Phosphorus trends and hot spots—a spatio-temporal data analysis of phosphorus derived from Everglades Agricultural Area (EAA) farms (Florida, USA)

Abstract

The Everglades Agricultural Area (EAA) in South Florida (USA) is a recognized source of total phosphorus (TP) that has impacted downstream oligotrophic Everglades marshes. Treatment wetlands, called stormwater treatment areas (STAs), were constructed and subsequently expanded to remediate EAA-derived TP, ideally yielding long-term outflow concentrations to the Everglades Protection Area (EvPA) of 13 µg/L TP or less. To date, TP-remediation has been insufficient relative to TP loads discharged from some EAA basins. We assessed 20 years of EAA basin-level and farm-level TP concentration and outflow data with the goal of understanding trends over time and identifying TP hot spots. Using monitoring data from water year (WY) 2000 through WY 2019, TP discharged from EAA farms averaged 74.68 ± 38.87 µg/L and was as high as 269.38 µg/L. We identified spatial and temporal variations in TP concentration, farm outflow, TP load, and TP flow-weighted mean concentration from the EAA farms. EAA basins TP concentration showed the presence of a decreasing trend between WY 2000 and WY 2012 and increasing trends for the more recent period WY 2010 to WY 2019. Using a nine-parameter Analytic Hierarchy Process (AHP), we observed that 31% of EAA farms posed above-average pollution risk, including 22 farms in the S-5A and 17 farms in the S-6 basins. These hot spot farms are primary candidate sites for targeted interventions aimed at reducing TP runoff, alleviating the TP burden on downstream STAs, and offsetting the recent increasing trend.

Supplementary information

The online version contains supplementary material available at 10.1007/s10661-025-13794-0.

Introduction

Agricultural nutrient runoff—primarily nitrogen and phosphorus—contributes to a wide range of adverse environmental impacts, diminishes ecosystem services, and threatens public health (Dang et al., 2021; Dodds & Smith, 2016; Foley et al., 2005). The effect of nutrient pollution also extends to fisheries, tourism, and real estate (Carpenter et al., 1998; Compton et al., 2011; Mekonnen & Hoekstra, 2018). Harmful algal blooms, driven in part by agricultural runoff, are a critical environmental concern and are increasing in frequency and severity (Anderson et al., 2008, 2021; Czajkowski et al., 2021; Karlson et al., 2021; Litke, 1999).

To address this growing problem and cost-effectively mitigate nutrient pollution in landscapes dominated by agricultural pollutants, studies increasingly focus on identifying agricultural pollutant hot spot sources across landscapes at various spatial scales (Czajkowski et al., 2021; Groffman et al., 2023; Li et al., 2022a; Schoumans et al., 2014; Sharpley et al., 2013). Studies have identified several critical factors in defining agricultural pollutant hot spots, including legacy nutrient concentrations in the soil, rainfall-runoff relationships, soil type, farm management practices, and discharge behavior—many of which reflect the benefits of targeted farm-level management practices (Caruso et al., 2013; Khare et al., 2021; Packett et al., 2009; Reddy et al., 2020; Robotham et al., 2021; Upadhaya et al., 2023). For instance, accumulated legacy soil nutrients can contribute to runoff concentration, and the timing of fertilizer application can significantly impact nutrient levels in water bodies (Carpenter et al., 1998; Jarvie et al., 2013; Panagopoulos et al., 2011; Sharpley et al., 2013). Additional factors such as topography, hydrological connectivity, vegetative cover, climate variability, and water management infrastructure also play key roles in influencing nutrient transport and release (Gray et al., 2016; Pionke et al., 2000). This complexity of agricultural nutrient pollution means that identifying the prominent sources of pollution can be challenged by the availability of data needed to identify trends and differentiate hot spot farms. Despite the complexity, identifying agricultural pollutant hot spots using comprehensive monitoring data analysis and watershed modeling or combining empirical data and modeling allows for targeted interventions (Chen et al., 2019; Khare et al., 2020, 2021). These interventions, by enabling the effective identification of hot spot nutrient sources and highlighting potential targets for intervention, offer a cost-effective way to mitigate nutrient pollution, reduce downstream nutrient loads, and protect water quality (Abimbola et al., 2020; Gray et al., 2016; Khare et al., 2021; Monaghan et al., 2008; Sharpley et al., 2013; Teshager et al., 2017).

The oligotrophic Everglades wetlands and coastal waters of South Florida (USA) are naturally limited by phosphorus availability (Rader & Richardson, 1992). The enrichment of Everglades marshes by P-rich runoff from the Everglades Agricultural Area (EAA) has been understood for decades (Davis, 1994; McCormick & O’Dell, 1996; Rader & Richardson, 1992). However, limited studies have been conducted to evaluate spatial trends within basins and to identify farms that disproportionately discharge phosphorus to the Everglades Protection Area (EvPA), hereafter referred to as “hot spot” farms. It has been demonstrated that when surface water total phosphorus (TP) concentrations in Everglades wetlands exceed ambient concentrations (5–10 µg/L) or when soil TP concentrations exceed 500 mg P/kg dry weight, a cascade of ecological changes, such as the loss of periphyton mat and expansion of cattail (Typha domingensis) relative to sawgrass (Cladium jamaicense), ensues (Belanger et al., 1989; Gaiser et al., 2005; McCormick & O’Dell, 1996; Noe et al., 2001). Historically, canals receiving runoff water mainly from the EAA had TP levels 30 times greater than ambient concentrations (Belanger et al., 1989). Current records of TP concentration in waters across the EAA were observed as high as 116 µg/L, more than 10 times ambient concentrations (Wang et al., 2022). This disparity highlights the influence of agricultural activities on phosphorus levels in the Everglades.

