Association of land use and its change with beach closure in the United States, 2004–2013

Science of the Total Environment 571 (2016) 67–76

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Association of land use and its change with beach closure in the United States, 2004–2013 Jianyong Wu a,⁎, Laura Jackson b a b

Oak Ridge Institute for Science and Education (ORISE) Fellowship Participant at US EPA, Office of Research and Development, Research Triangle Park, Durham, NC 27711, USA US EPA, Office of Research and Development, National Health and Environmental Effects Research Laboratory, Research Triangle Park, Durham, NC 27711, USA

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• The impact of land use and its change on beach closure was examined in 10 years. • Urbanization and agriculture development may impair beach water quality. • Afforestation may reduce beach closure frequency by protecting water quality.

a r t i c l e

i n f o

Article history: Received 31 May 2016 Received in revised form 15 July 2016 Accepted 16 July 2016 Available online xxxx Editor: D. Barcelo Keywords: Beach water quality Land cover Land use change Urbanization Ecosystem health

a b s t r a c t Land use and its change have great influences on water quality. However, their impacts on microbial contamination of beach water have rarely been investigated and their relationship with beach actions (e.g., advisories or closure) is still unknown. Here, we analyzed beach closure data obtained from 2004 to 2013 for N 500 beaches in the United States, and examined their associations with land use around beaches in 2006 and 2011, as well as the land use change between 2006 and 2011. The results show that the number of beach closures due to elevated indicators of health risk is negatively associated with the percentages of forest, barren land, grassland and wetland, while positively associated with the percentages of urban area. The results from multi-level models also indicate the negative association with forest area but positive association with urban area and agriculture. The examination of the change of land use and the number of beach closures between 2006 and 2011 indicates that the increase in the number of beach closures is positively associated with the increase in urban (β = 1.612, p b 0.05) and agricultural area including pasture (β = 0.098, p b 0.05), but negatively associated with the increase in forest area (β = −1.789, p b 0.05). The study suggests that urbanization and agriculture development near beaches have adverse effects on beach microbial water quality, while afforestation may protect beach water quality and reduce the number of beach closures. © 2016 Elsevier B.V. All rights reserved.

1. Introduction ⁎ Corresponding authors at: 109 T.W. Alexander Driver, Research Triangle Park, Durham, NC 27711, USA. E-mail addresses: [email protected] (J. Wu), [email protected] (L. Jackson).

http://dx.doi.org/10.1016/j.scitotenv.2016.07.116 0048-9697/© 2016 Elsevier B.V. All rights reserved.

Swimming in natural waters (e.g., oceans, lakes, or rivers) is one of most popular recreational activities in the United States. It was

68

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

estimated that there were 88 million people aged 16 years or above who swam in natural waters annually (NSRE, 2004; Collier et al., 2015). However, exposure to pathogens (e.g., Salmonella spp., Shigella spp., Cryptosporidium, Giardia, adenovirus and norovirus) in recreational waters can lead to a variety of adverse health outcomes (Craun et al., 2005; Heaney et al., 2009; Soller et al., 2010; Hlavsa et al., 2014). Beach sand can serve as a vehicle for microbial transport from and to beach water, thus affecting human health (Solo-Gabriele et al., 2016; Vogel et al., 2016; Whitman et al., 2014). A common ailment is acute gastrointestinal illness (AGI, e.g., diarrhea). Besides AGI, respiratory illness, rash, eye aliments and earaches have also been reported (Heaney et al., 2009; Sanborn and Takaro, 2013; Wade et al., 2013; Collier et al., 2015). According to data from the U.S. Centers for Disease Control and Prevention (CDC), the outbreaks associated with recreational waters (including swimming pools and small lakes) in the United States have varied from 10 to 90 each year and have displayed an increasing trend since the 1980s (Hlavsa et al., 2014). To protect public health and reduce the number of outbreaks associated with recreational waters, the Beaches Environmental Assessment and Coastal Health Act (BEACH Act) was passed in 2000, which required beach regulators to develop a formal plan to assess beach water quality and to notify the public if recreational waters are unsafe (e.g., water quality criteria are violated). Based on the Act, Enterococci as a microbial indicator for marine waters and either Escherichia coli (E. coli) or Enterococci as an indicator for freshwaters are monitored and assessed. These two types of bacteria are common indicators for fecal pollution (Ashbolt et al., 2001; Committee on Indicators for Waterborne Pathogens, 2004; U.S.EPA, 2003; Wu et al., 2011) and may be associated with waterborne illness (Cabelli et al., 1979; Wade et al., 2006). Accordingly, one class of beach advisories or closures is issued contingent on whether the concentrations of these indicators exceed recreational water quality criteria. Beach monitoring and advisories are important public health measures to protect beachgoers and swimmers from exposure to pathogens, thus reducing the number of recreational water-related illness (Nevers and Whitman, 2010; Rabinovici et al., 2004). High levels of microorganisms in water often follow extreme weather events. A clear relationship between heavy rainfall and deterioration of beach microbial water quality has been observed in southern California as well as in Florida (Ackerman and Weisberg, 2003; Brownell et al., 2007). Besides extreme weather events, the proximity of certain land uses to beaches may also have great influence on beach water quality. Microbial contaminants that lead to beach closures and human illness come mainly from land, either from discrete point sources or from diffuse non-point sources (Efstratiou, 2001; Marsalek and Rochfort, 2004). Several land use types, including forest, grassland and wetland, may reduce the transport of microbial contaminants from land to water through their function of filtration, an ecosystem service. For example, wetlands can effectively remove microorganisms from wastewater (Guan and Holley, 2003; Hill and Sobsey, 2001). Recently, studies have suggested there are positive associations between certain land uses (e.g., urban and agricultural areas) and microbial water contamination (e.g., Bradshaw et al., 2016; Cloutier et al., 2015; Didonato et al., 2009; Gotkowska-Płachta et al., 2016; Liang et al., 2013; Schreiber et al., 2015; Smith et al., 2001; Verhougstraete et al., 2015; Wu et al., 2016). It is also well known that the percentage of impervious cover (e.g., roads, parking lots, rooftops) in a watershed has impacts on water quality (Brabec et al., 2002; Long and Plummer, 2004), with observable water quality degradation when imperviousness is 10% or greater (Brabec et al., 2002). As a result, it is expected that land use will have considerable influence on beach microbial water quality. However, to date, studies on impacts of land use on beach microbial contamination are rare, and few researchers are aware of the relationship between land use and beach closures. In this study, we investigated 10 years of beach closures for N500 beaches in the United States. The objectives of this study were to characterize the spatial distribution and temporal trend of beach closures

in the United States, to examine the relationship between land use and beach closures, and to examine the association of beach closures with land use change. This is the first assessment of the impacts of land use and its change on beach closure decisions, which provides important information for understanding the potential benefit of ecosystem services and sustaining beach water quality through natural habitat management.

