Estimating Lake–Groundwater Exchange Using Hourly Water Level Fluctuations in Central Florida

Abstract

With mounting anthropogenic pressures on groundwater supplies, practical methods for quantifying lake–groundwater exchange are critical for water resources management. This is particularly important in karst environments where surface–groundwater connectivity is often high. The White method uses nighttime water level fluctuations to estimate groundwater flux. While the White method has been applied to flooded wetlands, published lake applications are rare. This study evaluated a modified White method for estimating leakage at 28 karst lakes in Florida. The method was modified to include evaporation correction, with both nighttime and all-day approaches evaluated. Using the nighttime correction approach, average annual groundwater flux (leakage) ranged from −2.4 to +1.9 m/y, with a mean of −0.5 m/y (negative indicates lake outflow). Without nighttime evaporation correction, leakage estimates would be erroneous by an average of −0.7 m/y. The results showed no significant difference from 138 leakage values compiled from previous studies that used diverse methods. The modified White method requires special attention to evaporation, filtering criteria, and hydrogeologic context. Overall, the method provides a useful complementary approach to other methods for estimating long-term annual lake–groundwater exchange with comparatively minimal data requirements.

1. Introduction

1.1. Lake–Groundwater Exchange

Surface–groundwater interactions play a crucial role in the ecohydrologic functions of many waterbodies [1,2,3,4,5]. Accurate quantification of this exchange is needed to support water resource management, such as understanding ecosystem responses to groundwater withdrawals, which are increasing in various regions around the world [6,7,8]. For example, in the Northern Tampa Bay region of Florida, intensive groundwater withdrawals for public supply historically caused considerable lowering of water levels at numerous lakes and wetlands [9,10,11,12,13]. Quantification of lake leakage and its relationship to regional pumping drawdowns have formed an important part of the regional regulatory agency’s efforts to assess and protect lake ecohydrology [14,15].

Lake–groundwater exchange is fundamentally governed by hydraulic head differences between the lake and surrounding aquifers, with water moving from areas of higher to lower hydraulic head [5,8,14,16]. This exchange can be quantified as groundwater flux (flow rate), often called leakage for lake outflow or seepage for lake inflow. However, the terms are often used interchangeably. In this paper, negative flux values indicate outflow from lake to aquifer, while positive values indicate inflow to the lake. Flow rates are largely controlled by the hydraulic conductivities and thicknesses of intervening lakebed sediments and confining units. These hydraulic properties and resulting groundwater fluxes can vary widely spatiotemporally both within and among lakes due to heterogeneities in subsurface geology, such as the presence of solution features.

1.2. Methods to Quantify Groundwater Flux

Methods for quantifying lake–groundwater exchange include water budget analyses, numerical models, seepage meters, isotopic tracers, Darcy’s law, and hydrograph analyses, among others [4,8,17,18]. Each approach includes assumptions, advantages, and limitations. Water budget methods are among the most widely applied [14]. However, groundwater flux is typically calculated as a residual term, accumulating errors from other terms. Numerical models allow characterization of more complex dynamics and development of scenarios [19]. However, these models can require extensive parameterization, data inputs, and calibration, and the spatial resolution of regional models may limit direct application to individual lakes. Seepage meters provide direct, in situ measurements of lakebed groundwater fluxes [20,21]. However, these point measurements must be spatiotemporally extrapolated and may not capture losses to deeper aquifers. Isotopic studies using conservative tracers within water molecules are ideal for regional-scale assessments, allowing simultaneous evaluation of multiple lakes without extensive well networks [22]. However, these methods require assumptions about the relative contribution and chemical makeup of water sources and require accurate evaporation and climate data. Darcy’s law allows estimation of lake groundwater exchange with different aquifer units [16]. However, this approach requires representative estimates of the thicknesses and hydraulic conductivities of lakebed sediment and relevant geologic units, often poorly known with plausible ranges spanning orders of magnitude. All methods are sensitive to time period, with wet periods often producing different results from dry periods [23].

