 |
|||||
| Ground-water conditions in southern Florida |
|
|||||
| Ground-water conditions in southern Florida |
This report has been reformatted for presentation on the World Wide Web. The official text of WRIR 01-4275 (6.3 MB download) is available in PDF format. The Adobe PDF Reader program is available, at no cost, from Adobe.
A real-time ground-water level monitoring network that contains only a fraction of wells in the existing ground-water monitoring network cannot provide the spatial coverage of the full network, but can still provide considerable insight into changes that occur during those intervals when data from the larger network are unavailable. Considerable care must be taken to ensure that this subset provides data that are as representative of changing aquifer conditions as possible. The subset of wells selected need to: (1) provide unambiguous and quantitative real-time information on unique and potentially damaging ground-water level events that are occurring and signal these events as early as possible; (2) represent ground-water conditions over a substantial area of the aquifer; (3) monitor specific areas where the aquifer may be more susceptible to water-level related problems; and (4) provide information that aids in the assessment of saltwater intrusion in those areas of the aquifer where such considerations are relevant.
Several quantitative and qualitative assessments need to be made when evaluating candidate wells for the real-time ground-water level monitoring network. The evaluation of well construction and period of record and the statistical approaches used to select real-time network wells are addressed in the subsequent sections of this report. Maps showing continuous ground-water monitoring wells used for this study are presented in figures 4 to 6. Figure 4 (96 K) shows the location of wells in southwestern Florida; figure 5 (146 K) shows the location of wells along the upper east coast of Florida in Palm Beach, Martin, and St. Lucie Counties; and figure 6 (132 K) shows the location of ground-water monitoring wells along the lower east coast of Florida in Miami-Dade and Broward Counties.
Well construction is an important consideration in the design of any ground-water monitoring network. It is the construction of the well that determines whether or not water levels from that well will be truly representative of the aquifer. Factors such as an indeterminate open interval, an insufficient annular seal, or an improper emplacement technique may adversely affect analysis of the water-level and water-quality data from the well. Although many characteristics of monitoring wells are set at the time of construction, others may change over time. For example, the part of a monitoring well left open to the aquifer may collapse over time, or sand may be forced up into the casing of a well by hydrostatic pressure in the aquifer. Well casings also can corrode, which in turn, may result in leakage from other parts of the same aquifer or other aquifers.
Network monitoring wells have been installed using a variety of methods. These methods vary because of differences in cost, aquifer lithology, types of drilling equipment, changes in available technology, or evolution of monitoring techniques. Some network wells were originally installed as water-supply wells and were designed to provide maximum water yield rather than to monitor the aquifer at discrete depths. The network also includes wells that have relatively long open intervals, short open intervals, and short screened intervals. For each candidate network well, well construction has been considered to determine if the data obtained from that well will yield unambiguous results (app. I).
Although well construction is important, period of record is one of the most important considerations in selecting representative wells and determining reasonable statistical results. The period of daily water-level record available for analysis differs from well to well. Several monitoring wells have less than 2 years of daily maximum water-level data, whereas others have greater than 60 years of daily value data. One area where data from many of the recorders do not span the full period of evaluation is southwestern Florida; continuous water-level monitoring of many of these wells did not start until the mid-1980's. Many of these water-level recorders were either removed or relocated to other wells in 1996, based in part on a statistical evaluation of well coverage (Hosung Ahn, South Florida Water Management District, written commun., 1996). Water-level recorders have been periodically removed and replaced in many aquifers throughout the network. In some cases, continuous recorders were replaced by periodic instantaneous manual measurements made by using a steel tape and chalk on a quarterly or monthly basis; in others, no data were collected during the intervening years.
There are several problems with using ground-water level data prior to 1974 for this analysis:
From 1974 to present, the data from almost all continuous ground-water level monitoring wells consist of daily maximum water-level elevation referenced to sea level.
For a long-term period of analysis, it is infeasible to accurately compare periodic instantaneous manual water-level measurements (collected monthly) to those obtained by a continuous water-level recorder (recording hourly) because of the large daily water-level cycles in many southern Florida ground-water wells. Computation of daily maximum water levels using hourly data essentially smooths these data by sampling the peak of each cycle. Conversely periodic, instantaneous manual measurements taken at different times of the day sample the water levels at different points in the daily cycle. These daily cycles are most pronounced in southwestern Florida.
Thus, daily maximum water levels provided by continuous hourly recorders or monthly means of these daily values have been used for all analyses, and period of record remains a critical component for interpreting the results from these analyses. Well construction and period of record information for each candidate well is presented in appendix I. The period of record information pertains only to the maximum daily value data entered in the NWIS database.
The distribution-free, nonparametric Seasonal Kendall trend test was used to test for the existence of trends in water-level data in candidate wells. This test, modified from the Mann-Kendall test (Helsel and Hirsch, 1992, p. 338), measures the monotonic association between two variables, determines whether these variables increase or decrease with time, and compares relative ranks of data values from the same season.
Monthly means of daily maximum water levels were used for the Seasonal Kendall trend tests. Because the period of record for each well was different, Locally Weighted Scatterplot Smoothing (LOWESS) of hydrographs from wells in each aquifer was used to help determine break points for the trend analysis for each aquifer, and trends were then analyzed for these shorter periods. For those wells where sufficient data exist, trends have been analyzed for the full period (1974-99) in addition to the shorter periods for each aquifer. The period of record available for each well was weighed against the number of wells that could be evaluated for that period. Trend tests were then run separately for each of these periods so that results could be compared. For example, if the period 1974-99 were selected, then only the wells that had record for all 26 years (83 wells) were analyzed for trends during this period. In this way, the trend results would be comparable between these 83 wells. Results of the trend analysis are presented in appendix II.
The monthly means of daily maximum water levels were compiled for each candidate well in the ground-water level monitoring network for the 1974-99 period. Summary statistics (mean, standard deviation, minimum, maximum, median, first quartile, third quartile, and interquartile range) were derived from these values for each well and are presented in appendix III. The summary statistics can be used to show those areas of each aquifer that have the greatest variation or lowest minimum values. These are the areas most susceptible to drought-related problems, depending on the physical characteristics of the aquifer and the effect of long term trends in water levels.
Water levels in some monitoring wells may not closely correspond to changes in rainfall. This is because monitoring wells are commonly located near areas where withdrawals of water from the aquifer are extensive. In some cases, water levels in a monitoring well may reflect local changes in withdrawal rates, rather than changes that affect large parts of the aquifer such as sustained reduction of recharge to the aquifer during a meteorologic drought. Changes in withdrawal rates may in turn be caused by crop cycles, population cycles related to tourism, and/or mandated decreases in pumpage as water restrictions are implemented. As previously discussed, however, the balance between recharge and withdrawals is most tenuous during droughts. Thus, it is important to identify monitoring wells where the relation between meteorologic droughts and water levels is clear.
A frequency analysis is commonly used to quantify variation in environmental data. For example, severe hydrologic events, such as floods, are said to have a 100-year recurrence interval. That is, historical water-level data from a stream are used in conjunction with a frequency analysis to determine which water level has a 1-percent chance of occurring each year. When water levels in the stream rise above that water level, a "100-year flood" has occurred.
A similar frequency analysis can be used to examine the relation between rainfall and water-level minimums. Theoretically, if extreme lows in water levels for a given monitoring well directly correspond to extreme lows in precipitation (meteorologic droughts), the well then could be used to assess the effect of meteorologic droughts on the aquifer. Conversely, if these extreme values do not correspond, the monitoring well may only provide useful information concerning a small area within the aquifer. One possible analysis is to compare the lowest 5 percent of monthly rainfall and water-level values for the same period to determine if they occur for the same periods. This concept was used herein to compare rainfall deficiencies and water levels in monitoring wells.
Monthly rainfall and water-level data usually are not comparable using frequency analysis without first performing some mathematical preprocessing of the data. The factors that were assessed before a correlation analysis was made between water level and rainfall from network wells, include: (1) long-term trends in data, (2) cumulative effects and lags, (3) recharge area uncertainty, and (4) seasonal cycles in data.