2. Method 2.1. Beach closure data Beach closure data were obtained online through EPA's Beach Advisory and Closing Online Notification (BEACON) system. In the BEACON database, beach closure information has been recorded for N1000 beaches since 2000 and up to 6000 beaches by 2015. Each beach was given a geographic location, identification number, beach status, length and other information. For each beach action, the action type, start date, end date and reasons were recorded. In this study, we selected all data on beach closures due to elevated bacteria from 2004 to 2013. Because most beach closures occur during summer season, we focused particularly on beach closures during June, July and August to examine the associations of beach closures and land use and its change. In total, 536 beaches were included in the study (Fig. 1).

2.2. Land use data The National Land Cover Database (NLCD) 2006 and NLCD 2011 were used to obtain land use information for these beaches. NLCD datasets were generated based on the classification of Landsat imagery with a spatial resolution of 30 m (Fry et al., 2011; Homer et al., 2015). For these beaches, 8 major classes (water, developed land, barren land, forest, shrubland, herbaceous, planted/cultivated, and wetlands), and 15 subclasses were identified. Considering different ecosystem functions of these classes, we reclassified the land use into 9 classes, including water, open land (open developed land, e.g., golf courses, airports), urban, barren land, forest, shrubland, grassland, agriculture and wetland, by dividing developed land into open land and urban area (Table 1). Agriculture includes cultivated crops and pasture, which may comprise areas of grasses used for livestock grazing. In the grassland class, the areas of grasses are not subject to intensive management but may be used for grazing.

2.3. GIS processing for land use calculation A point layer was created for the beaches of interest based on their centroids using ArcGIS 10.3 (ESRI, CA) and then matched with NLCD 2006 and NLCD 2011 images with appropriate map projection (Albers Conical Equal Area projection). Two methods were used to calculate land use components around each beach. First, around each beach centroid, buffers with the radii of 2 km, 5 km and 10 km were created and used to clip both NLCD 2006 and NLCD 2011 images. The percentage of each land use class was calculated in each buffer. These three radii were chosen because land use components may vary with area and these radii present small, medium and large landscapes, respectively, around beaches. In the second method, we linked each beach to a sixth-level hydrologic unit using its unique hydrologic unit code (HUC). We then calculated the percentage of each land use type in that hydrologic unit. A sixth-level hydrologic unit (HUC12) is a subwatershed delineated by USGS according to surface hydrologic characteristics using a standard hierarchical system. The HUC12 map was obtained from the National Hydrography Dataset (NHD) Plus, Version 2 (Moore and Dewald, 2016).

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

69

Fig. 1. The location of beaches in the United States. The red dots indicate all the beaches, and the green dots indicate the target beaches, which have ever been closed during 2004–2013 summertime (June–August) due to elevated bacteria. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

trend test were conducted to statistically examine whether the trends exist using the “trend” package in R program.

2.4. Data analysis 2.4.1. Spatial distribution and temporal trend of beach closures The number of summertime beach closures during 10 years for each beach was illustrated in a GIS map. From low to high, the frequency of beach closures for each beach was classified into five categories by roughly equal quintile. To examine the trend of beach closures, the 10-year beach closure data (including all months) were initially aggregated by month and then decomposed into three components, including trend, seasonal component, and residual component by seasonaltrend decomposition analysis, using the STL function in R program, an open source statistical software package (Cleveland et al., 1990). The nonparametric Mann-Kendall trend test and seasonal Mann-Kendall

Table 1 Land use classification and reclassification. Class

Code

Classification Description

Reclassification

Water Developed

11 21 22 23 24 31 41 42 43 52 71 81 82 90 95

Open water Developed, Open Space Developed, Low Intensity Developed, Medium Intensity Developed, High Intensity Barren Land (Rock/Sand/Clay) Deciduous Forest Evergreen Forest Mixed Forest Shrub/Scrub Grassland/Herbaceous Pasture/Hay Cultivated Crops Woody Wetlands Emergent Herbaceous Wetlands

Water Open land Urban

Barren land Forest

Shrubland Herbaceous Planted/Cultivated Wetlands

Barren land Forest

2.4.2. Association between land use and beach closures Negative binomial regression models were selected to examine the relationship between land use and beach closures. The dependent variable was the number of beach closures for each beach and the predictors were the percentages of land use classes around each beach. Specifically, we combined 5-year beach closure data (2004–2008 and 2009–2013), then linked the beach closure data during 2004–2008 to the land use data in 2006, and linked the beach closure data during 2009–2013 to the land use data in 2011. We also linked 3-year combined beach closure data (2005–2007 and 2010–2012) to the land use data in 2006 and 2011, respectively, and linked single-year beach closure data to land use data (beach closures in 2006 with land use data in 2006, and beach closures in 2011 with land use data in 2011). Initially, Poisson regression models were tested because the dependent variable, the number of beach closures, is count data and may follow a Poisson distribution, which assumes the expected mean and variance are equal. However, examination of the beach closure data showed they are over-dispersed, namely, the variance is larger than the mean. Therefore, negative binomial models were chosen because they allow a variance larger than the expected mean value. A general equation of the model is: logðμ i Þ ¼ β0 þ β1 x1i þ β2 x2i þ … þ βn xni

Shrubland Grassland Agriculture Wetland

ð1Þ

where μ is the expected number of beach closures, log is the natural logarithm, i is the index of beaches, and x1, x2, …, xn are the percentages of land use types. β0 is the intercept of the model, β1, β2, …, βn are regression coefficients. Multicollinearity among predictors was examined by Pearson correlation analysis. If predictors were highly correlated (e. g,

