Summary Introduction Methods Geologic Setting Results Rock Analysis Water Chemistry Ground Water Contamination QC/QA Conclusions Future Studies Acknowledgments References Appendices

Blank samples taken included eight equipment blanks and two field blanks. Their field identification includes two numbers separated by a hyphen with the first number giving the order of blanks taken during a sampling round and the second number giving the number of the sampling round (rounds 1 to 4). The equipment blanks tested the 5/8-in-ID plastic tubing, silicone tubing and filter units by taking a sample of DI water using the same procedures as were used for environmental samples (see section on sampling protocol). The field blanks were a test of the DI water after pouring DI water directly from its container in the field into a sample bottle of the same type as used for environmental samples. The two field blanks are blanks 1-3 and 2-4.

All of the nutrient analyses of the blank samples gave values at or below the detection limit. Contamination with total and dissolved organic carbon (TOC) and (DOC) is apparent in some of the blank samples, although blanks 3-3, 1-4, 2-4, and 3-4 have very low values. Blank 2-3 could have been contaminated with alcohol used to sterilize the tubing prior to sampling, as discussed below. Contamination of blank 2-2 with fecal coliform was found during the second sampling round.

A total of 13 duplicate samples was taken with at least three samples per sampling round. All of these samples were taken at groundwater sites with 10 from wells located offshore or in the bay and three from wells on land. Duplicate samples always followed the original sample, and they were taken using the same procedures as used for the original sample after changing the filters in both filter units and starting over by repurging the well and tubing with the electric impeller pump.

Five box plots were made for comparing duplicate samples with their original samples for each parameter measured (Figs. 22-26).

Parameter variable names with a "D" added on the end represent the duplicate samples. The number of analyses for original and duplicate samples is the same for each parameter (for some parameters this is less than 13).

Salinity data are shown in Figure 22, including specific conductance, dissolved solids concentration, and chloride concentration. Data for three different nitrogen parameters are shown in Figure 23, including ammonia (NH₄), dissolved ammonia+organic nitrogen (NH₄+ORG-N, dissolved), and total ammonia+organic nitrogen (NH₄+ORG-N, total). Data for nitrite+nitrate, dissolved (NO₂+NO₃), are shown in Figure 24. Data for the three different phosphorous parameters are shown in Figure 25. They are orthophosphate, dissolved phosphorus and total phosphorus. Dissolved and total organic carbon (DOC and TOC) are shown in Figure 26. Generally, the repeatability of the duplicate sample measurements as shown by these box plots is good. Comparison of the two box plots for dissolved phosphorus indicates the most variation.

Figure 22. Salinity data, specific conductance, dissolved solids and chloride concentration compared with duplicate samples. Duplicate samples have "D" added. [larger image]

Figure 23. Comparison of ammonia (AMM), dissolved ammonia+nitrogen (NH4 = ORG-N total) (AMMO), and total ammonia + organic nitrogen (NH4 + ORG-N total) (AMMOT) with duplicates. Duplicates have "D" added. [larger image]

Figure 24. Comparison of nitrite + nitrate, dissolved (NO2+NO3) in mg/L as N with duplicate. Duplicate has a "D" added. [larger image]

Figure 25. Comparison of data for orthophosphate (PORTHO), dissolved phosphorous (PDIS), and total phosphorous (PTOT) with duplicates. Duplicates have "D" added. [larger image]

Figure 26. Comparison of dissolved and total organic carbon (DOC and TOC) in mg/L as C with duplicates. Duplicate has "D" added. [larger image]

Contamination of the filter apparatus used in the field analysis of fecal coliform occurred during the second sampling round in May 1993. A sample for well SB-1A was the first to be analyzed and fecal coliform in this sample apparently continued to be present in the apparatus during the filtering of the rest of the samples from the Lower Keys (SBB and SB samples) and blank 2-2. However, no fecal coliform colonies were found in the samples from SB-3DUP and SB-3SW. After this problem was encountered, the filter apparatus was thoroughly sterilized between samples.

In an effort to prevent bacterial contamination of sampling equipment between samples, the plastic and silicone tubing were sterilized during the first part of the third sampling round (August 1993). This was done by pumping a 25 to 30% mixture of isopropyl alcohol in DI water through the tubing with the peristalic pump. Flushing of the alcohol from the system was done while purging with at least 15 gallons of sample water prior to taking the next sample. However, inadequate flushing of the alcohol was suspected about half way through the round because very few bacteria were found in the initial samples. The onshore well KLI-2B was resampled on 8/12/93 when alcohol was no longer used, and 22 fecal coliform colonies/100 ml were found as opposed to 1 and none for the samples taken on 8/11/93 when alcohol was used. If a residue of alcohol was still present when taking the samples, the concentrations of DOC and TOC would also have been affected. During this round, sterilization with alcohol was done at all the sites in the Lower Keys area (except at SBB-1, SBB-1SW, and SBB-2), the two KLI-2 wells, the two ORO-1 wells, KL-1, KL-1SW, KL-2, and blank 2-3.