Long-term monitoring data has documented changes in Everglades ecosystem conditions, and litigation ultimately prompted the State of Florida to take aggressive steps to reduce nutrient loads into the downstream Everglades. This litigation and subsequent legislation of the Everglades Forever Act (EFA) in 1994 included the establishment of an EAA-wide phosphorus load compliance methodology, implementation of farm-level best management practices (BMPs), and construction of stormwater treatment areas (STAs) to reduce TP runoff from farms within the EAA before discharging to the downstream Everglades (Johns, 2002). Despite efforts over the past two decades, including the implementation of BMPs, excess TP loads continue to prevent the achievement of long-term water quality goals at the outflow points of the STAs (NASEM, 2022). Upon completion of components associated with the state’s Restoration Strategies Program (expected by the end of 2025), the regulatory water quality–based effluent limit (WQBEL) will become effective. The WQBEL is a regulatory compliance tool with the objective of being protective of the downstream Everglades. The WQBEL is a two-part regulatory test that requires individual STA discharges that are not to exceed 13 µg/L TP more than 3 out of 5 years and 19 µg/L TP annually as an annual flow–weighted mean (FWM) (SFWMD, 2012).

In a recent biennial review, the National Academies of Sciences, Engineering, and Medicine (NASEM) Committee on Independent Scientific Review of Everglades Restoration Progress (CISRERP) identified among their recommendations that EAA TP load reductions through source controls are one option for attaining the WQBEL for downstream Everglades wetlands (NASEM, 2022). Comparatively, the long-term plan for achieving water quality goals in EvPA and tributary basins suggests interventions that include identifying hot spot farms within the EAA and implementing source control measures in these specific areas (Burns & McDonnell, 2003; Piccone et al., 2003). This necessitates a more resolute (i.e., farm level) approach to understanding TP load contributions, as loads are generally reported for the entire EAA and at the basin scale. Therefore, the objective of this study was to establish a structured, data-driven protocol for the identification of TP hot spot farms within the EAA. These results have the potential to aid in identifying areas to target nutrient reduction strategies, reducing farm TP runoff, STA loading, and downstream impacts to Everglades wetlands.

Methods

Site description

Located downstream of Lake Okeechobee, the EAA (Fig. 1) encompasses 283,000 ha of mostly agricultural land. Lake Okeechobee is the prominent source of water to irrigate the EAA farms. Prior to drainage and agricultural operations throughout the first half of the twentieth century, the EAA was a mix of swamp forest and sawgrass plains with more than 4-m peat soil deposits covering carbonate deposition predominated by a limestone platform (Aich et al., 2013; Dreschel et al., 2018; Hohner & Dreschel, 2015; Lodge et al., 2023). Extensive drainage and subsequent soil oxidation have led to the loss of more than 2 m of soil elevation across this area (Dreschel et al., 2018). Presently, the EAA is characterized by flat terrain with a thinning layer of organic soil atop a limestone bedrock.

Fig. 1.

Maps showing A the extent of the greater Everglades ecosystem and South Florida Water Management District, B the Everglades Protection Area (EvPA), comprised of Everglades National Park and the water conservation areas (WCAs) which are immediately downstream of the Everglades Agricultural Area (EAA), and C a zoomed-in view of EAA farms and farm infrastructure including flow and water quality sampling locations, drainage canals, and treatment wetlands: stormwater treatment areas (STAs) and flow equalization basin (FEB)

Due to drainage and soil loss, farming in the EAA relies heavily on agricultural intensification practices, including soil amendments, intensive fertilizer use, and a wide array of BMPs (Belanger et al., 1989; Izuno et al., 1999). According to the South Florida Water Management District’s (SFWMD) permits geodatabase (https://geo-sfwmd.hub.arcgis.com/), the EAA primarily cultivates sugarcane, with some vegetable farming, covering a total of 187,069 ha. Approximately 93% of EAA farmlands are dedicated to agricultural uses, including sugarcane and other crops, while the remaining land, particularly along the southern edge near the Everglades, has been transformed into water management features such as treatment wetlands (Fig. 1C).

Four primary canals drain farm runoff and Lake Okeechobee water into the downstream STAs: the Miami Canal (serving farms in the S-8 basin), North New River Canal (S-7 basin), Hillsboro Canal (S-6 basin), and West Palm Beach Canal (S-5A basin). Each basin routes runoff into the downstream STAs (Fig. 1C). For this study, we grouped the EAA farms according to the specific canal system they drain into (Fig. 1C). The STAs ultimately discharge water into the Everglades Protection Area (EvPA; Fig. 1A), which covers approximately 1,000,000 ha and includes the water conservation areas (WCAs) and Everglades National Park (ENP).

Data sources

We obtained daily recorded data from the SFWMD that included daily records of structure discharge volume, precipitation, and TP concentration measured at 315 distinct control structures within the EAA by the SFWMD (Fig. 1C). Of the 315 structures, some of these were either renamed or replaced by new structures over time. Additionally, certain farm structures were merged or split. After a thorough evaluation of the permit history spanning 20 years, we identified 249 unique structures, which were then aggregated into 174 farms. In this study, the data spanning from water year (WY) 2000 to WY 2019 is analyzed. In addition, the SFWMD provided us with basin-level annual TP loads (TPL), outflow, and TP flow–weighted mean concentration (TPFWMC) records. These EAA basin-level data capture the inflow from Lake Okeechobee to the EAA basins and the corresponding discharge generated from EAA basins.

Conceptually, water from Lake Okeechobee flows into the EAA basins via the four main canals. A portion of this water is diverted into farms within the EAA basins through secondary and tertiary canals for agricultural uses. Another portion flows through the canals (as flow-through) to the respective STAs located downstream. In addition to the diversion from the main canals, EAA farms generate significant runoff, especially during the wet season. The net water discharged from the EAA farms (hereafter called farm-generated) includes the runoff generated from farms and the excess water diverted from the main canals. The overall outflow from the EAA basin comprises this farm-generated outflow, lake flow-through, and other basin inflow sources with negligible contributions. This study focused on lake inflow, total outflow, and farm-generated values of flow volume, TPL, and TPFWMC.

