Water was collected from January through December of 2021 from five North American lakes including Lake Agawam, Southampton, NY (40.88148, −72.39256; Figure 1), the Lake in Central Park, New York City, NY (40.77458, −73.97073; Figure 1), Mill Pond, Watermill, NY (40.915906, −72.353753; Figure 1), Wainscott Pond, Wainscott, NY, (40.9282, 72.2416; Figure 1), and Maumee Bay within the western basin of Lake Erie (41.823817, −83.331580; Figure 1), all eutrophic systems, prone to CHABs (Davis et al., 2010; Jankowiak and Gobler, 2020; Jankowiak et al., 2019; Flanzenbaum et al., 2022; Kramer et al., 2024). While Lake Agawam, the Lake in Central Park, Mill Pond, and Wainscott Pond are all small (0.2–0.5 km²), shallow (2–4 m), and well-mixed systems, the western basin of Lake Erie is deeper (~11 m maximum depth) and becomes seasonally stratified (Watson et al., 2016). The Lake in Central Park, Mill Pond, Lake Agawam, and Wainscott Pond were sampled weekly-to-monthly during this study while Lake Erie was sampled for several days in August 2001 during a Microcystis bloom. On site, surface temperature, dissolved oxygen, and pH were measured with a YSI 556 ProQuatro multiparameter sonde and a 10% HCl-washed, 20-L polycarbonate carboy was collected and transported to the Stony Brook – Southampton Marine Science Center or the Ohio State Stone Laboratory on Lake Erie. In vitro chlorophyll-a concentrations of planktonic diatoms, chlorophytes, and cyanobacteria were measured using a bbe Moldaenke Fluoroprobe, yielding fluorescence-based biomass estimates of each group expressed in μg chlorophyll a L⁻¹ (Beutler et al., 2002; Jankowiak et al., 2019). Duplicate samples were collected for the analysis of whole water (dissolved and particulate) microcystin. Toxin samples were analyzed by an ABRAXIS® Microcystin/Nodularians test kit according to the manufacturer’s (Gold Standard Diagnostics) procedures which included a 3X freeze/thaw cycle.

Over the sampling year, nutrient amendment experiments were performed twice for Lake Agawam, three times for Mill Pond, once for the Lake in Central Park, once for Wainscott Pond, and once for Lake Erie (Table 1). Water was collected in triplicate, 20 L polyethylene carboys and transported to Stony Brook University’s Marine Sciences Center in Southampton, NY. Water was distributed into 2.75 L 10% HCl-washed polycarbonate Nalgene bottles and placed in an outdoor 300 L table receiving high flow-through (100 L min⁻¹) water from Old Fort Pond, Southampton, NY, creating temperature and light conditions similar to all NY field sites. For the Lake Erie experiment, bottles were incubated in a wire mesh enclosure and floated in Lake Erie at the Stone Lab of Ohio State University.

