Indian River Lagoon spans 250 km along the central east coast of Florida with five inlets to the Atlantic Ocean. The barrier island is continuous, without any inlets, along ∼140 km of the Central IRL, North IRL and Banana River Lagoon (BRL) from Sebastian Inlet north through our study area to Ponce Inlet. In this study, we investigated four tributaries, one in North IRL (Eau Gallie River) and three in Central IRL (Crane Creek, Turkey Creek and South Prong Saint Sebastian River) (Figure 1). Three of these four tributaries, excluding Eau Gallie River, have no causeway between them and Sebastian Inlet. The fourth tributary, Eau Gallie River, is ∼5 km north of Crane Creek (Figure 1). Surface water flow into the IRL and BRL north of Eau Gallie River is primarily via smaller creeks, outfalls, ditches and diffuse runoff. Detailed investigations for 2016 and 2017 were carried out for all four tributaries. Two tributaries, South Prong Saint Sebastian River and Crane Creek, with contrasting land use and land cover, are discussed in detail here. South Prong Saint Sebastian River (hereafter referred to in the text as South Prong) has the largest drainage basin in the overall Saint Sebastian River system and is a representative subbasin with respect to land use and soil type (Gao and Rhew, 2013). South Prong and Crane Creek have received special attention by the State of Florida for total maximum daily loads (TMDLs) of nutrients (Gao and Rhew, 2013; Harper and Baker, 2013). Data for all four tributaries were compiled in a report by Trefry et al. (2019); references to Turkey Creek and Eau Gallie River are made when applicable to the Discussion.

South Prong and Crane Creek (Figure 1) are separated by ∼36 km and have similar geomorphology and average land elevations ∼9 m. The South Prong drainage basin in Indian River County occupies 146 km² with median annual daily flow of 2.9 ± 0.8 m³ s^–1 relative to 32.6 km² and 0.59 ± 0.20 m³ s^–1 for Crane Creek in Brevard County (Bergman et al., 2002; Gao and Rhew, 2013; United States Geological Survey [USGS], 2020b,d). The Turkey Creek drainage basin spans 254 km² with median annual daily flow of 4.1 m³ s^–1 relative to 24 km² and 0.34 m³ s^–1 for Eau Gallie River (United States Geological Survey [USGS], 2020a,c). Despite their small sizes, the four IRL tributaries have an average annual water runoff [flow/basin area = (km³ y^–1)/km²); m y^–1] of 0.55 m y^–1, double the average for all continents except South America with 0.68 m y^–1 (Berner and Berner, 2012).

Land use in the South Prong drainage basin is 17% residential-commercial, 33% agricultural, and 50% undeveloped wetlands, forested and open land whereas Crane Creek is 67% residential-commercial, <2% agricultural and 31% undeveloped wetlands, forested and open land (Gao and Rhew, 2013). Land use in Turkey Creek and Eau Gallie River is >59 and >71% urban, respectively (Trefry et al., 2019). Nutrients transported from South Prong to Central IRL have an increased likelihood of being exported to the coastal Atlantic Ocean because of proximity of South Prong to Sebastian Inlet (Figure 1). The wet season in the IRL is presently defined as June 1-October 31 and the dry season as November 1-May 31. A fertilizer ban from June 1 to September 30 has been in effect since 2014 for both drainage basins; P-containing fertilizers are not permitted for turfgrass at any time in both areas. Large inputs of freshwater and nutrients from other tributaries and outfalls in the North IRL and BRL have monthslong residence times where they can promote HABs (Steward et al., 2005).

