Temporal Stability of Escherichia coli Concentrations in Waters of Two Irrigation Ponds in Maryland

ABSTRACT

Fecal contamination of water sources is an important water quality issue for agricultural irrigation ponds. Escherichia coli concentrations are commonly used to evaluate recreational and irrigation water quality. We hypothesized that there may exist temporally stable spatial patterns of E. coli concentrations across ponds, meaning that some areas mostly have higher and other areas mostly lower than average concentrations of E. coli. To test this hypothesis, we sampled two irrigation ponds in Maryland at nodes of spatial grids biweekly during the summer of 2016. Environmental covariates—temperature, turbidity, conductivity, pH, dissolved oxygen, chlorophyll a, and nutrients—were measured in conjunction with E. coli concentrations. Temporal stability was assessed using mean relative differences between measurements in each location and averaged measurements across ponds. Temporally stable spatial patterns of E. coli concentrations and the majority of environmental covariates were expressed for both ponds. In the pond interior, larger relative mean differences in chlorophyll a corresponded to smaller mean relative differences in E. coli concentrations, with a Spearman's rank correlation coefficient of 0.819. Turbidity and ammonium concentrations were the two other environmental covariates with the largest positive correlations between their location ranks and the E. coli concentration location ranks. Tenfold differences were found between geometric mean E. coli concentrations in locations that were consistently high or consistently low. The existence of temporally stable patterns of E. coli concentrations can affect the results of microbial water quality assessment in ponds and should be accounted for in microbial water quality monitoring design.

IMPORTANCE The microbial quality of water in irrigation water sources must be assessed to prevent the spread of microbes that can cause disease in humans because of produce consumption. The microbial quality of irrigation water is evaluated based on concentrations of Escherichia coli as the indicator organism. Given the high spatial and temporal variability of E. coli concentrations in irrigation water sources, recommendations are needed on where and when samples of water have to be taken for microbial analysis. This work demonstrates the presence of a temporally stable spatial pattern in the distributions of E. coli concentrations across irrigation ponds. The ponds studied had zones where E. coli concentrations were mostly higher than average and zones where the concentrations were mostly lower than average over the entire observation period, covering the season when water was used for irrigation. Accounting for the existence of such zones will improve the design and implementation of microbial water quality monitoring.

INTRODUCTION

The microbial quality of irrigation water has recently attracted substantial attention. Approximately 76 million people in the United States become ill from foodborne diseases annually, and over 40% of these cases are linked to fresh produce (1). Irrigation water can be a significant source of pathogenic microorganisms in produce, and hence, assessing potential contamination from water sources is important for human and animal health (2–7).

Regulations for control of the microbial quality of irrigation water use generic Escherichia coli as the indicator organism of the potential human exposure to pathogens. In the United States, the Food Safety Modernization Act (FSMA) (8) has empowered the U.S. Food and Drug Administration (FDA) to promulgate rules to improve the safety of produce. One of the rules developed by the FDA, the Produce Safety Rule (9), specifies regulations for surface irrigation water based on two metrics: the geometric mean (GM) of E. coli concentrations and the statistical threshold value (STV) of those concentrations. The GM reflects the central tendency of water quality, and its threshold value is 126 CFU E. coli per 100 ml. The STV reflects the variability of the water quality caused by adverse conditions, such as extreme precipitation or high streamflow, and represents the concentration at 90% probability. No more than 10% of samples should exceed the STV threshold, which is 410 CFU E. coli per 100 ml (8, 9).

Along with many natural ponds, there are 9 million artificial ponds in the United States (10), with a large number of them used for irrigation. The microbial quality of water in these ponds, as in other sources of irrigation water, has been mostly unknown (4). The concentrations of E. coli in ponds are spatially and temporally variable. The statistical distributions of those concentrations are often skewed, with low values found more often than large ones and standard deviations exceeding mean values (e.g., see reference 11).

Spatial variability of concentrations creates uncertainty in E. coli monitoring results in freshwater sources. Quilliam et al. (12) demonstrated that microbial water quality on two opposite river banks could suggest very different levels of pollution moving downriver. Jenkins et al. (13) reported that outflow concentrations of fecal indicator bacteria were significantly lower than inflow concentrations in a pond with perennial flow in Georgia, whereas no such difference was found for ponds with ephemeral flow. If some parts of ponds have E. coli concentrations consistently higher than the average E. coli concentration across the pond, and other parts of the pond have concentrations that are consistently lower than the average concentration, then a temporally stable pattern of E. coli concentrations exists.

