Trees and Streets as Drivers of Urban Stormwater Nutrient Pollution

30 Jul 2017 - However, few studies have addressed the full range of potential impacts .... total impervious area (TIA), fraction, 0.44 (0.20–0.80), ...
0 downloads 0 Views 5MB Size
Subscriber access provided by UNIVERSITY OF THE SUNSHINE COAST

Article

Trees and Streets as Drivers of Urban Stormwater Nutrient Pollution Benjamin D. Janke, Jacques C Finlay, and Sarah E. Hobbie Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.7b02225 • Publication Date (Web): 30 Jul 2017 Downloaded from http://pubs.acs.org on July 30, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 37

Environmental Science & Technology

1 2 3 4 5 6 7 8 9 10 11 12

Title: Trees and Streets as Drivers of Urban Stormwater Nutrient Pollution Authors: Benjamin D. Janke*, Jacques C. Finlay, Sarah E. Hobbie Affiliation (all authors): University of Minnesota Department of Ecology, Evolution and Behavior 140 Gortner Laboratory 1479 Gortner Ave St. Paul, MN, USA 55108 *Corresponding Author: Phone: 612-716-6012 Email: [email protected]

13

benefits provided to people, and potentially to water quality through reduction of

14

stormwater volume by interception. However, few studies have addressed the full range

15

of potential impacts of trees on urban runoff, which includes deposition of nutrient-rich

16

leaf litter onto streets connected to storm drains. We analyzed the influence of trees on

17

stormwater nitrogen and phosphorus export across 19 urban watersheds in Minneapolis-

18

St. Paul, MN, USA, and at the scale of individual streets within one residential

19

watershed. Stormwater nutrient concentrations were highly variable across watersheds

20

and strongly related to tree canopy over streets, especially for phosphorus. Stormwater

21

nutrient loads were primarily related to road density, the dominant control over runoff

22

volume. Street canopy exerted opposing effects on loading, where elevated nutrient

23

concentrations from trees near roads outweighed the weak influence of trees on runoff

24

reduction. These results demonstrate that vegetation near streets contributes substantially

25

to stormwater nutrient pollution, and therefore to eutrophication of urban surface waters.

26

Urban landscape design and management that account for trees as nutrient pollution

27

sources could improve water quality outcomes, while allowing cities to enjoy the myriad

28

benefits of urban forests.

Abstract: Expansion of tree cover is a major management goal in cities because of the substantial

29 1 ACS Paragon Plus Environment

Environmental Science & Technology

30

TOC Graphic:

31 32

2 ACS Paragon Plus Environment

Page 2 of 37

Page 3 of 37

Environmental Science & Technology

33

Introduction

34

Urban ecosystems are characterized by high levels of nutrient inputs associated with

35

humans123 and by amplified hydrologic transport due to extensive impervious surfaces

36

and storm drains. Aquatic ecosystems within and downstream of cities are subject to

37

excessive stormwater loading from the landscape, leading to flooding, loss of ecosystem

38

function, and degradation of habitat4–7. The most pervasive effect of excessive

39

stormwater nutrient loading to lakes, streams, and coastal waters is eutrophication, which

40

results in abundant algal growth including harmful cyanobacterial blooms, as well as low

41

oxygen, fish kills, and noxious odor, leading to degradation of aquatic habitat, recreation,

42

and water supply8.

43 44

Efforts to improve water quality of urban lakes and streams have focused heavily on the

45

reduction and treatment of stormwater runoff, typically through installation of end-of-

46

pipe management structures such as detention ponds and infiltration trenches. However,

47

widespread improvement of urban water quality has not been achieved, despite the

48

substantial resources invested in stormwater management9. Therefore, there is increasing

49

interest in strategies both for reducing non-point source nutrient pollution within

50

watersheds and for restoring more natural hydrologic regimes10–13. Particular emphasis is

51

placed on the expansion of “green” infrastructure14, often defined as engineered

52

structures such as bioswales and vegetated rooftops, but also including urban vegetation

53

in lawns, parks, and boulevards. Green infrastructure is appealing for stormwater

54

management because it provides reduction of runoff volume and peak flows via

55

interception of rainfall, infiltration of stormwater, and evapotranspiration, which

3 ACS Paragon Plus Environment

Environmental Science & Technology

56

potentially decrease associated runoff nutrient loads12,14,15. Green infrastructure, and trees

57

in particular, also have co-benefits, improving flood control, air quality, mental health,

58

recreational opportunities, property and aesthetic values, and climate change resiliency16–

59

22

.

