PLS-Pm for Ecological Integrity Mapping: Case Study of the Ave River Basin, Portugal

Abstract

The bonds between pollution sources, contaminant concentrations and ecological damage are quite complex to access, but the proper management of river basins requires the full understanding of those interactions. In the present study, four SEM-PLS models were used to set up environmental cause-effect relationships in a heavily polluted urban catchment: the Ave River Basin. Data were collected within this watershed relative to point source discharges, diffuse emissions, diffuse indicators (e.g. land use), contaminant concentrations and a Portuguese index of macroinvertebrates diversity (IPtIN). The same dataset was used in the four SEM-PLS models, but each run was based on a different combination of latent variables. In all cases, it was possible to recognize that nitrogen discharges from livestock production, industrial and urban treatment plants are the dominant threats to water quality. The model results were robust, characterized by high, adjusted R Squared values-from 0.744 to 0.931. The models were also used to draw water quality maps. Now, the purpose was to test their prediction capability. Since more than 80% of the water courses were correctly classified for IPtIN, the predictions were considered feasible.

Soil microbial communities in the face of changing farming practices: A case study in an agricultural landscape in France

“…This method is similar to the structural equation modelling but use partial least square instead of maximum likelihood as a criterion of assessment. It permits to obtain robust explanatory model [ 64 ]. It does not require any assumption about variables or error distribution.…”

Exploring the Effects of Landscape Metrics in Water Quality, Ave River Basin Case Study

“…In a previous study, the application of different statistical models was tested to evaluate which could be the most suitable form for prediction purposes [54], concluding that point source emissions and livestock production were the threats to water quality. In that study [54], landscape metrics were not included in the statistical models, and since such variables were not explored, it surged the necessity to study them. Resorting to a simpler method, by using Spearman Rank's Coefficient, it was possible to trace the effect of many variables.…”