1.3. The White Method

Water resource management and regulatory applications can benefit from techniques to estimate lake–groundwater exchange without extensive aquifer parameterization, costly field campaigns, or complex modeling efforts, balancing practicality with understanding of limitations. One simple technique, the White method [24], is a hydrograph analysis method that has been widely applied to estimate phreatophyte groundwater evapotranspiration in wetland and riparian areas [25,26,27,28,29,30,31,32,33]. During periods lacking precipitation, channel inflow, and channel outflow, this method isolates evapotranspiration and groundwater fluxes by comparing daytime to nighttime water table level changes. Accordingly, subdaily water level monitoring is required.

Using the White method, nighttime changes are taken to represent groundwater fluxes. Normalizing to a 24 h period, nighttime changes are subtracted from daytime changes to estimate evapotranspiration. Groundwater flux is assumed to not differ between day and night. Specific yield, representing aquifer characteristics, is multiplied by water level change to account for accessible pore space. The White method can characterize multiple hydrologic processes with a relatively simple monitoring approach. However, the method is highly sensitive to data noise, filtering or smoothing methods, nighttime length, and specific yield [25,26,27,28,29,30,31,32]. Furthermore, barometric correction is essential when water levels are calculated using pressure transducers not exposed to the atmosphere [34]. Additionally, plants may continue to access groundwater at night, challenging the method’s foundational assumption [32,35]. For these reasons, numerous extensions of the White method exist [33], although they often require fitting or parameterization processes that may be less accessible to time-constrained water managers.

The White method has also been applied, though less frequently, to open water systems, typically flooded wetlands [28,29,30,36]. Published applications to lakes are even more rare. In the open water adaptation of the method, specific yield is taken as 1.0, simplifying the problem. However, this assumption is only perfectly valid at a cylindrical waterbody, so its appropriateness depends on bathymetry in the region of stage changes [30,37]. At waterbodies where water level changes result in large area changes, the stage–volume relationship is non-linear, necessitating stage-dependent specific yield [28,30,37,38]. At larger waterbodies, wind and seiche effects may introduce further challenge. Critically, at large waterbodies like lakes, considerable evaporation occurs at the night owing to heat storage and other factors [39,40,41,42,43].

1.4. Study Aims and Contributions

While several studies have evaluated the advantages and limitations of the White method for groundwater evapotranspiration and wetland fluxes, a significant research gap exists in its application to lake environments. This gap is critical because of a fundamental difference in lake behavior: large open water bodies continue evaporating at night due to stored thermal energy, unlike terrestrial systems where nighttime evapotranspiration is minimal. Failure to account for nighttime evaporation at lakes would systematically produce more negative groundwater flux (higher leakage) using the White method. This study addresses this research gap through: (1) adaptation of the White method by incorporating nighttime evaporation correction to account for thermal storage effects in lake systems, (2) application to 28 karst lakes providing regional leakage estimates across west-central Florida, and (3) validation demonstrating statistical consistency with literature values from established methods. This work represents among the first systematic evaluations of the White method for lake–groundwater exchange. The study focuses on a sample of Florida lakes, a region with well-constrained leakage estimates from previous studies, providing an ideal testbed for a method validation.

2. Materials and Methods

2.1. Study Area

The study focused on Florida, particularly west-central Florida within the jurisdiction of the Southwest Florida Water Management District (SWFWMD) (Figure 1). The region is broadly characterized by flat topography, subtropical humid climate, and mantled karst terrain [44,45]. The carbonate Upper Floridan Aquifer underlies a variable thickness sand surficial aquifer, separated in places by a confining unit of variable thickness and integrity [46,47,48,49]. The state averages around 111 to 165 cm of precipitation per year, most arriving between June and September [44,45,50].

Florida contains over 7000 freshwater lakes, primarily formed by karst dissolution processes creating depressions that intersect the water table [44,45]. Central Florida lakes gain water from precipitation, runoff, and the surficial aquifer, and lose water through evaporation and leakage to the Upper Floridan aquifer (Figure 2) [14,23,45,51]. Additionally, various Florida lakes have channel inflow or channel outflow, although over 70 percent are closed basin systems [14,45].