70

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

r N 0.6, p b 0.01), only one was included in the model. The best fit model was selected based on Akaike's Information criterion (AIC). Generally, a model with a smaller AIC value is better (Bozdogan, 1987). The model was conducted with the R program. The associations are indicated by regression coefficients and p values. A lower p value (p ≤ 0.05) suggests that there is significant evidence to tell the association is true (different from zero), while a higher p value (p N 0.05) suggests that there is insufficient evidence to tell the association is true (Goodman, 2016). Since these beaches are located in 14 U.S. states, it is possible that there is a state-level effect on beach closure because beach management practices vary from state to state. Therefore, multi-level negative binomial regression models were also applied to take the variation of beach closures attributable to state-level effect into account in the association between the number of beach closures and land use (Gelman and Hill, 2006). Here, the data have two levels: individual beach (level 1) and U.S. state (level 2). The model is built on the negative binomial model described above. Differently, we let the intercept vary in each state and held the regression coefficient constant. Therefore, the model is also a mixed-effect (random intercept and fixed slope) model. A general mathematical expression of the model is shown below:

(Gelman and Hill, 2006). The model was carried out with SAS GLIMMIX procedure (SAS Institute., Cary, NC) and the details on model procedures are elsewhere (Wang et al., 2011; Zhu, 2014). 2.4.3. Association of land use change with beach closures Linear regression models were used to examine the relationship between land use change and beach closures. The dependent variable was the difference in the beach closure frequency between 2011 and 2006, which showed a normal distribution according to the histogram (Fig. 2). The predictors were the differences in land use class percentages from 2006 to 2011. A general equation for the model is shown below: ðy11i −y06i Þ ¼ β0 þ β1 ðx11I −x06i Þ þ …βn ðx11n −x06n Þ

ð4Þ

where y11 and y06 are the numbers of beach closures in 2011 and 2006, respectively; i is the index of beaches, x111, …, x11n are the percentages of land use types in 2011, X061, …, x06n are the percentages of land use types in 2006, β0 is the intercept of the model, and β1, …, βn are regression coefficients. A backward selection procedure was applied to find the best fit model. 3. Results

  log μ ij ¼ β0 j þ β1 x1ij þ β2 x2ij þ … þ βn xnij þ Rij

ð2Þ

β0 j ¼ γ00 þ U 0 j

ð3Þ

where j is the index of each state, Rij is the residual effect for individual beaches, γ00 is the mean of state-level intercepts, and U0j is the residual effect for individual states. Other symbols are as described above. To quantify the proportion of the variance of beach closures from the state-level effect, we calculated the intraclass correlation coefficient, which is the ratio of the state-level variance to the total variance

3.1. Description of beach closure data During 2004–2013, 13,803 beach closures were recorded for active beaches. Among them, 53% were due to elevated bacteria, 39% were due to rainfall. Among the beaches closed in summertime due to elevated bacteria, most are located on the coastlines of the Great Lakes and the Atlantic Coast (Massachusetts, New York, New Jersey and Maryland), and a few are located on the Pacific coastlines of California and Washington states. For the closures due to high levels of bacteria, the highest number of closures was found for the Great Lakes coasts, while on the

Fig. 2. The distribution of the number of beach closures. A: The distribution of the total number of beach closures during 2004–2013. B: The distribution of the total number of beach closures during 2004–2008. C: The distribution of the total number of beach closures during 2009–2013. D. The distribution of the change of beach closure numbers between 2011 and 2006.

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

Pacific coastline of California and Washington states, beaches were seldom closed (Fig. 3). The seasonal-trend decomposition analysis revealed that the number of closures due to elevated bacteria had a clear seasonal variation. Specifically, beaches were more likely to be closed in June, July and August, accounting for 93% of closures. No clear annual trend was found for the number of beach closures (Fig. 4). The Mann-Kendall trend test also showed insufficient evidence to support the existence of an annual trend among beach closure data (Kendall's tau = −0.089, p-value is 0.128), while there was sufficient evidence to support that the number of beach closures had a seasonal trend based on the seasonal Mann-Kendall trend test (Kendall's tau = −0.27, p-value b 0.001).

71

of water area is much smaller than that calculated based on the buffers (46.44–48.03%). Open land, agriculture and wetland account for 11.66%, 9.38% and 7.87%, respectively. The percentages of barren land, shrubland and grassland remain b2% (Table 2). By 2011, the average percentages of water, open land, barren land, shrubland and grassland barely changed (b0.1%) after 5 years. On average, urban area increased slightly (0.26% to 0.40%), depending on how the analysis was conducted. The percentage of forest area decreased slightly (0.07% to 0.26%). Agriculture increased by 1.67% when land use was calculated using the radius of 2 km but barely changed when other methods were used. The percentage of wetland decreased by 1.81% when land use was calculated using the radius of 5 km but barely changed when other methods were used (Table S1).

3.2. Land use characteristics 3.3. Relationship between land use and beach closure According to the NLCD 2006, major land use types around beaches are urban and forest (besides water, which is near half of the total area in three buffer radii). On average, the percentage of urban area is 21.59% in the radius of 2 km, and gradually decreases as the buffer radius increases (19.16% and 17.85% in radii of 5 km and 10 km, respectively). Forest area accounts for 9.43% and gradually increases as the buffer radius increases (to 11.51% in the radius of 10 km). The mean percentages of open land and wetland vary from 7.18% to 7.51% and from 4.47% to 5.85%, respectively. The percentage of agricultural area increases from 3.35% to 7.56% when the buffer radius increases from 2 km to 10 km. Barren land, grassland and shrubland account for b2% on average (Table 2). When land use components were calculated based on hydrologic unit, major land use types are still water (27.12%), urban (24.01%) and forest (16.39%), but the mean percentage

The number of beach closures (5 years combined) has a significant positive correlation with the percentage of urban area (r N 0, p b 0.05) but a significant negative correlation with the percentage of forest area (r b 0, p b 0.05) in 2006 and 2011 at all three buffer radii (2 km, 5 km and 10 km). The number of beach closures also showed negative correlations with the percentages of open land, barren land, shrubland, grassland and wetland, but the significance of correlations varied depending on the year of the land use dataset and the buffer radius. When hydrologic units were used, the number of beach closures had significant correlations only with the percentages of urban (r N 0, p b 0.05) and forest (r b 0, p b 0.05) (Table 3). The results from negative binomial regression models showed that the number of beach closures (5 years combined) had positive

Fig. 3. The spatial distribution of the number of beach closures during 2004–2014. A: The Great Lakes; B: the East Coast; C: the California coast; D: the coast of Washington state.

