Lake Okeechobee (27°N, 81°W) is a large (1803 km²), shallow (mean depth 2.7 m) subtropical lake in South Florida at the center of the Kissimmee-Okeechobee-Everglades ecosystem and the Central and Southern Florida Project (Figure 1; Aumen, 1995). Water levels within Lake Okeechobee are influenced by the subtropical climate of South Florida combined with water management that focus on regulatory controls for water supply, flood protection and the environment (Qiu and Wan, 2013).

We evaluate the lake as three regions: (1) littoral, (2) nearshore, and (3) limnetic zones. The shallow nearshore and littoral region comprises a third of the total lake surface area and contains a diverse plant community of emergent and submerged aquatic vegetation. The limnetic zone or open water region of the lake makes up the remaining two-thirds of the surface area. This limnetic portion in the central, north, and east has predominately flocculent mud sediments (Fisher et al., 2001; Julian et al., 2023), which are a persistent source of turbidity and nutrients in the lake (Phlips et al., 1993; Harris et al., 2007).

Water quality and hydrodynamic data were retrieved from the South Florida Water Management District Online Environmental Database (Table 1) for locations across Lake Okeechobee (Figure 1) between October 1987 and September 2023 (Federal Water Years [WY] 1988 to 2023; Table 1). Nutrient and chlorophyll-a data were collected as surface water grab samples during the period of record. Daily average phycocyanin concentration data were collected from a subset of long-term, near-surface (0.5 m below water surface) monitoring locations (Figure 1) between October 2016 and September 2023. Any data associated with a fatal qualifier indicating a potential data quality problem was removed from the analysis. For data analyses and summary statistics, values reported below the method detection limit (MDL) were assigned a value of one-half the MDL.

Scenario modeling was conducted using the Regional Simulation Model – Basins (RSM-BN). The RSM-BN model uses past climatology in a link-node framework to simulate hydrologic conditions by using a Hydrologic Simulation Engine (Chin et al., 2005) combined with a Management Simulation Engine (Bras et al., 2019). The RSM-BN model has a 52-year simulation period of record from January 1, 1965 to December 31, 2016. For purposes of restoration planning, the RSM-BN model uses observed variability in climatology across the SFWMD boundary as input to simulate water management changes. Therefore, over the 52-year simulation period, the model includes extreme events such as prolonged drought, hurricanes, tropical storms, and prolonged periods of high and low rainfall. The RSM-BN is a robust simulation tool that has been used in project planning efforts for several projects including the Central Everglades Planning Project, Lake Okeechobee Watershed Restoration Project, Western Everglades Restoration Project, Everglades Restoration Transition Plan, and Combined Operational Plan (South Florida Water Management District, 2020).

This study used the Lake Okeechobee System Operating Manual (LOSOM; US Army Corps of Engineers, 2024a) and the Lake Okeechobee Component a Reservoir (LOCAR; US Army Corps of Engineers, 2024b) RSM-BN based modeling efforts. Both efforts included unique baseline conditions and preferred/selected operational or restoration alternatives. Moreover, project planning and modeling were conducted at different times. In the LOSOM modeling effort, the intention was to evaluate changes in Lake Okeechobee’s regulation schedule, therefore, the no-action 2025 (NA25; equivalent to a future without project baseline) baseline condition assumes the Lake Okeechobee Regulation Schedule of 2008 (LORS08) lake operations. The infrastructure included in the NA25 baseline contained the completion of capital projects and foundational Comprehensive Everglades Restoration Plan (CERP) projects including the rehabilitation of the Herbert Hoover Dike (HHD), operation of the C-44 reservoir (to the east of Lake Okeechobee), and operation of stormwater treatment areas (STAs) downstream of Lake Okeechobee that treat lake water in addition to agricultural runoff. The final alternative was the preferred alternative 2025 (PA25) that represented the completed infrastructure outlined above for the NA25 alternative but also included C-43 reservoir (downstream of the western outlet), and the A-2 stormwater treatment area (downstream of southern outlets) with water management being guided by the new LOSOM regulation schedule (Table 2).