In this study, we analyzed datasets from 37 farms in the S-5A basin, 57 in the S-6 basin, 47 in the S-7 basin, and 35 in the S-8 basin, all of which had sufficient records for analysis. Each farm may have one or multiple structures where inflow and outflow from the farm are managed, and daily data is recorded. To establish the daily farm-level data, we aggregated the control structures’ data into a monthly scale. While the farm-level TP concentration and rainfall were calculated by averaging the structure-level records, the total farm-level flow volume (Q) was calculated as the sum of the flows recorded at individual structures. The farm-level daily TPL was determined by multiplying the daily farm-level TP with the farm-level total flow volume (TPL = TP × Q). While synthesizing the daily farm-level data from structure-based records, we evaluated historical changes in structure operations and farm area changes from permit documents available in the SFWMD repository. Historical farm-level permit information is available in the SFWMD database located at https://my.sfwmd.gov/ePermitting. Finally, we also aggregated the daily TP, Q, and TPL farm-level data into a monthly scale. The farm-level monthly TPFWMC is calculated based upon this monthly farm-level TPL and Q. Using farm-level monthly data, we evaluated the presence of temporal trends and correlation among variables and calculated descriptive statistics for each, including mean, median, standard deviation, range, and quantiles.

Trend analysis and hot spot mapping

To account for potential changes in TP concentration discharged over time from the EAA farms, we analyzed data for the periods WY 2000 to WY 2020, WY 2000 to WY 2012, and WY 2009 to WY 2020. This division was determined following a thorough exploratory data analysis. A summary of methodology can be found in Table 1. At the basin level, we evaluated the annual average TP concentrations using ordinary least squares trend to determine the best-fit trend lines, with statistical significance assessed using p-values < 0.05 to determine whether the observed trends were likely due to random variation.

Table 1.

Summary of analyses, scale, methodology, and date ranges

Description	Method	Date range
The sub-basin-scale trend analysis for TP used annual averages	Ordinary least squares (OLS) trend	WY 2000 to WY 2020, WY 2000 to WY 2012, and WY 2009 to WY 2020
The farm-scale trend analysis, using monthly data for TP, Q, TPL, and TPFWMC	Seasonal Mann–Kendall (MK)	WY 2000 to WY 2019, WY 2000 to WY 2010,  and WY 2010 to WY 2019
The farm-scale hot-spot analysis	Analytic Hierarchy Process (AHP)	WY 2000 to WY 2020

For farm-level assessments, the seasonal Mann–Kendall (MK) test was applied to evaluate monotonic trends for monthly TP, Q, TPL, and TPFWMC. We used the seasonal Mann–Kendall (MK) test package developed by the Department of Natural Resources and Parks, Science and Technical Services Section in Seattle, WA (Burkey, 2012). Widely used by researchers in the fields of hydrology and water resource systems evaluations, this function computes non-parametric monotonic trend tests, including Kendall’s Tau and Sen’s slope, with months identified as seasons (Abdi, 2007; Hirsch & Slack, 1984; McLeod, 2005; Teng et al., 2024). Just like the basin-level trend tests, the MK test was categorized into three time domains. However, to maintain a complete dataset that reflects consistency with the seasonality test, the second decade was adjusted to cover WY 2010 to WY 2019. In this study, trends were evaluated at a significance threshold (alpha) level of 0.05.

Hot spot farms—those with an above-average propensity, along with an increasing trend, to contribute TP pollution to the EvPA—were identified using an Analytic Hierarchy Process (AHP) approach (Saaty, 1980; Wind & Saaty, 1980). While other models could have been considered, the AHP approach was selected for its robustness in combining diverse inputs and offering a clear prioritization of hot spots. The strength of AHP lies in its ability to incorporate both quantitative data and intangible variables, making it suitable for complex decision-making scenarios where multiple factors must be considered. While it may not be traditionally associated with environmental trend analysis, AHP has proven effective in identifying and prioritizing hot spots in various contexts, including environmental management and planning (see Rawat et al., 2022; Guerriero et al., 2022; Jabbar et al., 2019; Borah et al., 2025, Sivrikaya and Küçük, 2022, Barman et al., 2024; Berhanu et al. 2021; Shinde et al., 2024; Shinga et al., 2024; Júnior and Rodrigues, 2012). Unlike deterministic models, the AHP approach does not calculate absolute contributions by specific variables, but rather determines how a criterion compares in importance to other factors within the specific AHP evaluation. The AHP method involves breaking down a complex decision problem into a hierarchy of simpler sub-problems, assigning weights to each criterion, and calculating a ranking of alternatives based on these weights (Saaty, 1980, 2005; Wind & Saaty, 1980).

After considering several factors and applying expert judgment, we identified nine elements that effectively assess a farm’s potential pollution risk. These criteria include farm-level TP, Q, and TP FWMC trends, along with average annual values for TP, Q, and TP FWMC. Additionally, the strength of correlation (r) between pairs of variables—specifically TP and Q, TP and precipitation, and Q and precipitation—was incorporated as part of the nine criteria for AHP evaluation. The strength of correlation was used as an indirect measure of BMP performance. It is important to note that we did not have access to the farms, nor could we directly assess the BMPs implemented on these farms. However, if BMPs were indeed implemented, the correlations between these variables could offer insights into the relationships and potential impacts that BMPs may have on water quality and flow dynamics. For example, a weak correlation between TP (total phosphorus) and Q (discharge) might suggest that BMPs are effectively controlling phosphorus concentrations in relation to water flow.

The relative weights and magnitudes of these criteria are calculated mathematically as follows:

$$\text{Ho}=\sum _{\text{i}=1}^{\text{n}}{W}_{\text{i}}\times R_{\text{i}}$$ 1

where Ho is the hot spot index of an individual farm i, $W_{\text{i}}$ is the relative weight of each indicator’s criteria, and $R_{\text{i}}$ represents the magnitudes of these indicator criteria.