Tributary surveys were carried out monthly from December 2015 to March 2018 and episodically during 1- to 7-day rain events (i.e., when rainfall was predicted to be >2.5 cm). South Prong and Crane Creek were sampled from bridges near USGS flow gauges (Figure 1). During each survey, vertical profiles for conductivity, temperature, pH and dissolved oxygen were obtained prior to sampling using a YSI 6600 V2. The sonde was calibrated at the beginning of each day following manufacturer’s specifications and successfully intercalibrated with a YSI ProDSS when it replaced the 6600 V2 in 2017. Sampling locations were upstream of their respective salt wedges; the water was typically <2 m deep and well mixed. Water samples were collected in acid-washed, low-density polyethylene bottles. The open bottles were placed in a weighted plastic holder and lowered from a bridge to the surface of the water and then quickly (1–2 s) lowered to a water depth of ∼0.5 m using a marked rope. Samples were placed in coolers until returned to the Marine & Environmental Chemistry Laboratories at Florida Institute of Technology (FIT) where filtration was carried out in a laminar flow hood within 2–3 h of collection. Supporting data for stations in the IRL (n = 8) and BRL (n = 2; Figure 1) were obtained from the St. Johns River Water Management District (St. Johns River Water Management District [SJRWMD], 2020a,b). Data and methods for biomass calculations of phytoplankton are from Phlips et al. (2021). These data for water chemistry and phytoplankton biomass were used to discuss the consequences of stormwater runoff to the IRL after Hurricane Irma in September 2017. All station coordinates are listed in Supplementary Table 1.

Water for analysis of most dissolved and particulate chemicals was vacuum filtered through polycarbonate filters (Poretics, 47-mm diameter, 0.4-μm pore size) in a Class-100 laminar-flow hood in an FIT clean room with controlled temperature and relative humidity. Prior to the field effort, filters were acid washed in 3N HCl, rinsed three times with deionized water, dried and then weighed to the nearest μg. Each filter was weighed twice in random order, with a minimum of 5% of the filters being weighed in triplicate as described by Trefry and Trocine (1991). Samples for particulate organic carbon (POC) were filtered in a Class-100 laminar flow hood through pre-combusted (450°C) Gelman Type A/E glass fiber filters mounted on HCl-washed filtration glassware. All particle-bearing filters were sealed in separate, acid-washed petri dishes, labeled, double-bagged in plastic and stored dried and re-weighed to a constant weight.

Concentrations of ammonium, nitrate + nitrite, TDN, DIP, and TDP in the filtrate were determined using a SEAL AA3 HR Continuous Segmented Flow AutoAnalyzer following manufacturer’s method G-218-98. For TDN analysis, organic and inorganic N compounds were converted to nitrate using UV irradiation and persulfate digestion. Nitrate was then reduced to nitrite using a cadmium column; analysis followed by UV-visible spectrometry. For TDP analysis, UV and persulfate digestion were used to free organically bound P for analysis by UV-visible spectrometry. We recognize that DIP is measured as soluble reactive phosphorus; we use DIP as a common term in marine sciences. The National Institute of Standards and Technology (NIST) traceable Dionex 5-Anion Standard was analyzed as a reference standard with each batch of samples to ensure accuracy; results were within established 95% confidence limits. Values for analytical precision as relative standard deviation (RSD) were as follows: ammonium (3%), nitrate + nitrite (6%), TDN (2%), DIP (3%) and TDP (3%). Procedural blanks were processed with each batch of samples. Concentrations of DON and DOP were calculated by subtracting the inorganic forms of each element from TN and TP, respectively. Precision was 4% for DON and 16% for DOP. Precision for DOP was weak when DIP was the dominant component of TDP. Detection limits (μM) were as follows: ammonium (0.5), nitrate + nitrite (0.1), TDN and DON (0.04), DIP (0.04), TDP and DOP (0.04).

Analysis for DOC was carried out using a Shimadzu TOC-5050A instrument with non-dispersive infrared detection. Concentrations were calculated by subtracting inorganic carbon from total carbon. Accuracy was determined using an NIST-traceable standard solution of potassium hydrogen phthalate (83.3 μM) from Inorganic Ventures; results were within 95% confidence limits. Precision averaged 1.9% (RSD) and the detection limit was 15 μM.

Suspended particles captured on polycarbonate filters, as well as separate milligram quantities of standard reference material (SRM) #2704, a river sediment issued by the NIST, were digested in stoppered, 15-mL Teflon test tubes using Ultrex II HNO₃ and HF as described by Trefry and Trocine (1991). Concentrations of Al and Fe were determined by flame atomic absorption spectrometry using a Perkin-Elmer 4000 instrument and concentrations of P were determined by inductively coupled plasma mass spectrometry using a Varian 820 instrument. Concentrations of the three elements were within the 95% confidence intervals for the SRM. Analytical precision ranged from 1 to 5% as a function of sample mass and elemental concentrations. Detection limits (% of TSS) were as follows: Al (0.2), Fe (0.03), and P (0.01).