For the LOCAR modeling effort, the baseline condition (FWOLL) assumes LOSOM water management and regulatory operational guidance for Lake Okeechobee and included the Everglades Agricultural Area (EAA) Reservoir and stormwater treatment area. In the preferred alternative for LOCAR (LCR1), a 200,000 acre-ft storage reservoir located north of Lake Okeechobee was also included. The northern reservoir was modeled to attenuate flows to the lake by capturing wet season flows, and provide supplemental flows during the dry season (USACE, 2024; Figure 1; Table 2).

Generally, LORS08 was a prescriptive operational plan with numerous water level-based management bands that dictated recommended discharge volumes to the major lake outlets. These discharges primarily benefited water supply while providing highly damaging discharges to the Northern Estuaries (Caloosahatchee and St Lucie River estuaries) and starving the Everglades ecosystem to the south (Julian and Reidenbach, 2024). Meanwhile, LOSOM was developed to provide operational flexibility, providing ample water supply benefits while improving salinity regimes for the Northern Estuaries and increasing ecological flows to the southern Everglades (Julian and Welch, 2022; Julian and Reidenbach, 2024).

Results from the NMDS demonstrated a clear separation of ecological regions for in-lake characteristics, in-lake load estimates, stage and lake inputs (inflow discharge volume and nutrient loading; Figure 2) with a stress value of 0.14 for the first two dimensions with a relatively high non-metric and linear fit R² values of 0.98 and 0.91, respectively (Supplementary Figure S1). Inflow discharge was linked to inflow TP and TN loads, while stage, volume, WRT, TN, temperature, and chlorophyll-a were also interrelated. Meanwhile, TP, SRP, and DIN concentrations were generally associated. Ecological regions were stretched across the first NMDS axis with the littoral zones clustering and pelagic and nearshore zones forming separate, distinct groups (Figure 2). This observed grouping of ecological zones was confirmed by ANOSIM, showing a significant global difference between groups and significant pairwise differences between pelagic/nearshore and littoral zones, while littoral zones were not significantly different from one another (Table 3).

Chlorophyll-a concentrations varied across space, time, and stage (Figures 3A–D) with the GAM explaining 75% of the deviance (relative to a null model) and an R² of 0.69 (Table 4). The spatial effect of the model (Figure 3D) varied across the lake with the largest spatial effect located in a region of the southwestern edge of the lake while the lowest spatial effect was adjacent to the southern part of the pelagic zone (Figures 1, 3D). Temporally, within year (i.e., monthly) effects were greatest in the summer months peaking around July with significant increases being detected between April and June and significant decreases between October and later November (Figure 3B). Between-year effects (Figure 3C) varied with some notable fluctuations after major hurricanes and other climatic events. For example, the annual effect significantly declined after 2008 following a tropical storm that relieved regional drought conditions, but increased, while not significant following 2017 (year of Hurricane Irma). The effect of stage elevation varied across the observed range of stage conditions during the 2008 to 2023 POR, with a significant increase between 3.57 and 3.96 m NGVD29 and a significant decrease between 4.91 and 5.15 m NGVD29 (Figure 3A).

Despite the much shorter period of record and spatial coverage, variation in phycocyanin concentrations across space, time, and stage were also detected (Figures 3E–H) with the model explaining 93% of the deviance and an R² of 0.92 (Table 4). The spatial effect of the model (Figure 3H) varied across the lake with the greatest spatial effect along the northern littoral zone and at the center of the lake and the lowest adjacent to the northern and southern littoral zone edge (nearshore zone). Temporally, within-year (i.e., DOY) effects were greatest at the start and end of the year with a double peak in within-year effect with the first smaller peak occurring near the end of July/beginning of August followed by a significant increase and the second peak near the beginning of December. A significant increase in annual effect was detected between late 2020 and late 2022 (Figure 3G). Despite the relatively short time series, significant changes in model effect were also detected relative to stage elevation with a consistent relative decline starting after 4.25 m NGVD29 (Figure 3E). However, due to the short time-series, the stage effect curve was truncated relative to the longer POR for the chlorophyll model therefore the relationship between phycocyanin concentrations and stage is uncertain when stages are less than 3.81 m NGVD29.