To determine the relative weight (Wi) of each indicator, we employed the standard pairwise comparison method as suggested by Saaty (1980). This methodology uses a pairwise comparison matrix where each element is evaluated against others for relative importance on a scale ranging from 1/9 (extremely unimportant) to 1 (equally important), to 9 (extremely important). The comparison helps quantify the relative importance of indicators, with increasing values indicating greater importance (Saaty, 1980, 2005; Wind & Saaty, 1980). To ensure the reliability and coherence of the weightings, a consistency ratio was calculated and maintained below the threshold of 0.1, in accordance with Saaty’s recommendations (Saaty, 1980).

The magnitudes of these indicator criteria were calculated data for WY 2000 through WY 2020, and each variable is used to rank values for each farm. The ranking (R_i) of individual farms was calculated based on the magnitude of each variable’s long-term observed data at the EAA farms level. After calculating the relative weight and ranking for each variable at the individual farm level, we presented the final AHP result as a normalized index ranging from 0 to 1, termed the “AHP Index.” Farms with above-average AHP Index, value greater than 0.5, are identified as hot spots. These hot spots are characterized by a higher potential and increasing trend for releasing phosphorus-polluted water, which can disproportionately impact downstream STAs and Everglades wetlands compared to the overall EAA farms.

Results

Basin-scale TP, farm-generated outflow, and TPFWMC

During the period of record (WY 2000 to WY 2019), outflow discharge volume, TP load, and TPFWMC consistently exceeded inflow values from Lake Okeechobee (Table 2), reflecting the contribution of rainfall and local sources of water and TP to the downstream STAs within each basin. Similar statistics for the interval decades can be found in Supplemental File 3. The statistical summary in Table 2 indicates that outflows from the EAA basins frequently exceeded inflows from Lake Okeechobee, underscoring the contribution of local sources. In the S-5A basin, the average annual inflow from Lake Okeechobee was 175.39 ± 118.41 Mm³/year, while farm-generated outflow averaged 257.84 ± 94.09 Mm³/year, yielding a mean total outflow of 358.57 ± 152.48 Mm³/year. Similar patterns were observed in the S-6 basin, where the average inflow from the lake was 94.92 ± 46.64 Mm³/year, while farm-generated and total outflows were 298.60 ± 65.64 Mm³/year and 347.64 ± 81.31 Mm³/year, respectively. The S-7 basin experienced mean lake inflow and total outflow volumes of 177.52 ± 87.23 Mm³/year and 303.51 ± 131.35 Mm³/year, respectively. By comparison, the S-8 basin had a notably high mean outflow volume of 487.56 ± 195.22 Mm³/year, with a farm-generated outflow of 304.05 ± 95.27 Mm³/year. These results demonstrate that farm-generated outflow discharge constituted a substantial portion of the EAA basins’ total outflow, accounting for approximately 71.9% in the S-5A basin, 85.9% in the S-6 basin, 79.5% in the S-7 basin, and 62.4% in the S-8 basin.

Table 2.

Descriptive statistics for TP, flow, and load conditions into and out of the S-5A, S-6, S-7, and S-8 basins. Annual averages for the period of record WY 2000 to WY 2019

Annual basin data	Lake inflow volume (Mm3)	Total outflow (Mm3)	Farm generated outflow (Mm3)	Lake inflow TPL (t)	Total outflow TPL (t)	Farm-generated outflow TPL (t)	Lake inflow TPFWMC (µg/L)	Total outflow TPFWMC (µg/L)	Farm-generated outflow TPFWMC (µg/L)
S-5A basin
Count	20	20	20	20	20	20	20	20	20
Mean	175.39	358.57	257.84	35.80	64.66	40.41	209.23	178.36	149.91
Std	118.41	152.48	94.09	23.04	32.45	23.23	37.30	44.95	54.77
Min	25.30	154.60	131.22	5.20	24.77	6.69	146.07	102.98	45.60
25%	105.27	242.70	186.65	23.50	38.36	23.48	186.67	137.10	113.19
50%	156.25	347.36	238.38	31.50	62.03	41.41	200.66	171.85	149.06
75%	218.23	444.46	322.50	43.70	73.25	49.97	243.42	214.98	191.85
Max	526.73	777.19	431.65	100.88	132.23	97.08	283.29	254.20	232.33
S-6 basin
Count	20	20	20	20	20	20	20	20	20
Mean	94.92	347.64	298.60	12.39	38.70	31.84	127.65	109.16	103.59
Std	46.64	81.31	65.64	7.37	16.71	14.98	46.62	29.94	32.41
Min	18.39	213.79	179.83	2.28	15.39	11.80	68.67	62.47	51.50
25%	77.82	278.16	246.81	7.16	26.91	23.04	92.96	82.47	80.57
50%	89.87	351.84	304.29	12.33	37.77	27.78	114.28	108.52	102.11
75%	114.28	396.44	342.63	17.18	45.75	39.10	154.03	122.75	119.48
Max	205.90	496.86	417.90	25.19	87.98	73.62	260.21	177.08	176.15
S-7 basin
Count	20	20	20	20	20	20	20	20	20
Mean	177.52	303.51	241.37	23.18	31.42	25.08	127.65	103.42	99.64
Std	87.23	131.35	80.59	13.78	16.75	14.84	46.62	37.44	42.60
Min	34.40	149.66	129.68	4.26	9.57	1.88	68.68	43.76	14.50
25%	145.54	244.49	177.07	13.39	20.33	16.06	92.96	76.74	80.51
50%	168.08	280.24	227.67	23.06	27.35	24.07	114.28	96.09	94.97
75%	213.73	328.04	296.82	32.13	39.87	32.94	154.03	123.99	116.42
Max	385.08	775.03	407.05	47.11	80.98	69.41	260.21	188.44	191.72
S-8 basin
Count	20	20	20	20	20	20	19	20	20
Mean	192.03	487.56	304.05	22.46	46.45	26.20	113.22	97.86	86.98
Std	109.57	195.22	95.27	15.31	20.76	14.09	40.08	31.72	40.89
Min	0.00	193.93	167.25	0.00	15.31	7.22	62.50	59.93	41.34
25%	121.15	349.80	217.23	10.36	34.72	17.80	83.19	65.25	48.80
50%	172.69	450.81	292.39	15.61	40.70	22.16	103.24	92.13	80.83
75%	250.15	640.94	392.92	35.29	61.98	34.16	133.33	120.52	114.53
Max	415.06	954.25	462.92	56.84	89.85	57.87	217.04	152.24	173.09