The effects of different nutrient sources were analyzed during experiments. Raw and sand-filtered wastewater was collected from the New York State Center for Clean Water Technology Wastewater Research and Innovation Facility in Stony Brook, NY, United States, that receives wastewater from Suffolk County, NY, pump station #10 prior to sending the wastewater to the Stony Brook University sewage treatment plant. While the ‘raw wastewater’ was collected after settling in a septic tank, the sand-filtered wastewater was collected after the septic tank effluent was passed through a 3 × 3 m cylinder filled with C33 grade sand. These two forms of wastewater represented direct discharge into an ecosystem and discharge after passage through a riparian zone (Freeze and Cherry, 1979), respectively. All wastewater was passed through a sterile, 0.2 μm filter prior to use to remove bacteria and significant amounts of particulate organic carbon. Nutrient solutions were analyzed for total N, total P, total Kjeldahl N (TKN = ammonium plus organic N), nitrate, and orthophosphate on a Lachat Quikchem flow injection system using US EPA methods 351.2, 353.2, 365.1 (US EPA, 1993a, 1993b, 1993c). Raw wastewater contained 3,980 μM N and 221 μM P with all of the N present as ammonium and organic N. The sand filtered wastewater contained 3,780 μM N and 121 μM P with most of the N present as nitrate (80%) and the remainder as ammonium and organic N. There were also differences in the concentrations of chemical oxygen demand (COD) as measured via US EPA method 410.4 (US EPA, 1993d) between wastewater sources. COD is a proxy for dissolved organic carbon in wastewater (Khan et al., 1998; Rieger et al., 2004) and concentrations were 132 ± 49.8 mg L⁻¹ in the raw wastewater and 9.9 ± 5.8 mg L⁻¹ in the sand filtered wastewater (Table 2). A low phosphorus, ammonium-N commercial fertilizer solution was made to roughly match the concentrations of N and P from the wastewater additions and solutions of ammonium and orthophosphate were made to approximately match the wastewater and fertilizer additions. All nutrient solutions were 0.2 μm filtered and stored frozen as aliquots until use during experiments. The final concentrations of N and P added to experimental vessels were 80 μM N and 4.4 μM P (1.15 mg N L⁻¹; 0.13 mg P L⁻¹) for raw wastewater, 76 μM N and 3.4 μM P (1.06 mg N L⁻¹; 0.10 mg P L⁻¹) for sand-filtered wastewater, 74.4 μM N and 5.16 μM P (1.06 mg N L⁻¹; 0.16 mg P L⁻¹) for the fertilizer, and 75 μM N and 4 μM P (1.05 mg N L⁻¹, 0.12 mg P L⁻¹) for the inorganic nutrient (ammonium and orthophosphate) treatments (Table 2). N additions were within 6% of each other as there was likely some denitrification within the sand beds (Waugh et al., 2020; Gobler et al., 2021) and P additions were within 25% as the sand filtration binds and removes some P (Wang et al., 2024). After 48 h, bottles were processed for the measurement of phytoplankton pigments, microcystin, δ¹⁵N content of particulate organic nitrogen (δPO¹⁵N), and cyanobacterial rRNA (16S SSU) sequencing as described above and below. Differences across the levels of measured parameters were compared by one-way ANOVAs followed by Tukey multiple comparison tests using SigmaPlot v15.

For some experiments, δ¹⁵N content of particulate organic nitrogen (δPO¹⁵N) was analyzed on plankton collected on precombusted (4 h at 450°C) glass fiber filters that were dried (24 h @ 60°C), pelleted in tin discs, and analyzed by an Elementar vario Micro Cube elemental analyzer (Elementar Analysensysteme GmbH, Hanau, Germany) interfaced to a PDZ Europa 20–20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, United Kingdom) at the U.C. Davis Stable Isotope Facility (Davis, CA). Samples were corrected based on batch-specific calibrated reference materials and final δ¹⁵N values are expressed relative to international air standards for N.

For some experiments, deoxyribonucleic acid (DNA) samples were obtained on 0.2 μm polycarbonate filters. Samples were frozen in liquid N after collection and then stored in a-80°C freezer prior to DNA extraction. Samples were extracted with a DNeasy® Power Water Kit (Qiagen), modified for cyanobacteria. The quality of DNA was evaluated on a NanoDrop Microvolume Spectrophotometer (ThermoFisher) and the quantity of DNA in extracts was evaluated on a Qubit Fluorometer (ThermoFisher), allowing the concentration of DNA for sequencing to be normalized across all samples. Samples were sent to the Molecular Research Laboratories in Shallowater, TX, for 16S sequencing of 20,000 reads per sample. Sequences were processed using the Quantitative Insights Into Microbial Ecology QIIME 2 (v. 2021.4.0) microbiome analysis software package following the “Moving pictures” pipeline (Bolyen et al., 2019). Briefly, paired-end reads were trimmed of their primers and barcodes using the Cutadapt plugin and then merged by Dada2 to produce a table of exact (100%) amplicon sequence variants (ASV) (Callahan et al., 2016). The sequences clustered into unique ASV’s with 100% sequence identity, with mitochondrial/chloroplast and heterotrophic bacterial ASVs not considered for further analysis. For taxonomic identification of the 16S dataset, the 99% 16S only rep set FASTA and majority consensus seven-level taxonomy files of the SILVA rRNA (16S SSU) release v138 database (Quast et al., 2012) were used to characterize sequences. Samples from sequenced experiments were analyzed microscopically to confirm identifications made via sequencing the 16S rRNA gene.