Due to the limited temporal and spatial coverage of phycocyanin concentrations and the truncated stage conditions, a predictive equation was not developed. The effect of lake stage (constrained to 3.5 m) on summer mean chlorophyll-a concentrations was significant, with a coefficient of 0.178 (SE = 0.089; t-value = 2.0; p-value<0.01; Table 5; Figure 4). The random effect of the ecological zone was significant with respect to random effect on chlorophyll-a concentrations (F = 519.1; p-value <0.01) and chlorophyll-a and stage (F = 332.9; p-value <0.01) with a variance of 0.279 and 0.035, respectively. The deviance explained by the entire model (i.e., conditional deviance explained) was 0.47 (Table 5) meanwhile the deviance explained by just the fixed effects (i.e., marginal) was 0.12. The average MAE, RMSE, and MBE of the five-fold validation were 7.24, 9.24, and 0.27, respectively.

The effect of lake stage (constrained to 3.5 m) on summer f_Chla20 was not significant, with a coefficient of 0.263 (SE = 0.1973; z-value = 1.33; p-value = 0.18; Table 5; Figure 4). However, the ecological zone had a significant random effect on f_Chla20 (χ² = 14,542,575; p-value <0.01) and f_Chla20 and stage (χ² = 13,042,007; p-value <0.01) with a variance of 1.68 and 0.235, respectively. The deviance explained by the entire model (i.e., conditional deviance explained) was 0.42 (Table 5) meanwhile the deviance explained by just the fixed effects (i.e., marginal) was 0.05. The average MAE, RMSE, and MBE of the five-fold validation were 0.14, 1.81, and 0.02, respectively.

The effect of lake stage (constrained to 3.5 m) on summer f_Chla40 was not significant, with a coefficient of 0.25 (SE = 0.15; z-value = 1.68; p-value = 0.09; Table 5; Figure 4). The ecological zone had a significant random effect on f_Chla40 (χ² = 2,825; p-value <0.01) and f_Chla40 and stage (χ² = 1754; p-value <0.01) with a variance of 0.660 and 0.100, respectively. The deviance explained by the entire model (i.e., conditional deviance explained) was 0.27 (Table 5) meanwhile the deviance explained by just the fixed effects (i.e., marginal) was 0.08. The average MAE, RMSE, and MBE of the five-fold validation were 0.10, 0.13, and 0.02, respectively.

Summer average chlorophyll-a varied significantly between ecological regions, with water depth being a significant effect for nearshore and littoral west ecological zones (Supplementary Table S4). Additionally, TP and DIN parametric terms were significantly different from zero (Supplementary Table S4), while the interaction of TP and DIN did not significantly vary between ecological regions, including it in the model did improve the predictive capabilities of the model. However, TP and DIN interaction and mean depth by ecological zone resulted in a non-linear relationship for most ecological regions (based on effective degrees of freedom (edf) values; Supplementary Table S4). For the testing dataset, the chlorophyll-a model resulted in an R² of 0.51 (observed vs. predicted during the training period), KGE of 0.59, MAE of 5.61 μg L⁻¹, RMSE of 7.45 μg L⁻¹_, and MBE of −2.83 μg L⁻¹. For validation, the MAE, RMSE, and MBE were 8.75, 11.29, and −0.47 μg L⁻¹, respectively as estimated by the LOOCV procedure. The resulting model produced a model fit (R²_adj) of 0.70 and a deviance explained of 83.5% (Supplementary Table S4).