The region’s highly variable karst hydrogeology results in marked differences in surface–groundwater connection among waterbodies, even those in geographic proximity [13,50,51,52]. Accordingly, some lakes show greater sensitivity to groundwater levels and withdrawals, and in some cases, augmentation has been used to reduce withdrawal impacts [23,53]. At individual lakes, the direction and magnitude of lake–groundwater exchange vary seasonally and annually with hydrologic conditions. Surficial aquifer fluxes commonly reverse direction, creating gaining, losing, and flow-through conditions [14,23,54].

2.2. Data Sources

Twenty-eight lakes within west-central Florida were selected for this study based on availability of active hourly stage data (Figure 1 and Table S1). The lakes range in surface area from 0.04 to 18.3 square kilometers (km²), with a median of 0.5 km². Continuous water level data were obtained from SWFWMD’s existing monitoring network [55]. As SWFWMD’s protocol for stage monitoring specifies venting of pressure transducers, barometric correction was not applied to raw stage data [56]. No new field installations or groundwater monitoring wells were required for this study.

Radar rainfall estimates were acquired from SWFWMD for each pixel overlying study lake centroids. These data are produced on a 2 km grid using the Next Generation Weather Radar (NEXRAD) network and are available from 1995 to present [55]. For this reason, a study period of 1995 to 2024 was used. However, few lakes had hourly monitoring for this entire period. Data were further reduced by filtering, discussed later.

Mean monthly lake evaporation values for west-central Florida were acquired from SWFWMD [14,57,58]. This dataset was developed using an energy budget method and indicates an average annual lake evaporation of 150 cm, with an absolute range of ±7% about the mean over 15 years (y). The data are used in SWFWMD’s standard water budget model process, and, consistent with SWFWMD’s approach, were disaggregated into daily values assuming uniform monthly distributions [14]. Interannual and spatial variability in lake evaporation have been previously shown to be generally modest within the west-central Florida study area [14,57,58]. Over the range of study area, latitudinal differences are expected to alter annual lake evaporation by up to about 5% [59].

Lake channel outflow elevations were identified for 25 lakes from publicly accessible SWFWMD minimum lake level and structure reports [60,61,62,63,64,65,66,67,68,69,70]. For lakes with operable structures with a range of potential outflow elevations, the low end of the operational range was used. Time-varying operations were not readily available for this study. Channel inflow was qualitatively assessed using aerial imagery, national digital elevation model data, National Hydrography Dataset information, and review of SWFWMD minimum lake level reports. Six lakes were identified as having potential substantial channel inflow (Table S1). Augmentation was assumed negligible at study lakes during the study period. Channel inflow, channel outflow, and augmentation were either explicitly filtered or were assumed to have negligible effects on long-term leakage estimates. In general, filtering criteria, discussed later, likely removed many stage changes affected by these variables.

All analyses were conducted using the R statistical software version 4.2 [71].

2.3. Modified White Method for Lakes

The White method for groundwater flux can be represented as

G = Sy × (24 × h),

(1)

where G is daily groundwater flux, Sy is specific yield, and h is the average hourly nighttime water level change, which can be positive or negative. Positive values of G indicate inflow to the lake, and negative values indicate outflow. As previously discussed, this method is only applicable during periods with no precipitation, surface inflows, or surface outflows, that is, when all fluxes besides groundwater exchange can be assumed negligible.

In this study, a modified White method was implemented for application to lakes. A specific yield of 1.0 was used for open water, assuming stage–volume functions are linear over the regions over which stage changes occur during any given night. Additionally, as heat storage in lakes powers nighttime evaporation, nighttime evaporation must be removed from water level changes to isolate groundwater flux. The method was therefore modified to:

G = 24 × h − 24 × E,

(2)

where E is the average hourly nighttime evaporation and other terms are as previously defined.

