Marine Pollution Bulletin Vol. 42, No. 11, pp.

1150±1154, 2001
Ó 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S0025-326X(01)00130-8 0025-326X/01 $ - see front matter

E€ect of Sampling Frequency on

Shoreline Microbiology Assessments
MOLLY K. LEECASTER 1, STEPHEN B. WEISBERG *
Southern California Coastal Water, Research Project, 7171 Fenwick Lane, Westminster, CA 92683, USA

More than 80,000 shoreline bacteriological samples are frequency of monitoring (Schi€ et al., 2001a). A sur-
collected annually in southern California to protect vey of beach monitoring throughout the country
beachgoer health, but sampling frequency varies from (USEPA, 2000) found similar kinds of programmatic
daily to monthly among sampling sites. To assess the ef- di€erences.
fectiveness of various sampling frequencies, we used ®ve One of the greatest di€erences among programs is
years of data from 24 Los Angeles area sites that have sampling frequency. Most monitoring is conducted on a
been monitored daily to simulate ®ve alternative sampling weekly to monthly basis, although no studies have been
strategies: ®ve weekdays, ®ve days per week including a conducted to evaluate the e€ectiveness of these di€erent
weekend day, three days per week, weekly, and monthly. temporal strategies. Twenty-four beach sites in Los
For each of these sampling strategies, we included in the Angeles County are monitored daily, providing a un-
simulation the local custom of adaptive sampling, in which ique opportunity to assess the e€ectiveness of less fre-
a site is resampled the following day if bacterial concen- quent sampling strategies. Here we sample from those
trations exceed the State of California's beach water data to assess the extent to which public health is pro-
quality standards. We found that sampling ®ve times per tected using ®ve lesser measurement frequencies pres-
week resulted in observing about 80% of the events in ently being employed at other beaches in southern
which State standards were exceeded. This frequency California.
dropped to 55%, 25%, and 5% for three times per week,
weekly, and monthly sampling, respectively. Adaptive
Methods
sampling was ine€ective because nearly 70% of the water
quality exceedences were single-day events, even at the Five years of data (1 January 1995±31 December
most frequently contaminated sites. This high frequency of 1999) from 24 Los Angeles County beach sites that are
single-day events is of concern because the public is typi- sampled seven days per week were used in the analysis.
cally noti®ed about water quality conditions 24±48 h after Eleven of these sites are located on coastal beaches 50
samples are collected, meaning that most warnings are yards from freshwater outlets that drain urban runo€,
out-of-date when they are issued. Ó 2001 Elsevier Sci- eleven are located on coastal beaches away from urban
ence Ltd. All rights reserved. drains, and two are on a protected beach located
inside of the Los Angeles Harbor breakwater. Total
coliforms were measured using membrane ®ltration
Introduction (APHA, 1995) at each of these sites, while fecal coli-
forms were measured using membrane ®ltration at 20 of
Water quality is measured at beaches throughout the
the sites.
country to protect swimmers from the potential in-
We simulated four sampling frequencies: ®ve days per
¯uence of human activity, such as surface runo€,
week, three days per week, weekly, and monthly. The
wastewater discharge, and industrial input. Although
®ve days per week simulation was conducted using two
there are federal initiatives to standardize these e€orts,
di€erent selectors: (1) all weekdays and (2) Monday±
most of the programs are presently conducted inde- Thursday plus Saturday. The three-days-per-week sim-
pendently by numerous county health departments,
ulation was conducted using Monday, Wednesday, and
resulting in inconsistencies in program implementa-
Friday as sampling days. The once-per-week sample was
tion. A review of southern California beach monitor-
taken on Thursday and the monthly sample was taken
ing programs found di€erences in the number of
on the ®rst day of every month. The sampling days were
indicators measured, measurement methods used, and
chosen to mimic sampling strategies employed in
southern California, although examination of the data
*
revealed little di€erence among the day of the week in
Corresponding author.
1
Present Address: IDAHO National Engineering and Environmen- the frequency of bacterial standard exceedences. In each
tal Labs, P.O. Box 1625 IDAHO Falls, ID 83415, USA. of these simulations, we mimicked the local custom of