Concentrations of POC and PN were determined by first treating particles on the glass-fiber filter with 10% (v/v) hydrochloric acid to remove any inorganic carbon, then dried. Organic N was not lost during this process based on our laboratory testing. Filters with 200–800 mg of pre-treated suspended sediment were weighed into ceramic boats and combusted with pure oxygen at 950°C using a LECO TruMac C/N/S system. Quantification of the resultant CO₂ and N₂ gases was via infrared and thermal conductivity detectors, respectively. Concentrations of C and N in sediment SRM #2704 and LECO reference sample 502–309 were within 95% confidence intervals for certified values. Analytical precision was 1.5% for POC and 2% for PN. Detection limits (% of TSS) were as follows: POC (0.2) and PN (0.3).

Data and graphical analyses were carried out using Systat 12 and SigmaPlot 10 (Systat Software, Inc.), Excel 2016 (Microsoft) and ArcGIS (Version 10.2.2.3552, Esri, Redlands, CA, United States). An alpha value to define statistical significance was set at 0.05 for statistical tests and regressions. Least squares linear regressions were calculated to determine relationships between individual pairs of parameters (e.g., DON versus flow rate). Equations, 95% prediction intervals, correlation coefficients (r) and p values were determined for each relationship. General descriptors used for r values were as follows: >0.4-0.6 (moderate), >0.6-0.8 (moderately strong), >0.8-0.9 (strong) and >0.9 (very strong). Comparisons of two independent groups of data were carried out using two-tailed t-tests assuming equal variance. Independent groups of data with p-values >0.05 were considered not significantly different from one another. Statistics for the 2016-2017 data were calculated using results for monthly samples to keep sampling days the same for both tributaries. Sampling for rain events took place on a variety of different days for each tributary, with some overlap.

Annual mass fluxes of nutrients and other chemicals from each tributary during 2016 and 2017 were calculated using two approaches. When a significant correlation (p < 0.05) was obtained for concentrations of a chemical for monthly and storm samples versus log flow rate (Q), the pertinent equation was used to calculate the concentration for each day using daily flow rate data from the United States Geological Survey [USGS], 2020b,d. Log flow rate was used in Eq. 1 because non-linear changes in concentration versus flow rate were observed and log flow rate has been successfully used in linear regressions for flow rate versus concentrations of dissolved organic C (Smith and Kaushal, 2015) and nitrate (Duncan et al., 2017). No differences in calculated concentrations were found by using natural (ln) versus base 10 logarithms (log) for our flow rates. The mass flux in t y^–1 (J) was determined from the sum of daily transport (Eq. 1).

where [a ^∗ log Q + b] is the general equation for concentrations of a chemical versus flow rate and Q is the average daily flow rate from the United States Geological Survey [USGS], 2020b,d. When a regression equation for a chemical versus flow rate was not significant, and the data grouped around the median (M) for the chemical, Eq. 2 was used with median concentrations and daily flow rates.

Equations for chemicals versus flow rates are listed in Supplementary Table 2. We used water flow at the time of sampling on graphs and in calculating fluxes for consistency. A few outlying data points were omitted, as noted, in calculating chemical versus flow rate equations because they were outside the mean by two standard deviations and/or had a clear explanation for the anomaly.

Two major weather events highlighted our 2016-2017 study period. First, an extended drought spanned November 2016 through May 2017 (Figures 2A-D). Second, this drought was followed by excess rainfall, runoff and flooding that began with passage of Hurricane Irma in early September 2017. Rainfall continued through October and flooding did not abate until late November 2017 (Figures 2C,D). The tributaries, outfalls and IRL all experienced these weather events.

Annual rainfall during 2016 in the South Prong and Crane Creek drainage basins was 162 and 166 cm, respectively (Figures 2A,B), relative to 30-year averages of 142 cm for South Prong and 122 cm for Crane Creek (National Oceanic and Atmospheric Administration [NOAA], 2020). During 2017, rainfall was 7 and 30% higher, respectively, for Crane Creek and South Prong than in 2016 (Figures 2C,D). Greater rainfall in South Prong during 2017, relative to Crane Creek, occurred because its larger drainage basin extended farther inland and closer to the storm track of Hurricane Irma (Cangialosi et al., 2018). Rainfall during both years in both drainage basins was above the 80th percentile for the past 20 years (National Oceanic and Atmospheric Administration [NOAA], 2020). In 2017, only ∼13% of rainfall in both basins occurred during 5 months of drought (January through May) relative to 40 and 53% in South Prong and Crane Creek, respectively, during September and October (Figures 2C,D).