Several mathematical methods have been proposed to express temporal stability (14). Temporally stable patterns across various spatial extents have been observed for various environmental variables, such as soil water content (15), crop yields (16), soil nutrients (17), etc., but temporal stability in concentrations of E. coli in irrigation ponds has not been studied.

Knowledge of temporally stable patterns appears to be critical for the design and interpretation of environmental monitoring. Microbial water assessment will depend on the choice of sampling locations in the case of well-expressed temporal stability. The objective of this work was to test the hypothesis that the E. coli concentrations in irrigation ponds exhibit temporal stability.

RESULTS

Spatiotemporal variability of E. coli concentrations and environmental covariates.

Statistics of E. coli concentrations and environmental covariates are presented in Tables 1 and 2. No trends of monotonic increase or decrease with time were found for the variables monitored. Temperature, pH, and nitrate concentrations in pond P1 and temperature in pond P2 were found to have low spatial variability, with coefficients of variation (CVs) between 0 and 10%. Nitrate and ammonium concentrations in pond P1 and dissolved oxygen (DO) concentrations, E. coli concentrations, turbidity, and orthophosphate concentrations in pond P2 displayed medium temporal stability characterized by coefficients of variation between 10 and 30%. High levels of variation (CVs of >30%) were found for log E. coli concentrations and ammonium concentrations in pond P1 and for DO concentrations, turbidity, and nitrate and ammonium concentrations in pond P2. The concentrations in pond P1 were generally lower than those in pond P2 (see Fig. S1 and S2 in the supplemental material). Microbial water quality was found to be satisfactory in each location of both ponds, since the geometric mean concentrations (Fig. S2) and the estimated STVs (data not shown) were below the threshold values of 126 CFU (100 ml)⁻¹ and 410 CFU (100 ml)⁻¹, respectively.

TABLE 1.

Spatiotemporal variability of E. coli concentrations and environmental covariates in pond P1

Parameter	Avg value ± SE on sampling datea:
	31 May 2016	13 June 2016	27 June 2016	11 July 2016	25 July 2016
Log(E. coli concn) (CFU 100 ml−1)	0.11 ± 0.38	0.59 ± 0.26	1.36 ± 0.30	1.00 ± 0.30	1.10 ± 0.26
Temp (°C)	26.56 ± 1.00	24.85 ± 0.39	25.94 ± 0.27	27.15 ± 1.21	26.48 ± 0.53
Conductivity (μS cm−1)	154.0 ± 3.73	137.2 ± 0.27	130.0 ± 3.67	136.8 ± 7.68	173.7 ± 1.34
pH	8.56 ± 0.25	8.86 ± 0.21	8.49 ± 0.37	8.24 ± 0.50	9.10 ± 0.57
Dissolved oxygen (ppm)	8.42 ± 0.88	9.17 ± 1.32	9.04 ± 0.83	13.10 ± 1.50	9.94 ± 1.97
Turbidity (NTUb)	NC	5.94 ± 0.70	6.58 ± 1.84	8.32 ± 1.88	NC
Nitrate (ppm)	NC	0.87 ± 0.02	0.64 ± 0.01	0.82 ± 0.01	1.19 ± 0.06
Ammonium (ppm)	NC	0.16 ± 0.15	0.01 ± 0.00	0.00 ± 0.02	0.02 ± 0.01
Orthophosphate (ppm)	NC	0.12 ± 0.03	BDL	BDL	BDL

^aNC, samples not collected; BDL, below detection limit.

^bNTU, nephelometric turbidity unit.

TABLE 2.

Spatiotemporal variability of E. coli concentrations and environmental covariates in pond P2

