Briefly, (1) the number of birds of a given species was multiplied by a pair adjustment that adjusts for the fact that each bird seen or heard is likely a member of a pair, (2) the number of birds of a given species was multiplied by a pair adjustment that adjusts for the fact that detection of birds varies by time of day, and (3) the area over which a species was effectively detected was adjusted by applying a detection distance estimated for each species. After making all three adjustments, the route data on species abundances were used to make estimates of global population size. We tested possible effects of detectability on MASS, CI, and PC by calculating these responses with and without these three adjustments.

The effect of detectability – estimating how many species were not observed in a survey - on species richness is often approached through the use of asymptotic species estimators (Chao 2005) (Gotelli and Colwell 2011) that use information on rarely encountered species (e.g., species counted once (singletons) or twice (doubletons)) to estimate whether there are other rarely encountered species that have not been observed in a survey. Based on the results of comparative analyses of the biases associated with different estimators (Hortal et al. 2006, Reese 2012), we selected the ACE (abundance coverage estimator) (Colwell 2009) to account for lack of detection of species in estimates of richness along the BBS routes.

Estimating how detectability might affect CS, the Combined Score averaged over five species, is challenging. The question we asked here is whether the most highly threatened species, those with the highest CS score, are more likely not to be detected in BBS surveys than species with lower CS scores and, if so, how might that affect CS trends. We asked first whether CS scores were negatively correlated with bird counts along BBS routes. If so, this would be an indication that birds with high CS scores were less frequently detected along BBS routes than were birds with lower CS scores. Second, we simulated how not detecting high CS birds might affect the results in our analyses. For this simulation, we eliminated the highest CS bird from our five species CS average and replaced it with the sixth highest CS scoring species. In essence, this examined how the results would change if we had not detected the highest CS species in the BBS surveys we examined. Not detecting such species should, however, not be common because seven years of data were examined for detection of each species and a single detection within that period would include that species on the detected list for a route.

Results

FigB1

Fig. B1. Pearson correlations between Combined Score of individual species and average yearly counts of those species (number birds counted per BBS route per year from 1997–2004).


 

Starting with CS, we see in Appendix B, Fig. 1 above that the Continental Combined Score for individual species does tend to be negatively correlated with frequency of observation, although the correlations are relatively weak. This suggests that species with high CS may be infrequently counted, indicating that detectability might affect the CS response variable, the average Continental Combined Scores for the five observed species with the highest Continental Combined Scores per BBS route (mean number of species per route = 60, range = 15–102).

In the analyses in the paper, detectability adjustments were made to CI and RICH. The denominator of CI (Eq. 1 in paper), the global population of each species, is based on the BBS counts adjusted for pairs, time of day, and distance of detection so a similar adjustment was used in the numerator (the abundance of the species on the route).

Richness is typically adjusted for detection in large scale studies of avian richness patterns (e.g., (Culbert et al. 2012)).

We see in Appendix B Table 1(D) that adjusted and non-adjusted values for each response were highly correlated (r > 0.9) with the lowest correlation for MASS and the highest for RICH. This suggests that incorporating detectability into MASS calculations might have a relatively greater effect on MASS than incorporating detectability into calculation of the other responses. Certainly, as calculated, MASS will underestimate total avian biomass along a route because MASS does not incorporate non-detected birds in the biomass calculation.

Correlations between responses are similar when detectability is incorporated or is not incorporated ((B) and (C) in Table 1). Correlations between responses and longitude differ very little among the five responses. Correlations between responses and latitude are also similar between adjusted and non-adjusted responses, with the possible exception of MASS for which the correlation with latitude changes from 0.03 when detectability is included in MASS calculations versus -0.19 when detectability is not incorporated. The similarity of correlation patterns between adjusted and non-adjusted responses, gives us an indication that the observed patterns of spatial congruence presented in the text are not substantially altered by differences in detectability but that spatial patterns of MASS might be most affected by detectability among the five response.

Table B1. Pearson correlations among the five responses and between the responses and latitude and longitude, (A) as presented in paper, (B) including detectability adjustments, (C) excluding detectability adjustments. (D) Correlation between detectability corrected and not corrected values of each response.