1150
Volume 42/Number 11/November 2001

temporal adaptive sampling, in which a site is resampled 45% of the threshold exceedences observed when
the following day if bacterial concentrations exceed the sampling daily. Sampling once per week resulted in
State of California's beach water quality standards. missing approximately 75% of the exceedences. Sam-
Daily sampling continues until the standards are no pling once per month missed approximately 95% of the
longer exceeded. exceedences.
To assess the e€ectiveness of these di€erent temporal These results were relatively consistent across sta-
sampling strategies, we compared the number of ex- tions and indicators (Tables 1 and 2). The number of
ceedences of State of California Health Department actual exceedences at the stations used in our analysis
beach water quality thresholds that were detected with ranged from 5 yr 1 to more than 100 yr 1 . For only
daily sampling (100% detection of threshold excee- two of these sites did weekly sampling identify even
dences) to that detected using the other temporal sam- 40% of the actual State threshold exceedences that
pling schemes. We also compared the number of were identi®ed with daily sampling. Monthly sampling
exceedences that were detected using the adaptive detected less than 10% of the actual exceedences at 23
sampling strategy with that which would be detected by of the 24 sites.
chance alone (i.e., random sampling at the speci®ed
sampling frequency without adaptive sampling). The
State thresholds used to de®ne an exceedence were the
Discussion
following single sample values: (1) total coliform The percent of water quality threshold exceedences
>10,000 organisms per 100 ml, or (2) fecal coliform correctly identi®ed with less than daily sampling fre-
>400 organisms per 100 ml. quencies was larger than would be expected by chance
alone (Tables 1 and 2). This eciency is due to inclusion
of the adaptive strategy for follow-up sampling in the
Results event of an observed exceedence. Even with adaptive
Sampling ®ve times per week resulted in missing sampling, though, lesser sampling frequencies were in-
approximately 20% of the total or fecal coliform ex- e€ective, yielding only an approximately 25% identi®-
ceedences (Tables 1 and 2). The e€ect of sampling one cation rate when sampling weekly.
of the ®ve days on a weekend, as is the practice at some The ine€ectiveness of all sampling frequencies ap-
southern California sites, was negligible. Sampling pears to result because approximately 70% of the water
three times per week resulted in missing approximately quality excursions lasted only a single day, with fewer

TABLE 1
Percent of California total coliform water quality standard exceedances detected at Santa Monica Bay stations (1995±1999) using several sampling
frequencies.

Station name Type of site Number of Percent of water quality standard exceedances detected
exceedances in
®ve years 5/Week 5/Week 3/Week 1/Week 1/Month
(weekend)

Inner Cabrillo Harbor 21 81 76 52 38 0

Inner Cabrillo and 34th St. Harbor 47 83 74 51 30 11
Malibu Creek Outlet 70 83 77 61 19 4
Topanga Point Beach 44 77 80 39 25 2
Pulga Outlet 37 78 84 57 24 3
Santa Monica Canyon Outlet 47 83 79 53 21 0
Santa Monica Pier Outlet 51 80 76 57 24 4
Pico-Kenter Outlet 144 84 79 69 25 6
Ashland Outlet 189 76 80 52 23 5
Windward Outlet 34 79 82 62 21 0
Marina del Rey Beach Beach 88 80 84 60 25 6
Ballona Creek Outlet 99 77 83 60 27 2
Culver Blvd. Outlet 89 82 84 62 26 2
Imperial Hwy. Outlet 77 81 73 55 25 4
Manhattan Beach Beach 11 91 73 55 27 0
Manhattan Beach Pier Beach 4 100 75 75 25 0
Hermos Redondo Pier Beach 34 76 82 47 21 6
Avenue I Outlet 17 71 71 35 24 0
Palos Verdes Estates Beach 2 100 100 50 10 0
Long Point Beach 21 71 62 57 50 0
Abalone Cove Beach 2 50 100 0 50 0
Portuguese Bend Beach 6 83 83 67 3 0
Cabrillo Beach Beach 3 67 67 33 33 0
Overall 1154 80 79 53 26 2
Expected by chance 71 71 43 14 3

1151
 Marine Pollution Bulletin

TABLE 2
Percent of California fecal coliform water quality standard exceedances detected at Santa Monica Bay stations (1995±1999) based on various
sampling frequencies.