72

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

Fig. 4. Seasonal-trend decomposition analysis of the monthly number of beach closures due to elevated bacteria during 2004–2013. The data were decomposed into three components (seasonal, trend, remainder) using the STL function in R package.

associations (β N 0, p b 0.05) with the percentage of urban area in both 2006 and 2011, and with the percentage of agricultural area in 2011 and under some of the study designs in 2006 as well (Table 4). It had a negative association with the percentage of barren land in both 2006 and

Table 2 Land use components around target beaches in 2006 and 2011 (unit: %). Dataset

NLCD 2006

NLCD 2011

Land use types

Radius = 2

Radius = 5

Radius = 10

km

km

km

Hydrologic unit

Mean SD

Mean SD

Mean SD

Mean SD

Water Open land Urban Barren land Forest Shrubland Grassland Agriculture Wetland

48.02 7.51 21.59 1.62

13.41 6.95 18.20 2.03

48.03 7.43 19.16 0.95

12.96 6.34 16.01 1.13

46.44 7.18 17.85 0.68

11.24 5.76 15.13 0.87

27.12 11.66 24.01 1.47

33.49 10.16 25.26 2.79

9.43 0.49 0.71 3.35 4.47

11.95 1.44 1.22 7.10 7.92

10.81 0.57 0.74 5.93 5.85

12.09 1.34 1.06 9.90 6.00

11.51 0.67 0.81 7.56 5.28

11.54 1.42 1.21 11.04 4.81

16.39 0.95 1.02 9.38 7.87

18.42 2.32 1.65 18.70 9.93

Water Open land Urban Barren land Forest Shrubland Grassland Agriculture Wetland

47.99 7.49 21.89 1.61

13.41 6.97 18.35 2.02

48.02 7.43 19.44 0.95

12.95 6.36 16.15 1.14

46.12 7.25 18.21 0.68

11.80 5.82 15.27 0.89

27.19 11.70 24.41 1.46

33.49 10.21 25.52 2.82

9.32 0.50 0.72 3.30 6.94

11.84 1.43 1.24 7.05 9.23

10.67 0.58 0.75 5.85 5.81

11.96 1.34 1.06 9.85 6.03

11.44 0.69 0.83 7.49 5.31

11.51 1.45 1.24 10.99 4.97

16.12 0.99 1.01 9.25 7.84

18.25 2.38 1.60 18.55 9.96

2011 at the three buffer radii. It also had a negative association (β b 0, p b 0.05) with the percentage of forest area in 2006 but not in 2011. The number of beach closures (5 years combined) had a negative correlation with the percentage of open land at all three radii in 2006 but only at the 10 km radius in 2011. Similar results were obtained when 3-year numbers of beach closures were combined or just single-year data were used to link land use data in the regression models (Table S2 and S3). The results of multi-level models show that the number of beach closures during 2004–2008 was positively associated with the percentages of urban and agricultural areas but negatively associated with the percentage of forest area, though the associations with the percentages of forest and agricultural areas were not always significant at all three buffer radii. The number of beach closures during 2009–2013 had significant positive associations with the percentages of urban and agricultural areas at all three radii (Table 5). Additionally, the state-level effect had a significant contribution (β N 0, p b 0.05) to the variations of the number of beach closures during 2004–2008 as well as during 2009–2013. According to intraclass correlation coefficients, the state-level effect accounted for 30% and 23% of the variance of the number of beach closures during 2004–2008 and during 2009–2013, respectively (Table S4). 3.4. Relationship between land use change and beach closure Pearson correlation analysis showed that the change in beach closures had significant positive correlations with the changes in agricultural area calculated at the radius of 2 km (r = 0.088, p = 0.039) and based on hydrologic unit (r = 0.117, p = 0.025). At the radius of 5 km, the change in beach closures had a significant positive correlation

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76 Table 3 Pearson correlation analysis between the number of beach closures and land use types. Dataset

Land use types

NLCD 2006

Open land Urban Barren Forest Shrubland Grassland Agriculture Wetland

NLCD 2011

Open land Urban Barren Forest Shrubland Grassland Agriculture Wetland

r p r p r p r p r p r p r p r p r p r p r p r p r p r p r p r p

Radius = 2 km

Radius = 5 km

Radius = 10 km

Hydrologic units

−0.148 0.001 0.310 b0.001 −0.076 0.078 −0.256 b0.001 −0.116 0.001 −0.097 0.023 −0.062 0.148 −0.079 0.066 −0.035 0.418 0.271 b0.001 −0.174 b0.001 −0.131 0.002 −0.036 0.400 −0.066 0.122 0.031 0.468 −0.063 0.143

−0.140 0.001 0.388 b0.001 −0.112 0.008 −0.274 b0.001 −0.132 0.002 −0.087 0.043 −0.079 0.063 −0.161 b0.001 −0.030 0.481 0.209 b0.001 −0.085 0.047 −0.135 0.002 −0.045 0.295 −0.067 0.116 0.010 0.815 −0.080 0.060

−0.141 b0.001 0.394 b0.001 −0.138 0.001 −0.296 b0.001 −0.136 0.001 −0.048 0.267 −0.069 0.106 −0.201 b0.001 −0.072 0.093 0.191 b0.001 −0.121 0.004 −0.146 b0.001 −0.090 0.036 −0.088 0.040 0.034 0.434 −0.081 0.058

−0.054 0.303 0.465 b0.001 −0.057 0.275 −0.170 0.001 −0.090 0.085 −0.014 0.783 0.034 0.521 −0.090 0.086 0.034 0.512 0.230 b0.001 −0.051 0.331 −0.021 0.696 −0.029 0.584 −0.001 0.983 0.070 0.185 0.013 0.806

For land use data in 2006, the total number of beach closures during the summer of 2004– 2008 was used in the correlation analysis. For land use data in 2011, the total number of beach closures during the summer of 2009–2013 was used in the correlation analysis. The bold value indicates the correlation is statistically significant (p ≤ 0.05).