Parameter	Avg value ± SE on sampling datea:
	8 June 2016	22 June 2016	6 July 2016	20 July 2016	4 August 2016	10 August 2016
Log(E. coli concn) (CFU 100 ml−1)	1.43 ± 0.25	3.07 ± 0.11	1.01 ± 0.50	0.69 ± 0.75	1.13 ± 0.24	0.77 ± 0.67
Temp (°C)	23.79 ± 0.71	24.18 ± 0.82	27.43 ± 0.91	28.94 ± 1.67	28.08 ± 0.43	28.12 ± 1.53
Conductivity (μS cm−1)	158.4 ± 1.64	172.3 ± 2.46	163.8 ± 0.77	151.3 ± 0.84	155.3 ± 2.25	164.0 ± 1.40
pH	7.99 ± 0.47	6.71 ± 0.51	6.54 ± 0.63	8.19 ± 1.20	7.41 ± 1.44	7.36 ± 1.05
Dissolved oxygen (ppm)	9.19 ± 1.83	5.64 ± 2.72	7.04 ± 3.34	10.16 ± 4.08	6.13 ± 1.26	10.54 ± 3.94
Chlorophyll a (μg liter−1)b	NC	137.13 ± 48.19	NC	71.65 ± 7.20	96.20 ± 30.09	320.61 ± 96.15
Turbidity (NTU)c	27.49 ± 18.04	28.68 ± 21.83	14.92 ± 20.24	72.26 ± 138.35	NC	NC
Orthophosphate (ppm)	NC	0.43 ± 0.05	0.37 ± 0.02	0.21 ± 0.05	0.15 ± 0.02	0.15 ± 0.05

^aNC, samples were not collected.

^bMeasured in the interior of the pond.

^cNTU, nephelometric turbidity unit.

Temporal stability of E. coli concentration patterns.

The mean relative differences (MRDs) of the logarithms of E. coli concentrations for both ponds are shown in Fig. 1, with rankings color coded as described in the legend.

FIG 1.

Temporal stability of spatial patterns of E. coli concentrations in the two ponds studied. (A) Pond P1 with sampling locations. Color coding shows ranking of mean relative differences (MRDs) of logarithms of E. coli concentrations as follows: blue, lowest third; yellow, middle third; red, highest third. (B) MRDs of logarithms of E. coli concentrations at sampling locations of pond P1 ordered by rank in ascending order. (C) Pond P2 with sampling locations color coded by MRD rank as described for panel A. (D) MRDs of logarithms of E. coli concentrations at sampling locations of pond P2 ordered by MRD rank. Sampling dates are in Tables 1 and 2. (The images in panels A and C are from Google Maps [imagery ©2017 Google, map data ©2017 Google].).

The temporal stability pattern at pond P1 was well discernible (Fig. 1A and B). Locations that had high MRDs, i.e., consistently higher than the geometric mean concentrations, were in the recreation area (Fig. 1A, locations 7, 8, and 14), the direct runoff entrance area (locations 1 and 2), and the inlet-outlet area (location 6). The latter location had the highest MRDs, with E. coli concentrations on average 30 times higher than the geometric mean. The interior of the pond was represented by locations that had close to zero or negative MRDs. Locations 11 and 13 had the lowest MRDs, corresponding to E. coli concentrations 10 times lower than the geometric mean.

In pond P2, samples close to banks had medium or high MRDs and, therefore, concentrations mostly higher than the median values, whereas concentrations in the samples from the pond interior were mostly lower than the median values (Fig. 1C and D). The highest MRDs were found in samples taken near the banks close to the pond inlet and outlet. The difference between the highest and the lowest MRD for pond P2 was 1.35. At location 12 near the bank at the pond inlet, the E. coli concentrations were on average 5 times higher than the geometric mean concentration. At the same time, the E. coli concentrations were on average 5 times lower than the geometric mean concentration at location 31, which is in the pond interior relatively close to location 12. Locations along the banks further from the inlet and outlet had E. coli concentrations that were on average close to the geometric mean (Fig. 1C and D, color coded in yellow).

Relationships between temporal stabilities of E. coli and environmental covariates.

Temporal stability patterns were found for all environmental covariates (Fig. S3 and S4). The spread of MRD values, shown in Fig. S3 and S4, was much smaller than that for the logarithms of E. coli concentrations, shown in Fig. S1. The smallest differences were found for temperature, with MRD values varying between 0.04 and 0.05 at pond P2 and between −0.02 and 0.15 at pond P1. For other environmental variables, the spreads were also higher in pond P2 than in pond P1. For example, the DO MRD ranged from −0.4 to 0.4 at pond P2 and from −0.2 to 0.15 at pond P1. Turbidity had the highest spreads.

The distribution of sampling locations by the MRD rank groups is shown in Fig. 2. The two last columns for each pond in this figure contain the average ranks of locations close to the banks and in the pond interior. On average, locations close to the banks had slightly higher ranks of log E. coli concentration MRDs than the interior locations in both ponds. The difference between the average bank and interior MRD ranks was much higher for other variables observed. Temperature, pH, and DO near the banks had average ranks that were about 1/3 of the average ranks in the interiors, and the opposite was true for turbidity. This means that temperature, pH, and DO were substantially greater near the banks than in the interiors of the ponds, and the turbidity was substantially higher in the pond interiors.