A frequent problem occurred with the organic-carbon measurements. The DOC concentrations were often higher than the TOC concentrations (Fig. 21), which is physically not possible. In order to show this problem, a plot of TOC by POC was made for the groundwater samples (Fig. 27). Negative values for POC occurred in all rounds except the first. This problem may have resulted from laboratory error. However, when encountered in the laboratory, the values were verified by repeat analyses. Another explanation is that the filter unit used for DOC was not adequately cleaned between samples, using the procedure described in the sampling protocol section, and an unseen buildup of organic carbon resulted. Some support for this explanation can be found in the DOC and TOC values for the blank samples.

Figure 27. Plot of total organic carbon (TOC) against particulate organic carbon (POC) calculated by subtracting DOC from TOC. [larger image]

The driving force behind this study were the questions: are nutrients from disposal wells being transported underground and subsequently being released in areas of coral growth and thus stimulating algal growth; and what is/are the pathway(s) and driving force(s) by which flow could be achieved?

When this study was first conceived, it was thought that the relatively impermeable top few inches of the Q3 unconformity could serve as a confining layer as previously discovered at the Dade County Landfill (Shinn and Corcoran, 1988). Our core drilling demonstrated, however, that although the Q3 is present and may be effective beneath and near the Keys, it is not generally present offshore. What this study shows is that the Holocene sediment cover, especially where it is lime mud, creates a very efficient and widespread confining layer. More than half the study area off Key Largo, essentially in all of Hawk Channel, is covered by a blanket of impermeable lime mud. The confinement effect of this lime mud was especially well demonstrated at KL-3, OR-2, and OR-3, where tidal pumping only became evident when the drill bit penetrated the limestone beneath the layer of Holocene lime-mud sediment.

Furthermore, to a large degree, the upper surface of the Pleistocene itself has a confining effect due to development of soils, soilstone crusts and general infilling of pore networks with sediment and precipitated carbonate. The confining effect of these infillings and crusts was demonstrated at SB-1A, SB-1B, KL-1, OR-1A and OR-1B. Tidal pumping was especially strong at those sites (Fig. 4). We suspect that tidal pumping, as indicated by the work in sediments by Simmons (1992), occurs in all the offshore monitoring wells installed thus far. Therefore, tidal pumping is considered to be a potential mechanism for transporting ground waters into the overlying water column.

Figures 28, 29 and 30 show the average levels of NH₄, NO₂+NO₃ and PO₄ in onshore and offshore ground waters. The bar graph in Figure 28 represents the transect of wells in the Lower Keys, and Figures 29 and 30 represent middle and northern Key Largo transects. The Key Largo bar-graph transects are keyed to generalized cross sections showing depth of wells, underlying limestone and the confining Holocene sediment layer. These graphs, especially for Key Largo (KL and OR transects), show increasing NH₄ levels offshore. The highest levels off Key Largo occur in ground waters under the outer reefs farthest from shore (Figs. 29 and 30). The only onshore wells with such high values are MO-171, 173, 175, and the 160-ft-deep (49 m) SCF well (Fig. 18). The MO wells are near two disposal wells. However, there are no disposal wells within miles of the SCF well. Thus, any correlation between NH₄ and disposal wells is inconsistent.

Figure 28. Average concentration of four sampling-round analyses for three major nutrients, shown as bar graphs (with error bars). Graph is arranged with north (onshore) to left and south (offshore) to right. Well names followed by SW are surface-water samples from same location. Ammonia (NH4) in ground water is 4 to 7 times higher than in surface water. [larger image]

Figure 29. Average concentration of four sampling-round analyses for three nutrients, shown as bar graphs with error bars arranged above simplified geologic cross section showing depth of wells. Section extends across Florida reef tract off middle Key Largo. Note increase in ammonia (NH4) offshore. High level onshore is from deep well below Q3 unconformity. High nitrite+nitrate (NO2+NO3) occurs in two shallow onshore wells where water lacks H2S and is not anoxic. Sources of N include septic tanks and fertilizers. High levels of NO2+NO3 do not extend offshore. [larger image]

Figure 30. Average concentration of four sampling-round analyses for three major nutrients, shown as bar graphs with error bars, arranged above a simplified geologic cross section off north Key Largo. Note change in scale from Figure 29. Ammonia level for offshore well OR-5 is actually greater than KL-5 in previous figure. Offshore wells are shallow. Levels of NO2+NO3 and PO4 in shallow ground-water in onshore well are much higher than in onshore wells in previous figure. Well is located near 50 disposal wells but is also situated near a golf course and a tree farm, which are regularly fertilized. [larger image]

Additional evidence against a disposal-well origin of NH₄ is the trend of increasing levels offshore and consistently high NH₄ levels in ground waters at KL-5 and OR-5 farthest from shore. The trend is the opposite of the trend expected if the source were onshore.