with the change in urban area (r = 0.149, p b 0.001) but a negative correlation with the change in forest area (r = − 0.145, p b 0.001) (Table S5). Based on the results from linear regression models (Table 6), the change in beach closures had a positive association with the change in urban area at the radius of 5 km (β = 1.612, p = 0.015) and 10 km (β = 1.085, p = 0.044), as well as that calculated based on hydrologic unit (β = 1.147, p = 0.035). The change in beach closures also had a positive association with the change in agricultural area when land use was calculated by the radius of 2 km (β = 0.098, p = 0.013) and by hydrologic unit (β = 2.093, p = 0.004). Negative associations were found between beach closures and the change in forest area in the radius of 2 km (β = −1.789, p = 0.018), 5 km (β = −2.860, p = 0.002), and 10 km (β = −0.923, p = 0.044). 4. Discussion This study highlights the potentially important role of land use and its change in beach closure caused by microbial contamination. By analyzing national beach closure and land cover datasets, we found that the percentages of urban and agricultural land have positive associations with beach closure and forest has a negative association with beach closure. These results suggest that urbanization, agriculture development and deforestation may increase the degree of microbial contamination of beach water, thus posing a potential risk to human health. Though many studies have indicated that rainfall is a major factor that compromises microbial water quality and leads to beach closure (e. g, Crowther et al., 2001; Shehane et al., 2005; Ackerman and Weisberg, 2003; Brownell et al., 2007; He and He, 2008; Kleinheinz et al., 2009; Bush et al., 2014), our study is the first one to show that land uses and their

73

Table 4 Association between the number of beach closure and land use types examined with negative binomial regression models. For NLCD 2006, the dependent variable is the number of beach closures during 2004–2008 summertime, the predictors are the percentage of each individual land use type in 2006. For NLCD 2011, the dependent variable is the number of beach closures during 2009–2013 summertime, the predictors are the percentage of each individual land use type in 2011. Model ID

Dataset

Land use calculation

Predictors

Regression coefficient

p

1

NLCD 2006

Radius = 2 km

Open land Urban Barren land Forest Shrubland Open land Urban Barren Forest Open land Urban Barren land Forest Grass land Agriculture Urban Barren land Agriculture Urban Barren land Forest Agriculture Urban Barren land Agriculture Open land Urban Barren land Agriculture Urban Barren land Forest Agriculture

−0.041 0.041 −0.124

b0.001 0.001 b0.001

−0.044 −0.107 −0.040 0.029 −0.064 −0.007 −0.034 0.055 −0.793

b0.001 0.051 b0.001 b0.001 b0.001 b0.001 0.007 b0.001 b0.001

−0.023 0.211 0.003 0.035 −0.087

0.003 b0.001 b0.001 b0.001 0.038

0.025 0.011 −0.219

b0.001 b0.001 0.004

−0.011 0.021 0.029 −0.174

0.069 0.015 b0.001 0.002

0.025 −0.027 0.033 −0.395

b0.001 0.013 b0.001 b0.001

0.024 0.024 −0.058

b0.001 b0.001 0.102

0.011 0.017

0.011 b0.001

2

NLCD 2006

Radius = 5 km

3

NLCD 2006

Radius = 10 km

4

NLCD 2006

Hydrologic unit

5

NLCD 2011

Radius = 2 km

6

NLCD 2011

Radius = 5 km

7

NLCD 2011

Radius = 10 km

8

NLCD 2011

Hydrologic unit

The bold value indicates the correlation is statistically significant (p ≤ 0.05).

change have significant effects on beach closure. These factors have been neglected in beach closure modeling and prediction (Nevers and Whitman, 2005; Olyphant, 2005; He and He, 2008). Humans and animals are two major sources of microbial contaminants in beach water. Human fecal matter sheds large amounts of pathogens, which can be transported to beaches through sewage overflows or leaky septic systems under certain circumstances such as heavy rainfall (Crowther et al., 2001). Impervious surfaces (e.g., rooftops, parking lots and roads) in the neighborhood of beaches may lead to degradation of beach water quality as their extent increases with coastal development. Our results show that urbanization has a positive association with beach closure, which is aligned with previous findings that urbanization has negative effects on beach water quality (Walters et al., 2011; Honda et al., 2016). For example, on the central California coast, the concentrations of fecal indicator bacteria and Salmonella were positively associated with the percentage of urban area (Walters et al., 2011). In the Chaophraya River, Thailand, a high ratio of antibiotic-resistant E. coli was linked to urban land use (Honda et al., 2016). The positive association between beach closures and agricultural land is also reasonable. Runoff from agricultural areas (including pasture) may contain animal feces with high numbers of bacteria, viruses and parasites, which can cause microbial contamination of beach

74

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

Table 5 Association between the number of beach closure and land use types examined with multi-level negative binomial regression models. For NLCD 2006, the dependent variable is the number of beach closures during 2004–2008 summertime, the predictors are the percentages of land use types in 2006. For NLCD 2011, the dependent variable is the number of beach closures during 2009–2013 summertime, the predictors are the percentages of land use types in 2011. Model ID

Dataset

Land use calculation

Predictors

Regression coefficient

p

1

NLCD 2006

Radius = 2 km

2

NLCD 2006

Radius = 5 km

Urban Forest Agriculture Urban Forest Agriculture

0.009 −0.022 0.124 0.016 −0.014 0.019

0.038 0.001 0.153 0.003 0.033 0.013

3

NLCD 2006

Radius = 10 km

Urban Barren land Agriculture

0.023 −0.218 0.034

b0.001 0.062 b0.001

4

NLCD 2006

Hydrologic unit

Urban Agriculture

0.017 0.018

b0.001 b0.001

5

NLCD 2011

Radius = 2 km

Urban Agriculture

0.017 0.025

b0.001 0.011

6

NLCD 2011

Radius = 5 km

Urban Agriculture

0.023 0.024

b0.001 0.003

7

NLCD 2011

Radius = 10 km

Urban Agriculture

0.022 0.033

b0.001 b0.001

8

NLCD 2011

Hydrologic unit

Urban Agriculture

0.012 0.023

b0.033 b0.001

The bold value indicates the correlation is statistically significant (p ≤ 0.05).

water. Positive associations have been found between bacteria (e. g, Salmonella) and cattle density in coastal waters of central California (Walters et al., 2011). Forest land has been shown to have a protective effect on water quality through its ecosystem services (Neary et al., 2009). First, it can minimize soil erosion on site, and reduce sediment transport to water bodies (Hartanto et al., 2003). Studies have shown that sediments played an important role in microbial fate and transport in waterbodies (Rehmann and Soupir, 2009; Wu et al., 2009). Reducing the transport of sediments from land to beach water is likely to attenuate microbial contamination of beach water. Second, forests can trap or

Table 6 Relationship between the change in the number of beach closure and land use change between 2011 and 2006 examined with stepwise (backward) linear regression models. The dependent variable is the change in the number of beach closures between 2011 and 2006, the predictors are the change in the percentage of each individual land use types. Model ID