Annual runoff in South Prong during 2017 was 50% higher than in 2016 due to greater rainfall and flooding associated with Hurricane Irma (Figures 2A,C). In contrast, annual flow from Crane Creek in 2017 was only 11% higher than in 2016 because rainfall was only ∼6% greater and high flow was during September and October, relative to June-December for South Prong (Figures 2A-D).

Total dissolved solids (TDS) peaked at ∼1,400 mg L^–1 in South Prong and ∼900 mg L^–1 in Crane Creek from infiltration of groundwater during the November 2016-May 2017 period of drought (Figures 3A,B). Groundwater in artesian wells and the upper Floridan Aquifer in the study area have TDS values of 1,000-4,000 mg L^–1 (Kroening, 2004). Lowest TDS of 300-500 mg L^–1 in both tributaries occurred during high flow, including Hurricanes Matthew in October 2016 and Irma in September 2017 (Figures 3A,B). During storm events, TDS in both tributaries often decreased by half within the first 12 h via dilution with rainwater (Supplementary Figure 1). A return to median values typically occurred within ∼1 day after rainfall stopped at Crane Creek, but >4-day in the larger and less well drained South Prong basin (Supplementary Figure 1). Statistically significant equations with correlation coefficients equal to −0.86 (South Prong) and −0.90 (Crane Creek) were calculated for TDS versus flow rate (Figures 3C,D). Correlations for dissolved chemical species versus TDS, when significant, were used to help assess the relative importance of surface runoff versus baseflow (groundwater) to nutrient concentrations and fluxes. Higher TDS and a positive correlation with dissolved chemical species suggested a substantial contribution from groundwater during low flow (O’Connor, 1976).

Both IRL tributaries had median concentrations of DON of 30-35 μM (Table 1) and ranges of <20 μM to >60 μM (Figures 3E,F). DON was the dominant form of TN in South Prong (57%) and Crane Creek (48%), a common, but not universal feature of rivers (Pellerin et al., 2006; Wiegner et al., 2006; Lorite-Herrera et al., 2009). Concentrations of DON in South Prong, but not Crane Creek, showed a strong positive correlation versus log flow rate (Figures 3G,H). Concentrations of DON in South Prong, but not Crane Creek, were well above the median during rain events (Supplementary Figure 1). A strong negative correlation for DON versus TDS was obtained for St. Sebastian (r = −0.88, p < 0.001), but not for Crane Creek (Supplementary Figure 1).

Mean values for nitrate + nitrite and ammonium were about twofold higher in Crane Creek than in South Prong (Table 1). Nitrate + nitrite accounted for 29% of TN in Crane Creek, but only 13% in South Prong. The average % of TN that was ammonium was 6 ± 2% for South Prong and 8 ± 5% for Crane Creek. Highest concentrations for these N species in both tributaries were generally found at median flow; lowest concentrations were commonly observed during high flow (Figures 4A–D). Concentrations of nitrate + nitrite (Figure 4C) and ammonium did not correlate significantly with flow rate for South Prong, likely because of complex relationships with flow and other variables. For Crane Creek, a significant negative correlation was obtained for nitrate + nitrite versus log flow rate (Figure 4D), but not for ammonium versus log flow rate (Supplementary Table 2).

Total P and DIP showed parallel patterns versus flow rate in both tributaries with greater values during high flow (Figures 4E,F). Offsets between concentrations of TP and DIP shown on graphs were larger during high flow because of increases in particulate P concentrations (Figures 4E,F). DIP comprised 48% of TP in both tributaries. Concentrations of DIP increased above the median during rain events in both tributaries (Supplementary Figure 1). Following Hurricane Irma, concentrations of DIP were >30 μM in South Prong (Figure 4E). During drought conditions, mean DIP values of 1.3 μM in South Prong were threefold lower than the monthly average and >20 times lower than maximum values during rain events and hurricanes (Figure 4E). DIP values in both tributaries correlated positively with log flow rate; however, the slope for South Prong was ∼5 times greater than for Crane Creek (Figures 4G,H). DIP correlated significantly and negatively with TDS for South Prong, but not for Crane Creek.