Rainfall was examined for long-term trends (1974-97) using the Seasonal Kendall trend test, but in almost every case, the trend determined was not statistically significant (p-value greater than 0.05). Rainfall data consisted of monthly rainfall totals from all cooperatively supported National Climatic Data Center stations that had data for the 1974-97 period. Twenty stations were available for analysis of rainfall in southeastern Florida, and eight stations were available in southwestern Florida (table 1). Because data from the National Climatic Data Center were unavailable for the 1998-99 period, monthly rainfall totals from SFWMD rainfall stations were used to estimate rainfall for these years. One set of rainfall stations showed a statistically significant trend, but the trend indicated was very small. As a result, rainfall data were not trend adjusted prior to use.
Water-level data may have significant long-term trends that can be upward or downward, and either monotonic or not. If these trends were not removed before performing the frequency analysis, then the results would be seriously skewed. For example, a severe drought may cause a 5-ft decline in water levels at a well, but this same amount of decline could be caused by a long-term (1 ft/yr) decline in the water levels within 5 years. Therefore, water-level data collected 5 years after the drought could be at the same elevation or lower on average as that collected during the drought (fig. 7 (54 K)). If the water-level data from wells affected in this way were directly compared to rainfall, the correlations between rainfall and water-level minimums would then be poor.
To compensate for the effect of monotonic or nonmonotonic long-term trends in water-level data, three mechanisms were used to trend adjust the data from each monitoring well examined: (1) linear regression, (2) second degree polynomial regression, and (3) LOWESS smoothing with an f-value of 1/5. (The f-value indicates the fraction of the data used to compute each point.) The residuals from these approximations were used for the subsequent analysis, rather than the raw data. Data that had less than 6 years of record were not trend adjusted.
Declines in water levels of aquifers tend to show the net or cumulative effect of decreases in rainfall over time and tend to lag behind these decreases (fig. 8 (55 K)). The amount of lag differs from aquifer to aquifer and also from well to well. To estimate the cumulative effects of rainfall deficiencies in ground-water levels, four different f-values were used for the LOWESS smoothing of rainfall data. The f-values used were 1/26, 1/52, 1/78, and 1/104. For 26 years of monthly rainfall data, these values correspond to using 1 year, 6 months, 4 months, and 3 months of data, respectively, to compute each point in the LOWESS smooth. The preprocessing of rainfall data using an f-value of 1/26 is shown in figure 9 (75K). Smaller f-values result in less smoothing of the rainfall data. Lag was addressed by mathematically lagging the water-level data by 0, 1, 2, 3 and 4 months relative to the rainfall data.
Recharge areas are poorly defined in the confined aquifers in southern Florida because of karstification and confining units of highly variable thickness and composition. Therefore data from one rainfall station may not necessarily correspond with changes in water levels at a water-level monitoring well, even if that well is near a rainfall station. An initial attempt was made to compare water levels in an aquifer with only those rainfall stations that were in each aquifer's recharge area, but there were too many uncertainties regarding confinement of the aquifers in southern Florida. Even the surficial aquifer system in Martin, Palm Beach, and St. Lucie Counties, though not differentiated into separate aquifers based on confinement, includes semipermeable units that slow direct recharge. Thus, two different rainfall models; one for southeastern Florida and one for southwestern Florida were used in this assessment (table 1). These models consisted of the average of rainfall data from all monitoring stations in each area.
A simple frequency analysis of rainfall and water level fails to account for normal seasonal fluctuation. Decreased water levels in monitoring wells can be produced by either reduced precipitation in the dry season or by lower than normal precipitation during the wet season as well as by increased municipal pumpage or increased drainage. To address the issue of seasonal cyclicity in rainfall and water-level data, monthly values were compared to the normal monthly mean values and monthly standard deviations for the period of record. The long-term monthly mean was subtracted from the value for that month, and the difference was expressed in number of standard deviations above or below the mean.
Water-level data consisted of monthly means of daily maximum ground-water levels from candidate monitoring wells, and rainfall data consisted of monthly rainfall totals from the previously mentioned National Climatic Data Center and SFWMD stations. An analysis of water-level and rainfall data was performed using S-plus Statistical Software. S-plus scripts were written to perform the following steps:
fa = vc / vt
where vc is the total number of minimum rainfall and water-level values that are concurrent in time between the two data sets, and vt is the total number of minimum values. The total number of minimum values is a direct function of the frequency analysis and the total number of values available from the data sets. For example, if 100 monthly values were available in both the rainfall and water-level data sets, 10 values in each data set would be less than or equal to the 10th percentile (+ 1 because of potential rounding issues). To account for rounding, the total number of values in the final rainfall file and the final water-level file were counted and divided by 2.
These 10 steps were performed iteratively for each well to include the 3 types of water-level trend adjusting, amounts of lag (0 to 4 months), and the rainfall models available for each area (LOWESS smooths with f-values of 1/26, 1/52, 1/76, and 1/104). Additionally, the lowest 5, 10, 15, and 20 percent of water-level and rainfall values were compared for each iteration. This analysis provided the fa for 240 combinations of these factors for each well examined to determine the closest relation between the rainfall data for the region and the water level for each candidate monitoring well.
Analyses were made for 246 wells. The best fa for each candidate monitoring well was determined, and results were arranged by period of record and completeness of record. The best fa for each well ranged from 29 to 100 percent and averaged 52 percent. Most of the wells that showed very high fa (80 percent or higher) had very short period of records (about 8 years, on average, after considering missing data). Of the 104 wells examined that had periods of record greater than 20 years (after considering missing data), only 15 wells showed an fa of 57 percent or greater. No fa for each of these 15 wells was greater than 62 percent. The period of record and completeness of record are the most important considerations in this analysis because agreement between water-level and rainfall minimums over a few years may show good statistical agreement, but they may not necessarily indicate what would occur during severe droughts. Final results from this analysis are presented in table 2.
Stepwise polynomial regressions were used to compare the water-level data from 203 candidate wells in each aquifer to determine which wells were most representative of the ground-water level monitoring network. The stepwise polynomial regression determines the best fit for the water-level data from one well relative to the water-level data from a comparable well, the best fit with the water level squared as an additional explanatory variable, and the best fit with time as another explanatory variable. For each iteration of this stepwise regression, a coefficient of determination (R2) is computed. The R2 value is a measure of the amount of variation in the dependent variable that is determined by the explanatory variable and must be from 0 to 1. The greatest value of R2 represents the best fit that can be provided using the explanatory variables available. A mean R2 value was computed for each candidate well based on an average of the greatest R2 values that were determined for all of the comparative wells; this process was repeated for every candidate well in each aquifer. The candidate wells with the greatest mean R2 value from each aquifer were considered to be the most representative of the water-level monitoring wells in the aquifer.
The interval of time over which the data from the wells can be compared is an important consideration. If one well has 20 years of water-level data and another well has only 1 year, the wells can be compared only for the 1 year of overlapping record. The R2 value for that comparison may be close to 1; however, because the two wells are only compared for 1 year, the result would not be very significant. If the water-level data from the two wells can be compared for a long-term period, differences can be assessed more thoroughly. This factor has been considered in conjunction with the regression analysis performed for this study.
The most representative wells (termed index wells) were selected based on the values of mean R2 and period of record. Index wells were required to have a minimum of 10 years of continuous water-level data. To evaluate the network coverage that these index wells would provide, the R2 values from individual well comparisons were considered. If the comparison of water-level data at the index well and a network well resulted in an R2 value of 0.64 or better, then the index well was considered to provide a fair estimate of water levels for that network well. (An R2 value of 0.64 corresponds to a correlation coefficient of 0.80.) This evaluation was repeated for each potential index well. Results of the analysis are discussed in subsequent sections of this report.
Although the primary goal of the real-time ground-water level monitoring network is to monitor water levels, the network also has the potential to provide water managers with early information about saltwater intrusion and upconing of saltwater as a result of aquifer withdrawals. To help assess the relation between changes in water level and chloride concentration, two factors were considered: (1) long-term trends in chloride concentration, and (2) correlation between water levels and chloride concentrations. The locations of the monitoring wells considered in these analyses are shown in figure 10 (98K).