Evaporation correction theoretically eliminates the White method’s need for nighttime-only analysis. This allows use of more data in daily estimates, which should increase reliability. However, nighttime periods offer lower absolute evaporation rates and typically reduced measurement noise from wind, thermal effects, and anthropogenic disturbances. To assess whether theoretical advantages of either all-day or nighttime analysis translate to superior empirical performance, both applications were evaluated.

2.4. Data Filtering

Hourly water lake level data were filtered using the following steps to meet key assumptions of the White method.

• Rainfall: Days with >0 rainfall, including 1 d before and 2 d after, were excluded to isolate groundwater-driven fluctuations. The buffer period allows for temporal mismatch between daily rainfall data and hourly lake data and post-rainfall runoff.
• Outliers: Hourly water level changes exceeding 0.1 m were excluded. This helped to remove measurement error and reduced the potential influence of unidentified channel flows, wind effects, or seiche effects.
• Channel flows: For each lake, periods during which lake stage exceeded the lake’s channel outflow elevation were excluded from analyses. For lakes for which outflow elevations were not identified or for which channel inflow was possible but not measured, it was assumed the lake is closed basin, that overflow periods were generally infrequent, or that outlier removal excluded large inflow or outflow events.
• Nighttime: For the nighttime evaluation, nighttime was defined as the period between midnight and 4:59 am local time. Only hourly change values within this time period were used in analyses. This time period was selected based upon standard practice in the literature, e.g., [28]. This filtering criterion was not used for the all-day evaluation.

After filtering, the median lake had 2.0 y of data, with a range of 0.04 to 7.7 y (Figure 3).

2.5. Groundwater Flux Calculation

Lake groundwater fluxes were calculated as shown in Equation (2) using evaporation estimates and observed water level changes.

For the nighttime evaluation, nighttime lake evaporation was estimated for the 5 h nighttime measurement period using a simple approach. The 5 h period, representing ~20% of the day, was assigned 10% of daily evaporation, yielding nighttime hourly rates approximately 50% of the average daily rate. Given the general absence of published diel evaporation data for Florida lakes, this percentage was estimated using heat flux measurements from Florida lakes showing nighttime rates approximately 50% of daytime values [72], assuming similar patterns for evaporation. This is also generally consistent with the 15–40% range reported for nighttime evaporation contributions in warm, though arid, climates [41,42] where relative nighttime evaporation would be expected to be higher than in humid climates. The nighttime evaporation component was extrapolated to a 24 h equivalent rate, approximately 50% of the average daily rate.

Daily groundwater flux was estimated by multiplying the mean water level change by 24. Then, the 24 h evaporation was subtracted, as shown in Equation (2).

Mean daily groundwater fluxes were annualized using all available daily values, due to limited data availability after filtering.

2.6. Literature Review

Annual net leakage rates for Florida lakes were compiled from the literature. These values provided a regional reference dataset to evaluate and contextualize sample-level leakage estimates derived using the modified White method.

Most studies report net groundwater flux (net leakage). Some studies differentiate between surficial and Upper Floridan aquifer fluxes, which were added to obtain net flux. If annual average values were not reported (e.g., monthly), values were annualized.

Overall, net leakage data were extracted from 37 studies, resulting in 138 values representing over 100 unique lake systems across 15 counties [16,20,21,23,53,54,57,59,61,65,66,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98] (Figure 1 and Table S2). Methods employed in evaluated studies include water budget analyses, Darcy’s law calculations, numerical models, seepage meter measurements, and combinations of these methods. By number of values, the water budget approach was the most common method in these studies (50%), used alone or in conjunction with other methods.

To evaluate method performance, net leakage distributions from the literature review and the modified White method were compared using Kolmogrov–Smirnov tests for both evaluation approaches (nighttime and all-day). Seven study lakes also had values reported in the literature, allowing direct comparisons for both approaches using paired t-tests after confirming difference distributions were normal.

2.7. Influence of Channel Inflow and Outflow

To evaluate whether potential channel inflow or the absence of outflow elevation data influenced results, Kolmogorov–Smirnov tests were applied to compare leakage distributions between lakes with and without potential channel inflow and with and without identified outflow elevations.