Station name Type of site Number of Percent of water quality standard exceedances detected
exceedances
in ®ve years 5/Week 5/Week 3/Week 1/Week 1/Month
(weekend)

Inner Cabrillo Harbor 73 75 78 51 30 1

Inner Cabrilloand 34th St. Harbor 536 82 84 69 27 7
Malibu Creek Outlet 514 84 88 70 34 8
Topanga Point Beach 161 80 80 53 23 7
Pulga Outlet 119 76 87 58 26 3
Santa Monica Canyon Outlet 221 80 83 56 24 2
Santa Monica Pier Outlet 539 85 85 67 26 8
Pico-Kenter Outlet 264 81 87 63 21 4
Ashland Outlet 117 81 79 55 18 6
Windward Outlet 47 70 81 43 13 2
Marina del Rey Beach Beach 278 81 82 59 33 9
Ballona Creek Outlet 127 85 83 65 28 6
Culver Blvd. Outlet 110 78 81 58 24 4
Imperial Hwy. Outlet 56 77 75 43 41 5
Manhattan Beach Beach 14 86 64 50 14 0
Manhattan Beach Pier Beach 13 85 77 46 31 8
Hermosa Beach Pier Beach 16 88 75 63 25 0
Redondo Pier Beach 123 75 76 52 14 3
Avenue I Outlet 25 92 88 56 16 0
Palos Verdes Estates Beach 19 89 79 47 32 5
Overall 3372 82 81 59 26 5
Expected by chance 71 71 43 14 3

than 10% lasting more than three days (Fig. 1). The by daily changes in meteorological patterns. Moreover,
high percentage of single-day events was relatively 13 of the 24 sites were located away from runo€ outlets,
consistent across seasons and indicators. The adaptive and no discernible di€erence in the temporal patterns
sampling strategy is based upon a paradigm of detecting was observed between these sites and those located near
chronic problems, such as sewage spills. Less than 0.1% outlets.
of the water quality excursions observed during the
study period were attributable to sewage spills (Schi€
et al., 2001b).
To assess whether a lesser sampling frequency with an
adaptive component would be e€ective at capturing
chronic events, we repeated our analysis, but rede®ned
an excursion to include only those events that lasted for
at least two days. Thus, if sampling occurred on the
third day of a four-day event, adaptive sampling would
correctly identify 50% of the event. Adaptive sampling
was more e€ective in this reanalysis, with three-times-a-
week sampling correctly identifying 75% of the water
quality exceedence days (Table 3). Adaptive sampling
would be increasingly e€ective if the analysis were lim-
ited to events lasting even longer, though events of more
than a three-day duration were rare during our 5-year
study period.
Numerous explanations might account for the short
duration of the events. Almost half of the sampling sites
were located 50 yards downcoast from urban runo€
outlets, which are known sources of bacterial contami-
nation, and shifts in current direction possibly could
cause runo€ to alternately ¯ow toward and away from
the sampling site on successive sampling days. However,
shifts in current direction are rare and are more medi- Fig. 1 Duration of exceedances at Santa Monica Bay daily monitor-
ated by seasonal shifts in oceanographic conditions than ing stations (1995±1999).

1152
Volume 42/Number 11/November 2001

TABLE 3
Percent of California water quality standard exceedances from Santa Monica Bay daily monitoring stations from 1995 to 1999 based on various
sample frequencies for standard, multi-day and high magnitude triggers for adaptive follow-up sampling.

Five weekday Five samples per Three samples per One sample per One sample per
samples week, one weekend day week week month

California water quality standard exceedance trigger

Fecal coliforms 82 81 59 26 5
Total coliforms 80 79 53 26 2
Mulit-day California water quality standard exceedance trigger
Fecal coliforms 85 91 73 36 5
Total coliforms 87 85 71 36 3
Exceedance of upper limit of laboratory variability trigger
Fecal coliforms 81 82 52 26 4
Total coliforms 78 77 56 19 2