Median concentrations of DOC for both South Prong and Crane Creek averaged ∼1,000 μM (Table 1) with a range of 300-2,600 μM; higher values occurred during increased flow (Supplementary Figure 2). An equation for DOC versus log flow rate for South Prong had a stronger positive correlation and a fourfold greater slope than the equation for Crane Creek (Supplementary Figure 2). A moderately strong negative correlation for DOC versus TDS (r = −0.78; p < 0.001) was obtained for South Prong, but not for Crane Creek.

Mean values for total suspended solids (TSS) in both tributaries during monthly surveys were <10 mg L^–1 (Table 1). Such low TSS values were found in <5% of rivers globally; moreover, 40% of world rivers have mean TSS >500 mg L^–1 (Meybeck et al., 2003). South Prong had significantly higher TSS than Crane Creek (Table 1). TSS generally increased with higher water flow in both tributaries (Supplementary Figure 3). Common to many rivers, higher TSS values were sometimes found during the onset of very high flow (i.e., “first flush” effect) with decreases over the duration of the rain event (Supplementary Figure 3). TSS correlated significantly with log flow rate for both tributaries (Supplementary Figure 3).

Concentrations of POC (μM) in South Prong averaged ∼80% greater than in Crane Creek because of higher TSS in South Prong (Table 1). Despite overall low POC concentrations (μM) in both tributaries, the percentage of TSS that was POC averaged 15.5% (South Prong) and 17.9% (Crane Creek), much higher than a global mean of 1.7% (Kandasamy and Nath, 2016). PN values (μM) in South Prong averaged 52% higher relative to Crane Creek (Table 1) and ∼10-fold greater (% of TSS; Table 1) than a mean value of 0.3% for world rivers (Paolini, 1995). No significant difference was found for particulate P (μM) in South Prong versus Crane Creek (Table 1); however, South Prong and Crane Creek contained 2.3- and 3.7-fold higher amounts of particulate P (% of TSS, Table 1), respectively, than an average of 0.3% P for global rivers (Viers et al., 2009). PN accounted for 24 and 15% of TN in South Prong and Crane Creek, respectively, and particulate P accounted for 35% of TP in both tributaries. The high C, N, and P content of tributary particles showed that, despite low TSS, both tributaries carried nutrient-rich particles to the IRL.

The grand mean for aluminum (Al) in suspended particles from both tributaries (3.6 ± 1.4% of TSS, n = 51) was within 20% of the average for IRL sediments (Al = 4.4 ± 1.4%, n = 23; Trefry and Trocine, 2011). In contrast, average values for particulate iron (Fe % of TSS) were very high at 6.6% in South Prong and 10.7% in Crane Creek relative to 2.6 ± 1.3% in IRL sediments (Trefry and Trocine, 2011). Therefore, Fe/Al ratios for tributary particles were 3-5 times higher than a mean of 0.6 reported for IRL sediments (Trefry and Trocine, 2011).

Monthly patterns for water flow to the IRL from both tributaries in 2016 were distinctly different than in 2017 (Figure 2). South Prong and Crane Creek transported only 10% of their annual water flow during the drought of January-May 2017, relative to 46% (South Prong) and 30% (Crane Creek) of annual flow from January-May 2016 (Figure 2). In 2017, >50% of the annual fluxes of water from both tributaries were exported to the IRL during September-October 2017, relative to <25% during the same period in 2016 (Figure 2). Similar trends during drought and extreme runoff in 2017 were observed throughout the Central IRL and North IRL (Trefry et al., 2019; United States Geological Survey [USGS], 2020a–d). The combination of 5 months of drought and 2 months of excess runoff and flooding in 2017 set up conditions for large-scale remobilization of nutrients from soils and large fluxes to the IRL.

Our calculated fluxes of TN and TP for 2016 from South Prong to the IRL were 106 t and 16.7 t, respectively (Table 2), similar to multi-year averages of 114 t for TN and 19 t for TP by Gao and Rhew (2013). Our fluxes for TN and TP in 2017 for South Prong were 60 and 80% greater, respectively, than in 2016 (Table 2). Gao and Rhew (2013) reported average fluxes from Crane Creek for TN (29 t) and TP (5.4 t) that were higher than our 2016 values and more like our 2017 values (Table 2). To better compare the tributary data, yields (fluxes divided by drainage basin area) were calculated. TN yields for both IRL tributaries matched within 6% for 2016; however, the TN yield in 2017 was 31% greater for South Prong than Crane Creek (Table 2 and Figures 5A,B). Similar annual trends were observed for yields of TP for both tributaries (Table 2 and Figures 5C,D).