60 61

Trees are a crucial component of green infrastructure, and the expansion of tree cover has

62

been widely promoted in cities23,24. Trees potentially improve water quality by decreasing

63

nutrient export when used in bioswales and planter boxes25–27, and by reducing

64

stormwater volumes and peak flows (and presumably nutrient export) at watershed scales

65

28–31

66

waters of expanded tree cover. While trees and other vegetated areas near streets promote

67

nutrient uptake27, large pools of nutrients in plant biomass and soils could serve as

68

sources of nitrogen (N) and phosphorus (P) transported to stormwater systems via

69

erosion, litterfall, and leaching32–35. If trees, as an integral part of green infrastructure,

70

contribute nutrients to stormwater, then disentangling the opposing influences of runoff

71

volume reduction and increases in stormwater nutrient concentrations is essential.

72

Furthermore, incomplete understanding of nutrient sources to streets and storm drains,

73

including vegetation as well as atmospheric deposition36,37, pet waste3, and fertilizer and

74

erosion from lawns38,39, is a major impediment to development of effective nutrient

75

pollution management strategies11, and to understanding the potential water quality

76

consequences of increasingly “green” urban environments.

. However, few studies have quantified a nutrient reduction benefit to downstream

77

4 ACS Paragon Plus Environment

Page 4 of 37

Page 5 of 37

Environmental Science & Technology

78

In this study, we assessed the role of vegetation, and trees adjacent to streets in particular,

79

on urban stormwater runoff quality by analyzing factors that control stormwater nutrient

80

levels across a large urban area, the Minneapolis-St. Paul metropolitan area, Minnesota,

81

USA (TCMA). We used extensive stormwater monitoring datasets based on over 2,300

82

measurements taken from 2005 to 2014 in 19 watersheds to compare nutrient

83

concentrations and loading across gradients of tree, vegetation, and impervious cover

84

typical of urban residential watersheds. We used these robust datasets to address the

85

following questions: (1) How does the cover of vegetation, and especially trees adjacent

86

to streets, affect nutrient loads and concentrations in stormwater? (2) Does the volume

87

reduction provided by street trees offset the potential enhanced nutrient inputs to streets

88

from leaf litter (e.g. leaves, seeds, pollen, flowers)? and (3) How important are “green”

89

nutrient sources relative to other factors associated with nutrient inputs to urban areas,

90

such as atmospheric deposition?

91 92

Background and Methods

93

Study Sites, Data Acquisition and Collection

94

We focused on understanding nutrient sources at two spatial scales dominated by urban

95

land use in the Twin Cities Metropolitan Area of Minneapolis-St. Paul (TCMA),

96

Minnesota, USA (Fig. 1). We used an extensive, multi-year dataset for 2,362 stormwater

97

runoff events across 19 urban sub-watersheds of the Mississippi River (Table S1) along

98

with high-resolution land cover data to assess the influence of urban vegetation and other

99

potential drivers of nutrient pollution (Tables 1, S2). We complemented these analyses

100

with investigations at the scale of individual streets with varying street tree canopy cover

5 ACS Paragon Plus Environment

Environmental Science & Technology

101

within a single residential watershed. Study watersheds were small, ranging in size from

102

4 ha to 3,170 ha, with generally mixed urban land use dominated by low-density

103

residential. In some watersheds, remnant surface water features (lakes, ponds) were

104

present. Development age across sites ranged from roughly 20 years old in the outer

105

suburbs, where street tree canopy tended to be minimal due to development in former

106

agricultural lands, to 100 years or older at sites in the urban core, where tree canopy was

107

older and denser. Drainage infrastructure was on average older in the urban core than in

108

areas of younger development; however, storm drain systems in the study watersheds

109

have been completely separated from sanitary sewers since 1996, and both storm and

110

sanitary systems are tested for leaks and maintained by municipalities and watershed

111

managers. These features, along with the lack of evidence for gross contamination of

112

sewage at sites with baseflow40 suggest that leaking sanitary sewers did not influence our

113

study sites. Use of P in lawn fertilizer has been restricted for individual household use

114

since 2004, while N fertilizer use is not regulated.

