A Virtual Museum on the State's Fish Biodiversity

Bioassessment Analyses

This page is a preliminary description of an ongoing analysis involving the application of species distribution models towards bioassessment of fish communities.

Our poster (download link) presented February 11, 2012 at the annual meeting of the Texas Chapter American Fishery Society is our latest installment on this aspect of our research.



Recent literature reviews of bioassessment methods question the common use of reference sites to define environmental benchmarks, pointing out that use of least-disturbed sites as references in developed landscapes is subjective and supports management of steadily declining ecosystem health. We explore an alternative approach to bioassessment by using species distribution models as environmental filters to construct a fish community model across a large and environmentally diverse landscape to serve as a reference state benchmark of taxonomic completeness.


Texas, USA.


We used a maximum entropy algorithm to construct distribution models for 100 species in a fish community to compare model-based predicted composition against empirical fish community data from 269 sites sampled by four independent surveys (Figure 1 below); two with robust approximations of presence and absence allowing for model evaluation, and two with associated multimetric-based index of biotic integrity (IBI) scores and methodologies typical of state agency bioassessment efforts. We then compare observed/predicted ratios to IBI scores and an independent measure of human influence, the National Fish Habitat Action Plan’s (NFHAP) index of cumulative disturbance to river fish habitats from landscape anthropogenic activity.


Numbers of species observed were moderately to strongly correlated with predictions among survey datasets (see Figure 2 below). Deviations between model predictions and survey observations were correlated with predictions in ways that indicate influence of species-specific prevalence, environmental quality, geographic patterns of species richness, and sampling intensity. We found significant, though weak, relationships between observed/predicted ratios and IBI scores, and site-specific values from the NFHAP index showed very weak and non-significant relationships to both our observed/predicted ratios and to the IBI scores.

Main Conclusions

Our results suggest that this model-based approach to bioassessment yields similar results to traditional methods based on reference sites while addressing many of the weaknesses of those methods and offering a promising heuristic platform for efficient investigation and application of assessments at large spatial scales.


This project grew out our use of species distribution models to reconstruct a historical fish community composition benchmark for Barton Creek (see Labay et al. 2011). We are exploring statewide application and adaptation of the concept, reconstructing site-specific (expected) taxonomic composition to be compared with newer data not included in the models (observed composition) to assess health of stream fish communities. The result would be a product analogous to commonly used regional multi-metric-based indices of biotic integrity (IBIs) evaluations as used in natural resource management in the U.S., but what we are attempting is in many ways more like the River InVertebrate Prediction and Classification System (RIVPACS, Wright et al. 1984) approach commonly used in Europe and Australia.

The need for such assessment has been pointed out in recent literature reviews on biological indicator systems in floodplain rivers (Dziock et al. 2006), bioassessment of freshwater biota (Dolédec & Statzner 2010), and most importantly the mis-application of reference conditions (Hawkins et al. 2010). These reviews all concur that new approaches for bioassessment need to allow assessment at large scales to match environmental policy and to move towards predictive assessment of the deviation of fish community condition from natural states. We believe these goals are near impossible to attain, especially long term for developed landscapes, with the current practice of using least-disturbed reference sites as benchmarks. The conclusions we derive regarding ecological health of a system depend entirely on how we estimate benchmarks, and the frequent use of least-disturbed sites in bioassessment supports management of steadily declining system health (Pauly 1995). Additional criticism of the application of the reference site approach include difficulty and cost associated with finding sufficient quantity (minimum 20 sites/grouping, recommended 30 - 50; (Bowman & Somers 2005)) and consistent quality of reference sites (Chessman et al. 2008; Hawkins et al. 2010), inconsistent reference definitions (Stoddard et al. 2006; Herlihy et al. 2008), inability to identify specific stressor mechanisms in typical multimetric (Hawkins et al. 2010) and multivariate (Dolédec & Statzner 2010) approaches, and non-scalability or non-transferability of indices across regions and studies (Cao & Hawkins 2011).

Much of the criticisms of reference site-based methodologies of bioassessment (Suter II, 1993; Karr and Chu, 1999; Norris and Hawkins, 2000; Seegert, 2000) hinge on the particular methods of estimating a reference site benchmark. The two most common approaches include regionalization (or classifications) and discriminative modeling using environmental covariates. Hawkins et al. (2010) points out that the former approach has insufficient precision or numerical criteria to detect ecologically meaningful deviations, and that the latter approach exemplified by The River Invertebrate Prediction and Classification System (RIVPACS; Wright et al., 1984) and its derivatives (e.g., AUSRIVAS; Turak et al., 1999) still relies on reference sites for calibration and does not allow extrapolation beyond the range of calibration data, limiting transferability. Recent bioassessment literature reviews (Dolédec and Statzner, 2010; Hawkins et al., 2010) point to Chessman and Royal’s (2004) approach as an alternative that circumvents the need for reference sites (also see Speight and Castella, 2001; Chessman, 2006; Labay et al., 2011). Chessman and Royal (2004) started with a known pool of potential colonists and used coarse-scale environmental parameters and known tolerances of species to them as filters to predict community composition for any site. Since the environmental variables they used are independent of anthropogenic influence, so too should be their community predictions. Hawkins et al. (2010) considered Chessman and Royal’s (2004) approach conceptually appealing, but point out that it has “yet to be tested thoroughly.” While this might be true in applications in the stream bioassessment literature (but see Chessman, 2006; Labay et al., 2011), major advances in the spatial modeling of biodiversity at the community level (Austin, 2002; Scott, 2002; Guisan and Thuiller, 2005; Ferrier and Guisan, 2006; Mateo et al., 2012) can now be leveraged to aid construction of meaningful ecological benchmarks without use of reference sites. Within the general field of spatial modeling, thanks in large part to advancements in digitization of biological occurrence databases and large-scale environmental coverages, species distribution models (SDMs) have seen particularly rapid growth and diverse applications (Peterson et al., 2011). We suggest that bioassessment without reference sites can be achieved using SDMs as environmental filters. This amounts to what Ferrier and Guisan (2006) identify as a “Predict first, assemble later” approach to modeling communities. Compared to other community modeling approaches, this method has the advantage of producing individualistic species responses and allows for disparate surveys to be combined (Gioia and Pigott, 2000; Guisan and Theurillat, 2000; Lehmann et al., 2002), whereas existing bioassessment methods require consistent community sampling methodology across all sites to be assessed. Conceptually this approach is not novel (see Guisan and Theurillat, 2000; Olden, 2003; Peppler-Lisbach and Schröder, 2004; Gelfand et al., 2005; Leathwick et al., 2005; Baselga and Araujo, 2010; Mateo et al., 2012), however, much research is needed toward assessing how this applies in stream bioassessment and compares to commonly used techniques such as the classic multimetric-based indices of biotic integrity (IBI) assessments (Karr, 1991).