In contrast with TN, yields for DON from South Prong were 32 and 115% greater in 2016 and 2017, respectively, than found for Crane Creek, even though water flow from Crane Creek was 28% greater in 2016 and only 5% less in 2017 (Table 2 and Figures 5E,F). DON comprised an average of 46% of the TN yield for Crane Creek versus 70% for South Prong (Table 2).

Average DIN (nitrate + nitrite + ammonium) yields for Crane Creek were about double values for South Prong for both 2016 and 2017 (Table 2 and Figures 5G,H). Annual yields of nitrate + nitrite averaged ∼2.6 times greater for Crane Creek than South Prong (Table 2). In contrast, ammonium yields were the same in both basins in 2016 and 44% higher for Crane Creek in 2017. PN accounted for 15-24% of TN yield for the two IRL tributaries (Table 2).

Average annual yields of DIP and particulate P for South Prong were higher than those for Crane Creek (Table 2). In contrast, yields for DOP were higher for Crane Creek (Table 2). However, particulate P comprised only ∼28% of the TP flux to the IRL relative 76% for global rivers.

Total suspended solids yields from South Prong averaged 27% greater than for Crane Creek (Table 2) because of higher TSS concentrations and greater increases in TSS with increased flow in South Prong. TSS values in both tributaries were very low; therefore, yields for TSS for both tributaries were only 3-5% of fluxes for world rivers (Table 2). Yet, we now know that IRL suspended matter was rich in C, N, and P.

Data from monthly surveys by St. Johns River Water Management District [SJRWMD] (2020b) were used to investigate changes in water chemistry in the IRL following Hurricane Irma. Salinity at 10 stations throughout Central IRL, North IRL and BRL decreased from an average of 34 ± 3 in July-August 2017 to <20 in September-October 2017 following the hurricane (Figures 1, 6A). The lowest salinities of 14-17 occurred during October 2017 at each of four IRL stations located near tributaries to the IRL: station 21 (Eau Gallie River), station 23 (Crane Creek), station 24 (Turkey Creek) and station 28 (South Prong) (Figures 1, 6A). Salinity subsequently increased over time with some spatial variations (Figure 6A). Water temperatures in the IRL were >25°C during September-November 2017 and again after April 2018; lowest temperatures during January 2018 were 15-17°C (Figure 6B).

Concentrations of DIN increased from <5 μM to >20μM during October-December 2017 at all IRL stations except 28 (Figure 6C). Highest DIN values of 50-60 μM were obtained at northern stations 13 and 16, whereas values of 26-35 μM were recorded at stations 21, 23, and 24, adjacent to tributaries (Figure 6C). Average concentrations of DOC and DON also increased from 600 to > 1,000 μM and ∼30 to ∼50 μM, respectively, at IRL stations near tributaries following rainfall during fall 2017 (Figures 6E,F). Following larges fluxes of DIN, DOC, and DON to IRL, concentrations of chlorophyll a increased to >40 μg/L, beginning as early as October 2017 at station 24, near Turkey Creek (Figure 6D). Corresponding data for the biomass of Aureoumbra lagunensis and other nanoeukaryotes showed the onset of a bloom in December 2017-January 2018 (Figures 7A,B).

Concentrations of chlorophyll a remained high from January to May 2018 (Figure 6D), with cell counts often > 1 × 10⁶ per mL (Judice et al., 2020; Phlips et al., 2021). The nanoeukaryotic bloom peaked during January-April 2018 (Figures 7A,B). Concentrations of TSS during the bloom ranged from 16 to 24 mg L^–1 (Figure 7C). Salinity decreased to 15-20 during late spring and summer 2018 when concentrations of dissolved inorganic N and P were low and DOP values were decreasing (Figures 6, 7; St. Johns River Water Management District [SJRWMD], 2020b). During the same period, concentrations of chlorophyll a and phytoplankton biomass diminished (Figures 6D, 7A,B).

