1 Département de Géomatique Appliquée, Université de Sherbrooke, Sherbrooke, QC J1K 2R1, Canada; Département de Géomatique Appliquée, CARTEL - Centre d'Applications et de Recherche en TELédétection, Université de Sherbrooke, Sherbrooke, QC J1K 2R1, Canada; Département de Sciences Biologiques, GRIL - Groupement de Recherche Interuniversitaire en Limnologie, Université de Montréal, Campus MIL, Montréal, QC H3C 3J7, Canada. Electronic address: anais.oliva@usherbrooke.ca.
2 Département de Sciences Biologiques, GRIL - Groupement de Recherche Interuniversitaire en Limnologie, Université de Montréal, Campus MIL, Montréal, QC H3C 3J7, Canada; Department of Biology, Concordia University, Montréal, QC H4B 1R6, Canada.
3 Département de Sciences Biologiques, GRIL - Groupement de Recherche Interuniversitaire en Limnologie, Université de Montréal, Campus MIL, Montréal, QC H3C 3J7, Canada; Department of Biology, Concordia University, Montréal, QC H4B 1R6, Canada; Environment

Lakes are sentinels of environmental changes within their watersheds including those induced by a changing climate and anthropogenic activities. In particular, contamination originating from point or non-point sources (NPS) within watersheds might be reflected in changes in the bacterial composition of lake water. We assessed the abundance of potentially pathogenic bacteria (PPB) sampled in 413 lakes within 8 southern Canadian ecozones that represent a wide diversity of lakes and watershed land use. The study objectives were (1) to explore the diversity of PPB; (2) to build a fecal multi-indicator from a cluster of co-occurring PPB; and (3) to predict the fecal multi-indicator over thousands of lakes. We identified bacterial taxa based on 16S rRNA amplicon sequencing and clustered 33 PPB matching taxa in the Canadian ePATHogen database using a Sørensen dissimilarity index on binary data across the sampled lakes. One cluster contained Erysipelothrix, Desulfovibrio, Bacteroides, Vibrio and Acholeplasma and was related to the NPS fraction of agriculture and pasture within the watershed as its main driver and thus it was determined as the fecal multi-indicator. We subsequently developed a fecal multi-indicator predictive model across 200 212 southern Canadian lakes which explained 55.1% of the deviance. Mapping the predictions showed higher fecal multi-indicator abundances in the Prairies and Boreal Plains compared to the other ecozones. These results represent the first attempt to map a potential fecal multi-indicator at the continental scale, which may be further improved in the future. Lastly, the study demonstrates the capacity of a multi-disciplinary approach leveraging both datasets derived from remote sensing and DNA sequencing to provide mapping information for public health governmental policies.