Summer average TP concentration varied between ecological zones, outflow discharge volume, and lake volume for most ecological zones (littoral south, littoral west, and nearshore; Supplementary Table S4). Based on the edf values of the lake volume by ecological zone smoothing terms, summer TP was linear and not significantly different from zero for littoral north and pelagic zones, the nearshore zone was linear but significantly different from zero (with a slight positive slope based on first derivative values) and the littoral south and littoral west were non-linear. For the testing dataset, the TP model resulted in an R² of 0.45, KGE of 0.49, MAE of 0.040 mg L⁻¹, RMSE of 0.044 mg L⁻¹_, and MBE of 0.038 mg L⁻¹. For validation, the MAE, RMSE, and MBE were 0.042, 0.052, and −0.0002 mg L⁻¹, respectively as estimated by the LOOCV procedure. The resulting model produced a model fit (R²_adj) of 0.64 and a deviance explained of 70.8% (Supplementary Table S4).

Summer average DIN concentrations varied between ecological zones, outflow discharge volume, and WRT, specifically within the littoral north zone. For the testing dataset, the DIN model resulted in an R² of 0.32, KGE of 0.22, MAE of 0.039 mg L⁻¹, RMSE of 0.048 mg L⁻¹_, and MBE of 0.025 mg L⁻¹. For validation, the MAE, RMSE, and MBE were 0.078, 0.100, and −0.009 mg L⁻¹, respectively as estimated by the LOOCV procedure. The resulting model produced a model fit (R²_adj) of 0.58 and a deviance explained of 61.2% (Supplementary Table S4).

Generally, in the northern hemisphere, algal biomass, expressed as chlorophyll-a concentrations, tends to peak between July and October (mid-summer to mid-fall) due to a complex interaction of between and within-season factors with significant intra- and inter-annual variability in peak season biomass (Singh and Singh, 2015; Wilkinson et al., 2022; Beal et al., 2023). This period was characterized by warmer temperatures and increased sunlight (longer days) that contributed to algal productivity. However, algal biomass is also driven by local hydrology and regional weather modulated to some degree by global climate teleconnections (weather and climate patterns that connect regions globally) (Beal et al., 2021). The hydrologic driver of algal productivity has been linked to the transport of external nutrients to lake ecosystems (Vollenweider, 1976; Schindler, 1977; Elliott et al., 2006) with watershed land cover composition influencing nutrient concentration and loading intensity (Bachmann et al., 2012; Xiong and Hoyer, 2019). Moreover, internal loading (nutrients fluxing from sediment and to some degree plant biomass decomposition) has resulted in a significant influence on the overlying water column and has been known to fuel algal blooms (Albright et al., 2022; Waters et al., 2023).

In addition to specific drivers of algae dynamics within discrete ecological zones, algae pigments varied spatially and temporally (Figure 3). While direct comparison of algal pigments (chlorophyll-a vs. phycocyanin) dynamics was problematic with the current analyses due to spatial and temporal limitations between datasets, it was apparent that stage elevation, season, and location affected pigment concentrations. Some drivers were different (e.g., seasonal and spatial; Figures 3B,F,D,H, respectively). Meanwhile, there appeared to be consistency with the influence of stage to some degree (Figures 3A,E; Supplementary Figure S2). The spatial differences observed in chlorophyll-a concentrations (Figure 3D) as demonstrated by the fine scale spatial effect, were consistent with zonal descriptions of algal dynamics in prior studies (Havens et al., 1994; Havens, 2003).

