Report on uncertainty in phytoplankton metrics

A central tenet of the Water Framework Directive (WFD) is that attributes of biological communities are used to make inferences about the ecological quality of rivers, lakes and coastal/transitional waters. For lakes, the phytoplankton are a key biological community, or Biological Quality Element (BQE), to be used for this purpose. Phytoplankton communities are complex, containing many species responding to changes in many different environmental factors. In order to use phytoplankton to indicate ecological quality, it is necessary to develop metrics that describe high-level properties of the phytoplankton community that are sensitive to environmental pressures, such as nutrient enrichment. In the earlier stages of the WISER project, six candidate metrics were proposed as either phytoplankton composition or bloom metrics, to be considered alongside the established metric of chlorophyll a (chl), which is used to represent phytoplankton abundance.If such metrics are to be useful indicators of differences in ecological quality among lakes, it is necessary to know the extent to which they can be affected by sampling and sample processing procedures. Phytoplankton communities vary spatially across lakes and so we could expect phytoplankton metrics to show similar variation among sampling locations. Also, the results of phytoplankton sample processing depend upon procedures used to analyse samples in the laboratory and the expertise of the analyst processing the sample (people vary in their abilities to discriminate between species). If phytoplankton metrics vary more with differences in sampling and sample processing procedures within a lake than they do among lakes of different pressure then these metrics are unlikely to provide a sensitive means of describing differences in the biological impacts of an environmental stressor among lakes. On the contrary, little variation with sampling and sample processing compared to variation among lakes would indicate the potential for relationships between metrics and environmental pressures to be detected after accounting for sample/sample processing "errors". Here we analyse the results of a multi-scale field campaign of 32 European lakes, to resolve the extent to which chl and the six proposed WISER phytoplankton metrics vary among lakes and with sampling/sample processing. We also relate these metrics to different environmental variables, including total phosphorus concentration as an indicator of eutrophication. The seven metrics analysed show some similarities in their variation among countries, lakes and samples, indicating that some of them reveal similar information on changes in phytoplankton communities. However correlations among metrics are not perfect, suggesting that each metric is potentially also showing a unique aspect of phytoplankton community change. For all seven metrics, between 65% and 96% of the variability in metric scores is due to variability among lakes, much higher than variability occurring due to sampling/sample processing. After the among-lake variation, errors associated with sub-sampling of field samples (2-13%) and variation in analysts (2-19%) are more important than spatial variation between locations within a lake (<1-4%). Importantly, among-lake variation in most of the metrics can be related to differences in total phosphorus concentrations, though this effect is often better described when also taking account of some aspects of lake geography (longitude and altitude). These variables typically explain only a modest proportion of total among-lake metric variability, indicating that metrics are additionally sensitive to unmeasured environmental factors. These positive findings on metric robustness and sensitivity to pressure are especially true for four candidate phytoplankton metrics being considered for Intercalibration: chlorophyll, PTI, MFGI and cyanobacterial blooms, for which > 85% of the variability in metric scores was attributed between lakes, and, total phosphorus concentration was the best single predictor of variation in these metrics. This WISER study, therefore, concludes that these four proposed metrics are robust metrics for ecological status assessment and are suitable for adoption in the Intercalibration process. The study was carried out following clear sampling, counting and identification guidelines and it indicates that rigorous standardisation of procedures helps minimise sampling and analytical variability and makes possible more meaningful comparisons of ecological status among lakes.