Stormwater inputs of freshwater and nutrients can be trapped behind the barrier islands of the IRL for weeks to months (Steward et al., 2005; Rosario-Llantin and Zarillo, 2021), thereby promoting intense HABs, including an Aureoumbra lagunensis bloom in 2017-2018 (Galimany et al., 2020; Judice et al., 2020). Lapointe et al. (2020) highlighted the importance of land-based nutrients as a driving force for eutrophication and HABs in the IRL, especially noting a significant negative correlation between DIN and salinity. Furthermore, high TDN:TDP ratios (i.e., >70) in the IRL favored Aureoumbra lagunensis that can form dense blooms at low concentrations of TDP (Kang et al., 2015; Lapointe et al., 2020). Paerl et al. (2001) showed that stormwater input during three sequential hurricanes delivered at least half the annual N load to Pamlico Sound, subsequently promoting three to fivefold increases in chlorophyll a for 5-6 months. Predicting and tracking flow rates, concentrations and ratios of pertinent nutrient species in IRL tributaries may foreshadow pending increases in chlorophyll a and HABs, thereby initiating or supporting appropriate responses.

As previously mentioned, Hurricane Irma was a major event during our study. It brought rain and flooding that began in early September 2017 and continued episodically through October 2017 (Figures 2C,D). In these 2 months of 2017, >50% of the annual fluxes of freshwater and ∼60% of annual fluxes of TN, and TP were transported to the IRL from South Prong, Crane Creek and other tributaries and outfalls to the IRL (Figures 2, 5; Trefry et al., 2019). By comparison, <30% of the annual fluxes of freshwater, TN and TP were exported to the IRL from South Prong, Crane Creek and other tributaries during September and October 2016 (Figures 2, 5; Trefry et al., 2019).

Freshwater runoff (salinity ∼0.1) during fall 2017 caused salinity to decrease from 34 to <20 and diluted lagoon water by at least 40% [((34–20)/34)^∗100]. Salinity <20 persisted from <1 to 4 months at the 10 stations (Figure 6A). Subsequent increases in salinity over time varied considerably, consistent with flushing times (R₅₀ = 50% renewal time) that ranged from as low as ∼10 days in parts of Central IRL to as high as ∼100 days in the North IRL and >150 days in BRL (Steward et al., 2005; Rosario-Llantin and Zarillo, 2021). For example, salinity increased to ∼24 in less than 1 month and to 36 by April 2018 at station 28, near Sebastian Inlet that connects IRL to the Atlantic Ocean (Figure 6A). Salinities at stations 23, 24, and 26, located ∼20-30 km north of Sebastian Inlet remained below 18 until at least January 2018 (Figure 6A). All stations had salinities >20 by February-March 2018.

A negative relationship for DIN versus salinity in the IRL was observed following inputs of freshwater and DIN from Hurricane Irma (Figures 6A,C), confirming results by Lapointe et al. (2020). DIN values at IRL stations 13 and 16 were equal to or greater than a maximum of 52 μM in Crane Creek. Such high DIN concentrations in the IRL were not likely supported by tributaries such as Crane Creek and South Prong because incoming freshwater accounted for only ∼40% of the water in the lagoon during late 2017 (Figure 6A). This observation supports previously discussed sources of DIN to the IRL, north of stations 18 and B9, that likely included groundwater, sewer overflows, leakage from septic systems and benthic fluxes of ammonium (Lapointe et al., 2015, 2020; Barile, 2018; Fox and Trefry, 2018; Tetra Tech Inc and Closewaters LLC, 2021).

Ammonium accounted for an average of 60% of DIN in the IRL during the period of peak DIN concentrations, consistent with an average 64 ± 21% of DIN as ammonium in the IRL for 2015-2018 (St. Johns River Water Management District [SJRWMD], 2020b). Ammonium concentrations in nearby Crane Creek were ≥50% of values for DIN during high flow in fall 2017. This fraction of DIN that was ammonium in Crane Creek during high flow was much greater than monthly averages of 22% for Crane Creek and 30% for South Prong. Enhanced ammonium in the IRL tributaries during high flow and flooding (Figures 4A,B) was likely due to lower redox potentials in the drainage basins. Most importantly, the large fraction of DIN that was ammonium in the IRL and the tributaries during very high flow was especially favorable for Aureoumbra lagunensis, an organism that is unable to use nitrate as source of N (Deyoe and Suttle, 1994). In addition to DIN, concentrations of DOC and DON increased in the lagoon between September 2017 and January 2018 (Figures 6E,F).