2.8. Time to Converge on Leakage Estimate

To better understand data requirements for the modified White method, time needed for leakage estimates to converge on the final estimate was evaluated. Lakes with ≥730 days of filtered data (n = 14) were assessed. A convergence threshold of 10% was selected, reflecting a pragmatic tolerance level commonly used in water management contexts. For each lake, 100 bootstrap resamples were drawn from daily leakage estimates. Cumulative leakage means were calculated to identify the minimum data length needed for the resample estimate to converge within 10% of the final estimate as calculated from all available filtered data. The median convergence time across bootstrap resamples was calculated for each lake. This time to converge was compared to days of filtered data available and to standard deviation of daily leakage estimates using Kendall’s tau.

2.9. Sensitivity to Nighttime Evaporation

For the nighttime evaluation, sensitivity to the nighttime evaporation fraction was tested by varying the nighttime fraction of 24 h evaporation from 5% to 20% in 1 percentage point increments. The resulting change in mean annual leakage estimates, relative to the 10% value utilized in the study, was then calculated.

3. Results

3.1. Long-Term Leakage Estimates

For the nighttime evaluation, estimated annual net groundwater flux ranged from −2.4 to +1.9 m/y across the 28 study lakes, with a mean of −0.5 m/y and a standard deviation of 1.0 m/y (Figure 4 and Table S1).

For the all-day evaluation, estimated annual net groundwater flux estimates ranged from −1.9 to +1.9 m/y, with a mean of −0.2 m/y and a standard deviation of 0.8 m/y (Figure 4 and Table S1).

3.2. Comparison to Literature Values

Literature values ranged from −4.6 to +4.6 m/y, with a mean of −0.5 m/y and a standard deviation of 1.1 m/y (Figure 4 and Table S2). Leakage distributions were not significantly different between the literature values and the nighttime evaluation (p = 0.13) or the all-day evaluation (p = 0.19) (

Table 1. Summary of statistical tests of comparisons between modified White method estimates and literature values and water budget estimates, modified White method estimates using nighttime-only and all-day evaporation corrections, and modified White method estimates for lakes based on outflow and stream status.

Group 1 *	Group 2 *	Mean Difference (m/y)	Test	p-Value
Nighttime (n = 28)	Literature (n = 138)	−0.0	Kolmogorov–Smirnov	0.13
All-day (n = 28)	Literature (n = 138)	0.3	Kolmogorov–Smirnov	0.19
Nighttime (n = 7)	Water Budget (n = 7)	−0.3	Paired t-test	0.32
All-day (n = 7)	Water Budget (n = 7)	0.1	Paired t-test	0.20
Nighttime (n = 28)	All-day (n = 28)	−0.3	Paired t-test	0.03
Outflow Data, NT (n = 17)	No Outflow Data, NT (n = 11)	0.2	Kolmogorov–Smirnov	0.75
No Channel Inflow, NT (n = 22)	Potential Channel Inflow, NT (n = 6)	0.4	Kolmogorov–Smirnov	0.18
Outflow Data, AD (n = 17)	No Outflow Data, AD (n = 11)	0.4	Kolmogorov–Smirnov	0.31
No Channel Inflow, AD (n = 22)	Potential Channel Inflow, AD (n = 6)	−0.4	Kolmogorov–Smirnov	0.00

* NT = nighttime evaluation; AD = all-day evaluation.

).

Seven study lakes had leakage estimates available in previous literature studies, allowing direct comparisons. For these lakes, estimates from both the nighttime (p = 0.32) and all-day (p = 0.20) evaluations showed no significant differences from published values (Table 1). However, this mainly reflected the low sample size and low statistical power. Individual estimates varied considerably between this study and previous work. Compared to the literature values, root mean square error was 0.62 m/y for the nighttime evaluation and 0.26 m/y for the all-day evaluation (Figure 5).

3.3. Daily Estimate Variability

Daily groundwater flux estimates showed substantial variability for both approaches. For the nighttime evaluation, standard deviations for individual lakes ranged from 0.2 to 12.2 m for the nighttime evaluation (median of 3.6 m). For the all-day evaluation, standard deviations range from 0.4 to 11.0 m (median of 2.2 m) (Figure 6).