The relative abundances of cyanobacterial populations from sequences were converted to estimated abundances of cyanobacterial genera following the methods of Ladds et al. (2021). The method specifically applies the relative abundances of cyanobacterial genera identified via sequencing to the total cyanobacterial biomass as determined by the bbe Fluoroprobe to provide estimates of the biomass of individual cyanobacterial genera. While this approach has been used successfully in other cyanobacterial studies (Ladds et al., 2021; Lusty and Gobler, 2023; Kramer et al., 2024), it does have limitations. Cyanobacteria can differ in their pigment content as well as their copy number of the 16S rRNA gene (Klappenbach et al., 2000; Schirrmeister et al., 2012) and this variance could skew the absolute outcome of this approach. Still, given the fluoroprobe data provides data on the biomass of the total cyanobacterial population, the approach here allows for estimates of biomass at the genera level rather than simply the relative abundance data from the 16S rRNA gene sequencing alone. We emphasize this method is one of five approaches used to evaluate experiments (in addition assessments of the absolute fluorometric analyses of phytoplankton groups, the relative abundance of cyanobacteria, microcystin analyses, and δPO¹⁵N) and that they provide estimates of cyanobacterial biomass at the genus level.

Cyanobacteria concentrations in Lake Agawam were < 25 μg cyano chl-α L⁻¹ from January through April in 2021 but were higher May through early December (25.0–719 μg cyano chl-α L⁻¹), peaking in mid-July (Figure 2A). Microcystin concentrations followed a similar trend, remaining below 15 μg L⁻¹ from January through April before increasing above 15 μg L⁻¹ for the remainder of the year, peaking in July at ~27,400 μg L⁻¹ (Figure 2A). Green algae concentrations at Lake Agawam ranged undetec_table to 60.1 μg green algal chl-α L⁻¹ being detectable during winter and spring (Figure 2A).

In the first Lake Agawam experiment (8/3/21–8/6/21), all treatments had higher (Tukey’s HSD; p < 0.05) cyanobacteria concentrations (69.4–81.7 cyano chl-α L⁻¹) compared to the unamended control (50.0 μg cyano chl-α L⁻¹; Figure 3A), but there was no significant difference in green algae concentration across treatments (Figure 3A). The relative abundance of cyanobacteria was higher (Tukey’s HSD; p < 0.05) in the sand-filtered wastewater, fertilizer, and inorganic nutrient treatments, compared to the control (Figure 3B). There were no statistically significant differences in microcystin concentrations across treatments (Figure 3C). Cyanobacterial biomass was lowest in the control and highest in the fertilizer treatment, with Microcystis, Caenarcaniphilales, and Pseudanbaena being the most prominent taxa found across all treatments (Figure 3D). Microcystis and Pseudanbaena biomass levels were higher (Tukey’s HSD; p < 0.05) in the raw wastewater, sand-filtered wastewater and fertilizer treatments, compared to the control and inorganic nutrient treatments, while Caenarcaniphilales biomass was higher (Tukey’s HSD; p < 0.05) in the inorganic nutrient treatment compared to all other treatments (Figure 3D). δPO¹⁵N was lower (Tukey’s HSD; p < 0.05) in the fertilizer treatment (mean ± S.D. = 4.83 ± 0.04‰) compared to all other treatments (8.87–9.73 ‰) (Figure 3E).