To provide information concerning saltwater intrusion or upconing, long-term trends in chloride concentration were determined. Determination of water-quality trends for specific water-quality constituents requires that extraneous variation caused by natural phenomena (such as seasonality, streamflow, or precipitation) be compensated for so that temporal changes resulting from anthropogenic activities may be discerned. One of the principal causes of variation in water quality is seasonality. Many water-quality constituents may vary seasonally as a result of biological reactions, climactic changes, or changes in land use or water-management practices. This is also true for salinity monitoring based on chloride concentrations. The water-management system in southern Florida is regulated by control structures along the east coast canals. These structures are closed during the dry season to prevent saltwater intrusion and are opened during the wet season to discharge excess water to prevent flooding during heavy rainfall events. During the dry season, when lowered freshwater heads prevail in the aquifer systems, encroachment or upconing of the saltwater interface is more likely and may be reflected by increases in chloride concentration. Conversely, during the wet season, when higher freshwater heads are maintained, there is a likelihood of seasonal retardation of the saltwater interface, with resultant lower chloride concentration. However, the movement of the saltwater interface may not be immediate and may lag water-level changes. The seasonal variation in chloride concentration from well G-1351 in Miami-Dade County is shown in figure 11 (17K). Negating the variation caused by seasonality on chloride concentration enables an investigator to determine the long-term changes that have taken place over the years. The principal statistical tool used for trend detection was the Seasonal Kendall trend test.
Tests for trends in chloride concentration based on two seasons, wet and dry, were conducted on data from water years 1974 to 1998 for 50 wells in the Biscayne aquifer in Miami-Dade and Broward Counties and for 14 wells in the surficial aquifer system in Palm Beach and Martin Counties. In Lee and Collier Counties, tests for trends in chloride concentration were conducted on data from 1974 to 1998 for 9 wells in the water-table aquifer (west coast), 22 wells in the lower Tamiami aquifer, 15 wells in the mid-Hawthorn aquifer, and 3 wells in the sandstone aquifer. Statistically significant results are presented in table 3.
One aspect of the real-time ground-water network project was the examination of the relation between chloride concentrations and water levels. Wells that exhibit a statistically significant correlation between instantaneous water levels and chloride concentrations might logically be wells selected for high priority monitoring during drought periods depending on their proximity to the saltwater/freshwater interface. A correlation analysis between instantaneous water levels and chloride concentrations were performed for 114 wells in southern Florida during water years 1974-98.
Correlation coefficients measure the strength of the association between two variables but do not indicate a causal relation between the two. Variables may be correlated with each other in either a linear or nonlinear manner (Helsel and Hirsch, 1992, p. 210). For this aspect of the study, Spearman's r and Pearson's r correlation coefficients were employed, both of which measure monotonic (as x increases, y either increases or decreases) relations between two variables. Spearman's r is based on ranks, is resistant to outliers, and measures both linear and nonlinear monotonic associations. Pearson's r measures only linear monotonic associations between variables (Helsel and Hirsch, 1992, p. 210). Those wells for which a statistically significant correlation (p-value less than 0.025) was determined by Spearman's r were also analyzed by Pearson correlation coefficients. Correlation coefficients for these analyses are presented in table 4.
Parker and others (1955, p. 611) indicate that there are both seasonal and long-term changes in the saltwater/freshwater interface that occur over time, with long-term changes in the position of the interface lagging water-level changes. Merritt's (1996) assessment of saltwater intrusion in Broward County also documents seasonal as well as long-term changes in the position of the saltwater interface, and indicates that long-term water-level changes are more responsible for changes in the position of the saltwater interface than seasonal fluctuations. The significant, but relatively weak, correlations between chloride concentrations and instantaneous water levels documented by this study may be representative of short-term seasonal variations and not long-term changes in the position of the saltwater/freshwater interface. The South Florida Water Management District (1998) found that persistently lowered water levels, for greater than 6 months duration, resulted in a permanent inland movement of the saltwater interface, as opposed to lowered water levels over shorter time periods.
In order to examine the possible correlation between chloride concentrations and lagged water levels, three wells for which sufficient long-term monthly instantaneous water-level and chloride concentration data exist were tested for correlation between incrementally lagged chloride concentrations and water levels. The three wells (G-1179, G-1180, and G-1251) are located in the Biscayne aquifer in southern Miami-Dade County near the city of Homestead. At all three wells, chloride concentration data were lagged at monthly increments, from 1 to 60 months, and then correlated with instantaneous water-level data. No significant correlations were found for well G-1179, located slightly east of the approximate extent of the saltwater interface, as determined by Sonenshein (1997). Well G-1180, located at the saltwater interface, showed a significant but weak inverse correlation (Spearman correlation coefficient of -0.24) at a lag of 54 months. Well G-1251, located west and inland of the saltwater interface, showed significant but weak correlations at lags of 35, 36, 37, 38, 39 and 48 months, with Spearman correlation coefficients of -0.27, -0.36, -0.44, -0.40, -0.36 and -0.20, respectively.
Of the correlation coefficients determined from the water-level and chloride concentration data, 50 percent of Spearman's r coefficients were higher than Pearson's r coefficients. Results indicate that the water-level and chloride concentration relations between the wells were both linear and nonlinear in nature.
Selection of wells for the real-time ground-water level monitoring network was based on a combination of the factors discussed in preceding sections. In this section, an "initial" real-time network is presented for each aquifer that is based on regression analysis alone. A "preferred" real-time network is then proposed that considers additional factors (for example, analysis of water-level trends and minimums, chloride concentration trends, and so forth). Only data from the "preferred" or final network is presented in tables.
Wells that are already equipped with satellite telemetry were examined to see if they could be substituted into potential index well networks without seriously reducing the statistical validity of the network. In some instances, however, these real-time wells have little data available for analysis and as such, are not very useful. Because of the dynamic nature of the ground-water usage and changes in the drainage system, real-time ground-water level network wells should be frequently reevaluated to determine if they are still representative of regional aquifer conditions.
The lowest monthly mean water level recorded in the water-table aquifer (west coast) of southwestern Florida was 1.70 ft below sea level (well L-954 in app. III). Mean water levels in the water-table aquifer (west coast) are about 29 ft above sea level at well C-1075 (located about 5 mi northeast of Immokalee) and decrease toward the coast, particularly in the southern part of the study area. Several monitoring wells in the water-table aquifer (west coast) near the coast have recorded minimum monthly mean water levels that are near or below sea level. Wells C-969 and C-1063 have recorded minimum water levels less than 1 ft above sea level (app. III). Well L-954 in Cape Coral, well L-1403 on Sanibel Island, and well C-1065 near the Tamiami Trail in southernmost Collier County all have recorded monthly mean water levels below sea level. Variation in this aquifer tends to be small. Interquartile ranges vary from 0.75 ft at well C-981 to 3.58 ft at well C-1071.
Wells C-392 and C-496 showed water-level increases of 0.04 and 0.03 ft/yr, respectively, for the full period analyzed during 1974-99 (app. II). No statistically significant trends toward decreased water levels were found in any of the wells examined for the various time periods. Large increases in water levels were observed at wells C-131, C-496, C-953, C-1071, L-1403, L-1997, and L-2195 during 1989-95, ranging from 0.19 to 0.98 ft/yr (app. II). The largest increases during this period occurred at wells C-131 and C-1071 (0.57 and 0.98 ft/yr, respectively) southwest of Immokalee in northeastern Collier County.
Chloride concentrations in water at monitoring wells C-953 and C-1063 have shown increases of about +1 mg/L (milligram per liter) per year over the last 15 and 12 years, respectively (table 3). Despite this trend, chloride concentrations in both wells are still very low (less than 50 mg/L). Well C-953 is about 14 mi from the coast. The mostly likely sources of saltwater are leakage from other aquifers or upconing of connate water. Well C-1063 is much closer to the coast (less than 4 mi). The proximity of well C-1063 to the coast, combined with the minimum water levels that are near sea level in the vicinity of well C-1063, suggest that this well may be more susceptible to lateral saltwater intrusion. Well C-1063 only has about 3 years of daily maximum water-level record.