Another possible explanation for the short duration equal argument can be made that issuing incorrect
of these events is that many of the exceedences were warnings 70% of the time will serve to undermine con-
barely above State of California thresholds and could ®dence in the warning system as a whole. A more cogent
have resulted from normal laboratory measurement er- plan might be to limit the warning system to exceedences
ror. To assess this possibility, we repeated our analysis of weekly or monthly average values, unless a known
using thresholds beyond those which could have resulted sewage leak has occurred.
from laboratory measurement variability, as established The longer term strategy for addressing this concern
through repeated laboratory analyses from the same appears to lie in development of more rapid detection
sample (McGee et al., 1999): (1) total coliform >21,875 techniques that would allow swimmers to receive
organisms per 100 ml and (2) fecal coliform 875 or- warnings based upon samples taken the day of their
ganisms per 100 ml. The results were very similar when potential swimming event. Recent advances in molec-
the analysis was repeated using the higher thresholds. ular-based techniques will potentially lead to improved
For total coliforms, the average percent of exceedences speed and sensitivity in testing for microbial patho-
identi®ed with weekly sampling increased from 24% to gens (Tani et al., 1998; Sharma et al., 1999). Fur-
25% using the higher threshold, but decreased for fecal ther research will be necessary to adapt molecular
coliforms from 27% to 22% (Table 3). techniques for use in routine monitoring, but our
A more likely explanation for the short duration of ®ndings suggest that application of these techniques
the water quality exceedences is that the primary sources will result in better health protection for the beach-
of contamination come from urban in¯uences, which going public.
can be episodic in nature. Southern California is the
most urbanized coastal area in the country and also has We wish to thank the City of Los Angeles Environmental Monitoring
a very dry climate. Rainfall, particularly during the Division and the Los Angeles County Sanitation Districts for sharing
their data and Jessica Morton for assisting with the compilation of
summer months, is rare; runo€ typically results from these data. We also wish to thank Al Dufour and Jack Petralia for
human activity, such as washing of city streets, over- their constructive comments on the manuscript.
watering of lawns, etc. Moreover, some of the contam-
ination may result from animal sources, such as birds, American Public Health Association (APHA) (1995) In Standard
whose presence can also be episodic. Methods for the Examination of Water and Wastewater, In eds. A.
Regardless of the source, our ®ndings raise concerns D. Eaton, L. S. Clesceri and A. E. Greenberg, 18th ed. Washington,
DC.
about the e€ectiveness of present public warning sys- McGee, C. D., Leecaster, M. K., Vainik, P. M., Noble, R. T., Walker,
tems. Most beach monitoring in southern California is K. O. and Weisberg, S. B. (1999) Comparison of bacterial indicator
limited to weekly sampling (Schi€ et al., 2001a), which measurements among southern California marine monitoring
programs. In eds. S. B. Weisberg and D. Hallock, Southern
likely results in a high percentage of missed water California Coastal Water Research Project Annual Report 1997±
quality excursions. Moreover, because of laboratory 1998, Westminster, CA, pp. 187±198.
processing time, the public is typically noti®ed about Schi€, K. C., Weisberg, S. B. and Dorsey, J. H. (2001a) Microbiolog-
ical monitoring of marine recreational waters in southern Califor-
water quality conditions 24±48 h after the sample is nia. Environmental Management 27, 149±157.
collected. Thus, 70% of the warnings are out-of-date Schi€, K. C., Morton, J. and Weisberg, S. B. (2001b) Retrospective
when they are issued. This calls into question the de- evaluation of shoreline water quality along Santa Monica Bay
beaches. In eds. S. B. Weisberg, and D. Elmore, Southern California
sirability of issuing warnings on the basis of a daily Coastal Water Research Project 1999±2000 Annual Report, West-
threshold exceedence, unless a clear source for con- minster, CA, pp. 248±252.
tinuing contamination has been identi®ed. While the Sharma, V. K., dean-Nystrom, E. A. and Casey, T. A. (1999) Semi-
automated ¯ourogenic PCR assays: Taq Man for rapid detection of
argument can be made that the public has a right to Escherichia coli 0157:H7 and other Shiga toxigenic E. coli.
know when water quality exceeds a State threshold, an Molecular Cell Probes 13, 291±302.

1153
 Marine Pollution Bulletin

Tani, K., Kurokawa, K. and Nasu, M. (1998) Development of a US Environmental Protection Agency (USEPA) (2000) EPA's
direct in situ PCR method for detection of speci®c bacteria in BEACH Watch Program: Release of 1999 Swimming Season Data.
natural environments. Applied Environmental Microbiology 64, EPA-823-F-00-013, National Technical Information Service,
1536±1540. Spring®eld, VA.

1154