A new model linking macroinvertebrate assemblages to habitat composition in rivers: development, sensitivity and univariate application
Kiesel, Jens; Schröder, Maria; Hering, Daniel; Schmalz, Britta; Hörmann, Georg; Jähnig, Sonja C.; Fohrer, Nicola
published: Feb 1, 2015
ArtNo. ESP141018601006, Price: 29.00 €
Habitat models are frequently used to simulate species occurrence and abundance in rivers and lakes. While most of these models target individual species, there is yet no spatially distributed macroinvertebrate community model whose output can be directly linked to ecological stream assessments. These assessments are usually based on metrics calculated with a list of occurring taxa and their abundances. Such a model would allow simulating the effect of environmental changes, e.g. due to climate or land use change, on macroinvertebrate assemblages. This paper describes the development, sensitivity analysis, calibration and application of the empirical, statistical macroinvertebrate community model HET (Habitat Evaluation Tool). The model requires three types of input data: (1) Reference data: Habitat-specific, quantitative macroinvertebrate lists of taxa for reference sites, which are grouped into a combination of environmental variables (Habitat Sensitivity Classes, HSCs). (2) Environmental data: Spatial distribution of HSCs for the sites where the macroinvertebrate assemblages are simulated. (3) Biotic data: Macroinvertebrate abundances for the calibration of the HET model. The model is applied in two steps: First, the species abundances in each HSC of the test site are calculated from the reference data. Transformation formulas are used to find an optimum relation between reference and test species lists. Suitability values are calculated for each HSC and each species, from which species master lists per HSC are derived. These master lists are then randomly 'sampled' using bootstrapping to mimic the randomness of real-world sampling with the option to reproduce model results through high bootstrapping repetitions. Second, the species assemblage within the targeted stream section is calculated through applying an electronic Multi-Habitat-Sampling (MHS), which is frequently used in river assessment. The resulting species lists can be used to calculate metrics and ecological status. A first univariate application of the tool was carried out in Schleswig-Holstein (Germany). A sensitivity analysis showed that the transformation formula settings have the highest influence on model performance, followed by quality and filtering of the input data. Model simulations reached a Renkonen Index (RI) of 56 between simulated and observed species abundance within the HSCs of the test site, which indicates a high agreement. Based on these results we carried out two MHS, using substrate maps from 2008 and 2009. For each MHS application, we ran 1,000 repetitions to evaluate the impact of random sampling on the overall MHS result. Random sampling variance impacted ecological status classes and metrics, but did not mask the influence of the substrate changes from 2008 to 2009. We thus conclude that the model is useful for assessing the impact of substrate changes on macroinvertebrate assemblages and ecological status.