3.4. Influence of Outflow and Streamflow Conditions

For the nighttime evaluation, leakage distributions were not significantly different between lakes with and without outflow elevations (p = 0.75) or between lakes with and without potential streamflow (p = 0.18) (Table 1). For the all-day evaluation, leakage distributions were not significantly different between lakes with and without outflow elevations (p = 0.31) but were significantly different between lakes with and without potential streamflow (p < 0.01) However, these tests were limited by low sample sizes and low statistical power.

3.5. Time Needed for Leakage Estimate

Convergence analysis of 14 lakes with sufficient data showed leakage estimates converged to within 10% of the final estimate for only 8 lakes for the nighttime evaluation and 9 lakes for the all-day evaluation, with 5 lakes common between them. For the nighttime evaluation, the median number of years required ranged from 0.6 to 4.5 y (median of 1.9 y) (Figure 7a). For the all-day evaluation, the median number of years required ranged from 0.2 to 4.4 y (median of 0.8 y) (Figure 7b). The time to converge was not significantly correlated to the standard deviation of daily estimates or to number of filtered days of data available for either group (p > 0.10 for all tests), although the tests have low statistical power.

3.6. Sensitivity to Nighttime Evaporation

For the nighttime evaluation, leakage estimates changed linearly with nighttime evaporation assumptions. Each 1 percentage point change in the nighttime evaporation fraction (relative to the 10% baseline assumption) resulted in 0.07 m/y change in mean annual leakage estimates (Figure 8). The sensitivity coefficient directly reflected the regional annual evaporation rate of 1.5 m. The assumption that 10% of total evaporation occurs at night, then extrapolated from 5 to 24 h, yielded an annual nighttime correction of 0.7 m/y. Thus, each 1% change in the nighttime fraction corresponded to a 0.07 m/y shift in leakage estimates.

4. Discussion

4.1. Nighttime and All-Day Evaluations

Annual net leakage estimates from the all-day evaluation were significantly more positive than the nighttime evaluation. Neither showed a clear advantage with respect to literature values. Central tendencies were more similar between the nighttime evaluation and the broader literature values. Qualitatively, the nighttime evaluation’s leakage rates agree better with general reference ranges suggested by some Florida water managers [14], discussed later. However, the all-day evaluation offered lower variability, much better fit for the limited direct comparison lakes, and shorter times to converge.

Nighttime evaluation offers lower absolute error and typically fewer confounding variables. Given 1.5 m of annual lake evaporation, 20% error propagates 0.3 m error in leakage estimates for the all-day evaluation. For the nighttime analysis, when evaporation is plausibly half of daytime rates, the same relative error propagates only about 0.15 m of error. However, the nighttime requirement aggregates fewer hourly water level changes into daily water level change estimates, making each daily estimate less reliable. For the all-day evaluation, the increased sample size of hourly data supporting each daily estimate resulted in decreased variance and shorter time to converge.

In practice, the mean difference in estimated daily water level change before evaporation correction between the two approaches averaged 1 mm across lakes, with the nighttime evaluation generally more negative. However, this can aggregate into large differences over long periods, and annualized, explains the approximately 0.3 m more negative mean annual value for the nighttime evaluation.

Further discussion refers only to nighttime evaluation results.

4.2. Typical Leakage Rates

Through analysis of 28 lakes and literature review compiling 138 annual leakage estimates, the study produced one of the most comprehensive treatments of leakage for Florida lakes. This study’s application of the modified White method resulted in annual net leakage estimates in statistical agreement with previous studies, which used different methods, lakes, and time periods. Qualitatively, the most negative leakage rate estimated in this study was associated with Pasco Lake in Pasco County, known to have extreme leakage due to its unique hydrogeology (Table S1) [70]. This demonstrates the method’s potential to support regional and individual lake studies.