During the second Lake Agawam experiment (9/17/23–9/19/23), cyanobacteria concentrations were higher (Tukey’s HSD; p < 0.05) in the sand-filtered wastewater and fertilizer treatments compared to the control (Figure 4A). Green algae concentrations did not differ across treatments (Figure 4A). Only the fertilizer treatment contained a higher (Tukey’s HSD; p < 0.05) relative abundance of cyanobacteria compared to the control (Figure 4B). Similarly, microcystin concentrations were higher (Tukey’s HSD; p < 0.05) only in the fertilizer (32.7 ± 3.24 μg L⁻¹) treatment compared to the control (17.9 ± 2.11 μg L⁻¹) (Figure 4C).

Cyanobacteria concentrations in Mill Pond ranged ~8–16 μg cyano chl-α L⁻¹ in late May and early June, increased from mid-June through mid-November (~20 to ~72 μg cyano chl-α L⁻¹) and then declined below ~6 μg cyano chl-α L⁻¹ through the rest of 2021 (Figure 2B). In contrast, green algae concentrations at Mill Pond peaked in early June-2021 at ~334 μg green algal chl-α L⁻¹ and ranged from ~17–101 μg green algal chl-α L⁻¹ thereafter (Figure 2B). Microcystin concentrations ranged below detection to ~3 μg L⁻¹ from May through December at Mill Pond, apart from 9/14/21 when concentrations reached ~12 μg L⁻¹ (Figure 2B).

During the first Mill Pond experiment (7/27/21–7/29/21), cyanobacterial concentrations were higher (Tukey’s HSD; p < 0.05) in the raw wastewater (20.29 μg cyano chl-α L⁻¹), fertilizer (16.9 μg cyano chl-α L⁻¹) and inorganic nutrient (25.8 μg cyano chl-α L⁻¹) treatments compared to the control (9.18 μg cyano chl-α L⁻¹) (Figure 5A). Green algae concentrations were higher (Tukey’s HSD; p < 0.05) in the fertilizer treatment than in the control only (Figure 5A). Similarly, diatom concentrations were higher (Tukey’s HSD; p < 0.05) only in the fertilizer treatment compared to the control (Figure 5A). The relative abundance of cyanobacteria was higher (Tukey’s HSD; p < 0.05) in the raw wastewater (15.4%) and inorganic nutrient (22.0%) treatments, compared to the control (8.26%) (Figure 5B). The inorganic nutrient treatment contained a higher relative abundance of cyanobacteria compared to all treatments (Tukey’s HSD; p < 0.05; Figure 4B). There was no difference in microcystin concentrations across treatments (Figure 5C).

During the second Mill Pond experiment (8/10/21–8/13/21), cyanobacterial concentrations and the relative abundances of cyanobacteria were higher (Tukey’s HSD; p < 0.05) in the fertilizer, raw wastewater, and inorganic treatments than in the control (Figures 6A,B). There was no difference in green algae, diatom, or microcystin concentrations between treatments (Figures 6A,C). Cyanobacterial biomass was lowest in the control and highest in the fertilizer treatment, with Microcystis, Cyanobium, and Nodosilinea, being the most prominent genera found across all treatments (Figure 6D). Microcystis levels were higher (Tukey’s HSD; p < 0.05) in the raw wastewater and fertilizer treatments than in the control (Figure 6D). Cyanobium concentrations were higher (Tukey’s HSD; p < 0.05) in the raw wastewater, fertilizer, and inorganic nutrient treatments compared to the control treatment, while Nodosiliniea concentrations were higher (Tukey’s HSD; p < 0.05) in all treatments compared to the control (Figure 6D). δPO¹⁵N was higher (Tukey’s HSD; p < 0.05) in the raw wastewater (12.1‰) treatment compared to all other treatments and lower (Tukey’s HSD; p < 0.05) in the fertilizer (4.02‰) treatment compared to all other treatments (Figure 6E).

During the third Mill Pond experiment (8/25/21–8/27/21), cyanobacterial concentrations were higher (Tukey’s HSD; p < 0.05) in the raw wastewater (42.8 μg cyano chl-α L⁻¹) and inorganic nutrient (53.3 μg cyano chl-α L⁻¹) treatments compared to the control (27.4 μg cyano chl-α L⁻¹) (Figure 7A). However, there was no significant difference in green algae nor diatom concentrations between any treatment (Figure 7A). The relative abundance of cyanobacteria in the raw wastewater (52.3%), sand-filtered wastewater (47.5%) and inorganic nutrient (53.6%) treatments were higher than in the control (43.4%; Tukey’s HSD; p < 0.05; Figure 7B).