FIG 2.

Ranking of mean relative differences for observed variables at sampling locations and average ranks near banks (B) and in the interior (I) of ponds in this study. Ranks are color coded as follows: blue, lowest third; yellow, middle third; red, highest third.

Spearman's rank correlation coefficients did not show any strong relationships between the ranks of log E. coli concentrations and the ranks of environmental variables (Table 3). At pond P1, the MRD ranks of pH and DO were significantly correlated. The turbidity MRD ranks were negatively correlated with the MRD ranks of pH and dissolved oxygen. Conductivity was found to have significant positive correlations with temperature, pH, and DO, but only in pond P1. Conductivity showed a significant negative correlation with turbidity in pond P1. At pond P2, significant positive relationships were found between DO concentrations and both temperature and pH levels. Nitrate concentrations showed a significant but moderate correlation with temperature. The values for pH and temperature were found to have a significant positive relationship.

TABLE 3.

Spearman's correlation coefficients for E. coli concentrations and environmental covariates

Variable	Spearman's ρ for indicated covariatea
	Log (E. coli concn)	Temp	pH	DO	Turbidity	Nitrate	Ammonium
Log(E. coli concn)		−0.247	−0.267	−0.246	−0.293	0.015	−0.220
Temp	−0.026		0.765**	0.764**	−0.109	−0.529*	−0.253
pH	−0.211	0.686*		0.919**	−0.138	−0.117	−0.213
DO	0.053	0.528	0.807**		−0.114	−0.028	−0.268
Turbidity	0.472	−0.389	−0.602*	−0.660*		0.096	0.423
Nitrate	−0.256	0.300	0.312	0.196	−0.211		0.297
Ammonium	−0.033	−0.018	−0.221	−0.114	0.011	0.377

^aData in lightface are for pond P1, and data in boldface are for pond P2. **, P < 0.001; *, P <0.01.

Temporal stability of chlorophyll a concentrations.

Chlorophyll a was measured along the transect between locations 25 and 34 in the interior of pond P2. The concentrations ranged from 2.4 to 865.1 μg liter⁻¹ over the observation period and exhibited a temporally stable spatial pattern (Fig. 3a). The chlorophyll a MRD ranks increased along the transect in the outlet-to-inlet direction, which means that the amounts of algal and cyanobacterial biomass tended to be larger as the distance to the inlet decreased. The data in Fig. 3b show a strong negative relationship between the ranks of E. coli MRDs and chlorophyll a MRDs along the transect, with a Spearman's correlation coefficient of 0.819. Higher E. coli concentrations relative to their average across the transect corresponded to smaller chlorophyll a concentrations relative to their average across the transect.

FIG 3.

Chlorophyll a concentrations in the interior transect of pond P2: temporal stability (a) and relationships between the sampling location ranks by logarithms of concentrations of E. coli and chlorophyll a (b).

Concentrations and prevalence of E. coli in P2 pond sediments.

On 10 August 2016 and 14 September 2016, sediments were sampled at pond P2 in the same locations where water samples were collected. Bank sediments mainly had a pale-colored coarse sandy texture, while interior sediments were dark and organic, with a muck-like consistency. On 10 August 2016, the E. coli concentrations were (1.11 ± 0.26) × 10³ (average ± standard error) and (6.71 ± 5.25) × 10¹ CFU 100 g⁻¹ for interior and bank samples, respectively. E. coli was detected in 90% of interior samples and in 21% of bank samples. For the 14 September 2016 sampling, E. coli concentrations were (1.13 ± 0.36) × 10³ and (2.38 ± 0.49) × 10² CFU 100 g⁻¹ for interior and bank samples, respectively. The prevalence of E. coli was 70% for interior samples and 21% for bank samples on the second sampling date. Sediment samples were not collected at pond P1.