Looking at the TPL record, over the 20-year study period, the sum of all lake inflow TPL values was 1876.52 tons, while the EAA basins discharged a total of 3624.48 tons. At the basins scale, in the S-5A basin, the mean TPL from lake inflow was 35.80 ± 23.04 tons/year, and the mean total outflow TPL was 64.66 ± 32.45 tons/year. The mean farm-generated TPL from all the farms in the S-5A basin was 40.41 ± 23.23 tons/year. For comparison, the S-6 basin received a mean TPL of 12.39 ± 7.37 tons/year, total outflow TPL of 38.70 ± 16.71 tons/year, and farm-generated TPL of 31.84 ± 14.98 tons/year, showing relatively low variability. Similarly, the S-7 basin received mean TPL values with lake inflow TPL at 23.18 ± 13.78 tons/year, total outflow TPL at 31.42 ± 16.75 tons/year, and farm-generated TPL at 25.08 ± 14.84 tons/year, exhibiting moderate variability. Meanwhile, the S-8 basin received lake inflow TPL of 22.46 ± 15.31 tons/year, total outflow TPL of 46.45 ± 20.76 tons/year, and farm-generated TPL of 26.20 ± 14.09 tons/year, also showing moderate variability. In terms of contributions, the farm-generated TPL comprised 62.49% of TP discharged from the S-5A basin. Similarly, the S-6 and S-7 farm-generated TPL accounted for 82.27% and 79.82%, respectively, of the total basin-level TPL discharged to their respective downstream STAs. For the S-8 basin, the farm-generated TPL contribution constituted 56.40% of the total TPL discharge from this basin. It is worth noting that the overall total TPL discharged from the EAA basins is an aggregate of lake flow-through, farm-generated TPL, and other sources, as calculated by the SFWMD.

The overall 20-year annual average farm-generated outflow TPFWMC across all EAA basins ranged from 14.50 to 232.33 µg/L (Table 2). However, the basin-specific average discharge from the S-5A basin was 149.91 ± 54.77 µg/L. This was higher than the farm-generated discharge TPFWMC from the S-6 basin, which was 103.59 ± 32.41 µg/L, and from the S-7 and S-8 basins, at 99.64 ± 42.60 µg/L and 86.98 ± 40.89 µg/L, respectively.

Correlation analysis revealed a strong relationship between farm-generated outflow volume and farm-generated TPL in all basins, with Pearson correlation coefficients r = 0.83 (p < 0.001) in the S-5A basin, r = 0.82 (p < 0.001) in the S-6 basin, r = 0.72 (p < 0.001) in the S-7 basin, and r = 0.75 (p < 0.001) in the S-8 basin. In all basins, the total outflow showed a strong correlation with farm-generated outflows (r = 0.76, p < 0.001 in the S-5A basin, r = 0.90, p < 0.001 in the S-6 basin, r = 0.80, p < 0.001 in the S-7 basin, and r = 0.75, p < 0.001 in the S-8 basin). Additionally, a robust correlation existed between farm-generated TPL and total outflow TPL in the S-5A (r = 0.80, p < 0.001), S-6 (r = 0.96, p < 0.001), S-7 (r = 0.92, p < 0.001), and S-8 (r = 0.81, p < 0.001) basins, and between farm-generated TPL and total outflow TPFWMC in the S-5A (r = 0.74, p < 0.001), S-6 (r = 0.89, p < 0.001), S-7 (r = 0.78, p < 0.001), and S-8 (r = 0.83, p < 0.001) basins. These correlations illustrate that farm-generated outflow volumes drive TPL and TPFWMC from EAA basins to downstream STAs.

Farm-level TP, farm-generated outflow, and TPFWMC

Across the entire study period, the EAA discharged an average annual TP concentration of 74.68 ± 38.89 µg/L. The long-term average TP concentration for each farm ranged from zero to 269.38 µg/L (Fig. 2), with half of the farms producing mean annual TP concentrations within the interquartile range: 48.24 µg/L (1st quartile) and 92.20 µg/L (3rd quartile). Additionally, approximately 80% of the farms (~ 140 out of 176) had TP concentrations ranging from 37.36 to 122.80 µg/L, indicating that most farms were discharging TP concentrations 3.7 to 12 times greater than the water quality standard of 10 µg/L. A small subset, comprising 10% of the farms (18 out of 176), accounted for much higher concentrations ranging from 122.80 to 269.38 µg/L. When looking at farm-level discharge in different basins, higher TP concentrations were generated from farms located in the S-5A and S-6 basins, whereas farms in the S-8 and parts of the S-7 basin released more modest TP concentrations (Fig. 2A). Farms in the S-5A basin generated a long-term average TP concentration of 106.48 ± 52.41 µg/L, whereas the S-6 basin farms discharged an average TP concentration of 72.43 ± 28.81 µg/L. The S-7 basin had an average TP concentration of 66.11 ± 32.14 µg/L, while the S-8 basin farms averaged 56.23 ± 23.25 µg/L.

Fig. 2.

Maps illustrating EAA farm-level annual averages from WY 2000 to WY 2019 for A TP concentration (µg/L), B farm outflow (Mm3/y), and C TPFWMC (µg/L). Map data are presented using Jenks Natural Breaks classification scheme. Sub-basin borders are outlined in various colors. Histograms in the lower row display the distribution of the respective metrics data, with key quartiles (Q1 and Q3 values in blue dashed lines) and percentile thresholds (red lines) marked

This variability in TP concentrations across the basins highlights the differences in phosphorus management and contributions from the farms within each basin. While evaluating the year-to-year TP discharge time series data for individual farms, significant variability in TP releases over time was observed, with a coefficient of variation (CV) ranging from 0.21 to 2.81. This farm-level TP discharge variability, indicating substantial fluctuations in TP concentration over the study period across farms, was particularly pronounced in the S-8 basin, with an average CV of 0.76. In contrast, the S-5A basin exhibited relatively lower variability, with an average CV of 0.61, suggesting more stable TP discharge patterns. These CV values reflected varying levels of control and consistency in water management, with higher CVs indicating greater fluctuations and lower CVs suggesting more consistent outflow behavior.