A downward trend in chloride concentration of -500 mg/L per year has been determined at well C-1065 over the 13 years evaluated. Although this is a very large decrease, chloride concentrations were initially well over 10,000 mg/L.
Regression analysis alone indicated that 4 index wells would be able to cover 89 percent of the water-table monitoring network (28 wells) in southwestern Florida, with an average R2 value of 0.80. These four wells are L-2195, L-1137, C-131, and C-997. Data from well L-2195 alone can be used to estimate water levels at 15 other wells (which would cover 57 percent of the network) with a mean R2 value of 0.66.
Analysis of water-level trends and minimums, chloride concentration trends, and correspondence of minimum water levels to droughts leads to the suggestion of a proposed network for the water-table aquifer (west coast) that includes index wells C-131, C-392, C-496, C-503, C-969, and L-2195 (tables 5 and 6). Considered together, these 6 wells could provide direct coverage or estimations of water levels at 89 percent of the 28 continuous monitoring wells in the network, with an average R2 value of 0.82 (table 6). Figure 12 (81K) shows parts of the network that are covered using these six index wells.
The first 4 wells of this network (C-131, C-503, C-969, and L-2195) can be used to estimate water levels at 19 other wells in the water-table network (table 5). Index well C-392 has been included in this network because it provides better coverage close to the coast where minimum water levels have been near sea level. Well C-969 was also selected because it is near well C-1063 (fig. 12(81K)) where an upward trend in chloride concentration has been determined (table 3). Considering the period of data available for comparison, water-level data from well C-392 showed one of the best agreements with extreme rainfall minimums. The fa determined was 0.57 (fig. 13(58K) and table 2).
The total depth measurements performed at potential index wells indicated that wells C-131, C-392, and C-496 are not at the full depths indicated in the construction logs. Well C-496, which should be 57 ft deep (app. I), is currently 36 ft deep (relative to land surface). This well is still open to the aquifer from 7 to 36 ft below land surface (app. 1). The casing of well C-392 ends at a depth of 24 ft below land surface. Several inches below the end of the casing of the well has filled with sand. Well C-131 may never have been drilled to the depth indicated in the construction logs (54 ft). The casing extends to a depth of 19 ft below land surface, and the bottom of the well is just below that depth (app. 1). There is a very thin layer of sediment at the bottom of this well, but the material beneath this sediment is hard and does not appear to have resulted from borehole collapse. Attempts to clear any loose material at the bottom of the well have been unsuccessful. Despite the well depth discrepancies, these wells are all open to the water-table aquifer (west coast) and appear to be good index wells. Basic well construction and analysis results for the index wells selected and the potential replacement well, C-997, are listed in table 7.
The lowest monthly mean of daily maximum water levels in the lower Tamiami aquifer was 5.07 ft below sea level at well L-1691 in Bonita Springs (app. III). During the period of record, water levels at wells in a large area of this aquifer have at times reached minimum values that are well below sea level. Twelve of the 20 continuous monitoring wells in this aquifer have had monthly means of daily maximum water levels that were below sea level (app. III). These wells are: C-391, C-409A, C-460, C-998, C-1004, C-1083, L-738, L-1691, L-2194, L-5727, L-5745, and L-5747. All are located in an area that extends from Naples in the south to Bonita Springs in the north (fig. 4 (96K)). The Pelican Bay and Naples Well Fields are located in this area (fig. 14 (105K)). The first quartiles of water levels at wells L-738 and L-5747 are near or below sea level for much of the period of record for each well (app. III; 1992 to present at well L-738; 1997 to present at well L-5747). Water-level variation in this aquifer, expressed as the interquartile range in feet, is about 3 ft on average and ranges from 1.11 ft at well C-600 to 5.60 ft at well L-1691 (app. III).
Water levels in a large part of the lower Tamiami aquifer increased by 0.15 to 0.80 ft/yr during the 1989-95 period. This increase is apparent in the water-level data from wells C-460, C-462, C-492, C-600, C-1004, C-1074, L-1691, and L-2194 (fig. 14 (105K) and app. II). This may have contributed to the decrease in chloride concentrations observed in four monitoring wells (table 3; wells C-975, C-1083, L-5725 and L-5727). Yet, chloride concentrations in water at five wells (C-489, C-492, C-525, C-528, and L-738) have increased (table 3). Chloride concentrations in wells C-489, C-525, and C-528 have increased by +1.5, +8.9, and +0.5 mg/L per year, respectively (table 3 and fig. 15A (71K)), despite the water-level increase of 0.11 ft/yr in nearby monitoring well C-391 during 1974-99 (app. II). Because of the proximity of these wells to the Gulf of Mexico, increased chloride concentrations in these wells may be caused by lateral intrusion of saltwater. Closer examination of the long-term water-level trend at well C-391 reveals that water levels have decreased by 0.17 ft per year for the 1989-99 period (app. II). This decrease may have contributed to the trend of increased chloride concentrations in the nearby wells.
The upward trend in chloride concentration observed at monitoring well L-738 (fig. 15B(71K)) may have been caused by leakage of saltwater from lower aquifers. Schmerge (2001) reports that a nearby well (L-2310) open to the Upper Floridan aquifer had a poorly sealed annular space, which may have allowed saltwater intrude into the lower Tamiami aquifer in this area. Well L-2310 was plugged in 1999.
The water-level and chloride concentration correlation analysis indicated that only three wells (C-528, C-975, and L-5747) in the lower Tamiami aquifer showed a significant correlation between chloride concentration and water level. Spearman's r for wells C-528, C-975, and L-5747 was -0.32, -0.80, and 0.32, respectively, whereas Pearson's r for these wells was -0.29, -0.64, and 0.24, respectively (table 4).
Regression analysis alone of water-level data indicated that three monitoring wells (C-391, C-492, and L-2194) in the lower Tamiami aquifer could be used to estimate water level at the remaining 20 monitoring wells in this aquifer, with an average R2 value of 0.78. In addition to the coverage that these three wells provide, the value of these wells is supported by the correspondence of minimum water levels at well L-2194 to droughts and the proximity of well C-391 to wells C-489, C-525, and C-528 where chloride concentrations have been increasing. The fa determined, using the 22 years of water-level data from well L-2194, is 0.54. The only wells in this aquifer where better agreements were indicated had less than 13 years of data available for examination.
Furthermore wells C-391 and L-2194 are nested with wells C-392 and L-2195, respectively (fig. 4 (96K)). These are two of the wells that have been selected to monitor the water-table aquifer (west coast) (table 5). The cost of installation of satellite telemetry can be reduced when monitoring nested wells because only one satellite transmitter is required.
Owing to problems with well C-492, another well (C-462) is the proposed substitute index well in a proposed network that also includes index wells C-391 and L-2194 (table 5). Well C-492 is already equipped with satellite telemetry, and construction records indicate that the well is 64 ft deep and cased to a depth of 60 ft below land surface (app. I and table 7). However, a total depth measurement using a borehole camera indicated that the well was only 21 ft deep. Attempts were made to clear this well. Borehole videos showed that the well is open to the water-table aquifer (west coast) 19 ft below land surface and has apparently collapsed to a depth of 21 ft. The casing may have separated and collapsed, or the construction records may have been incorrect. As a result, this well can no longer be considered to monitor only the lower Tamiami aquifer. Because of this, it is necessary to substitute monitoring well C-462 for well C-492 (table 7). The average R2 value for the network actually improves slightly to 0.83 because well C-492 is no longer considered (tables 5 and 6). Figure 16 (85K) shows (spatially) the parts of the network that could be estimated by using index wells C-391, C-462, and L-2194.