Cyanobacteria concentrations in Wainscott Pond ranged ~29–358 μg cyano chl-α L⁻¹ in 2021, peaking in mid-August (Figure 2C). Green algae concentrations ranged ~55–333 μg green algal chl-α L⁻¹ throughout the year, peaking in mid-June (Figure 2C). Microcystin concentrations ranged ~3 to ~59 μg L⁻¹ and peaked in late June (Figure 2C).

During the Wainscott Pond experiment (7/14/21–7/16/21), cyanobacterial concentrations were higher (Tukey’s HSD; p < 0.05) in all treatments (146–176 μg cyano chl-α L⁻¹) compared to the control (110 μg cyano chl-α L⁻¹; Figure 8A). Green algae concentrations across treatments were not different from the control. Diatom concentrations were higher (Tukey’s HSD; p < 0.05) only in the sand-filtered wastewater (15.2 μg diatom chl-α L⁻¹) treatment compared to the control (8.65 μg diatom chl-α L⁻¹) (Figure 8A). All treatments contained a higher (Tukey’s HSD; p < 0.05) relative abundance of cyanobacteria (24.3–29.2%) than the control (21.6%) (Figure 8B). The fertilizer (29.2%) treatment also had a higher (Tukey’s HSD; p < 0.05) relative abundance of cyanobacteria compared to all treatments (Figure 8B). There were no differences in microcystin concentration across treatments (Figure 8C). Cyanobacterial biomass was lowest in the control and highest in the fertilizer treatment, with Cuspidothrix, Cyanobium, Planktothrix, and Microcystis being the most prominent genera (Figure 8D). While Cuspidothrix, Cyanobium, and Planktothrix levels were not different between treatments, Microcystis concentrations were higher (Tukey’s HSD; p < 0.05) in all treatments compared to the control (Figure 8D).

During the summer of 2021, cyanobacteria concentrations at the Lake in Central Park averaged ~13,900 μg cyano chl-α L⁻¹ and reached ~29,900 μg cyano chl-α L⁻¹ on 21-July-2021 and ~ 140,000 μg cyano chl-α L⁻¹ on 10-August-2021 (Figure 2D). Green algae concentrations ranged ~0–82 μg green algal chl-α L⁻¹, peaking in early July, while microcystin concentrations ranged ~16–5,860 μg L⁻¹, with the peak occurring in mid-August (Figure 2D).

During the Lake in Central Park (7/21/21–7/23/21) experiment, cyanobacterial concentrations were higher (Tukey’s HSD; p < 0.05) in all treatments (76.8–94.4 μg cyano chl-α L⁻¹) compared to the control (43.1 μg cyano chl-α L⁻¹) (Figure 9A). There was no difference in green algae nor diatom concentration between treatments (Figure 9A). All treatments contained a higher (Tukey’s HSD; p < 0.05) relative abundance of cyanobacteria (36.4–42.8%) compared to the control (26.0%) (Figure 9B).

Finally, during the Lake Erie experiment (8/16/21–8/19/21), cyanobacteria concentrations, cyanobacterial relative abundance, green algae concentrations, and diatom concentrations did not differ between treatments (Figures 10A,B). Cyanobacterial biomass was lowest in the sand-filtered wastewater treatment (14.2 μg cyano chl-α L⁻¹) and highest in the inorganic nutrient (17.1 μg cyano chl-α L⁻¹) treatment, with Cyanobium and Microcystis being the most prominent taxa found across the treatments (Figure 10C). While Cyanobium concentrations were not different between treatments (Figure 10C), Microcystis concentrations were higher (Tukey’s HSD; p < 0.05) in the raw wastewater (4.85 μg cyano chl-α L⁻¹) treatment compared to the control (3.15 μg cyano chl-α L⁻¹; Figure 10C).