The farm-generated outflow volume record (Fig. 2B) indicated that the EAA farms discharged an average annual outflow of approximately 8.24 ± 11.58 Mm³/year. While records show that farm-generated outflows range from about zero to 94.30 Mm³/year, half of the farms have outflows within the 1.50 to 10.03 Mm³/year range. The highest farm-generated outflows, ranging from 23.30 to 94.30 Mm³/year, are contributed by only 10% of the farms. Despite observing a few high farm-generated outflows in the S-6 and S-7 basins, the overall average farm-generated outflow is highest in the S-8 basin, followed by the S-5A basin, accounting for 11.38 ± 11.32 Mm³/year and 10.07 ± 10.85 Mm³/year, respectively. On average, farm-generated outflows in the S-6 and S-7 basins accounted for 7.18 ± 9.49 Mm³/year and 5.74 ± 13.98 Mm³/year, respectively. Aside from these statistics reflecting variability in farm-generated outflows across different farms, there was significant variability in outflow over time for each farm, with CVs ranging from 0.23 to 4.29. The S-7 basin showed the highest variability, with an average CV of 0.67. In contrast, the S-8 basin had a lower average CV of 0.56, indicating more consistent outflows over time. The S-5A and S-6 basins displayed moderate variability, with average CVs of 0.58, reflecting differences in water management and discharge practices among the farms within each basin.

Overall, farm TPFWMC ranged from 36.86 to 385.76 µg/L, while half of the EAA farms released between 97.35 and 170.45 µg/L to the respective canals. The average TPFWMC of all farms in the EAA was 141.10 ± 63.72 µg/L. High TPFWMC values, ranging from 225.46 to 385.76 µg/L, were observed on only 10% of the EAA farms. Farms in the S-5A basin exhibited the highest average TPFWMC of 183.78 ± 69.86 µg/L, indicating significant TP concentrations compared to other basins. Farms in the S-6 basin also exhibited elevated levels, with an average TPFWMC of 143.44 ± 58.38 µg/L. The S-7 basin followed with an average of 132.73 ± 56.18 µg/L. In contrast, farms in the S-8 basin were characterized by an average TPFWMC of 103.56 ± 48.75 µg/L. These values highlight the differences in phosphorus management and discharge practices across the basins.

For completeness, the interval plots associated with the long-term trends discussed above can be found in Supplementary File 1.

Phosphorus trends in the EAA

The temporal trends in EAA discharges were evaluated at both the basin level and at the level of individual farms from WY 2000 to WY 2020 as well as the intervals from WY 2000 to WY 2012 and WY 2009 to WY 2020. Although basin-scale TP concentrations over the entire timeframe did not reveal significant trends for any basin, decadal analyses indicated the presence of statistically significant decreasing trends from WY 2000 to WY 2012 in the S-6 (p = 0.006), S-7 (p = 0.003), and S-8 (p = 0.012) basins. During this period, the S-5A basin did not show a significant trend (p = 0.150). In contrast, from WY 2010 to 2019, TP concentrations increased significantly within all basins (Fig. 3).

Fig. 3.

Annual average TP concentrations discharged from EAA basins from WY 2000 to WY 2019. Regression lines (with slopes and p-values) fitted to each period show a significant decrease in TP during the period from WY 2000 to WY 2012 and a significant increase in TP from WY 2009 to WY 2019

In addition to TP concentration, trends in farm outflow, TPL, and TPFWMC datasets for individual farms were evaluated using the seasonal Mann–Kendall trend analysis approach, the results of which are summarized in Table 3. The detailed insights discussed below can be obtained with a simple viewing and filtering of the raw data (See Supplementary File 2) that was used to create Table 3. The long-term trend (WY 2000 to WY 2019) revealed a large proportion of EAA farms exhibited no trends, including 127 farms (72.1% of 176) for TP concentration, 147 farms (83.5% of 176) for farm outflow, 148 farms (84.1% of 176) for TPL, and 66 farms (37.5% of 176) for TPFWMC. The absence of a trend indicates that nutrient reduction measures may not be working as expected. In addition, 23 farms (13.1% of 176) showed a long-term increase in TP concentration, 18 farms (10.2% of 176) exhibited increased farm outflow, 15 farms (8.5% of 176) exhibited increasing TPL, and 40 farms (22.7% of 176) showed increasing TPFWMC. Compare this to 24 farms (13.6% of 176) that have maintained a decreasing TP concentration trend and 68 farms (38.6% of 176) that have maintained long-term decreasing trends for TPFWMC.

Table 3.

The total number and percentage of EAA farms with an increasing trend, no change, or decreasing trend for TP concentration, flow, TP load (TPL), and TP flow-weighted mean concentration (TPFWMC) from WY 2000 to WY 2019. “No Trend” indicates no statistically significant trend over the study period. “Trend Not Detected” indicates that no discernible trend was identified, likely due to limitations or gaps in the data

Time frame	Decreasing trend	No trend	Increasing trend	Trend not detected	Total farms
TP concentration
2000 to 2019	24	127	23	2	176
2000 to 2010	51	110	11	4	176
2010 to 2019	5	109	59	3	176
Farm outflow
2000 to 2019	9	147	18	2	176
2000 to 2010	22	139	12	3	176
2010 to 2019	1	150	22	3	176
TPL
2000 to 2019	11	148	15	2	176
2000 to 2010	21	139	12	4	176
2010 to 2019	0	146	27	3	176
TPFWMC
2000 to 2019	68	66	40	2	176
2000 to 2010	58	75	39	4	176
2010 to 2019	26	85	62	3	176

A detailed analysis of long-term (WY 2000 to WY 2019) and decadal (WY 2000 to WY 2010 and WY 2010 to WY 2019) seasonal trends at the EAA farms also revealed significant shifts in TP concentration, farm outflow, TPL, and TPFWMC (Fig. 4). While many farms showed no trend or increasing trends over the long term, opposite patterns emerged in the second decade (WY 2010 to WY 2019). Out of 59 EAA farms (33.5% of 176) with increasing TP concentrations in the second decade, 39 had previously shown no trend over the 20-year period, and 26 had shifted from a decreasing trend in the first decade to an increasing trend in the second decade (Fig. 4). Similarly, farm outflow data showed that 147 farms (83.5% of 176) exhibited no trend over the long term, but 22 farms (12.5%) saw increasing outflows in the second decade. Notably, 10 of these 22 farms had previously shown a decreasing trend, indicating a reversal.