Although, regression analysis has shown that water levels at well L-2194 are most representative of the majority of continuous monitoring wells in the lower Tamiami aquifer (table 5), well L-738 has already been equipped with satellite telemetry. This well also is monitored for chloride concentration, and there has been a statistically significant (+6.7 mg/L per year) upward trend in chloride concentration over the last 25 years (fig. 15B (71K) and table 3). Geographically, well L-738 is one of the closest wells to well L-2194. Regression analysis indicated that water-level data from well L-738 could be used to estimate water levels at 14 other wells with an R2 value of 0.64 or greater. This included all of but one of the wells where water levels could be estimated using data from well L-2194. Well L-738 has daily water-level record for only about 8 years (app. I), and as such, was not evaluated for water-level trends. Despite this lack of long-term water-level data, the increasing chloride concentrations in this well provide a good reason for considering it to be a key indicator site. However, because this trend was likely caused by leakage in a well that has now been plugged (Schmerge, 2001), chloride concentrations in this area may decrease. Well L-2194 is, therefore, still considered to be the best of the two possible index wells. table 7 provides information concerning well construction and analysis results for the three proposed index wells and wells C-492 and L-738.
During the period of record, monthly mean water levels in the sandstone aquifer averaged 14 ft above sea level (app. III), but there is a broad area in north-central Collier County and southeastern Lee County where monthly mean water levels have reached as low as 0.26 to 31.81 ft below sea level (app. III; wells C-989, C-1079, L-731, L-1998, and L-2215). The lowest monthly mean water level (-31.81) was determined at well L-1998.
Water levels at wells L-731 and C-1079 increased during 1986-99, but decreased over time (during this same period) in a large part of the sandstone aquifer at wells L-729, L-1994, and L-1998 near Lehigh Acres (fig. 17 (98K)). Well L-1998 is close to the center of a major cone of depression in the sandstone aquifer. Effects of municipal water withdrawals are pronounced in monitoring well L-1998, which probably explains why water-level data from other monitoring wells cannot be used to estimate water levels in this well with any degree of certainty. Well L-1998 has the largest statistically significant water-level trend observed in this aquifer. The Seasonal Kendall Slope Estimator (SKSE) for well L-1998 was -1.02 ft/yr from 1974 to 1999 (app. II). From 1986 to 1999, the estimate was -1.50 ft/yr. The top of the sandstone aquifer near well L-1998 (app. III) is about 40 ft below sea level (estimated from a structural contour map by Wedderburn and others, 1982). Because monthly mean water levels of -31.81 ft have already been observed in well L-1998 (app. III), and if the trend of -1.5 ft/yr were to continue, water levels would drop below the top of the aquifer during droughts beginning in 2005.
Long-term analysis of chloride concentration trends at three wells in the sandstone aquifer indicated one downward trend (table 3; -6.67 mg/L per year at well C-303) and one upward trend (table 3; +0.56 mg/L per year in well C-688). No correlation between water levels and chloride concentrations was found at the wells analyzed in the sandstone aquifer.
In the sandstone aquifer, wells HE-556, L-727, L-731, L-1998, L-2186 and L-2215 are currently equipped with satellite telemetry. The network R2 value for the existing Data Collection Platforms (DCP) network is 0.83, and it would cover 85 percent of the wells in the continuous monitoring network.
Regression analysis, consideration of existing satellite telemetry, and analysis of water- level trends and minimums indicated that water-level data from 6 monitoring wells in the sandstone aquifer could be used to estimate water levels or provide direct coverage for 95 percent of the 20 continuous monitoring wells with an R2 value of 0.85 on average. This proposed network includes index wells C-1079, HE-556, L-729, L-731, L-1998, and L-2186 (tables 5 and 6). Implementing the proposed network by transferring the satellite telemetry from wells L-727 and L-2215 to wells L-729 and C-1079 improves overall network coverage by 10 percent, improves average period of record per monitoring index well from 20 years on average to 21 years on average, and improves the average network R2 value from 0.83 to 0.85 (tables 5 and 6).
Water-level data from well L-729 can be used to estimate water levels at well L-2215 with an R2 value of 0.95 (table 5). This relation is shown in figure 18 (32K). The advantage of using well L-729 in lieu of well L-2215 is that well L-729 has a much longer period of record (app. I) and indicates a significant downward trend in water levels during 1986-99 (app. II).
Water-level data from either well C-1072 or well HE-556 can be used to estimate water levels for the same group of wells. The average R2 value for these comparisons is much better for well C-1072 (0.83) than for well HE-556 (0.76); however, well HE-556 has 23 years of data, compared to the 13 years of data available for well C-1072 (app. I). Therefore, well HE-556 would be a better index well based on the longer period of record.
The satellite telemetry for well L-727 is redundant because the water-level data from well L-2186 can be used to estimate water levels at well L-727 with an R2 value for the regression of 0.89 (table 5). This relation is shown in figure 19 (68K).
Water levels at wells HE-517 and L-1998 could not be fit with an R2 value of 0.64 or greater by any of the other monitoring wells that had at least 10 or more years of data on which to base a comparison. Of all the wells in this aquifer that had substantial record, water levels at well HE-517 showed the clearest relation with extreme minimums in rainfall (table 2). The best fa (0.60) was obtained for the comparison of the lowest 20 percent of smoothed average rainfall deviations and trend adjusted water-level deviations. For this comparison, water-level data were trend adjusted using a linear regression and were not lagged relative to smoothed rainfall deviations (f-value = 1/26) (fig. 20 (59K)). Because of this relation and because water-level data from other wells in the aquifer could not be used to create a satisfactory estimate of water levels in well HE-517, it would be useful to install satellite telemetry for this well. However, regression analysis indicated that water-level data from well HE-517 was not representative of water levels in other wells. As such, well HE-517 was not proposed as an index well.
Monitoring well L-1996 was considered as a possible index well. Preliminary regression analysis indicated that this well would have been a better index well than well C-1079. However, investigation of the construction of well L-1996 using a borehole camera and examination of well construction records indicated that this well was probably open to multiple aquifers, and therefore, would not be a good candidate for an index well.
Well construction logs indicate that wells L-729 and L-1998 were constructed to have open-hole intervals 22 and 60 ft in length, respectively (app. I). However, the borehole camera logs revealed that the open interval for well L-1998 is actually 26 ft and that about 26 ft collapsed. Both wells are open to the aquifer and monitor the water-level changes that occur. Because of this, these wells are still considered to be valuable index wells.
Wells C-1079, HE-556, L-731, and L-1996 are all partially obstructed by floats, float tapes, or sampling equipment (table 7 and app. I). Borehole camera examination clearly shows that the majority of the screened interval of well HE-556 is still free of obstruction or defect, and the well is 20 ft deeper than indicated in construction notes. A sampling hose in well L-731 obstructs the well at the depth that the open-hole segment should occur. It is likely that the open- hole segment of this well has collapsed around the sampling hose, which has caused it to become lodged in the well. The borehole camera could not maneuver around the obstruction in well C-1079. If additional examination were to indicate that well C-1079 is not functional, monitoring well C-989 could be used as a replacement without impairing real-time network coverage. The open interval of well C-989 has almost completely collapsed, but the well is still open to the sandstone aquifer.
Figure 21 (120K) shows the parts of the network that could be covered by index wells C-1079, HE-556, L-729, L-731, L-1998 and L-2186. table 7 summarizes basic well construction and analysis results for the proposed index wells and wells C-989 and L-1996.
For the period examined (1974-99), the lowest monthly mean water level recorded was 76.02 below sea level at well L-742 (app. III). Monthly mean water levels averaged about 15 ft below sea level. The interquartile range of monthly mean water levels at monitoring wells in the mid-Hawthorn aquifer averaged about 11 ft and ranged from 5.00 ft in well L-2193 to 23.18 ft at well L-742 (app. III). Because the period of record for each well was highly variable, these statistics are probably not as representative of aquifer conditions as would be provided by longer records.
There are strong downward trends in water levels at wells L-581, L-742, L-1993 and L-2644 for the 1984-99 period (fig. 22 (146K)). During this period, water-level decreases at these wells averaged close to 1 ft/yr (app. II). The rate of decline was lowest at well L-1993 (-0.39 ft/yr) and highest at well L-742 (-1.34 ft/yr). The water-level recorder at monitoring well L-2703 was discontinued in 1996, and therefore, the well could not be analyzed for the full period (1984-99). However, from 1984 to 1995, this well also indicated a decline in water levels of 1.03 ft/yr (app. II).