115 116

Stormwater nutrient chemistry and hydrology data from five watershed management

117

organizations were integrated into our analyses (Fig. 1; Table S1). Data were collected as

118

part of regional stormwater monitoring programs initiated as early as 2001, but more

119

typically since 2005. Monitoring was usually conducted during the April to November

120

warm period of each year. Cross-site comparisons used only the data collected from 2005

121

– 2014, restricted to the warm season (April 1 – October 31) when the majority of annual

122

precipitation occurs (79% on average from 1981 - 2010)41. Monitoring protocols,

123

including sample collection, chemical analyses, and quality control procedures, were

6 ACS Paragon Plus Environment

Page 6 of 37

Page 7 of 37

Environmental Science & Technology

124

similar among organizations (Table S1). Nine sites did not have baseflow. For most of

125

the other 10 sites, the influence of baseflow on stormwater was small since runoff rates

126

were generally larger during storms than during baseflow by an order of magnitude or

127

more40; for sites with appreciable baseflow (MS1, MS2), sliding-interval baseflow

128

separation was applied to hydrologic data42.

129 130

Stormwater samples were primarily composite samples (n = 1895), combined from sub-

131

samples within an event to provide a single, volume-weighted composite. Roughly 17%

132

of the samples (n = 330) were grab samples; however, the potential bias of including grab

133

samples was minimal, as the significance of regressions (see below) were unchanged

134

when grab samples were excluded from the data set. Samples were analyzed for

135

concentrations of total phosphorus (TP), total dissolved phosphorus (TDP), nitrate- plus

136

nitrite-nitrogen (hereafter NOx-N), ammonium nitrogen (NH4-N), and total Kjeldahl

137

nitrogen (TKN). Total nitrogen (TN) was estimated as the sum of TKN + NOx-N, and

138

total organic nitrogen (TON) as TKN – NH4-N. The majority of samples were analyzed

139

by the Metropolitan Council Environmental Services Laboratory (St. Paul, MN), using

140

standard U.S. E.P.A. protocols43. Soluble reactive phosphorus (SRP) was not consistently

141

measured at most sites, so TDP was used in the data analysis. For the CRWD sites, for

142

which SRP was generally measured instead of TDP, we estimated TDP from SRP using a

143

linear regression applied to a subset of 641 runoff samples that had been measured for

144

both forms (TDP [mg/L] = 1.20*SRP [mg/L] + 0.012, R2 = 0.91; unpublished data).

145

7 ACS Paragon Plus Environment

Environmental Science & Technology

146

Stormwater event mean concentrations (EMC) observed in this study for N and P (Table

147

2) were typical of urban runoff44, and similar to previous observations in the TCMA45. TP

148

and TN greatly exceeded that measured in precipitation in the study area, including in

149

rainfall at the AHUG watershed during 2011-2013 (0.03 mg/L for TP, 1.05 mg/L for TN,

150

n = 27 samples; unpublished data), and in wet deposition measurements of TP across the

151

TCMA in a 1980 study46 (TP = 0.06 mg/L, n = 5 sites). Stormwater NOx-N (0.45 mg/L)

152

was similar to mean wet deposition at AHUG (0.25 mg/L) and in Payne et al.46 (0.46

153

mg/L), while NH4-N (0.24 mg/L) was much lower than observations in the two

154

precipitation data sets (0.69 mg/L at AHUG and 0.92 mg/L in Payne et al.46).

155 156

Continuous flow was recorded at all sites but quality-controlled data for stormwater

157

runoff volumes were available only for a subset of 12 sites. Nutrient yields (kg/km2) were

158

estimated for each sampled event at these sites by multiplying the observed volume by

159

the observed concentration (typically from a volume-weighted composite, but sometimes

160

represented by a grab sample), and normalized by watershed area3,40,42.

161 162

We also investigated the street scale in a small (17 ha) residential watershed in St. Paul,

163

MN (AHUG; Table S1). During several late spring (post leaf-out) and fall (post leaf-

164

drop) events from fall 2012 through spring 2015, we sampled street gutter runoff from 9

165

blocks within the watershed that varied in street canopy coverage due to differences in

166

tree species and age. Runoff was sampled using a 1-L plastic bottle by collecting water as

167

it entered the catch basins at the end of each major block. Water samples were analyzed

168

for major nutrient forms including TP, SRP, NOx-N, NH4-N, total dissolved N (TDN),

8 ACS Paragon Plus Environment

Page 8 of 37

Page 9 of 37

Environmental Science & Technology

169

and particulate N (PN) at the University of Minnesota (UMN) using similar laboratory

170

methods as MCES40. For these samples, TN was estimated as TDN + PN, and TON as

171

TN – NOx-N – NH4-N.

