Multivariate-statistics based selection of a benthic macroinvertebrate index for assessing water quality in the Paute River basin (Ecuador)

Abstract Multivariate statistics -Soft Independent Modelling of Class Analogies (SIMCA), Principal Components Analysis (PCA), Multiple Regression (MR)- were used to search for key biotic and water quality (WQ) variables within a dataset/matrix collected over a five-year period in the Paute River Basin (Ecuador). Benthic macroinvertebrates and 27 descriptive physical, chemical, microbiological, hydrological and geomorphological variables were collected from 64 monitoring sites across the basin. Nine macroinvertebrate biotic indices were calculated. The SIMCA method was applied to find the most accurate biotic index that best discriminated among less polluted (C1), moderately polluted (C2) and highly polluted (C3) sites. A cross-validation scheme was applied to evaluate the performance of the modelling process. Within the PCA that was further refined using a Kruskal-Wallis test, the key WQ variables that mostly contributed to the macroinvertebrate-based WQ classification were identified. The results showed that the Elmidae-Plecoptera-Trichoptera (ElmPT) index was the most accurate biotic classifier. Riparian vegetation and streambed heterogeneity were the best predictors of the C1 class, while the concentration of faecal coliforms, pH, temperature and dissolved oxygen, best predicted the C3 class. The reduction of the field monitored parameters could help designing more cost-effective but equally accurate future WQ monitoring schemes in the basin.