Prediction of taxon occurrence: a test on taxon‐specific change point values of stream benthic invertebrates

Abstract

Summary Ecological thresholds represent abrupt species responses to minor changes in the environment. However, as aquatic organisms are frequently exposed to multiple stressors, prediction of taxon occurrence is challenging and often associated with much uncertainty. In this study, we analysed the response of stream benthic invertebrates to selected environmental gradients and determined the frequency and predictability of threshold responses. Re‐analysis of taxon‐specific change point values for 213 riverine benthic invertebrate taxa and nine water quality variables showed that the predictive performance was acceptable for 30.4% and excellent for 9.6% of the total number of validated change points (n = 978). In total, 48.3% of the taxa that were predicted to occur (taxon presence predicted) and 67% of the taxa predicted not to occur (taxon absence predicted) were correct but with high variability across taxa and chemical variables. Predictive performance was the highest for stenoecious taxa and for certain chemical variables, such as ammonium nitrogen and the daytime dissolved oxygen concentration. Responses to peak level concentrations of water quality variables were less threshold‐like compared with relationships with mean values. However, model predictions were equal for both of the approaches. Successful prediction of taxon absence at stress levels beyond the change point values showed that the latter can well represent the upper limit of the taxon‐specific range of occurrence. Thus, change points are valuable for establishing and evaluating new water management tools, such as water quality criteria. By contrast, prediction of taxon presence at low levels of stress was associated with relatively high uncertainty, reflecting the limitation of taxon occurrence by stressors other than those investigated.

Spatial, temporal and experimental: Three study design cornerstones for establishing defensible numeric criteria in freshwater ecosystems

“…Water quality practitioners increasingly recognize that algal assemblage response measures are sensitive to nutrient pollution in streams and should be incorporated in studies designed to establish regional criteria for nutrients in streams (United States datasets for identifying algal and other biological responses to nutrient stressor gradients (Chambers et al, 2012;Dodds, Clements, Gido, Hilderbrand, & King, 2010). In particular, single-stressor models that are commonly used to establish criteria have the significant shortcoming of potentially confounding interactions among nutrient enrichment and natural environmental variables or other stressors (Leps, Leisner, Haase, & Sundermann, 2016;Smucker et al, 2013;Wagenhoff, Clapcott, Lau, Lewis, & Young, 2017). However, study designs that adequately represent regional gradients in enrichment and potentially confounding factors can quickly become prohibitively expensive.…”

“…Finally, we analysed the magnitude, direction and uncertainty of individual taxa and whole assemblage threshold responses to P‐enrichment gradients for each year separately using Threshold Indicator Taxa Analysis (TITAN; Baker & King, ). We performed TITAN with the TITAN2 package (Baker, King, & Kahle, ) in r using un‐transformed taxa abundances (King & Baker, ) for taxa with ≥3 occurrences. All partitions required a minimum number of three observations on both sides.…”

Setting Priorities in River Management Using Habitat Suitability Models

“…Bennetsen et al [22] indicate that there is a good relationship between the performance of the suitability models used and the inclusion of ecological knowledge in the construction of the models. For example, compared to the use of threshold indicator taxa analysis (TITAN) to detect change points [43], both the average sensitivity (0.57 vs. 0.48) and specificity (0.87 vs. 0.67) for the prediction of species occurrence are higher. Furthermore, using the habitat suitability models from Bennetsen et al [22], we assessed the predictive power for the MMIF of the models used.…”

Lost in translation: the German literature on freshwater salinization

“…0.04 -009 g l 21 could make the achievement of the good ecological condition unlikely in German rivers, although specific thresholds depend on the geology of the catchment (table 3). Other studies on the same dataset used change point analyses to identify that a mean Cl 2 concentration of around 25 mg l 21 marked a shifting point for invertebrate community composition [74,75], with most taxa being negatively affected once concentrations reached 50 mg l 21 . However, these values are still under debate: they are not only considerably lower than the toxicity thresholds revealed by laboratory and field studies so far, but they also seem to mismatch the geogenic hydrochemical situation of a number of running waters.…”