172 173

Data Analysis and Model Selection Approach

174

Land cover, land use, and hydrologic connectivity

175

In cities, primary new sources of N and P to the landscape include fertilizer, pet waste,

176

and atmospheric deposition from automobiles and industrial activities3, all of which may

177

be exported to stormwater during runoff events. Much of the N and P from these sources

178

may also be assimilated by plants and microbes, and bound to soil, where it can later

179

become a source of nutrients to runoff through leaching of vegetation and surface soils,

180

leaf (and other) litter and grass clippings that fall or are washed or blown into streets, and

181

eroded soils. Urban stormwater hydrology, which influences the magnitude of nutrient

182

loading, is primarily controlled by the extent and configuration of impervious

183

surfaces47,48, which also serve as accumulation areas for atmospheric deposition.

184

Although we did not have direct information to trace these sources, to gain insights into

185

the importance of potential nutrient sources to stormwater and the factors that influence

186

stormwater runoff volume, we analyzed relationships between stormwater nutrient (and

187

water) export and watershed characteristics related to streets, impervious cover, traffic,

188

population, housing density, and vegetation cover (Tables 1, S2). The variables used in

189

analyses, and the potential sources of nutrients and runoff that they represent, are

190

summarized in Table 1 and described briefly below (see SI for details on data sources and

191

calculation of characteristics). All spatial analyses were completed in ESRI ArcGIS 10.1.

9 ACS Paragon Plus Environment

Environmental Science & Technology

192 193

Land cover and land use attributes that potentially influence stormwater N and P included

194

vegetation and factors associated with human activities such as traffic volume (average

195

annual daily traffic), population density (people/km2), and low-density residential parcel

196

area (as a fraction of total watershed area). Vegetation was described by total vegetation

197

(open lawn + tree) cover, total tree cover, and tree canopy over the street as well as tree

198

canopy over and within 1.52 m and 6.10 m of the street (Table S2). Limitations of the

199

spatial data prevented the estimation of total or near-street turf grass cover (see SI).

200

Traffic density is related to the potential input of local inorganic N by deposition from

201

combustion by vehicles, and is concentrated near roadways36. Population density

202

(people/km2) is associated with nutrient inputs from pets and vehicles, and potentially

203

food waste or trash. Low-density (three families or fewer) residential parcel area is

204

closely associated with lawn area and with household nutrient inputs such as lawn

205

fertilizer or pet waste. Without explicit numbers on pet ownership or lawn fertilizer

206

application rates in the study watersheds, we acknowledge that residential parcel area

207

integrates the potential effect of both nutrient sources. A recent study3 found that the

208

largest new inputs of N and P to our study watersheds were fertilizer and pet waste,

209

respectively. However, past fertilization may have accrued in soils, which complicates

210

source tracing of P.

211 212

Drainage intensity, which exerts a dominant influence on stormwater runoff volumes,

213

was characterized by total impervious area, total street area, and street density (length per

214

unit area watershed). Street density was assumed to represent the most directly-connected

10 ACS Paragon Plus Environment

Page 10 of 37

Page 11 of 37

Environmental Science & Technology

215

impervious areas, as a true effective impervious area (EIA) could not be determined for

216

all watersheds due to limitations of spatial and hydrologic data. Additionally, incomplete

217

storm drain maps for many watersheds prevented the characterization of the extent of

218

storm sewer connectivity of the drainage areas.

219 220

Statistical Analysis

221

The influence of near-street tree canopy on stormwater nutrient concentrations, and its

222

importance relative to other human and landscape factors in the urban study area, was

223

assessed using two sets of analyses. First, the across-site relationships of stormwater

224

volumes and nutrient concentrations to individual watershed characteristics (Table 1)

225

were investigated with simple linear regression (SLR). Event mean nutrient

226

concentrations (EMC) by site were used in the regressions, with data restricted to the

227

typical monitoring season (i.e. April 1 – October 31) since not all sites were monitored

228

year-round. Mean event runoff and nutrient yields by site were used in the regressions for

229

the subset of sites with event hydrology data (n = 12; Table S1). Statistical significance is

230

reported at p < 0.05 and p < 0.001.

231 232

Multiple linear regression analysis (MLR) was used to assess the influence of street tree

233

canopy relative to the other watershed factors on nutrient concentrations and yields.

234

Candidate factors were assembled separately for each nutrient form by first selecting

235

those variables hypothesized to influence stormwater nutrients that also had high

236

correlation coefficients from SLR. For sets of highly collinear factors (Pearson |r| > 0.7),

237

such as street density and street area, the factors with the lowest correlation to the nutrient

11 ACS Paragon Plus Environment

Environmental Science & Technology

238

of interest were excluded. The full model for each nutrient was then tested exhaustively

239

for every combination of candidate factors (main effects only; no interactions), with sub-

240

models ranked by sample size-adjusted Akaike Information Criterion (AICc). Models for

241

which constituent factors exceeded a variance inflation factor (VIF) of 5.0 were rejected.