A possible, natural, explanation is that reduced tidal pumping offshore would reduce groundwater oxygenation and increase residence time of anoxic water. Increasing residence time of anoxic ground water would lead to increased levels of H₂S. NO₂+NO₃ would convert to NH₄ (Fig. 17); however, the only waters with significant NO₂+NO₃ were from shallow onshore wells.

An alternative explanation is a deeper hydrogeological source for the NH₄. The underlying Floridan aquifer is known to be artesian (Healy, 1962; Stringfield, 1966) and thus has the potential to leak upward into the shallow ground water, providing there are sufficient faults or permeable rock facies to transmit these fluids. Many geologists, summarized in Ball (1992), have speculated that the Florida platform margin is fault controlled. Rock facies along platform margins are usually composed of grainstone and reef deposits. These facies are generally more porous and permeable than platform-interior facies. We have no direct evidence for the existence of faults or permeable facies.

Another mechanism that could bring deep saline waters to the surface is "Kohout" convection (Kohout, 1967). According to the theory, Kohout convection occurs when a geothermal gradient (downward increase in rock and fluid temperature) heats cold saline waters, which flow into the base of the platform from an adjacent basin, causing the waters to become buoyant and rise. Kohout (1967) based this hypothesis on observations of temperatures and salinities in the Floridan aquifer penetrated by oil wells in south Florida and the Florida Keys. A cavernous zone of the Floridan aquifer, called the Boulder Zone, is also used for effluent disposal (Class I wells) throughout south Florida. An oil well drilled on north Key Largo in the 1950s (Coastal Williams) encountered a highly permeable 1,000-ft-thick (305 m) zone between -2,600 and -3,600 ft (793-1,097 m) below sea level, which contains a cavern 50 ft high (15 m; Kohout 1967). This cavernous zone of the Floridan aquifer is also incised by the Straits of Florida east of Key Largo. Thus, ample opportunity exists for cold nutrient-rich waters to enter (or escape) the aquifer beneath Key Largo. However, this explanation for high NH₄ in KL-5 and OR-5 is not considered likely for two reasons: 1) artesian pressure is likely to direct flow toward the platform margin to the southeast and into the Straits of Florida rather than upward into the platform margin, and 2) as discussed previously, analyses of Floridan aquifer waters from 26 wells (some from the Florida Keys) show that NH₄ as mg/L as N levels are considerably lower than levels found in our wells. Admittedly, the Floridan aquifer wells that were sampled may have been in areas with low NH₄.

The explanation we favor is that reduced tidal pumping causes anoxia (buildup of H₂S) denitrification of NO₂+NO₃./and or ammonification of organic N. There may also be a slight, net inflow of water near the platform margin. Net downward flow (inflow) through living porous and permeable Holocene reefs and other biotic communities would carry organic material that could be converted to H₂S by sulfate-reducing bacteria. In an investigation of a mechanism to explain marine cementation, Land et al. (1989) and Mcllough and Land (1992) observed net inflow in core holes drilled near the platform margin on reefs off the north coast of Jamaica. We can only speculate because we have not quantified tidal pumping in our platform-margin reef wells. We are encouraged by measurements conducted in reef-edge, reef-top and backreef sediments using mini-piezometers (Simmons, 1992). These studies demonstrated fluid cycling (both inflow and outflow of ground water) with outflow dominant in some sites and inflow in others. Nutrient levels in seepage water, relative to sea water, were found to be elevated slightly in some and reduced in others (Simmons, 1992). Because the piezometers consisted of perforated pipes driven ~3.3 ft (1 m) into the sediment, they could be deployed only where the sediment layer was thick and soft. Both conditions would tend to reduce water flow and dampen tidal pumping emanating from within the underlying rock. The piezometer measurements of flux, therefore, can only provide an indication of much greater flux in the underlying rock.

Coral reefs are very porous and tend to accumulate directly on rock highs (mud does not accumulate on highs). Thus, groundwater flux is more likely to occur through reefs than through sediment. Simmons (1992) unfortunately did not have the technology to install piezometers in reef rock. Recent work has also shown that Holocene reef accumulation on the seaward side of the platform margin (water depths between 35 and 60 ft (10.7-18.3 m) is very thin, <3.3 ft (1 m), or nonexistent. Reefs and areas of no coral accumulation are therefore considered the most likely areas for groundwater seepage. These are the areas currently undergoing algal infestation. Thus, we believe there is leakage in the platform-margin reef areas but have insufficient evidence to determine if the nutrients present in the ground water are derived from onshore disposal wells and/or septic tanks.