Lake Okeechobee experienced distinct wet and dry seasons where the wet season (May – October) corresponded with peak algal bloom periods. Moreover, the majority of surface water inputs occurred during the wet season. Wet season inputs combined with water management rules affected water storage volumes, water levels, and water retention times of the lake (Tarabih and Arias, 2021; Julian and Welch, 2022). Inflow discharges introduced the bulk of the surface water nutrient loads. However, internal lake loads were many times greater than surface water inputs (Havens and James, 2005) with some regions within the lake (i.e., pelagic) having a much higher sediment nutrient concentration, hence a nutrient reservoir (Julian et al., 2023). Internal loading effects on algal communities are not isolated to just Lake Okeechobee. In Lake Mendota (Wisconsin, United States), studies have demonstrated internal loading met phytoplankton demand for growth (Kamarainen et al., 2009b; Kamarainen et al., 2009a; Carpenter and Brock, 2024). This internal loading supplementation of phytoplankton simulation has been observed in other lake systems such as Lake Taihu (Ding et al., 2018; Yin et al., 2022), Lake Chaohu (Yang et al., 2020), Lake Rauwbraken (van Oosterhout et al., 2022), Lake Erie (Wang et al., 2021) to name a few.

Nearly all freshwater cyanobacteria contain phycocyanin, an accessory pigment that complements chlorophyll-a to produce energy, especially in low light conditions (Alegria Zufia et al., 2021; Kheimi et al., 2024). As observed in this study, chlorophyll-a and phycocyanin concentrations had distinct seasonality (Figure 3). This seasonal change corresponds with the typical algal biomass trends observed across the northern hemisphere in aquatic systems. However, phycocyanin concentrations within Lake Okeechobee exhibited a bimodal seasonality with the first peak occurring in the typical summer months followed by another peak later in the year (Figure 3). This bimodality was driven by either changes in light attenuation/water clarity, physiological changes in cyanobacteria due to changing seasonal conditions, shifts in species composition, or a combination of these factors (Zhang et al., 2016; Carpenter and Brock, 2024).

Seasonally, cyanobacteria blooms occurred in the summer months, predominately as Microcystis sp. (Supplementary Figure S3) between April to July. Meanwhile, later in the year between September and March there was a much more diverse cyanobacteria community composed of Microcystis sp., Cylindrospermopsis sp., Dolichopermum sp., and other species (Supplementary Figure S3). While phycocyanin pigments/sensors themselves cannot differentiate between species, they are more sensitive to detecting cyanobacteria specific biomass changes than just chlorophyll-a (Carpenter and Brock, 2024). However, in Lake Chaohu (Anhui Province, China), Zhang et al. (2016) observed seasonal changes in chlorophyll-a, phycocyanin concentrations, and species biomass. Additionally, Zhang et al. (2016) observed a bimodal seasonality in cyanobacteria biomass, with Anabaena sp. (now Dolichospermum sp.) as the dominant species contributing to this pattern, while Microcystis sp. was less dominant when bloom conditions peaked in summer and gradually declined throughout the rest of the year. While individual species demonstrated a single unimodal seasonality, together they exhibited a cyanobacteria biomass bimodal seasonality.

Generally, TP-chlorophyll-a statistical models have been used to predict chlorophyll-a concentrations, algal bloom conditions, and trophic conditions (Vollenweider, 1975, 1976; Nicholls and Dillon, 1978; Håkanson et al., 2005; Quinlan et al., 2021; Olson and Jones, 2022). However, these relationships typically had poor predictive capabilities due to stochastic dynamics, error, variability and uncertainty in predictors, simplified model structure and failure to meet model assumptions. Moreover, most of these models were developed using a single year or average values across several years thereby missing the year-to-year variability. Year-to-year variation in the TP-chlorophyll-a relationship has been significant, driven by a combination of internal and external processes that either maintain some disequilibrium or conflate inter-annual variation in relationships (Davidson et al., 2023). However, process-driven models, like the one presented by Olson and Jones (2022), can identify biological limitations and other factors driving variability in TP–chlorophyll-a relationships across space and time.