Fig. 4.

Mann–Kendall trend results at the farm level for A TP concentration, B farm outflow, C TPL, and D TPFWMC, evaluated over three time periods. Columns from left to right represent analyses for WY 2000–WY 2019, WY 2000–WY 2010, and WY 2010–WY 2019

TPL data also reflects these shifts. Although 148 farms (84.1%) exhibited no significant trend over the long term, 27 farms (15.3%) showed increasing trends in the second decade (Fig. 4). Of these, 7 had previously shown decreasing trends, and 20 had shifted from no trend to an increasing trend. The TPFWMC data reveals even more pronounced shifts: 62 farms (35.2%) showed increasing trends in the second decade, with 29 transitioning from a decreasing trend and 20 from no trend. Additionally, 20 farms consistently showed increasing TPFWMC trends across both the 20-year period and the second decade, signaling persistent challenges.

Phosphrous hot spots in the EAA

The nine-criteria pairwise comparison matrix for AHP analysis (Table 4.) suggests that farm-level TPFWMC holds the highest weight, with a priority vector (PV) of 21.28%. Other key parameters, such as farm outflow (17.88%), the trend of TP concentration (16.08%), and the trend of TPFWMC (12.99%), were also considered significant elements for identifying hot spot farms. Collectively, these four parameters constituted nearly 68.23% of the total AHP weight, highlighting the critical role of TP concentration and farm outflow in identifying hot spots within the EAA. In addition to these, the correlation coefficients between TP concentration and flow, TP concentration and precipitation, and flow and precipitation accounted for a significant portion of the remaining weight, with TP and flow correlation at 12.39%, TP and precipitation at 3.70%, and flow and precipitation at 1.80%.

Table 4.

Pairwise comparison matrix and calculated weight Priority Vector (PV) of the nine AHP evaluation criteria for identifying TP hotspot farms in the EAA. The parameters included (1) TP, (2) flow, (3) TP FWMC, (4) trends of TP, (5) flow, (6) TP FWMC, (7) correlation coefficients for TP and flow, (8) TP and precipitation, and (9) flow and precipitation

To assess the phosphorus pollution potential across the EAA, we normalized the AHP results, resulting in a normalized AHP value, the AHP Index (Fig. 5). This normalization allowed us to identify and focus on high-risk farms that exhibit an above-average pollution potential (AHP Index ≥ 0.5). These high-risk farms are considered to contribute disproportionately to environmental pollution, with higher AHP Index values indicating greater environmental concern. Results indicated that the mean farm level AHP Index is around 0.45 while the most frequent AHP Index values approximated 0.50.

Fig. 5.

Spatial distribution of AHP Index outlining the hot spot farms that disproportionately release TP into downstream waters and STAs

Based on our analysis, there were 55 high-risk farms in the EAA. At the basin level, the S-5A basin contained 22 (or 40%) of all high-risk farms. Similarly, there were 17 (or 31%) high-risk farms in the S-6 basin. In comparison, the S-7 and S-8 basins accounted for 20% (11 farms) and 9% (5 farms) of high-risk farms, respectively. Focusing on the upper end of the highest-risk category, there are 18 farms with an AHP Index ≥ 0.7, and notably, 10 of these (56%) are in the S-5A basin—reinforcing that farms in the S-5A basin constitute a disproportionately high source of phosphorus pollution within the EAA. For farms with an AHP Index value below 0.5, S-6 has the highest number, with 40 farms, followed by S-7, with 36 farms (Fig. 5).

Discussion

Hot spot analysis has long been recognized as a critical tool in environmental management, particularly in identifying areas that contribute disproportionately to pollution, and then targeting these areas for remediation. By focusing on hot spots, managers can prioritize limited resources and implement more effective mitigation strategies, ultimately leading to significant improvements in environmental quality (Abimbola et al., 2020; Ariano et al., 2024; Teshager et al., 2017; Yang et al., 2021). In various contexts, including agricultural landscapes, urban environments, and natural ecosystems, hot spot analyses have proven invaluable in identifying key sources of pollution (Carpenter et al., 1998; Foley et al., 2005). For instance, targeted interventions to manage point source pollution in the Chesapeake Bay watershed have led to a 58% reduction in phosphorus and a 28% reduction in nitrogen (Boesch et al., 2001; Pionke et al., 2000). The hot spot analysis of 2606 sub-basins in the Yangtze River Basin by Li et al. (2022b) identified croplands as the primary source of nutrient loading, accounting for 58.88% of TN and 79.15% of TP in the basin (Li et al., 2022b). These examples underscore the utility of hot spot analysis in environmental management, providing a framework for understanding and mitigating pollution sources across diverse ecosystems.