Lapointe et. al. (1990) installed monitoring wells at onshore locations in the Lower Keys and demonstrated a direct relation between septic tanks and elevated nutrients, especially ammonium, in adjacent ground water. Their study provided clear evidence of nutrient-rich ground water leaking into adjacent canal systems, especially at low tide when groundwater levels were higher than seawater levels in the adjacent canals.

Table V lists the wells where contamination by fecal coliform and fecal streptococci bacteria were found. In the Lower Keys transect, SB-1A&B, SB-2 and SB-3 tested positive for fecal bacteria. A high count was detected in SB-1A during the first sample round (200 colonies/100 ml) and a relatively high count (74/100 ml) was detected during the second round. We believe residual alcohol in the tubing prevented detection of bacteria during the third round (see explanation in quality control section). During the fourth round, fecal strep tested positive in this well. We cannot ignore these results. They were determined using a standard, universally accepted test method performed by four different trained technicians. There is the possibility, however, that some form of bacteria that normally lives in anoxic saline ground water can mimic fecal bacteria. The results are considered reliable, however, because they have been supported by an independent study of our Key Largo wells. The source of these bacteria in the Lower Keys offshore transect wells, assuming they are derived from human waste, is unknown because the location is remote from areas of large human populations. Water from the well closest to the disposal wells (SBB-1) at the north end of the transect contained fecal strep (2 colonies) only once during the study (round 4). There is a concentrated human community on the island east of the RV camp, but all the homes there use septic tanks.

Contamination in the SB-1A&B wells is likely to be from septic tanks. Effluent from septic-tank drain fields could be carried downward by hypersaline water through breaches in the Q3 layer. Hypersaline water was always present in these wells. An alternative is the possibility of south and westward flow within the Key Largo Limestone facies, which is constrained by less permeable oolite facies to the north and grainstone/packstone facies to the south. If such flow is taking place, we may be analyzing water that originated from as far away as Marathon, where there is a large community built on Key Largo Limestone.

During our first sampling round, we also found fecal bacteria in SB-2&3. SB-3 is more than 2 nmi from shore (Fig. 5). If the bacteria are not some unknown anoxic non-fecal non-human form indigenous to hypersaline ground water, then their presence suggests a land source and considerable offshore groundwater movement.

At Key Largo, both fecal coliform and streptococci bacteria were consistently present in the shallow ground water at KLI-2B. Fecal streptococci was detected below the Q3 unconformity in the deep well KLI-2A only once during sampling round 2. Presence of fecal bacteria in the shallow ground water was not surprising because the location is within a community using septic tanks and the well is less than 50 ft (15 m) from the drain field servicing the NOAA/NURC facility. It is only 60 ft (18 m) from a canal harboring live-aboard boats. In a recent study by Paul and Rose (in prep., 1994), fluoroscene dye flushed into the septic tank at the NOAA/NURC facility appeared in KLI-2B and in the canal in less than 5 hrs.

On one occasion during sampling round three, three red bacterial colonies of an unknown type were counted in the fecal coliform test at KL-3 and during round four, five fecal streptococci colonies were counted in well KL-4 (Table V). KL-4 is about 3.5 nmi offshore.

In the shallow ground water at ORO-1B, some form of fecal bacteria was present during all four sampling rounds (Table V). Only once was there a single colony of fecal strep below the Q3 unconformity in the deep well at ORO-1A. The 50 nearby disposal wells penetrate only 30 ft (9 m) and their casings are very shallow; thus, contamination was considered less likely in ORO-1A. Likewise, fecal bacteria were at no time detected in the offshore well OR-1A. One colony of fecal strep was encountered in the shallow well OR-1B during the fourth sampling round. Two fecal strep colonies were counted in OR-2 during the last round, and 1 fecal coliform colony was counted in OR-3 during the second round. During the third round, 100-200 red bacterial colonies of an unknown type were counted using the fecal coliform test in water from OR-4 along with 23 clear colonies of an unknown type using the fecal strep test. OR-4 is located approximately 5 nmi offshore. Fecal bacteria were not detected farther offshore, such as in wells OR-5 or KL-5.

Supporting evidence of offshore contamination by fecal bacteria is provided by recent unpublished data prepared by microbiology specialists Paul and Rose (in prep., 1994). Paul and Rose use a sophisticated technique that detects bacterial colonies concentrated from 20 L of water. The standard test we used analyzed only 100 ml. In addition, Paul and Rose tested for and found Clostridium perfringens and viruses specific to coliform bacteria, both additional indicators of sewage contamination. They sampled the middle Key Largo and north Key Largo wells (KL and OR transects) twice (four months between rounds) and detected evidence of contamination in the same wells discussed above.