With average lake leakage rates of around −0.5 m/y, the results generally agree with net leakage rates of −0.5 ± 0.3 m/y used by some Florida water managers as general rule for reasonable leakage rates [14]. However, with inter-lake standard deviations of about 1.0 m/y, the results also underscore the wide range in leakage among lakes. High inter-lake variability in leakage results from Florida’s highly variable karst hydrogeology and spatially disparate pumping impacts [13,51,52]. Thus, population-derived values provide valuable benchmarks for regional efforts such as regional groundwater models, while offering reasonable starting points for individual lake assessments.

4.3. Regional Significance of Lake–Groundwater Exchange

The results highlight the important role of lakes as conduits for recharging groundwater, the primary water supply source for much of Florida and its renowned spring systems. Extrapolating the mean leakage rate across Florida’s approximately 1.2 × 10¹⁰ m² of lake area [99] suggests that, as a rough order-of-magnitude estimate, lakes collectively contribute roughly 2.0 × 10⁷ m³/day of groundwater recharge. This recharge from lakes, combined with contributions from Florida’s extensive wetlands and other recharge pathways, helps sustain the aquifer systems that provide drinking water to most of the state’s population and discharge to hundreds of springs, rivers, and coastal waters throughout the region. By comparison, the state’s freshwater groundwater withdrawals total around 1.4 × 10⁷ m³/day [100]. By capturing surface runoff and facilitating downward flux, lakes represent an important component of the regional hydrologic cycle that supports both human water needs and ecosystem functions [45].

4.4. Need for Nighttime Evaporation Data

The success of the modified White method demonstrates that nighttime evaporation correction addresses a fundamental requirement for lake applications. Without this correction, annual leakage estimates in this study would be systematically lowered by an average of 0.7 m/y. This bias would result in outflows well beyond the range established by the literature. The general absence of published White method applications to lakes may reflect this limitation, as uncorrected estimates would typically exceed physically reasonable ranges established by other methods. Nighttime evapotranspiration correction has similarly been identified as crucial at some wetland systems [35].

The coarse ratio approach for nighttime evaporation utilized in this study was generally successful at matching literature values. However, nighttime evaporation at Florida lakes remains poorly understood and could help improve the accuracy of the modified White method. Indeed, a key finding of this study is the importance of nighttime evaporation for understanding lake hydrologic processes. Future research priorities should include seasonal characterization of diel evaporation patterns and their regional variability. The direct proportional relationship between sensitivity and regional evaporation rates means the method would be more sensitive to nighttime evaporation assumptions in high-evaporation climates and less sensitive in low-evaporation climates.

4.5. High Variability of Short-Term Estimates

Despite the modified White method’s promise, daily estimates showed substantial variability, with the median lake showing a standard deviation of 3.6 m. This shows expected limitations in subdaily flux precision. However, the overall agreement with literature values indicates that annual means converge to reasonable long-term estimates despite high short-term variability. This suggests the method can produce reliable long-term estimates but is inappropriate for short-term flux estimates. For the study lakes, approximately 2 y of filtered data appears necessary for stable annual estimates; however, this value will differ among lakes and regions. In Florida, lake regulators often focus on long-term (annual to decadal) fluxes for groundwater management purposes (e.g., [14]).

Quantifying groundwater flux at lakes is inherently challenging. High uncertainty in leakage estimates is typical of virtually all methods to estimate lake groundwater exchange, with errors and standard deviations routinely exceeding mean flux estimates [17,22,23,58,93,94]. High temporal variability of lake groundwater fluxes is a real and well-documented phenomenon, with values at individual lakes differing by multiples or orders of magnitudes among months and years [23,58,93,94,96,98].

4.6. Limitations and Considerations

The modified White method addresses a key need in water resource management by providing estimates of lake–groundwater exchange without the extensive parameterization or field effort requirements of many alternative approaches. However, like all methods for quantifying lake–groundwater exchange, it includes inherent limitations that require consideration during application and interpretation.

Data requirements include hourly water levels and rainfall, nighttime evaporation estimates, and knowledge of lake channel flows. While substantial, these requirements are not more onerous than other hydrologic methods. The method shares well-documented sensitivities with terrestrial and wetland White method applications, including sensitivity to measurement noise, short-term fluctuations, and nighttime length definitions [25,26,27,29,31].

