Considerations for analyzing ecological community thresholds in response to anthropogenic environmental gradients

Abstract

Abstract. The goal of this paper is to help managers better understand implications of using aggregate community metrics, such as taxon richness or Indices of Biotic Integrity (IBI), for detecting threshold responses to anthropogenic environmental gradients. To illustrate, we offer an alternative analytical approach, Threshold Indicator Taxa ANalysis (TITAN), geared toward identifying synchronous changes in the distribution of multiple taxa as evidence of an ecological community threshold. Our approach underscores the fundamental reality that which taxa are affected by stressors is important, both from a conservation standpoint and because taxon-specific life-history traits help us understand relevant mechanisms. First, we examine macroinvertebrate community response to an impervious cover gradient using a well-studied biomonitoring data set to show that representative community metrics are relatively insensitive to synchronous threshold declines of numerous individual taxa. We then reproduce these response relationships using a simulated community data set with similar properties to demonstrate that linear or wedge-shaped responses of community metrics to anthropogenic gradients can occur as an artifact of aggregating multiple taxa into a single value per sampling unit, despite strong nonlinearity in community response. Our findings do not repudiate the use of community metrics or multimetric indices, but they challenge assumptions that such metrics are capable of accurately reflecting community thresholds across a broad range of anthropogenic gradients. We recommend an alternative analysis framework that begins with characterization of the responses of individual taxa and uses aggregation only after distinguishing the magnitude, direction, and uncertainty in the responses of individual members of the community.

A critique of the use of indicator-species scores for identifying thresholds in species responses

“…In practice, we have observed a variety of responses to novel gradients, and we refer interested readers and potential users to our description in fig. 7 of King and Baker (2010).…”

“…Moreover, the term threshold appears to have different meanings for different people. We may have inadvertently contributed to the confusion because some of the simulations initially used to test TITAN involved underlying step-functions in the probability distribution of individual taxa (figs 3, 4 in Baker and King 2010), and we have documented that many taxa declining along anthropogenic disturbance gradients do show disproportionate change in frequency and abundance with an incremental change in the gradient (i.e., a threshold; King and Baker 2010. However, in TITAN, a change point really means what its name suggests (see next sections).…”

“…We developed more realistic simulations using negative binomial distributions with overdispersion in R (version 2.15.1; R Development Core Team, Vienna, Austria; Appendix S1; available online from: http://dx.doi.org/10.1899/12-142.1.s1). We set mean abundance across observations to range from 2 to 10 individuals, which approximates the range of empirical values we have observed in our analyses of various survey data sets (e.g., King and Baker 2010, Bernhardt et al 2012. We set mean abundance to be a function of the response models suggested by fig.…”

“…In empirical data, truncated distributions usually are preceded by low abundances and a marked increase in absences ( fig. 7 by King and Baker 2010), so TITAN typically identifies this zone rather than the point of intersection. Here, we differ philosophically with C&Q because we think that a point of sudden extinction along an environmental gradient is both statistically and ecologically a location of far greater change in a taxon's response than either a break in slope or a peak in a unimodal distribution, especially when compared to the subtle change in slopes used to define Type 2 thresholds.…”

Diatom taxa and assemblages for establishing nutrient criteria of lakes with anthropogenic hydrologic alteration

“…Linear stressor-response relationships between nutrients and chlorophyll a (Chl a) are often shown in numerous waterbodies, since Chl a is a surrogate for algal biomass and strongly affected by nutrient concentrations (Huo et al, 2013a(Huo et al, , 2014. However, nonlinear statistical methods that define ecological change points have become prevalent for deriving nutrient criteria, since the biological responses to environmental stressor are always non-linear, nonnormal, and heterogeneous (King and Richardson, 2003;King and Baker, 2010). Moreover, non-linear statistics are useful for visually facilitating interpretation, revealing data structures, and displaying interactions.…”