Chloride concentrations at the majority of locations sampled in the mid-Hawthorn aquifer have declined or do not show any significant trend. The largest declines in chloride concentration were found at monitoring wells L-735 (-35.0 mg/L per year over 19 years), L-2702 (-3.92 per year over 20 years), and L-2820 (-15.6 mg/L per year over 21 years) (fig. 23A-B (91K) and table 3). Chloride concentrations in water at well L-735 were initially near 900 mg/L and have declined to about 400 mg/L. Changes in chloride concentrations in well L-2820 have been more variable. Changes in chloride concentration between water samples of about 200 mg/L are common. However, between 1993 and 1999, chloride concentrations in water at well L-2820 increased from about 700 to 900 mg/L.
Chloride concentrations have increased at two wells; the largest increases were 10 mg/L per year at well L-1109 (with 23 years of data) and 2.5 mg/L per year at well L-2640 with 18 years of data (fig. 23C (91K) and table 3). Unfortunately part of the open interval of well L-1109 collapsed (apparently in the interval from 230 to 80 ft in 1996), which prevented additional chloride water samples from being collected. In both cases, chloride concentrations have been near or greater than the 250-mg/L limit for drinking water (Florida Department of Environmental Protection, 1993), so increases in chloride concentration in these areas is a concern. A minimal increase in chloride concentration (+1.0 mg/L per year) occurred at well C-1080, but this well only has 13 years of data.
Fitzpatrick (1986) indicated that a major cause of saltwater contamination in the mid-Hawthorn aquifer was leakage through nearly 8,000 (2-in. diameter) steel-cased wells that were drilled into the aquifer. As pumpage in the mid-Hawthorn aquifer increased and head in the aquifer fell below that of overlying aquifers, leakage in many of these wells permitted downward movement of saline water from overlying aquifers. Other wells were open to both the lower Hawthorn aquifer and deeper saline aquifers. Saltwater was allowed to flow freely between these aquifers. A pilot well plugging program was initiated by the SFWMD in 1979 and established criteria to plug all flowing wells by 1992 (Burns, 1983). The subsequent declines in chloride concentration observed may be the result of this well plugging program. La Rose (1990) documents a slight decrease in chloride concentrations in part of the mid-Hawthorn aquifer where well plugging had been completed. The increases found at wells L-1109 and L-2640 could be the result of movement of the existing saltwater contamination within the aquifer.
Water levels and chloride concentrations were positively correlated at wells L-735 and L-2244 (table 4). Spearman's r coefficients were 0.74 and 0.52 respectively. Pearson's r values for these wells were 0.74 and 0.53, respectively. At well L-735, this relation may be caused by corresponding long-term declines in both chloride concentration (table 3) and water level (Prinos and Overton, 2000). As previously discussed, the decline in chloride concentration may be related to the well-plugging program.
Regression analysis alone indicated that water levels for 78 percent of the mid-Hawthorn continuous ground-water level monitoring network (nine wells) could be approximated by two monitoring wells (L-581 and L-2644), with an average R2 value of 0.82. Data from well L-581 alone can approximate water levels or provide direct water-level measurements for 67 percent of the wells in the aquifer (table 5). In the mid-Hawthorn aquifer, wells L-581 and L-2644 are already equipped with satellite telemetry. Well L-4820 is equipped with a conductivity probe so that changes in chloride concentration in this area of the aquifer can be monitored in real time. Well L-4820 is about 1 mi east of well L-1109 and about 4 mi northwest of L-2640 (fig. 10 (K)) where increases in chloride concentration have been observed (table 3).
When minimum water levels in relation to the mid-Hawthorn and overall coverage are considered, well L-742 (instead of well L-2644) and well L-581 are the preferred choices for index wells in the proposed network. Well L-742 is the lowest point of the cone of depression in the mid-Hawthorn aquifer defined by the USGS water-level monitoring network. The lowest maximum daily water level recorded for this well was 78.61 ft below sea level on May 10, 1974, and water levels of about 70 ft below sea level were reached in the 1999 water year (Prinos and Overton, 2000). Estimating from the structural contour map of Wedderburn and others (1982), the top of the mid-Hawthorn aquifer is about 125 ft below sea level at well L-742. The top of the aquifer varies in depth from 125 to 175 ft below sea level at wells L-581, L-2644, and L-2703. Because minimum water levels at well L-742 are closer to the top of the aquifer, this is considered to be a more valuable index well than well L-2644. The percentage of network coverage is the same (78 percent), but the mean R2 value for the proposed network increases to 0.86 (tables 5 and 6). This is because the R2 value from the regression of data from well L-2644 against that of well L-742 is only 0.65, whereas data from well L-581 provided an excellent fit when regressed against the data of well L-2644 (R2 value = 0.93; table 5). Thus, overall coverage would be improved by the addition of well L-742, which cannot be estimated as accurately as well L-2644 using data from well L-581.
When well L-742 was constructed, it was cased to a depth of 136 ft below land surface and open to the mid-Hawthorn aquifer from 136 to 225 ft (app. 1); however, 25 ft of the open-hole section of this well has collapsed. Well L-581 is cased to a depth of 107 ft below land surface and was originally open to the aquifer from 107 to 177 ft. The bottom 7 ft of this open-hole interval has collapsed. In both of these wells, most of the open interval is still open and water levels still represent changes occurring in the aquifer; these problems should not adversely affect water-level monitoring at these wells. Figure 24 (81K)shows parts of the network that can be estimated using wells L-581 and L-742. table 7 includes basic well construction and analysis results for the wells discussed.
As indicated in the discussion of the hydrogeologic setting, the surficial aquifer system in this part of the study area is not differentiated into locally named aquifers. For the period of record examined (1974-99), the lowest monthly mean water level recorded in this aquifer system was 2.50 ft below sea level at well PB-1491 (app. III). In comparison to the deeper aquifers of southwestern Florida, the surficial aquifer system in Martin, Palm Beach, and St. Lucie Counties does not show much variation in the monthly means of daily maximum water levels. Interquartile ranges for these wells vary from 3.23 ft at well PB-1491 to only 0.33 ft at well PB-900 (app. III). Average monthly mean water level for the aquifer was 14.48 ft above sea level; however, this average includes data from many wells that do not have data for the full period examined.
Of the nine wells that had enough record for examination of water-level trends during 1974-99, four wells (44 percent) showed upward trends. Trends ranged from +0.05 to +0.13 ft/yr at wells PB-683, PB-809, PB-831, and PB-99 (app. II). None of the wells examined for this full period indicated any downward trends in water levels. Downward trends, however, were found at several wells for shorter periods. Water levels at wells PB-445 and PB-732 decreased by 0.02 and 0.28 ft/yr, respectively, during 1974-81 (app. II). Wells PB-565 and PB-900 showed water-level declines of 0.23 and 0.03 ft/yr, respectively, during 1981-91 (app. II). Water levels at well PB-445 declined by 0.07 ft/yr during 1991-99.
Fourteen wells located in the surficial aquifer system in Martin and Palm Beach Counties were tested for chloride concentration trends, with the results indicating that three (21 percent) of the wells showed downward trends (table 3). Trends of -3, -300, and -2 mg/L per year were determined at wells M-1052, PB-595, and PB-1669, respectively. There was not enough chloride concentration data from St. Lucie County to perform a trend analysis. No statistically significant correlation between water levels and chloride concentrations was found for the wells examined in the surficial aquifer system.
In the surficial aquifer system in the northeastern part of the study area, water levels at many of the wells could not be estimated as accurately as in other aquifers. More index wells were needed in this area than in other aquifers, and even with these additional index wells, the average R2 value for the resulting network was lower, and overall coverage was less. Considering the regression analysis alone, six monitoring wells (M-1004, PB-99, PB-689, PB-732, STL-41, and STL-125) could be used to provide or estimate water levels with an average R2 of 0.82 at 26 of the 36 network wells examined. This corresponds to about 72 percent of the wells examined.
The proposed real-time network also includes wells PB-565 and PB-1491 as index wells, which improves coverage by providing better estimates of water levels. With these two wells added, the average R2 value for the network is 0.85. The average period of record for these eight index wells is 23 years. The first four wells in this list actually provide the majority of the network coverage (table 5; 61 percent). The remaining four wells only add the ability to estimate water levels at a few additional wells (fig. 25 (169K); tables 5 and 6).