Land use calculation

Predictors

Regression coefficient

p

1

Radius = 2 km

Open land Forest Agriculture Wetland

−1.571 −1.789 0.098 −1.393

0.069 0.018 0.013 0.073

2

Radius = 5 km

3

Radius = 10 km

4

Hydrologic unit

Open land Urban Forest Grass land Agriculture Open land Urban Forest Agriculture Wetland Urban Barren land Agriculture Wetland

−1.830 1.612 −2.860 −1.822 1.128 −2.219 1.085 −0.923 1.513 1.436 1.147 2.102 2.092 1.473

0.088 0.015 0.002 0.119 0.156 0.064 0.045 0.044 0.108 0.066 0.035 0.132 0.004 0.059

filter nonpoint source contaminants suspended in surface runoff (Lowrance, 1998). This process directly lowers the microbial contaminant concentration in runoff before it enters beach water. As a result, beaches with a large percentage of forest nearby may be less likely to be closed for microbial risks. However, forest may be positively associated with fecal contamination of water in the vicinity if wild animals are abundant in the area because wild animals are one of potential sources of fecal contamination. For example, wild animals, especially deer and elk, sometimes cause high concentrations of fecal indicators in waters in pristine lands, such as forest (Niemi and Niemi, 1991). The different types of land uses are not randomly distributed in a landscape. Land cover pattern may affect the buffering capacity of natural habitats near urban beaches. In addition, changing the scale of contributed area will change the percentage of each individual land use class, which may greatly affect the results of the study (Gove et al., 2001). In this study, we assessed four spatial scales (buffer radii at 2 km, 5 km, and 10 km, and hydrologic unit) to calculate land use components. In general, the connection between two objects is stronger if the distance between them is shorter. It is expected that land use within 2 km has stronger influences on beach water quality. However, it is unknown at which distance the effect of land use on beaches is not significant. Therefore, land use components at multiple landscape scales were considered and the results are mostly similar with slight discrepancies. The results highlight the associations between beach closure and a few key types of land use classes (e.g., urban, forest and agriculture). In this study, we also applied multi-level models to examine the associations between land use and beach closure and found the state-level effect contributed significantly to the variation of beach closure frequency. This finding may be explained by different beach management procedures and criteria, different environmental characteristics, (e.g., sea water vs. freshwater/lake water), and different climates across states. There is still no universal standard for sample collection, measurement methods and results interpretation regarding beach microbial water quality monitoring (Nevers and Whitman, 2010). Therefore, incorporating a state-level effect into models will better quantify the relationship between land use and beach closure. The association between beach closure and land use may be confounded by some meteorological factors, such as rainfall and solar radiation. Rainfall often deteriorates beach microbial water quality, while solar radiation, particularly UV light, can inactivate fecal indicator bacteria in beach water (Whitman et al., 2004). To control the effects contributed by weather variability among different beaches, we analyzed a subgroup of beaches in the same region (the coasts of the Great Lakes) using the same statistical method. Our results still showed that the number of beach closures were positively associated with urban and agricultural areas, while negatively associated with forest areas, though the associations varied with the different land use calculation methods (Table S6). The type of beaches also affects beach microbial water quality and it was reported that sand beaches had higher concentrations of fecal bacteria than gravel beaches (Aragonés et al., 2016). In a future study, it would be interesting to take the type of beaches into account when the relationship between land use and microbial water quality is examined. Additional interesting issues are that most beaches examined in this study are located in the northern U.S., and that beach closures in the south (e.g., the Florida coast) were not included in our analysis. There are a few possible reasons. First, beaches in the north (e.g., the Great Lakes) are more extensively monitored; second, some beach closures in summer in the south coast may not be reported to the BEACH program; furthermore, the closures of beaches in summer in the south (e.g., South Florida) are more likely caused by other reasons (such as algal blooms), instead of bacteria. In summary, our results show that a few land use types, particularly urban, forest and agriculture, have significant impacts on beach closure. They suggest that conserving or restoring natural land cover near beaches may exert beneficial effects on beach microbial water quality,

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

thus reducing the risk of waterborne diseases among swimmers and beachgoers. Disclaimer The research described in this article has been reviewed by the National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency, and approved for publication. Approval does not signify that the contents necessarily reflect the views and policies of the Agency, nor does the mention of trade names of commercial products constitute endorsement or recommendation for use. Acknowledgements We thank Bill Kramer and John Wathen from the Office of Water, U.S. EPA, and Yang Shao from the Department of Geography, Virginia Tech, for their valuable comments. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2016.07.116. References Ackerman, D., Weisberg, S.B., 2003. Relationship between rainfall and beach bacterial concentrations on Santa Monica Bay beaches. J. Water Health 1 (2), 85–89. Aragonés, L., López, I., Palazón, A., López-Úbeda, R., García, C., 2016. Evaluation of the quality of coastal bathing waters in Spain through fecal bacteria Escherichia coli and Enterococcus. Sci. Total Environ. 566–567, 288–297. Ashbolt, N.J., Grabow, W.O.K., Snozzi, M., 2001. Indicators of microbial water quality. In: Fewtrell, L., Bartram, J. (Eds.), Water Quality: Guidelines, Standards and HealthRisk Assessment and Management for Water-related Infectious Disease. IWA Publishing, London, pp. 289–315 (Chapter 13). Bozdogan, H., 1987. Model selection and Akaike's information criterion (AIC): the general theory and its analytical extensions. Psychometrika 52 (3), 345–370. Bradshaw, J.K., Snyder, B.J., Oladeinde, A., Spidle, D., Berrang, M.E., Meinersmann, R.J., Oakley, B., Sidle, R.C., Sullivan, K., Molina, M., 2016. Characterizing relationships among fecal indicator bacteria, microbial source tracking markers, and associated waterborne pathogen occurrence in stream water and sediments in a mixed land use watershed. Water Res. 101, 498–509. Brabec, E., Schulte, S., Richards, P.L., 2002. Impervious surfaces and water quality: a review of current literature and its implications for watershed planning. J. Plan. Lit. 16 (4), 499–514. Brownell, M.J., Harwood, V.J., Kurz, R.C., McQuaig, S.M., Lukasik, J., Scott, T.M., 2007. Confirmation of putative stormwater impact on water quality at a Florida beach by microbial source tracking methods and structure of indicator organism populations. Water Res. 41 (16), 3747–3757. Bush, K.F., Fossani, C.L., Li, S., Mukherjee, B., Gronlund, C.J., O'Neill, M.S., 2014. Extreme precipitation and beach closures in the great lakes region: evaluating risk among the elderly. Int. J. Environ. Res. Public Health 11 (2), 2014–2032. Cabelli, V.J., Dufour, A.P., Levin, M.A., Mccabe, L.J., Haberman, P.W., 1979. Relationship of microbial indicators to health effects at marine bathing beaches. A. J. Public Health 69 (7), 690–696. Cleveland, R.B., Cleveland, W.S., McRae, J.E., Terpenning, I., 1990. STL: a seasonal-trend decomposition procedure based on loess. J. Off. Stat. 6, 3–73. Cloutier, D.D., Alm, E.W., McLellan, S.L., 2015. Influence of land use, nutrients, and geography on microbial communities and fecal indicator abundance at Lake Michigan beaches. Appl. Environ. Microbiol. 81 (15), 4904–4913. Collier, S.A., Wade, T.J., Sams, E.A., Hlavsa, M.C., Dufour, A.P., Beach, M.J., 2015. Swimming in the USA: beachgoer characteristics and health outcomes at US marine and freshwater beaches. J. Water Health 13 (2), 531–543. Committee on Indicators for Waterborne Pathogens, 2004. Indicators for Waterborne Pathogens. National Research Council of the National Academies. The National Academies Press, Washington, DC. Craun, G.F., Calderon, R.L., Craun, M.F., 2005. Outbreaks associated with recreational water in the United States. Int. J. Environ. Health Res. 15 (4), 243–262. Crowther, J., Kay, D., Wyer, M.D., 2001. Relationships between microbial water quality and environmental conditions in coastal recreational waters: the Fylde coast, UK. Water Res. 35 (17), 4029–4038. DiDonato, G.T., Stewart, J.R., Sanger, D.M., Robinson, B.J., Thompson, B.C., Holland, A.F., Van Dolah, R.F., 2009. Effects of changing land use on the microbial water quality of tidal creeks. Mar. Pollut. Bull. 58 (1), 97–106. Efstratiou, M.A., 2001. Managing coastal bathing water quality: the contribution of microbiology and epidemiology. Mar. Pollut. Bull. 42, 425–432. Fry, J., Xian, G., Jin, S., Dewitz, J., Homer, C., Yang, L., Barnes, C., Herold, N., Wickham, J., 2011. Completion of the 2006 National Land Cover Database for the conterminous United States. Photogramm. Eng. Remote. Sens. 77 (9), 858–864.