DISCUSSION

The E. coli concentrations in both ponds studied had temporally stable spatial patterns reflecting differences between sampling locations. In particular, sampling in the inner parts of the ponds provided pathogen concentrations that were consistently lower than the average. Davis et al. (18) monitored E. coli concentrations in a small (1.2 km²) monomictic reservoir in southeastern California and reported substantially higher concentrations in the shallow eastern area than in the rest of the water body. Jenkins et al. (13) used tracers to estimate the residence time of microorganisms in Bishop Pond, which had perennial flowthrough, and found that ideal complete mixing within Bishop Pond was never obtained. The long residence time meant that fecal bacteria were exposed to solar UV radiation and microbial predation; furthermore, long residence times selected for high algal and cyanobacterial concentrations. At the Bishop Pond outflow location, the concentrations of fecal indicator bacteria were significantly lower than the concentrations at the inflow. The ponds in our work did not have perennial flowthrough. Nevertheless, a concentration gradient along the inlet-outlet transect in the interior of pond P2 (locations 25 to 33) was observed, as evidenced by the sequence of ranks shown in Fig. 1B and in Fig. S1 in the supplemental material. No bacterial concentration gradient was found at pond P1, where the inlet and outlet concentrations were similar (Fig. 1C, location 6).

Recreational activity at the banks could affect the pattern of E. coli distribution in pond P1 near locations 6, 8, and 14, which had the highest-ranked concentrations (Fig. 1C and D). Francy et al. (19) observed that concentrations of E. coli were lower in nearshore samples collected 150 ft from the shoreline than in those collected within a swimming area in Lake Erie. Swimming area bed sediments appeared to be important reservoirs of E. coli in their system. There are indications that indicator microorganisms can move from sediments to the water column in the absence of substantial resuspension in streams (20–22), and the same process might affect concentrations in ponds. Differences in sediment composition in different parts of ponds along the banks also may matter. Sediment composition was shown to influence spatial variation in the abundances of human pathogen indicator bacteria within an estuarine environment (23). In this work, sediments had relatively low levels of E. coli compared to the levels found in other, previously observed freshwater systems (24), with fine sediments hosting elevated E. coli concentrations compared with the concentrations in coarse sediment bank areas.

The highest E. coli concentrations were found in locations near the inlets and outlets of the ponds, i.e., locations 12, 13, and 11 and 23, 24, and 1 at pond P2 and location 6 at pond P1 (Fig. 1). These locations had also high MRD ranks for turbidity. One can hypothesize that the high turbidity in the absence of flow may be caused by very fine particles or the presence of suspended organic flocs; the latter have been shown to improve the survival of E. coli (25). Determining the presence and contribution of such fine-grain, high-surface-area particles and flocs in inflow and outflow zones could be an interesting monitoring component for future work.

The existence of temporal stability of concentrations of E. coli can potentially be caused by differences in survival in different parts of the pond. The trees on the banks provide shade. However, recent results indicate that sunlight is not necessarily the dominant factor in E. coli survival. Indigenous microbiota and habitat influenced Escherichia coli survival more than sunlight in simulated aquatic environments in a study performed by Korajkic et al. (26). Furthermore, wind is known to be a driver of E. coli concentrations at beaches (e.g., see reference 27), providing fine material resuspension and E. coli release (28, 29). Monitoring wind may, therefore, shed additional light on the observed variations of E. coli concentration in ponds. Benjamin et al. (11) determined that wind speed and the distance to rangeland were the only environmental variables that could serve as predictors of microbial water quality in surface freshwater sources used to irrigate leafy greens in California. Dada and Hamilton (30) reported a correlation between wind speed and E. coli concentrations but not with the wind direction at the beaches of a large freshwater lake in New Zealand. These authors suggested that this might be evidence of the lake experiencing wind-driven resuspension of sediments and chronic high turbidity. The role of wind in the formation of spatial patterns of E. coli concentrations in ponds has not been studied, and investigations conducted in regions of the world with wind speeds different from those experienced in our region should provide further evidence of whether or not this factor plays a role in produce contamination (31, 32).

The standard deviations of the logarithms of E. coli concentrations were about 0.3 for pond P1 and from 0.1 to 0.8 for pond P2, which appears to be typical for E. coli variation in water bodies. For example, the standard deviation of the log E. coli concentration was about 0.6 in pond and reservoir water in central California (11). In our study, the spatial variation was smaller than the temporal variation (Tables 1 and 2). The latter provided the major proportion of the total variation of microbial water quality, as found in other systems (e.g., see references 33 and 34). In the work of Amorim et al. (27), spatial variation explained about 25% of the total spatiotemporal variability.