As salinity increased, concentrations of chlorophyll a and phytoplankton biomass increased (Figures 6A,D, 7A,B). Increases in chlorophyll a and the onset of a eukaryotic bloom began as early as October 2017, but lagged into February 2018 at some stations until salinities eventually increased to >20 (Phlips et al., 2021; Figures 6A,D, 7A,B). Previous laboratory studies showed that specific growth rates for Aureoumbra lagunensis slowed considerably at salinities of 10 and 15; however, specific growth rates at 25°C were not significantly different for salinities of 20-50 (Buskey et al., 1998). Water temperature in the IRL was favorable (>25°C), except during January and March, 2018 (Figure 6B). The bloom varied in intensity and duration in the North IRL and BRL (Figures 7A,B; Phlips et al., 2021). Biomass for Aureoumbra lagunensis plus other nanoeurkaryotes accounted for an average of 90 and 97% of total phytoplankton biomass, respectively, at phytoplankton stations 3 and 5 during the months of active blooms (December-May 2018; Figures 7A,B). Lower chlorophyll a was found at northern stations 13 and 16 as well as southern stations 26 and 28 (Figure 6D). Variations in chlorophyll a and phytoplankton biomass likely reflected variations in salinity and temperature. Continuous data for salinity and chlorophyll a obtained only for station B4 showed reasonably good correspondence with results from monthly surveys (St. Johns River Water Management District [SJRWMD], 2020a).

Several shifts in concentrations of nutrient species occurred during the bloom (January-May 2018). For example, concentrations of DIN decreased to <2 μM by January 2018 and remained low at all stations, except for a spike in April-May 2018 (Figure 6C) that was predominantly ammonium (St. Johns River Water Management District [SJRWMD], 2020b). This 2018 spike for DIN was probably related to benthic fluxes or recycling of organic N in the water column or at the sediment-water interface. The overall deficiency in DIN during the bloom likely led to uptake of components of DON, such as urea and amino acids, to support primary productivity (Kang et al., 2015; Cira and Wetz, 2019). DON in the IRL typically accounts for 80 to >90% of the TDN (Lapointe et al., 2020); but, during the bloom, DON comprised an average of 98% of TN, except for localized and short-term increases in DIN (Figure 6C; St. Johns River Water Management District [SJRWMD], 2020b).

Concentrations of DON and DOP varied considerably during the bloom (Figures 6F, 7D), likely resulting from cycles of assimilation and release of these species (Kang et al., 2015; Cira and Wetz, 2019). Concentrations of DOP decreased appreciably during April 2018, likely due to consistent uptake by Aureoumbra (Figure 7D). Ratios for TDN:TDP were as high as 100 (St. Johns River Water Management District [SJRWMD], 2020b); however, these totals do not accurately assess the bioavailable fractions of DON and DOP. Ratios for DIN:DIP were constrained by very low values for DIP, but were likely much >100 (St. Johns River Water Management District [SJRWMD], 2020b). Overall, decreases in salinity to 15-20 during late spring and summer 2018, along with low concentrations of inorganic N and P and decreasing DOP values, likely combined with biological processes to diminish the bloom by mid-2018 (Figures 6, 7; St. Johns River Water Management District [SJRWMD], 2020b).

Concentrations of TSS during the bloom (Figure 7C) were at least 10 mg L^–1 higher than typical values for the IRL (St. Johns River Water Management District [SJRWMD], 2020b). At 10 mg L^–1, the mass of excess TSS in 1 m³ of water would be 10 g. For a 20 km × 2 km area (equivalent to an area in the IRL from stations 16 to 21) and a depth of 1 m, the estimated mass of TSS would be 400 t [(20,000 m × 2,000 m × 1 m) × (10 g m^–3)] = 400 × 10⁶ g = 400 t]. By comparison, annual fluxes of less organic-rich suspended matter from Crane Creek and South Prong averaged ∼170 t y^–1 and 1000 t y^–1, respectively (Table 2). Therefore, decomposition of residual, organic-rich particles from the bloom provided a potentially large source of nutrients and a high biochemical oxygen demand in the IRL.