75

Gelman, A., Hill, J., 2006. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press. Goodman, S.N., 2016. Aligning statistical and scientific reasoning. Science 352 (6290), 1180–1181. Gotkowska-Płachta, A., Gołaś, I., Korzeniewska, E., Koc, J., Rochwerger, A., Solarski, K., 2016. Evaluation of the distribution of fecal indicator bacteria in a river system depending on different types of land use in the southern watershed of the Baltic Sea. Environ. Sci. Pollut. Res. 23 (5), 4073–4085. Gove, N.E., Edwards, R.T., Conquest, L.L., 2001. Effects of scale on land use and water quality relationships: a longitudinal basin-wide perspective1. J. Am. Water Resour. Assoc. 37 (6), 1721–1734. Guan, T.Y., Holley, R.A., 2003. Pathogen survival in swine manure environments and transmission of human enteric illness–a review. J. Environ. Qual. 32 (2), 383–392. Hartanto, H., Prabhu, R., Widayat, A.S., Asdak, C., 2003. Factors affecting runoff and soil erosion: plot-level soil loss monitoring for assessing sustainability of forest management. Forest Ecol. Manag. 180 (1), 361–374. Heaney, C.D., Sams, E., Wing, S., Marshall, S., Brenner, K., Dufour, A.P., Wade, T.J., 2009. Contact with beach sand among beachgoers and risk of illness. Am. J. Epidemiol. 170 (2), 164–172. He, L.M., He, Z.L., 2008. Water quality prediction of marine recreational beaches receiving watershed baseflow and stormwater runoff in southern California, USA. Water Res. 42 (10), 2563–2573. Hill, V.R., Sobsey, M.D., 2001. Removal of Salmonella and microbial indicators in constructed wetlands treating swine wastewater. Water Sci. Technol. 44 (11 −12), 215–222. Hlavsa, M.C., Roberts, V.A., Kahler, A.M., Hilborn, E.D., Wade, T.J., Backer, L.C., Yoder, J.S., 2014. Recreational water–associated disease outbreaks – United States, 2009–2010. MMWR 63 (1), 6–10. Honda, R., Watanabe, T., Sawaittayotin, V., Masago, Y., Chulasak, R., Tanong, K., Chaminda, G.T., Wongsila, K., Sienglum, C., Sunthonwatthanaphong, V., Poonnotok, A., 2016. Impacts of urbanization on the prevalence of antibiotic-resistant Escherichia coli in the Chaophraya River and its tributaries. Water Sci. Technol. 73 (2), 362–374. Homer, C.G., Dewitz, J.A., Yang, L., Jin, S., Danielson, P., Xian, G., Coulston, J., Herold, N.D., Wickham, J.D., Megown, K., 2015. Completion of the 2011 National Land Cover Database for the conterminous United States-representing a decade of land cover change information. Photogramm. Eng. Remote. Sens. 81 (5), 345–354. Kleinheinz, G.T., McDermott, G.M., Hughes, S., Brown, A., 2009. Effects of Rainfall on E. coli concentrations at Door county, Wisconsin beaches. Int. J. Microbiol, 876050 http://dx. doi.org/10.1155/2009/876050. Liang, Z., He, Z., Zhou, X., Powell, C.A., Yang, Y., He, L.M., Stoffella, P.J., 2013. Impact of mixed land-use practices on the microbial water quality in a subtropical coastal watershed. Sci. Total Environ. 449, 426–433. Long, S.C., Plummer, J.D., 2004. Assessing land use impacts on water quality using microbial source tracking. Am. Water Resour. Assoc. 40 (6), 1433–1448. Lowrance, R., 1998. Riparian forest ecosystems as filters for nonpoint-source pollution. In Successes, Limitations, and Frontiers in Ecosystem Science. Springer, New York, pp. 113–141. Marsalek, J., Rochfort, Q., 2004. Urban wet weather flows: sources of fecal contamination impacting on recreational waters and threatening drinking water resources. J. Toxicol. Environ. Health 67, 1765–1777. Moore, R.B., Dewald, T.G., 2016. The road to NHDPlus – advancements in digital stream networks and associated catchments. J. Am. Water Resour. Assoc. DOI http://dx.doi. org/10.1111/1752-1688.12389. Neary, D.G., Ice, G.G., Jackson, C.R., 2009. Linkages between forest soils and water quality and quantity. For. Ecol. Manag. 258 (10), 2269–2281. Nevers, M.B., Whitman, R.L., 2005. Nowcast modeling of Escherichia coli concentrations at multiple urban beaches of southern Lake Michigan. Water Res. 39 (20), 5250–5260. Nevers, M.B., Whitman, R.L., 2010. Policies and practices of beach monitoring in the Great Lakes, USA: a critical review. J. Environ. Monit. 12 (3), 581–590. Niemi, R.M., Niemi, J.S., 1991. Bacterial pollution of waters in pristine and agricultural lands. J. Environ. Qual. 20 (3), 620–627. NSRE (National Survey on Recreation and the Environment), 2004. National Survey on Recreation and the Environment 2000–2002 (The Interagency National Survey Consortium Coordinated by the USDA Forest Service Recreation). Olyphant, G.A., 2005. Statistical basis for predicting the need for bacterially induced beach closures: emergence of a paradigm? Water Res. 39 (20), 4953–4960. Rabinovici, S.J., Bernknopf, R.L., Wein, A.M., Coursey, D.L., Whitman, R.L., 2004. Economic and health risk trade-offs of swim closures at a Lake Michigan beach. Environ. Sci. Technol. 38 (10), 2737–2745. Rehmann, C.R., Soupir, M.L., 2009. Importance of interactions between the water column and the sediment for microbial concentrations in streams. Water Res. 43 (18), 4579–4589. Sanborn, M., Takaro, T., 2013. Recreational water–related illness office management and prevention. Can. Fam. Physician 59 (5), 491–495. Schreiber, C., Rechenburg, A., Rind, E., Kistemann, T., 2015. The impact of land use on microbial surface water pollution. Int. J. Hyg. Environ. Health 218 (2), 181–187. Shehane, S.D., Harwood, V.J., Whitlock, J.E., Rose, J.B., 2005. The influence of rainfall on the incidence of microbial faecal indicators and the dominant sources of faecal pollution in a Florida river. J. Appl. Microbiol. 98 (5), 1127–1136. Smith, J.H., Wickham, J.D., Norton, D., Wade, T.G., Jones, K.B., 2001. Utilization of landscape indicators to model potential pathogen impaired waters. J. Am. Water Resour. Assoc. 37 (4), 805–814. Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J., 2010. Estimated human health risks from exposure to recreational waters impacted by human and non-human sources of faecal contamination. Water Res. 44 (16), 4674–4691.