242

Adjusted R2 was then computed for all models within AICc 2.0 of the best model49. Best

243

model selection, including estimates of coefficients, significance, and effect size (as

244

provided by η2), is shown in the SI along with model fits to observations. R was used for

245

all statistical analyses (MLR and SLR).

246 247

Analyses of the net influence of trees on stormwater nutrient yields via effects on runoff

248

reduction and on stormwater EMC were complicated by our relatively small data subset

249

for nutrient loads (n=12 sites), and by covariance of street canopy cover with street

250

density (and with runoff volume) among these 12 sites (R2= 0.40; Table S2). To examine

251

the influence of tree canopy on nutrient loading via effects on both concentration and

252

runoff, we constructed a nutrient yield model from the MLR analyses for water yield and

253

for EMC of TP and TN (see Results and SI). Nutrient yields were estimated as a product

254

of predicted EMC and predicted water yield for hypothetical watershed configurations

255

(combinations of street canopy and street density within ranges present in our dataset).

256 257

Results and Discussion

258

Urban trees as a major source of nutrients to stormwater

259

Our results indicate that trees adjacent to streets were a dominant factor in determining N

260

and P concentrations in stormwater during the warm weather period (April – October),

12 ACS Paragon Plus Environment

Page 12 of 37

Page 13 of 37

Environmental Science & Technology

261

when typically 60-80% of annual runoff and nutrient loading from stormwater occurs in

262

our study sites (n=7 sites3). Analyses of stormwater concentration data provided strong

263

evidence for this conclusion; variation in event mean concentration (EMC) of TP across

264

sites was explained significantly in simple linear regression (SLR) by tree canopy over (r

265

= 0.84, p < 0.001) and near the street (Table S3), and TP in the watersheds with the

266

greatest street canopy cover was up to three-fold higher than in those with negligible

267

street canopy (Fig. 2). Street canopy was also the dominant influence on TP when

268

considered with other factors in multiple linear regression (MLR; Table 3), as all

269

candidate models within 2.0 AICc units (n=5) included street canopy. Similarly for

270

nitrogen, EMC of TN was strongly related to street canopy (r = 0.68, p < 0.05; Fig. 2,

271

Table S3). N was primarily delivered as organic N (71% of TN on average across sites),

272

which was even more strongly influenced by street canopy (r = 0.71, p < 0.001). Street

273

canopy, along with total impervious area (TIA) and residential area, were the dominant

274

influences on TN when all variables were considered (Adj. R2 = 0.69; Table 3). TON was

275

most closely associated with street canopy (present in all 3 models within 2.0 AICc of the

276

best model; Table 3), which along with residential area comprised the best model by

277

AICc (Adj. R2 = 0.55).

278 279

Concentrations of N and P in gutter runoff in the AHUG watershed showed strong

280

positive (but seasonally variable) relationships with street canopy (Fig. 3), confirming the

281

influence of street canopy on nutrient concentrations observed at the watershed scale

282

(Table S3). In fall, the influence of street canopy on stormwater N and P concentration

283

was especially strong (r = 0.95, p < 0.001 for TP; r = 0.96, p < 0.001 for SRP; r = 0.77, p

13 ACS Paragon Plus Environment

Environmental Science & Technology

284

< 0.05 for TN; r = 0.81, p < 0.05 for TON). For TP and SRP, the relationship was not

285

significant in late spring (leaf-out); however, TN and TON were positively related to

286

street canopy during this period (r = 0.75, p < 0.05 for TN; r = 0.73, p < 0.05 for TON).

287 288

Seasonal patterns in stormwater P and N concentrations at the watershed scale further

289

indicated trees and vegetation as major sources of nutrients to stormwater. These seasonal

290

patterns mirrored the phenology of urban vegetation with seasonal peaks in means of P

291

and N that coincided with autumn leaf drop and with spring leaf-out and flowering (Fig.

292

S1), and were strongly related to presence of street trees in the study watersheds. For

293

example, elevated spring TP and TN concentrations across sites (characterized by mean

294

May-minus-September difference in concentration) were significantly related to street

295

canopy (R2 = 0.38, p < 0.05 for TP; R2 = 0.26, p < 0.05 for TN). Less variable and

296

decreasing concentrations of TP and TN over summer (Fig. S1) are consistent with

297

establishment of lawns and trees during the growing season, accompanied by low rates of

298

litterfall34. The subsequent increase of mean event TP and TDP concentrations from

299

September to October were significantly correlated with street canopy (R2 = 0.30, p