Environmental variables also demonstrated patterns of temporal stability (Fig. S3 and S4). The ranking of sampling locations by these variables did not have significant relationships with the ranking of log E. coli concentrations, as shown by the data in Table 3. The relationships between ranks of locations by the environmental covariates were similar to the previously observed relationships between these covariates. Beutel and Larson (35) observed a weak but significant positive correlation between DO and pH and fecal coliform (FC) removal in biofilters, possibly because of the ability of oxygen and hydroxide to enhance sunlight-driven inactivation of pathogens. Additionally, during the day, algal photosynthetic activity converts dissolved CO₂ into organic matter and oxygen. This is accompanied by HCO₃ dissociation, increasing the pH (36). This occurs to a greater extent when water temperatures are between 20 and 35°C, which appears to be the optimum temperature range for the growth of many cyanobacteria and chlorophytes (37). The mean daily temperatures across sampling locations throughout the experiment ranged from about 24 to 29°C, which may explain the significant positive correlation of temperature with both pH and DO in the present study. The negative correlation between turbidity and dissolved oxygen could be due to the reduced light penetration, which would limit aquatic photosynthesis and reduce oxygen content (38). Positive correlations between DO and pH also explain why turbidity was negatively correlated with pH in our work.

The concentrations of chlorophyll a also demonstrated temporal stability. It is possible that the inverse relationship between the MRD ranks of chlorophyll a and bacteria is due to the effect of sunlight, which facilitated photosynthesis and impeded the survival of bacteria at the sampling depth of this work. However, this does not explain the rank gradient along the interior transect of locations 24 to 33 at pond P2. Davis et al. (18) indicated that some studies have shown a positive correlation between bacteria and chlorophyll a in freshwater systems (39, 40). This occurred, in part, because of the release of dissolved organic carbon and other nutrients back into the water column. Better survival of E. coli in waters enriched with organic matter was noted in the review by Blaustein et al. (41). However, the compounds released are a function of the species of algae and cyanobacteria present, and some may be stimulatory while others could be inhibitory. Reduction of water clarity and effectiveness in inactivating solar radiation was mentioned as another possible reason for the positive effect of algal biomass and chlorophyll a concentration on E. coli survival (42). Possible interactions of algae and cyanobacteria and E. coli, as well as the effect of bacterial attachment to solids on these interactions, present an interesting avenue to explore.

Other methods of temporal stability characterization exist and can be applied to this work and similar endeavors. For example, principal-component analysis can be used to find not a single but multiple spatial patterns if these exist (14, 43). Principal-component analysis is focused on absolute rather than relative differences between local observations and the average across observation points.

The existence of temporally stable spatial patterns creates multiple implications for monitoring and management of the microbial quality of freshwater sources. Sampling water from zones with predominantly elevated or predominantly lower pathogen concentrations may create a distorted microbial water quality assessment. One consequence of the temporal stability in indicator concentrations can be receiving false-negative results with composite samples. Kinzelman et al. (44) note in their analysis of sampling freshwater swimming sites that these false negatives can be caused by dilution effects that would potentially mask an individual high concentration when combined with those with lower levels.

The existence of zones with consistently different concentrations creates an interesting question of the effect of the water intake location on the microbial quality of water delivered to fields. The microbial quality of irrigation water may change over time as water from different parts of the pond is sent to fields. The existence of three-dimensional patterns needs to be researched over the irrigation periods. Further research on the presence of stable spatial concentration patterns as an interannual phenomenon, the seasonality in those patterns, and their response to specific weather and management conditions may eventually lead to mechanistic interpretation and site-specific explanation of the effects of various microbial sources on the microbial quality of irrigation waters. That eventually can make monitoring of the microbial sources an effective complement to the microbiological monitoring of waters themselves.

The existence of temporal stability of an environmental variable typically has been used to select a single sampling location that would represent the sampled area as a whole. Finding the representative location for microbial water quality was set as an objective of some microbial water quality research (45). An additional objective can be testing the representativeness of a single E. coli sample collected from a fixed, routine monitoring station for the overall E. coli levels within an irrigation event. It will be interesting to see whether there is one representative concentration for irrigation water coming from ponds to fields.

Conclusions.

Substantial relative differences in E. coli concentrations were observed between sampling points, and these differences were fairly stable over time. The pond interiors had persistently lower E. coli concentrations than the areas near the bank, even though interior bottom sediments had higher E. coli levels than the coarser-grain areas most typical of nearshore banks. Moreover, areas near the banks had their own stable differences. Furthermore, the limited chlorophyll data indicated potential algal and cyanobacterial controls on pathogen densities, and as phytoplankton patchiness is characteristic of many systems, biological controls on E. coli levels should also be assessed. Without knowing the temporal stability differences, there is a chance that water samples will have persistently lower or persistently higher concentrations than the average levels sampled across a pond. Hence, the relative contributions of water and associated E. coli from bank and interior areas and high- versus low-biomass locations for exported irrigation water should be known.