76

J. Wu, L. Jackson / Science of the Total Environment 571 (2016) 67–76

Solo-Gabriele, H.M., Harwood, V.J., Kay, D., Fujioka, R.S., Sadowsky, M.J., Whitman, R.L., Wither, A., Caniça, M., Da Fonseca, R.C., Duarte, A., Edge, T.A., 2016. Beach sand and the potential for infectious disease transmission: observations and recommendations. J. Mar. Biol. Assoc. U. K. 96 (01), 101–120. US Environmental Protection Agency, 2003. Bacterial water quality standards for recreational waters (freshwater and marine waters). EPA-823-R-03-008, Status Report. Office of Water, Washington, DC. Verhougstraete, M.P., Martin, S.L., Kendall, A.D., Hyndman, D.W., Rose, J.B., 2015. Linking fecal bacteria in rivers to landscape, geochemical, and hydrologic factors and sources at the basin scale. Proc. Natl. Acad. Sci. U. S. A. 112 (33), 10419–10424. Vogel, L.J., O'Carroll, D.M., Edge, T.A., Robinson, C.E., 2016. Release of Escherichia coli from foreshore sand and pore water during intensified wave conditions at a recreational beach. Environ. Sci. Technol. 50 (11), 5676–5684. Wade, T.J., Sams, E.A., Beach, M.J., Collier, S.A., Dufour, A.P., 2013. The incidence and health burden of earaches attributable to recreational swimming in natural waters: a prospective cohort study. Environ. Health 12, 67–77. Wade, T.J., Calderon, R.L., Sams, E., Beach, M., Brenner, K.P., Williams, A.H., Dufour, A.P., 2006. Rapidly measured indicators of recreational water quality are predictive of swimming-associated gastrointestinal illness. Environ. Health Perspect. 114, 24–28. Walters, S.P., Thebo, A.L., Boehm, A.B., 2011. Impact of urbanization and agriculture on the occurrence of bacterial pathogens and stx genes in coastal waterbodies of central California. Water Res. 45 (4), 1752–1762.

Wang, J., Xie, H., Fisher, J.F., 2011. Multilevel Models: Applications Using SAS®. Higher Education Press and Walter de Gruyter, Berlin/Boston. Whitman, R.L., Harwood, V.J., Edge, T.A., Nevers, M.B., Byappanahalli, M., Vijayavel, K., Brandão, J., Sadowsky, M.J., Alm, E.W., Crowe, A., Ferguson, D., 2014. Microbes in beach sands: integrating environment, ecology and public health. Rev. Environ. Sci. Biotechnol. 13 (3), 329–368. Whitman, R.L., Nevers, M.B., Korinek, G.C., Byappanahalli, M.N., 2004. Solar and temporal effects on Escherichia coli concentration at a Lake Michigan swimming beach. Appl. Environ. Microbiol. 70 (7), 4276–4285. Wu, J., Long, S.C., Das, D., Dorner, S.M., 2011. Are microbial indicators and pathogens correlated? A statistical analysis of 40 years of research. J. Water Health 9 (2), 265–278. Wu, J., Rees, P., Storrer, S., Alderisio, K., Dorner, S., 2009. Fate and transport modeling of potential pathogens: the contribution from sediments. J. Am. Water Resour. Assoc. 45 (1), 35–44. Wu, J., Yunus, M., Islam, M.S., Emch, M., 2016. Influence of climate extremes and land use on fecal contamination of shallow tubewells in Bangladesh. Environ. Sci. Technol. 50 (5), 2669–2676. Zhu, M., 2014. Analyzing Multilevel Models with the GLIMMIX Procedure. In Proceedings of the SAS Global Forum 2014 Conference (URL http://support. sas.com/resources/papers/proceedings14/SAS026-2014.pdf).