A -3 mg/L chloride concentration trend was indicated at well M-1052 (table 3). However, considering the normal variation in chloride concentration data from this well, the trend is minimal and does not require the addition of real-time monitoring.
This proposed real-time network includes wells to aid in the assessment of minimum water levels, agreement with decreased rainfall, and water-level and chloride concentration trends. This network also would involve considerably less effort to initiate than other possible networks because it makes use of the existing satellite telemetry. For example, well PB-565 is less than 0.5 mi from well PB-595 (figs. 5 (146K) and 10 (98K)) where chloride concentrations have substantially decreased (table 3).
During a drought period on April 14, 1989, maximum daily water levels at well PB-1491 reached an elevation of 3.04 ft below sea level (Prinos and Overton, 2000). As previously mentioned, during this same drought, monthly mean water levels declined to 2.50 ft below sea level (app. III). Well PB-1491 is near the coast adjacent to a supply well in the Boca Raton Well Field where sustained water levels below sea level can lead to saltwater intrusion. Of the wells that had more than 10 years of record, well PB-1491 showed the greatest agreement with reductions in rainfall. The fa determined for this relation was 0.63 (table 2). Even though there are a number of continuous monitoring wells near this well, none of the other wells in the surficial aquifer system with more than 10 years of record could be used to estimate water levels at well PB-1491 with an R2 value greater than 0.64 (table 5). As a result, well PB-1491 would be a useful index well.
Of the potential index wells in this area, four already have satellite telemetry installed. These are wells M-1004, PB-565, STL-125, and STL-175. All but one of these, well STL-175, is included in the alternate network. Water levels at well STL-175 can be estimated using data from well M-1004 with an R2 value for the regression of 0.70 (fig. 26 (30K) and table 5). Well M-1004 has a longer period of record than well STL-175 (app. I), and would, therefore, be the more useful of the two.
Water levels at well PB-565 could be estimated using data from well PB-689, but well PB-565 has a much more extensive period of record (table 5 and app. I). Data from well PB-565 can also be used to estimate water levels at some, but not all, of the wells estimated using data from well PB-99. Because of this, the satellite telemetry for well PB-565 is still beneficial.
Water levels at well STL-125 could not be estimated using data from any other well that had 10 years or more of data. Unfortunately, many wells in this area do not have much data available for analysis. Well STL-125 has about 9 years of data available (table 7 and app. I). Based on these limited data, water levels at several other wells may be estimated using data from well STL-125. Analysis showed relatively good agreement with wells M-1261, PB-683, PB-689, STL-185, and STL-213. Well STL-125 could potentially be used to increase network coverage to 72 percent (table 5). Although this well has only about 9 years of data, it already has satellite telemetry, and the water-level data are not redundant. Therefore, well STL-125 has been retained as a real-time index well in the proposed network. Basic well construction and statistical information for the proposed index wells is listed in table 7.
In 2001, index well STL-41 became plugged by clay from the formation (table 7 and app. I). Prior to this problem, the well was responding to changes in water levels in the surficial aquifer system and was representative of changes occurring in several wells. A replacement well is needed. A float and float tape were found to be obstructing well PB-1491 at a depth of 80 ft below land surface. These obstructions were removed, but the casing has filled with sand from the formation up to a depth of 80 ft (table 7 and app. I). This sand does not impede the changes in water levels at this well, but may reflect in a lower response to water-level changes. Therefore, the well is still functional.
The lowest monthly mean of daily maximum water levels in the Biscayne aquifer was 13.86 ft below sea level and was recorded at well G-2395 near the City of Fort Lauderdale Prospect Well Field (fig. 27 (157K)). Water levels at this well have dropped 14 ft below sea level several times since 1992 (Prinos and Overton, 2000). The average interquartile range of the monitoring wells examined in the Biscayne aquifer was 1.15 ft and ranged from 11.36 to 0.27 ft (app. III). Mean water levels tend to be highest in the north and in the water-conservation areas. They are lowest near the coast and in the southern part of the aquifer. Near most of the major well fields in the Biscayne aquifer, water-level minimums below sea level have been recorded (fig. 27(157K)). Of the 133 monitoring wells examined, 28 wells had recorded monthly means of daily maximum water levels that were below sea level (app. III). Ten of these wells are located in a broad area around Homestead (fig. 27 (157K))
Of the 60 wells that have water-level data for the1974-99 period, 67 percent showed a significant upward trend in water level (app. II and fig. 27(157K)). Only one of the wells tested for this period, G-1213, indicated a downward trend (app. II; -0.03 ft/yr). Generally upward trends of 0.01 to 0.02 ft/yr occur near the coast. In and around Everglades National Park, water levels have increased by 0.01 to 0.06 ft/yr. The largest increases in water levels occurred in northern Broward County (0.25 ft/yr in well G-853) and near the Hialeah-Miami Springs Well Field in Miami-Dade County (fig. 28A (87K)) where water levels increased by as much as 0.40 ft/yr. The actual trends near the Miami Springs-Hialeah Well Field were not monotonic, instead water levels increased sharply in late 1983.
Nine of the wells analyzed show strong influence from historical changes in pumpage at the Hialeah-Miami Springs Well Field. Between 1984 and 1992, pumpage in this well field was reduced because of industrial contamination in the supply wells (Sonenshein and Koszalka, 1996). As a result of this change in pumpage, water levels in surrounding monitoring wells follow a pattern that is unique to this area (fig. 28A (87K)).
Because many of the wells in the Biscayne aquifer had only partial record during 1974-99, more wells could be analyzed when trends were assessed during shorter periods. During 1974-83, it was possible to examine water-level data from 65 wells for trends (app. II). For this shorter period, fewer wells (28) indicated statistically significant trends, but the trends observed during this period were generally much larger than determined during 1974-99 (app. II). Twenty of the wells, which indicated statistically significant upward trends during 1974-99, generally also indicated a much larger upward trend during 1974-83. These wells increased by 0.21 ft, on average, more in the 10 years from 1974 to 1983 than during the 26-year period from 1974-99. At least some of these increases could probably be attributed to the implementation (1976-84) of an improved conveyance system that was designed to increase the quantities of water from Water Conservation Areas 3A and 3B into southern Miami-Dade County (Klein and Waller, 1985).
Thirty-four (68 percent) of the 50 wells in the Biscayne aquifer tested for chloride concentration trends showed significant trends, with 20 wells (40 percent) having upward trends and 14 (28 percent) having downward trends (table 3). Four of the wells (fig. 10 (98K); G-432, G-896, G-901, and G-1180) for which statistically significant increases in chloride concentration were recorded are located near the approximate location of the saltwater interface as determined by Sonenshein (1997).
Chloride concentrations in the Hialeah-Miami Springs Well Field area (fig. 27 (157K); wells G-548, G-571, G-1351, and G-1354) increased up until about 1976 because of contamination from the Miami-Tamiami Canal basin. In 1976, a salinity control structure was installed in the Tamiami Canal just east of Le Jeune Road. This structure and another structure on the Miami Canal at N.W. 36th Street allow higher heads to be maintained in this area (Klein and Ratzlaff, 1989). As a result, chloride concentrations in these four wells have been declining (fig. 28B (87K)).
The largest upward trends in chloride concentration in the Biscayne aquifer occurred at wells G-432, G-854, G-901, G-1241, and G-1435 (table 3). Wells G-432 and G-901 are located in Miami-Dade County, about 3 mi east of the Alexander Orr Well Field (fig. 27 (157K)). From 1976 to 1995, chloride concentrations in water at these two wells increased from about 30 to 2,200 mg/L (mostly between 1988 and 1995). Beginning in 1995, chloride concentrations in water at both wells began to decrease. By 1999, chloride concentrations in water at wells G-432 and G-901 decreased to 640 and 1,600 mg/L, respectively (fig. 29 (27K)). Although the trend analysis described in this report was only performed on data collected up to 1998, it is useful to note that during 2000 and 2001, chloride concentrations in water at both wells increased once again as drought conditions were experienced. At well G-432, a chloride concentration of 2,350 mg/L was measured in April 2001. This was higher than any previous measurements taken at this well. Chloride concentration in water at well G-901 increased to 2,050 in October 2001. Sonenshein (1997) suggested that the landward movement of the saltwater interface in this area of the Biscayne aquifer could be caused by: (1) decline in water levels at the Alexander Orr Well Field, (2) lowering of water levels in the Coral Gables Canal as a result of reconstruction of the tidal control structure, or (3) combination of both factors.