The implications of the temporal stability of E. coli concentrations for assessment of water's suitability for irrigation have not been substantial in this work, since most of the concentrations and the geometric mean water quality metric were below the FDA-set thresholds. However, these initial data suggest that temporal and spatial stability could govern exceedances above these levels, and hence, assessments should become routine in future use of pond water for irrigation.

The results of this work show that without a rigorous sampling program, the value of irrigation water source monitoring may be jeopardized. A similar conclusion was made previously for recreational waters, and collecting multiple samples was suggested to improve the estimate of true water quality (46). The appropriate monitoring design for irrigation ponds appears to be an important research avenue.

MATERIALS AND METHODS

Pond monitoring. (i) Site description.

Two ponds in Maryland were chosen for the current study. These ponds were selected to test the spatiotemporal stability of the microbial indicator organism Escherichia coli at approximately the same locations within the ponds throughout the summer of 2016.

(ii) Pond P1.

Pond P1, located on a private working farm, is an embankment pond providing irrigation water primarily for the surrounding strawberry fields in the summer (Fig. 1C). The pond is approximately 91 m long and 68 m in width at its widest points. The average depth is 2.7 m. Small shrubs and deciduous trees grow along the west bank, while other banks are grassed. The topography around the field results in the collection of some runoff from the fields during rainfall events. Runoff can enter the pond from the southwest and north sides, whereas the east side is bordered by constructed fill that diverts water down the backslope and away from the pond. Fields are treated with chemical fertilizers throughout the summer but do not receive animal manures. Irrigation water was drawn intermittently from the pond during prolonged periods of high temperatures at the best judgment of the farm operators. Irrigation never occurred on sampling days. Irrigation was normally applied for 2 to 6 h and did not generate runoff to the pond. Water was pumped from another creek-fed pond into pond P1 occasionally throughout the summer when pond levels were visibly low. Both the inflow and the outflow of the pond are at location 12 in Fig. 1C. Pond P1 also served infrequently as a recreational pond, with access on the southwest side.

(iii) Pond P2.

Pond P2 is an excavated pond located on the University of Maryland Eastern Shore's Wye Research Center. Throughout the observation period, irrigation water was drawn from this pond on nine separate dates at a rate of 155 gal min⁻¹ for durations ranging from 1 to 8 h. The irrigation dates were 15 June, 21 June, 21 July, 27 July, 5 August, 8 August, and 10 August and did not coincide with sampling dates except on 10 August, when irrigation water was drawn hours after sampling. The pond is approximately 200 m long and 22 m wide, with an average depth of 2.7 m. The pond is flanked by corn fields on the west side and agricultural supply storage facilities and a parking lot on the east side. The banks of the pond are covered by dense shrubs and grasses with some trees. Pond P2 is at a lower elevation than the surrounding area on the west, north, and east sides but relatively even with the land near the outflow location. The crops around pond P2 receive chemical fertilizers in March, and no animal manures are applied. The water level in the pond is naturally maintained by precipitation, as well as by an ephemeral creek that enters through a culvert at the north end inflow (Fig. 1A, location 12). This creek routes overland flow from the surrounding corn fields to the pond. The water level in pond P2 is restricted by a water level-dependent orifice outflow drain (Fig. 1A, location 24) that flows to a ponded marsh-like area that drains into a small creek. The latter transports water away from the system.

Sample collection, handling, and storage.

Water samples were collected biweekly from May to September 2016 (Tables 1 and 2). Sampling was conducted on a grid (Fig. 1A and C) at both ponds at a depth from 0 to 15 cm between 9 and 11 a.m. All sampling locations were geotagged using a handheld global positioning system (GPS) device (BE-2300; Bad Elf, Tariffville, CT). Orange flags were placed on the pond exteriors to maintain consistency of bank sampling. Bank samples were collected with field-disinfected (70% ethanol) 500-ml-capacity 6-foot grab samplers and then transferred to sterile Nasco Whirl-Pak bags and placed on ice. Interior pond samples were taken from a kayak. Water samples collected for chlorophyll a quantification were kept separately from water samples collected for fecal indicator bacterium enumeration, but both were collected simultaneously with disinfected gear from the same locations and at the same time. The positioning of the interior sampling locations was approximated via reference to bank flags, as well as with the assistance of a land crew. Environmental covariate measurements, including temperature (°C), dissolved oxygen (mg DO liter⁻¹), pH, and conductivity (μS cm⁻¹) were taken in conjunction with water samples using a handheld YSI 556 multiprobe system (MPS; YSI, Inc., Yellow Springs, OH), and turbidity (measured in nephelometric turbidity units [NTU]) was measured in the laboratory (LaMotte Company, Chestertown, MD). Water samples were placed on ice shortly after collection and transported to the laboratory for processing within a couple of hours after collection. Samples remained on ice and in the dark throughout processing. All sampling materials were disinfected with 70% ethanol solution before and after each sampling day.

