Comment: An alternative interpretation of the relationship between TN:TP and microcystins in Canadian lakes

Abstract

Landscape-scale patterns of the cyanobacterial toxin, microcystin, in Canadian lakes were recently analyzed by Orihel et al. (2012). The analysis of this comprehensive dataset was important for linking accelerated eutrophication and water quality issues affecting human health. The primary conclusion by Orihel et al. (2012) was that high microcystin concentrations occur only at low total nitrogen to total phosphorus (TN:TP) ratios in nutrient-rich lakes. Using these same data, however, we show that high microcystin concentrations across Canadian lakes of all trophic states are more likely to occur at intermediate TN:TP, where the relative availability of N and P aremore closely balanced with phytoplankton nutrient demand. Additionally, in the most nutrient-rich lakes, high microcystin concentrations were not related to TN:TP. Orihel et al. (2012) hypothesized that “the presence of microcystins in lakes should theoretically be higher under low N:P ratios if cyanobacteria dominate under conditions of relative N deficiency.” To test this hypothesis, the authors used arbitrary TN:TP categories of 60 (all ratios presented by mass) based on the presumption that lakes with TN:TP 22.5 (Guildford and Hecky 2000). Therefore, the TN:TP categories used by Orihel et al. (2012) did not include values at which phytoplankton N limitation was likely, even though their prediction of cyanobacterial dominance was predicated on the idea that N2-fixing cyanobacteria can overcome N limitation. Our reanalysis of these same data illustrate that when TN:TP categories are expanded to encompass the full range overwhich both N and P deficiency are probable (i.e., 50, and all increments of 5 in between), elevated microcystin concentrations are most frequently observed when the TN:TP ratio is between 15 and 20 (Fig. 1), which is the range for balanced phytoplankton growth identified by Guildford and Hecky (2000). Orihel et al. (2012) developed a regression tree model using the TREES module in SYSTAT 13 (Systat Software Inc., Chicago, Illinois) to quantify the relationship betweenmicrocystin concentrations and TN and TP concentrations and the TN:TP ratio. We were able to recreate their analysis identically using the MVPART library in R 2.9.2, but only after removing one outlier with a microcystin concentration >500 g·L−1, whichwe assumewas also done but not mentioned by Orihel et al. (2012). We also used nonparametric changepoint analysis in R 2.9.2 to calculate the statistical significance of the modeled thresholds, which is only possible with bootstrapping (Qian et al. 2003; King and Richardson 2003). The results of our regression tree analysis showed that TN was the strongest predictor of microcystin across all lakes (Fig. 2). Most lakes had low TN ( 23 (Fig. 2B). In lakes with TN > 2600 g·L−1 and TN:TP 219 g·L−1 (Fig. 2C). Thus, the regression tree shows that the highest microcystin concentrations across all Canadian lakes sampled occurred at TN:TP ratios between 12 and 23 because TN:TP was bound on the lower end by