Between and within lakes the TP-chlorophyll-a relationship can be highly variable. Several synthesis studies have suggested the relationship was non-linear across the TP concentration continuum and affected by other factors. These factors included lake morphology, lake depth, thermal regime, N vs. P growth limitations, inputs, outflow, sedimentation, and depth-dependent processes (Guildford and Hecky, 2000; Phillips et al., 2008; Søndergaard et al., 2017; Qin et al., 2020; Quinlan et al., 2021; Davidson et al., 2023). Depth-dependent and hydrodynamic processes either directly affect the lake nutrient mass balance (e.g., changes in inflow P or N loads) or in-directly affect lake mixing, biotic interactions (e.g., zooplankton grazing) or biogeochemical cycling (e.g., denitrification). Some TP-chlorophyll-a modeling approaches attempted to account for the potential influences of these other factors by including lake depth, WRT, and water balance to name a few (Vollenweider, 1975; Olson and Jones, 2022).

Lake water depth, water level, or stage elevation, sometimes all used interchangeably, have been applied as metrics by water managers for reservoir and natural system management to estimate the storage volume in a given lake and to inform management decisions. Moreover, lake water level is a relatively easy and effective metric to measure. Storage volume, outflow discharge rate, and WRT are key factors in nutrient cycling, collectively influencing ‘contact time’—the duration for which nutrients remain available for uptake (dissolved) or settling (particulate).

Early Lake Okeechobee studies generally focused on algal biomass, TP concentrations, and water level (Walker and Havens, 1995; Havens and Walker, 2002). Prior studies documented a direct effect of water levels on algal biomass and recognized it as a local phenomenon generally restricted to the near-littoral region. However, on a lake-wide basis the correlation of stage and bloom frequency was observed to be highly variable (Havens et al., 1994; James et al., 1994). Lake Okeechobee is a spatially heterogeneous ecosystem with variability in physical attributes, water chemistry (Phlips et al., 1993), sediment nutrient distribution (Julian et al., 2023), and hydrodynamic influences (Chen and Sheng, 2003, 2005). These factors combined with seasonal variability in stage drives algal biomass trends and bloom frequency. This variability was also apparent in the stage-based chlorophyll-a and bloom frequency models developed by Walker (2020) and in this study (Figures 4, 6). Therefore, including other factors that account for the unique spatial characteristics of the system should provide a more robust estimation of algal biomass.

While phosphorus (P) is predominantly the limiting nutrient in most lake ecosystems, in some cases, nitrogen (N) can be a limiting nutrient for algal growth (Guildford and Hecky, 2000; Jones et al., 2008). Therefore, incorporating N into the TP-chlorophyll-a model can improve model fit and better explain changes in primary productivity (Søndergaard et al., 2017). While Lake Okeechobee has received high TP loads from its watershed, to the point of reducing its assimilative capacity and contributing to high internal loading (Havens and James, 2005; Julian et al., 2023) resulting in abundant resources for algal growth, N has been an important driver of algal dynamics as well.

For example, Havens (1995) documented a shift from P limitation to a secondary N limitation in the mid to late 1980s, however some variability in this limitation status has been identified across ecological regions of the lake (Supplementary Figure S4; Julian Unpublished Data). Moreover, it has been documented through in-situ observations and incubation experiments that algal dynamics in Lake Okeechobee have been driven in part by changes in DIN concentrations (Phlips and Ihnat, 1995; Gu et al., 1997; James et al., 2011). Based on this dynamic and the interplay between P, N, hydrodynamic characteristics, and the overall heterogeneity of the lake among regions, the hierarchical predictive models (i.e., LOK HABAM) presented in this study were developed to understand changes in chlorophyll-a with an aim to evaluate changes in system management. The inclusion of DIN and the interactive effect with TP allowed for capturing greater variability in the pelagic zone. In contrast, the stage-only model captured very little variability in the various algal bloom metrics with relatively flat slopes (Figure 4; Supplementary Table S3), resulting in reduced predictive capabilities in restoration scenarios (Figure 6). This result demonstrated that important ecosystem drivers other than stage influence this part of the system, consistent with prior studies discussed above.

