2.3. Data analysis Analysis of 16S rRNA data was conducted using the microbial ecology community software program Mothur (http://www. mothur.org/wiki/Download_mothur) (Schloss et al., 2009). The reads were processed by removing tags and primers, only accepting reads with an average quality score above 20 and read lengths longer than 300 nt. Data analysis was carried out following the Schloss standard operating procedure (Schloss et al., 2009). Sample coverage, ACE richness estimators and Shannon diversity index were calculated at 97% similarity. A dendrogram describing the similarity of the samples was generated using Thetayc calculators. Sequences were classified using the greengene database and NCBI. A neighbor-joining phylogenetic tree of Cyanobacteria 16S rRNA genes was constructed in MEGA 4 using representative sequences at a 0.03 genetic distance (Tamura et al., 2007). Nearest relatives were retrieved from the NCBI database. A heatmap showing the relative number of sequences per sample for each operational taxonomic unit (OTU) was generated in iTol (Letunic and Bork, 2007). The relationship of environmental factors and the species composition of community (based on the relative abundance of top 250 OTUs) were analyzed by the constrained linear ordination technique redundancy analysis (RDA) using CANOCO v4.5 (Microcomputer Power, USA) and the BIOENV analysis provided in PRIMER 5 software (Primer-E Ltd, UK). RDA was used to test which of the environmental factors explain the majority of the variation in the bacterial/archaeal community composition (Monte Carlo permutation test, 1000 permutations) and show relationship between major bacterial families and environmental factors. The BIOENV analysis was applied to determine the correlation between the environmental factors and bacterial/archaeal communities with the application of a Spearman’s correlation coefficient.