Wells G-1241 and G-1435 are located in Broward County near Hallandale (fig. 10 (98K)). Chloride concentrations in water at well G-1435 have increased by 266 mg/L per year on average during 1974-98 (table 3). Koszalka (1995) indicated landward movement of the saltwater interface at well G-1435 and indicated that the interface near the Hallandale Well Field (fig. 27 (157K)) was between wells G-1435 and G-1473. The landward movement of this interface has continued, and chloride concentrations in water at well G-1435 have increased from 5,500 to almost 8,000 since 1990. Chloride concentrations in water at well G-1473, however, actually declined slightly during 1980-98 (table 3), so the interface remains between these two wells in this area.
Chloride concentration in water at well G-1241 increased from about 70 mg/L in 1981 to about 4,500 mg/L in 1993. Between 1993 and 1998, chloride concentrations in water at this well decreased to as low as 380 mg/L, but by 1999 had risen again to about 1,700 mg/L. Trend tests conducted on chloride concentration data from wells G-2410 and G-2478 also in the Hallandale area indicate increases of about 3 mg/L per year on average over their respective periods of record (fig. 30 (70K) and table 3). This trend for well G-2410, however, is not linear. Chloride concentrations in water at well G-2410, for example, increased much more during 1996-99 than during the preceding period.
Chloride concentrations in water at well G-854 near Fort Lauderdale have increased from about 1,000 mg/L in 1975 to about 2,400 mg/L in 1999 (fig. 31A (102K)). Other wells a little farther inland of well G-854 and closer to the Fort Lauderdale Well Field have also shown increases over their respective periods of record (fig. 31 (102K)). Chloride concentrations in water at wells G-1343, G-2125, G-2130, and G-2352 have increased by about 1.5, 0.3, 0.4, and 7.0 mg/L per year, respectively (fig. 31B-C (102K)and table 3).
Only 4 (10 percent) of the 39 wells located in the Biscayne aquifer in Broward County, for which correlation analyses were performed, demonstrated a statistically significant correlation between chloride concentrations and water levels (table 4). All showed a negative correlation with Spearman's r ranging from -0.20 to -0.51 and Pearson's r ranging from -0.22 and -0.51. Of 21 wells located in the Biscayne aquifer in Miami-Dade County, 7 wells (33 percent) showed a significant correlation between water levels and chloride concentrations (table 4). At five wells, chloride concentrations and water levels were inversely correlated; at two wells, chloride concentrations and water levels were positively correlated. For the inversely correlated wells, Spearman's r ranged from -0.21 to -0.65 and Pearson's r ranged from -0.15 to -0.55. For the two positively correlated wells, G-571 and G-548, Spearman's r was 0.24 and 0.42, respectively. Pearson's r, however, was negative for well G-571 and positive for well G-548.
Generally, during drought periods, the reduced freshwater head resulting from deficient rainfall and increased municipal pumping increases the potential for further inland encroachment of the saltwater/freshwater interface. This would logically be manifested in an inverse relation between chloride concentrations and water levels. As previously mentioned, however, not all of the correlations were negative. The two wells (G-571 and G-548), which exhibited positive correlations between chloride concentrations and water levels, are located in north-central Miami-Dade County (fig. 10 (98K))−an area that is highly influenced by municipal pumping at the Hialeah-Miami Springs and Northwest Well Fields (Sonenshein and Koszalka, 1996). Both wells are also in relatively close proximity to the coastal control structures at the Miami (S-26) and Tamiami (S-25B) Canals. The combined effects of municipal pumping and operation of the coastal control structures may account for the positive correlation existing between chloride concentrations and water level in these two wells.
Of the 11 wells in the Biscayne aquifer that showed correlation between water levels and chloride concentrations, 4 wells had median chloride concentrations under 100 mg/L, 5 wells had median concentrations from 110 to 210 mg/L, 1 well had a median concentration of 708 mg/L, and 1 well had a median concentration of 2,500 mg/L. Chloride concentrations greater than 100 mg/L are generally indicative of saltwater contamination (Sonenshein, 1997). Many of the wells that showed chloride/water-level correlations in Miami-Dade County were located near the approximate location of the saltwater interface as determined in 1995.
There are 92 active continuous ground-water monitoring wells that have 10 or more years of data in the Biscayne aquifer in Miami-Dade and Broward Counties. Because of this unusually extensive coverage, regression analysis was only performed on the wells that had 10 years or more of data. This analysis alone indicated that water-level data from 12 wells can be used to provide and estimate water levels for 88 percent of the 92 wells analyzed. The average period of record for each indicator well was 34 years, and the R2 value averaged 0.80.
Seven of the of the 12 index wells are already equipped with satellite telemetry (F-291, G-580A, G-620, G-975, G-1221, G-1260, and S-196A,). These 7 wells and 1 additional well, F-239, are the proposed index wells, which could be used to provide or estimate water levels for the 72 wells in the proposed network. The average R2 value for this network is 0.81, and 78 percent of the 92 long-term Biscayne aquifer continuous monitoring wells would be covered (tables 5 and 6). Of the 12 wells, the remaining 4 wells generally added the ability to estimate water levels at only 1 or 2 other wells in each case. Considering this, additional satellite telemetry would be more beneficial in other aquifers.
In addition to the spatial coverage provided by these eight wells, many can provide additional assessment of aquifer conditions during droughts. Well G-580A provided an fa of 0.55 for the comparison of rainfall and water-level minimums (table 2). This was not as high as the fa value obtained for well G-3264A (0.80). The fa determined for well G-580A, however, was based on 23 years of data, whereas well G-3264A only had 16 years of record available for this analysis (app. I).
Well S-196A provides coverage near Homestead where minimum water levels below sea level have been recorded in numerous wells during the 1970's. Index wells F-291, G-580A, and G-1221 provide coverage near the Hallandale, Alexander Orr, and Fort Lauderdale (Dixie) Well Fields, respectively, where chloride concentrations have increased dramatically during the period examined. Water levels at well F-239 are representative of water levels in the eight other wells affected by changes in pumpage at the Hialeah-Miami Springs Well Field.
Figure 32 (162K) shows parts of the network that are covered by index wells F-239, F-291, G-580A, G-620, G-975, G-1221, G-1260, and S-196A. table 7 provides basic statistical and well construction information for the potential index wells discussed.
Of the proposed index wells in the Biscayne aquifer, three are in fair but acceptable condition. The casing of well F-291 has corroded to the point that small roots are growing into the well from land surface to a depth of 9 ft below land surface. The casing of well G-1221 has filled with sand to within about 5 ft of land surface. About 10 ft of the open-hole interval of well G-580A (originally 18 ft long) has collapsed. In each of these cases, the wells are still responding to changes in water levels in the Biscayne aquifer, and these conditions should not substantially affect the ability of each well to properly record water levels.
Return to Table of Contents ||
Introduction ||
Selected References
Appendix I: Well Construction and Period of Record
Appendix II: Results of Seasonal Kendall Trend Tests of Continuous Water-Level Data
Appendix III: Summary Statistics of Water-Level Data from Candidate Monitoring Wells
Funding for the USGS to design and maintain this site has been provided through a cooperative agreement with the South Florida Water Management District (SFWMD). Water-level conditions are monitored by the USGS with support from Federal, State, and local cooperators.
| Real-time conditions (de-trended data) |
End-of-month conditions (de-trended data) |
FISC-WRS, home page |
| Accessibility FOIA Privacy Policies and Notices | |
Page Last Modified: Wednesday, 18-May-2005 09:18:20 EDT |
  |