We apply an environmental filters approach (Poff 1997) modeled after Chessman’s (2006) work predicting Australian fish assemblages, but our “filters” are much more complex and advanced species distribution models (SDMs) that are now widely accepted for predicting organismal occurrences and habitat suitability. To our knowledge this is the first use of SDMs in such a filters-based approach to assess fish assemblage condition.

Our analysis compares recent survey data to expectations derived from SDMs that approximate historical suitability by excluding variables influenced from human actions in model development (see Labay et al. 2011 for a detailed explanation of models). We can thus produce observed/predicted ratios similar to those of RIVPAC approaches. We differ from a RIVPAC approach, however, by avoiding use of reference conditions.

Responsive image

Figure 1. Distribution of sample localities for the 4 survey datasets.

Responsive image

Figure 2. Relationship between number of native fish species predicted and observed per site for all survey datasets. All regressions are significant at α = 0.05. The solid line represents equality of predicted and observed richness.

Literature Cited

Austin, M., 2002. Spatial prediction of species distribution: an interface between ecological theory and statistical modelling. Ecological modelling, 157(2-3), pp.101–118.

Bond, N. et al., 2011. Using species distribution models to infer potential climate change-induced range shifts of freshwater fish in south-eastern Australia. Mar. Freshwater Res., 62(9), pp.1043-1061.

Bowman, M. & Somers, K., 2005. Considerations when using the reference condition approach for bioassessment of freshwater ecosystems. Water Quality Research Journal Of Canada, 40(3), pp.347-360.

Cao, Y. & Hawkins, Charles P., 2011. The comparability of bioassessments: a review of conceptual and methodological issues1. Journal of the North American Benthological Society, 30(3), pp.680-701.

Chessman, B.C., 2006. Prediction of riverine fish assemblages through the concept of environmental filters. Marine and freshwater research, 57(6), pp.601–609.

Chessman, B., Muschal, M. & Royal, M., 2008. Comparing apples with apples: use of limiting environmental differences to match reference and stressor-exposure sites for bioassessment of streams. River Research and Applications, 24(1), pp.103–117.

Chessman, B.C. & Royal, M.J., 2004. Bioassessment without reference sites: use of environmental filters to predict natural assemblages of river macroinvertebrates. Journal of the North American Benthological Society, 23(3), pp.599–615.

Dolédec, S. & Statzner, B., 2010. Responses of freshwater biota to human disturbances: contribution of J-NABS to developments in ecological integrity assessments. Journal of the North American Benthological Society, 29(1):286-31.

Dziock, F. et al., 2006. Biological indicator systems in floodplains–a review. International review of hydrobiology, 91(4), pp.271–291.

Guisan, A. & Thuiller, W., 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters, 8(9), pp.993-1009.

Hawkins, C. P, Olson, J.R. & Hill, R.A., 2010. The reference condition: predicting benchmarks for ecological and water-quality assessments. Journal of the North American Benthological Society, 29(1), pp.312–343.

Herlihy, A.T. et al., 2008. Striving for consistency in a national assessment: the challenges of applying a reference-condition approach at a continental scale. Journal of the North American Benthological Society, 27(4), pp.860–877.

Kleinsasser, L. et al., 2009. Status of biotic integrity, water quality, and physical habitat in wadeable east Texas streams. Resources Protection Division, Texas Parks and Wildlife Department, River Studies Report;No. 19.

Labay, B. et al., 2011. Assessing Historical Fish Community Composition Using Surveys, Historical Collection Data, and Species Distribution Models. PLoS ONE, 6(9), p.e25145.

Pauly, D., 1995. Anecdotes and the shifting baseline syndrome of fisheries. Trends in Ecology and Evolution, 10(10), p.430.

Poff, N.L.., 1997. Landscape filters and species traits: towards mechanistic understanding and prediction in stream ecology. Journal of the North American Benthological Society, 16(2), pp.391–409.

Scott, J.M., 2002. Predicting species occurrences: issues of accuracy and scale, Island Pr.

Stoddard, J.L. et al., 2006. Setting expectations for the ecological condition of streams: the concept of reference condition. Ecological Applications, 16(4), pp.1267–1276.

Turak, E. et al., 1999. Assessment of river condition at a large spatial scale using predictive models. Freshwater Biology, 41(2), pp.283-298.

Wright, J.F. et al., 1984. A preliminary classification of running‐water sites in Great Britain based on macro‐invertebrate species and the prediction of community type using environmental data. Freshwater Biology, 14(3), pp.221-256.