Combining best professional judgement and quantile regression splines to improve characterisation of macrofaunal responses to enrichment
Keeley, NB and MacLeod, CK and Forrest, BM, Combining best professional judgement and quantile regression splines to improve characterisation of macrofaunal responses to enrichment, Ecological Indicators, 12, (1) pp. 154-166. ISSN 1470-160X (2012) [Refereed Article]
Many benthic quality indices rely on categorising impacts by assigning species to ecological-groups (EGs) that reflect their tolerance to pollution. This is usually based on best professional judgement (BPJ) by experts with access to relevant ecological and taxonomic information. However, international applicability of such indices is restricted in areas where the species taxonomy, biology and response to pollution are poorly understood. In this study we describe an approach that enables objective allocation of EGs in situations where species information is limited. This approach utilised BPJ to categorise the environmental condition of benthic habitats around fish farms in New Zealand in relation to defined enrichment stages (ESs). Quantile regression was then used to model distributions of select taxa. The experts assigned ES scores from 1 to 7, for stations that ranged from relatively natural to excessively enriched (i.e. near-azoic), respectively, with judgements based on a suite of quantitative and qualitative indicators of enrichment, but without reference to detailed species information. The individual BPJ estimates were highly correlated, with minimal bias, indicating good agreement among the experts. Forty key indicator taxa were identified and quantile regression models based on ES (derived as a continuous explanatory variable) were fitted for 34. Abundances of the same taxa were also modelled in response to a more traditional enrichment indicator (organic content, %OM) for comparison with the BPJ technique. The regression approach characterised enrichment responses and objectively identified both the upper and lower tolerance limits of a range of taxa according to their ES and %OM. The models discriminated a number of key indicator taxa, including several that were responsive to low-level changes in ES, but not necessarily %OM. There was reasonable agreement (59%) between EGs derived using the regression approach and those defined using the AMBI database (one of the most commonly applied benthic quality indices). Moreover, the regression method allowed the classification of 10 additional taxa for which our ecological understanding was limited. A key outcome of this study was the acknowledgement that EG characterisations for species need to be regionally validated, no matter how well defined they might appear to be. The combined BPJ/regression analysis approach described provides a valid means of both assigning and validating EG classifications, which will be particularly useful in situations where the taxa are poorly defined, and will enable existing biotic indices to be more broadly applied and interpreted.