Laboratory analysis.

Membrane filtration was used to enumerate E. coli. The filtration volumes varied throughout the experiment based on fluctuations of bacterial concentrations within the sampling period. Sample sizes ranging from 30 ml to 200 ml were filtered through 0.45-μm filters (Millipore Corp., Bedford, MA), which were placed onto modified mTEC (membrane thermotolerant E. coli) agar plates (Difco, Sparks, MD). The plates were placed in a 35°C incubator for 2 h and were then transferred to a 44.5°C incubator for 22 to 24 h. After the incubation period, red colonies were counted as E. coli. All counts were reported as CFU per 100 ml. Chlorophyll a was determined according to the Standard Methods for the Examination of Water and Wastewater (47). Nitrate and ammonia concentrations were obtained by flow injection analysis (FIA) on a Lachat QuikChem 8000 series FIA system (Lachat Instruments, Loveland, CO) using Omnion 3.0 software. The QuikChem methodology was modified by using water instead of KCl. Reagents, standards, and manifold settings were prepared according to the QuikChem 12-107-06-2-A and 12-107-04-1-B methods. Orthophosphate concentrations were determined in triplicate according to a modification of the method of Murphy and Riley (48), using a microplate reader and with the addition of internal standards to each sample.

Temporal stability assessment.

The mean relative difference (MRD) (49) is currently the most common method used to characterize temporal stability. The relative difference RD_ij between the x_ij, or observation of variable x at location i at time j, and the <x>_j, or spatial average of x at the same time, is defined as follows:

$${\text{RD}}_{ij}=\frac{x_{ij}-{〈x〉}_j}{{〈x〉}_j}$$ (1)

The MRD for the location i becomes

$${\text{MRD}}_i=\frac{1}{N_t}\underset{j=1}{\overset{j=N_t}{\mathrm{\Sigma }}}{\text{RD}}_{ij}$$ (2)

where N_t is the number of observation times and i = 1,2,…, N_i, where N_i is the total number of locations.

The standard deviation SDRD_i of the set RD_i,1, RD_i,2,…, RD_i,Nt of relative differences at the location i over the observation period is computed along with MRD_i as follows:

$${\text{SDRD}}_i=\sqrt{\frac{1}{N_t-1}{\displaystyle \sum _{j=1}^{N_t}}{({\text{RD}}_{ij}-{\text{MRD}}_i)}^2}$$ (3)

This value serves as a metric of the temporal stability for a specific location. The larger the value for SDRD_i, the less stable is the mean relative difference MRD_i in the location i.

Observation locations can be sorted by their MRD values. After locations are sorted in the ascending order, i.e., from the smallest MRD to the largest, each location receives a rank which is equal to the position of the location in the sorted MRD array. Location ranking can be used to compare patterns for different variables measured in the same locations. Assuming that locations received ranks R_X,i based on MRDs for the measured variable X and ranks R_Y,i based on MRDs for the measured variable Y, one can compute the correlation between these two sets of ranks and obtain the Spearman's correlation coefficient ρ. Values of ρ close to one indicate pattern similarity, whereas values close to −1 indicate the inverse ranking of locations; a large MRD for one of the measured variables corresponds to a small MRD for another variable and vice versa. The probability that the computed value of ρ will be significantly different from zero can be estimated for values of n from about 20 upwards based on the fact that at those n, the variable $\rho \sqrt{(n\hspace{0.17em}-\hspace{0.17em}2)/(1\hspace{0.17em}-\hspace{0.17em}{\rho }^2)}$ has an approximately Student's t distribution with n − 2 degrees of freedom. Microsoft Excel was used in all computations. E. coli concentrations, expressed as CFU (100 ml)⁻¹, were common log transformed for the statistical analyses.