Temporal precision is limited, with individual daily estimates showing high uncertainty due to the method’s reliance on subdaily fluctuations. This is consistent with the high temporal variability documented for lake groundwater fluxes across virtually all quantification methods [17,22,23,58,93,94]. However, annual means converge to reliable long-term estimates. This positions the method as more suitable for long-term assessments where it can provide semi-independent validation of other approaches.

Study-specific limitations include the absence of channel outflow elevations for some lakes, conservative overflow procedures, limited channel inflow assessment, and the coarse evaporation estimation approach. Data filtering necessarily excluded a higher proportion of summer wet season days when precipitation and overflow are common. The specific yield assumption of 1.0 is generally more appropriate for larger lakes than for shallow wetlands, where non-linear stage–volume relationships can introduce substantial errors, though complex bathymetry could still affect applicability in some lake systems. The study did not explicitly account for seiche oscillations or wind setup effects, though these processes would likely average out over the annual time scales analyzed and be partially captured by existing outlier filtering. Additionally, only northern and central Florida lakes are represented. Despite these constraints, the method produced results statistically indistinguishable from literature values, suggesting these limitations do not prevent meaningful application and represent opportunities for future research.

However, the validation approach has important limitations. Only 7 lakes were available for direct comparison, limiting statistical power. The direct comparison and the broader comparison using 138 literature values may both be affected by inter-method biases, as different approaches (water budget, Darcy’s law, etc.) carry different assumptions and error structures. They also reflect different time periods and, for the broader comparison, different lakes. As classically noted by Winter [17], all methods for estimating lake–groundwater exchange carry substantial uncertainty, with errors routinely exceeding mean flux estimates. The observed statistical agreement demonstrates that the modified White method produces regionally consistent results with existing approaches, rather than proving superior accuracy of any specific method.

4.7. Appropriate Applications

Overall, the modified White method is most appropriate as a complementary approach, particularly for regional assessments where it shows promise for developing reasonable distributions using large sample sizes with limited individual lake information. For individual lakes, the method provides an estimate of long-term net groundwater flux using only two measured terms and no aquifer parameterization assumptions, a significant advantage over more complex approaches. However, individual applications benefit from longer (multi-year) data collection periods, careful filtering protocols, and integration with hydrogeologic knowledge. As with any lake study, leveraging multiple lines of evidence and grounding interpretation in physical understanding of the system increases confidence in results.

4.8. Global Relevance

The modified White method offers broad global applicability. The approach can be applied to any lake system with hourly water level monitoring and daily or nighttime evaporation estimates, providing water managers worldwide with a cost-effective tool for assessing lake–groundwater exchange without extensive field campaigns or aquifer parameterization. This is particularly valuable in data-sparse regions or for rapid regional assessments where traditional methods would be prohibitively expensive or time-consuming.

5. Conclusions

While the White method has been extensively applied to terrestrial and wetland systems, applications to lakes are rare in the published literature. This study demonstrates that lake applications are feasible with proper accounting for nighttime evaporation. The modified White method successfully estimated annual net groundwater exchange at Florida karst lakes, producing results statistically indistinguishable from a collection of previous studies. The results indicate that typical annual leakage rates at Florida lakes are around −0.5 ± 2.0 m/y. Nighttime evaporation correction was essential, removing an average −0.7 m/y that would otherwise be erroneously attributed to leakage. All-day evaporation corrections also offer reasonable results. While daily estimates show substantial variability, necessitating several years of data collection, the method provides suitable long-term estimates for regional assessments. For individual lakes, the method is thus best applied as a complementary approach. Future research should prioritize characterizing nighttime lake evaporation in Florida, utilizing the approach in other regions, and identifying lake characteristics associated with greatest success for modified White method applications. Although hydrogeologic context is essential for proper interpretation, the method’s minimal data requirements make it a practical option for estimating groundwater fluxes at lakes around the globe, particularly when other approaches are not feasible.