Tools for bioindicator assessment in rivers: The importance of spatial scale, land use patterns and biotic integration

This study outlines an alternative, versatile and flexible procedure to the Assessment System for the Ecological Quality of Streams and Rivers throughout Europe using Benthic Macroinvertebrates (AQEM) protocol for selecting and assessing candidate bioindicators for Water Framework Directive (WFD) compliant monitoring programmes. Based on different forms of benthic macroinvertebrate data using relative abundance of family level taxonomic groups, metrics and traits collected during spring 2010 at 96 lotic sites across northern Portugal, the procedure employs components of top down and bottom up analytical processes and introduces the concept of niche breadth to assess biological quality element response to environmental and stressor parameters across different spatial levels. Random Forest classification revealed that fractal and non-fractal land use metrics at basin and local level were extremely important determinants of Water Framework Directive determination of “Good” ecological quality, particularly at the local scale. The amount of urbanization at the lower spatial level was a particularly important determinant of ecological quality, while the extent and type of forest (especially coniferous) was more important at higher, river basin scale. Distance-based linear models (DISTLM) and distance-based redundancy analysis (dbRDA) were used to determine associations between invertebrate data and non-redundant environmental predictors selected using the Akaike Information Criterion (AIC). Results revealed a stronger association between invertebrate traits and selected environmental predictors compared to the other types of invertebrate data, although some association between invertebrate relative abundance and eutrophication was detected. Principal Components Analyses (PCA) were run for the non-redundant sets of predictors for each macroinvertebrate data set to extract an environmental quality gradient along the first axis. Niche breadth, calculated for candidate indicators to avoid bias resulting from expert judgement, was distributed by rank along its respective PCA gradient. Five candidate indicators for each data type were selected for their preference for the most pristine sites and five were selected due to their close link with the most degraded streams. Candidate bioindicators for impacted sites tended to be stenobiotic in character, due to the impoverished structural and functional diversity associated with such conditions. Finally Partial Least Squares Regression was used to refine and validate selected candidate metrics, to produce a comprehensive final list of macroinvertebrate measures of ecological quality.