(A) ORIGINAL

MASS

RICH

CS

CI

Latitude

Longitude

MASS

 

 

 

 

-0.19

0.36

RICH (detection adjusted)

0.12

 

 

 

-0.06

0.62

CS

0.08

0.35

 

 

-0.52

0.01

CI (detection adjusted)

0.02

0.16

0.59

 

-0.27

-0.15

PC

-0.38

-0.69

-0.12

0.07

0.30

-0.74

 

 

 

 

 

 

 

(B) DETECTABILITY INCLUDED

MASS

RICH

CS

CI

Latitude

Longitude

MASS

 

 

 

 

0.03

0.33

RICH

0.12

 

 

 

-0.06

0.62

CS

-0.06

0.48

 

 

-0.49

0.03

CI

0.05

0.16

0.58

 

-0.27

-0.15

PC

-0.23

-0.75

-0.27

0.02

0.31

-0.75

 

 

 

 

 

 

 

(C) DETECTABILITY NOT INCLUDED

MASS

RICH

CS

CI

Latitude

Longitude

MASS

 

 

 

 

-0.19

0.36

RICH

0.12

 

 

 

-0.10

0.63

CS

0.07

0.39

 

 

-0.52

0.01

CI

0.13

0.13

0.52

 

-0.29

-0.16

PC

-0.38

-0.68

-0.12

0.07

0.30

-0.74

 

 

 

 

 

 

 

(D)

Detect:No Detect Correlation

 

 

 

 

 

MASS

0.91

 

 

 

 

 

RICH

0.99

 

 

 

 

 

CS

0.95

 

 

 

 

 

CI

0.92

 

 

 

 

 

PC

0.95

 

 

 

 

 

Literature Cited

Boulinier, T., J. D. Nichols, J. R. Sauer, J. E. Hines, and K. Pollock. 1998. Estimating species richness: the importance of heterogeneity in species detectability. Ecology 79:1018–1028.

Chao, A. 2005. Species richness estimation. Pages 7909–7916 in N. Balakrishnan, C. B. Read, and B. Vidakovic, editors. Encyclopedia of Statistical Sciences. Wiley, New York.

Colwell, R. K. 2009. EstimateS: Statistical estimation of species richness and shared species from samples. Version 8.20.  Available from http://purl.oclc.org/estimates (accessed 13 August 2009).

Culbert, P. D., V. C. Radeloff, V. St-Louis, C. H. Flather, C. D. Rittenhouse, T. P. Albright, and A. M. Pidgeon. 2012. Modeling broad-scale patterns of avian species richness across the Midwestern United States with measures of satellite image texture. Remote Sensing of Environment 118:140–150.

Gotelli, N. J., and R. K. Colwell. 2011. Estimating species richness. Pages 39–54  Biological diversity: frontiers in measurement and assessment.

Hortal, J., P. A. V. Borges, and C. Gaspar. 2006. Evaluating the performance of species richness estimators: sensitivity to sample grain size. Journal of Animal Ecology 75:274–287.

Reese, G. C. 2012. Simulating species assemblages and evaluating species richness estimators. Colorado State University, Colorado Springs, Colorado, USA.

Rittenhouse, C. D., A. M. Pidgeon, T. P. Albright, P. D. Culbert, M. K. Clayton, C. H. Flather, J. G. Masek, and V. C. Radeloff. 2012. Land-Cover Change and Avian Diversity in the Conterminous United States. Conservation Biology 26:821–829.

Rosenberg, K. V., and P. J. Blancher. 2005. Setting numerical population objectives for priority landbird species. Pages 57–67 in C. J. Ralph and T. D. Rich, editors. Bird Conservation Implementation and Integration in the Americas: Proceedings of the Third International Partners in Flight Conference. U.S. Department of Agriculture Forest Service, General Technical Report.

Sauer, J. R., and W. A. Link. 2011. Analysis of the North American Breeding Bird Survey using hierarchical models. Auk 128:87–98.

Thogmartin, W. E., F. P. Howe, F. C. James, D. H. Johnson, E. T. Reed, J. R. Sauer, and F. R. Thompson, III. 2006. A review of the population estimation approach of the North American landbird conservation plan. Auk 123:892–904.


[Back to A024-046]