Despite some variation in potential nutrient limitations (Supplementary Figure S4), prior studies have conducted nutrient limitation bioassays and determined that Lake Okeechobee is predominantly N-limiting, especially in summer when nitrate concentrations are typically the lowest (Supplementary Figure S5; Phlips and Ihnat, 1995; Philips et al., 1997; James et al., 2011). Additionally, Philips et al. (1997) noted regional occurrences of N and P co-limitation in the mid to late summer. These patterns suggest seasonal shifts in nutrient concentrations and biological cycling that align with variations in potential summer nutrient limitations observed over the period of record (Supplementary Figure S4). Meanwhile, ammonium concentrations remained relatively low and did not vary much between seasons (Supplementary Figure S5). Despite these low concentrations, Hampel et al. (2019) documented significant ammonium regeneration within Lake Okeechobee from internal sources suggesting rapid ammonium turnover rates have the ability to fuel and sustain algal blooms. Given this dynamic, including DIN concentrations as a predictor to the hierarchical predictive models not only resulted in an improved fit to the model it also captures the internal biogeochemical dynamics of the system.

Water management, regional flood control, and restoration infrastructure (e.g., reservoirs, flow-equalization basins, stormwater treatment areas, etc.) have exhibited significant influences on local and regional hydrology as well as the ecology of lakes and downstream waterbodies (Havens, 2002; Richter et al., 2003; Julian and Welch, 2022). In the context of this study, both water management and restoration infrastructure were evaluated by comparing in-lake water, biogeochemical, and algal dynamics between LORS08 and LOSOM followed by the effect of utilizing storage north (LOCAR) and south (EAA Reservoir). Changes in lake hydrodynamics (i.e., inflow, outflow, storage volume, HRT, etc.) have shown a substantial effect on lake water quality and algal dynamics, effectively regulating the bioreactor nature of the lake as suggested by LOK HABAM and other modeling efforts in other lake ecosystems (Wang et al., 2013; Olson and Jones, 2022; Hanson et al., 2023).

As a potential application of LOK HABAM, water quality improvement scenarios were evaluated to understand the response of summer mean chlorophyll-a concentrations across the lake under LOSOM with the EAA reservoir in place. These scenarios are not necessarily the “how” but the final results of nutrient reductions observed within the lake. While more robust water quality modeling is needed, this modeling exercise demonstrates that not all parts of the lake respond similarly. Moreover, the predicted chlorophyll-a concentrations responded disproportionately with greater changes relative to in-lake TP concentration reductions (ranging from 10 to 30%) relative to concurrent changes in DIN concentrations (Supplementary Figure S6).

Under LOSOM (PA25) algal bloom conditions and/or bloom risk could be greater than all other alternatives. This risk is driven by significantly higher stage elevations relative to the baseline conditions (Julian and Welch, 2022) and alternatives associated with LOCAR (Figures 5–7) in its current eutrophic state with consistent exceedance of the regulatory inflow load limit. Although with the addition of southern storage features (FWOLL) and the addition of northern storage (LCR1) with LOSOM in place, the benefits of LOSOM (i.e., reduce discharges to northern Estuaries, increased discharges to the Everglades) (Julian and Reidenbach, 2024) were achieved and algal bloom risk was reduced with added storage. This added storage aids in modulating water levels in the lake by buffering high flow events from the north and reducing periods of high-water levels thereby improving the lake’s ecology and building resilience in the system (Figure 5). The combined effect of water management utilizing the necessary storage infrastructure significantly improved the hydrodynamic condition of the lake, which can alleviate some ecological stress associated with high water levels (Julian and Welch, 2022). However, other issues remained that also contribute to the lake’s chronic algal bloom and eutrophic condition, including high nutrient inflow loads and the legacy effect of high nutrient storage in lake sediment. While an attempt was made to address inflow nutrient loads with water quality improvement projects and deployment of best management practices, efforts may need to be more aggressive to achieve the expected goal outlined in the Lake Okeechobee Basin Action Management Plan to ensure that the Total Maximum Daily Load can be achieved (Florida Department of Environmental Protection, 2020; Julian et al., 2023).