The EvPA continues to receive significant phosphorus inputs from EAA basins, particularly from hot spot farms, posing ongoing challenges for water managers in meeting WQBEL standards. Prior to BMP implementation (WY 1980–WY 1996), the TPFWMC from EAA averaged around 164 µg/L (Wade et al., 2022). During the study period (WY 2000–WY 2019), we found that approximately half of the EAA farms maintained mean annual TP concentrations between 47.8 and 92.3 µg/L. However, given that the environmentally permissible effluent limit to the Everglades is 13 µg/L or less and should never exceed 19 µg/L, this level of TP discharge from the EAA basins continues burdening the STAs’ cleaning capabilities. Particularly, excessive TP discharge from several large farms in the S-5A basin has heavily impacted STA-1E and STA-1W (NASEM, 2022). According to NASEM (2022), these two STAs have consistently exceeded the WQBEL since WY 2017–WY 2022. While STA-1E and STA-1W are dealing with inflow TP concentrations of 157 and 194 µg/L, respectively, they discharge TP concentrations of 29 and 33 µg/L into the EvPA. Our findings reflect the fact that these STAs are overburdened because of the hot spot farm concentration in S-5A and S-6 basins. In addition, the increasing trend of TP concentration discharge from the EAA basins over the last decade (WY 2010–WY 2020) explains why these STAs TP concentrations were reported high. To achieve the target outflow concentration of 13 µg/L, the NASEM report recommends reducing inflow TP concentrations into STA-1E and STA-1W by 50 to 60%, or alternatively, increasing their treatment efficiencies from the current levels of 81 and 83% to 92 and 93%, respectively. Similarly, to meet the same target outflow concentration, STA-5/6 must improve its treatment efficiency from 74 to 91%.

At the EAA basin level, as well as within several individual farms, data shows that the TP concentration has been increasing, with a distinct shift from decreasing in most basins from WY 2000 to WY 2012 to increasing from WY 2010 to WY 2020. The data illustrates that such discrepancies in trends over time are attributed to the presence of a high TP concentration during the initial decade. The observed shift from decreasing trends in TP to more recent increasing trends raises questions about the long-term effectiveness of current BMPs. This reversal may be linked to evolving agricultural practices, potential BMP fatigue, or climatic variations that have altered soil oxidation rates or runoff patterns, leading to increased phosphorus mobilization in recent years. Understanding these drivers is crucial for developing more resilient phosphorus management strategies.

The exceedance of effluent limits in these STAs suggests that targeted interventions in basins are important for reducing the phosphorus load and improving the effectiveness of the STAs. Without addressing the TP contributions from high-risk farms, the STAs will continue to struggle to meet regulatory standards. If the increasing trend in TP concentrations persists, the Everglades could experience increased eutrophication or perhaps even reduced flow to ensure the WQBEL is met. The growing burden on STAs may mirror the costly infrastructure upgrades required in the Lake Okeechobee Protection Program due to escalating phosphorus inputs (Flaig & Reddy, 1995) potentially leading to regulatory penalties and increased operational costs.

Limitations

This study used TP concentration and its derivatives in discharge from EAA farms as the primary indicators for identifying hot spots. While these metrics are effective for highlighting patterns of phosphorus discharge, successful BMP implementation should account for farm-specific factors, including soil properties, hydrological conditions, and local climate variability, as these aspects significantly impact TP discharge across different scenarios. Additionally, unit area load (UAL) was excluded as a metric, as many large farms actively cultivate only portions of their total land area. Including UAL could underestimate the phosphorus impact of larger farms by spreading the TP load over the entire land area rather than the actively farmed portion, potentially misrepresenting their true contribution to phosphorus pollution.

While this study was limited by the availability of data up to 2020, the methods and techniques applied are sound. Further review and analysis of newer data as it becomes available is recommended to assess any potential advancements, and to review any potential changes in previously identified hot spot farms.

Conclusion and recommendation

Our study suggests that farms in the S-5A and S-6 basins contribute disproportionately to nutrient loading in downstream STAs based on observed phosphorus concentrations. The evidence of varying TP and farm-generated outflow behaviors among EAA farms highlights the point that a collective EAA basin compliance evaluation fails to capture the environmental impacts of individual farms. Given the strong correlation between farm runoff and TP load, we infer that farm-level runoff is the significant source of effluent into the STAs. The increasing trends observed from WY 2010 to WY 2020 in large farms in the S-5A and S-6 basins suggest these basins are potential sources of concern. To address persistent phosphorus pollution from EAA runoff, it is crucial to manage phosphorus hot spot farms with a targeted approach. Our findings support the recommendations by NASEM (2022) that propose enhancing and expanding STAs while focusing on targeted source control programs to reduce phosphorus discharge from the EAA farms.

By applying the AHP approach to monitoring data, we effectively integrated time-series information into spatial evaluation, providing systematic and interpretable hot spot identification in an intensive agricultural system. This study highlights the importance of monitoring data in informing environmental restoration and presents a suite of data analysis protocols, including a segmented trend evaluation that revealed hidden trends. The combination of time-series analysis with AHP allowed precise hot spot identification, offering flexibility in incorporating various determiners. Our method, which is adaptable to other landscapes, demonstrates how area-specific water quality and hydrologic variables can be used to identify hot spots.

Supplementary information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, A.Z.A., S.E.D. III; Methodology, A.Z.A.; Software, A.Z.A.; Validation, A.Z.A., S.E.D. III, P.J.; Formal Analysis, A.Z.A.; Investigation, A.Z.A., S.E.D. III; Resources, A.Z.A., S.E.D. III; Data Curation, A.Z.A.; Writing – Original Draft Preparation, A.Z.A.; Writing – Review & Editing, A.Z.A., G.L.L., N.M.A., P.J., S.E.D. III; Visualization, A.Z.A.; Supervision, S.E.D. III; Project Administration, A.Z.A., S.E.D. III.

Funding

This research did not receive any specific external funding. It was conducted as part of the authors’ routine work responsibilities at The Everglades Foundation. The findings and views expressed in this paper are solely those of the authors and do not necessarily represent the views of The Everglades Foundation.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All authors have read, understood, and have complied as applicable with the statement on “Ethical responsibilities of Authors” as found in the Instructions for Authors.

Consent for publication

All listed authors have reviewed and approved the manuscript before submission, including the names and order of authors. This ensures that all individuals credited as authors agree with the content of the manuscript and its submission to the journal.

Competing interests

The authors declare no competing interests.