Water table fluctuation from green
2
infrastructure sidewalk planters in
3
Philadelphia, Pennsylvania
n
1
io
4 Min-Cheng Tu1, Ph.D.
6
Robert Traver2, Ph.D., P.E., D. WRE
7
1
8
University, Villanova, PA 19085 (corresponding author). Email:
[email protected]
9
2
ve
pt e
d
Postdoctoral Research Fellow, Department of Civil and Environmental Engineering, Villanova
ce
Professor, Department of Civil and Environmental Engineering, Villanova University, Villanova, PA
19085. Email:
[email protected]
Ac
10
rs
5
11
12
Abstract
13
Popularity of infiltration-based green infrastructure (GI) has spurred the concern of rising groundwater
14
tables and the potential detrimental effect on building foundations. This study examined water table
15
fluctuations adjacent to green infrastructure sidewalk planters from two winters (2014-2015 and 2016-
16
2017) in Philadelphia. Groundwater mounding was observed in the latter period but not the former.
17
Due to the proximity to a park, the water table rise is the combined effect from both the GI and the
park. For the first winter, the sidewalk planters were not fully vegetated or maintained. It was
19
hypothesized that the increased groundwater mounding in 2016-2017 was from the increased
20
infiltration rate caused by improved vegetation and maintenance. Groundwater mounding was not
21
observed beyond 3 meters from the GI except for large storms. Because the water table rise was
22
transient, groundwater mounding had minimal impact beyond 3 meters for current GI configurations
23
and soil conditions. The observations conformed with prior computer simulations. The results also
24
showed that intra-season water table fluctuations were far greater than those created by local
25
infiltration.
io
n
18
rs
26
Key words
28
Green infrastructure; Groundwater; Infiltration; Mounding; Philadelphia; Sidewalk planter; Stormwater;
29
Storm water; Vegetation; Water table
Ac
ce
pt e
d
ve
27
Introduction
31
Green infrastructure (GI) provides many benefits to urbanized areas by reducing combined sewer
32
overflow (Urbonas and Stahre 1993), reducing stormwater runoff, improving water quality, and/or
33
providing better ecosystems and human health (Tzoulas et al. 2007). Such benefits are important as
34
urbanization has become a global trend (Tu and Traver 2018). The objective of GI practices is to mimic
35
natural systems (e.g. vegetation, wetlands, open space) in urban areas to mitigate the impact of
36
impervious surfaces or compacted soil in urbanized areas to sensitive subjects such as the receiving
37
water bodies (Tu and Smith 2018; Tu et al. 2018).
38
Since one major benefit provided by GI practices is to promote water infiltration as the means to reduce
39
stormwater runoff, groundwater can be impacted. The balance between reduction (from lower
40
infiltration due to urbanization) and replenishment of groundwater is a topic that requires more
41
research to support a balanced perspective. In arid climates, increased recharge rates are encouraged,
42
and have shown not to impact groundwater quality (Stephens et al. 2012). However, the potential
43
impact of localized water table rise to subsurface infrastructure in humid climates is a concern (Endreny
44
and Collins 2009). Understanding the impact from GI infiltration is crucial in Philadelphia, Pennsylvania,
45
as an ambitious plan to “green” 40% of the city’s impervious area was initialized in 2011 and will
46
continue for the next 25 years (Maimone et al. 2011; Philadelphia Water Department [PWD] 2017).
47
Utilizing GI practices in such a grand scale has not been attempted before (Civic Federation 2007) except
48
in Boston to combat the declining groundwater table (Thomas and Vogel 2012), where infiltration from
49
GI practices is found to have a small but confirmed effect in raising the groundwater table depleted by
50
urbanization.
51
Maimone et al. (2011) explored the impact of GI installation in the local (block) scale and city-wide scale
52
through computer simulations. They determined that groundwater mounding caused by local water
Ac
ce
pt e
d
ve
rs
io
n
30
infiltration through GI practices drops off quickly a few meters away from the GI and dissipates over
54
several days. This study was performed to complement and validate the local-scale simulation results
55
and provide recommendations to improve the city-wide scale simulation results of Maimone et al.
56
(2011) by providing an analysis of data collected from groundwater wells adjacent to a GI in
57
northeastern Philadelphia.
58
Research Site and Instrumentation
59
The GI under investigation is built on the sidewalk at the northeastern side of Roosevelt Playground (on
60
Hellerman St., between Cottage St. and Walker St.) in Philadelphia, Pennsylvania. Philadelphia is in the
61
humid subtropical climate region according to the Koppen-Geiger climate classification system (Peel et
62
al. 2007) with an average annual precipitation of 1,054 mm (41.50 inches) from 1981 to 2010 (NOAA
63
2017). The monthly precipitation ranges from 66 mm (2.60 inches) in February to 109 mm (4.29 inches)
64
in July. As part of the annual precipitation, the average annual snowfall is 584 mm (22.99 inches), or
65
58.4 mm (2.30 inches) equivalent liquid water depth, typically occurring from December to April with a
66
peak in February.
67
Bore logs showed that the upper 1.52 meters (60 inches) of native soil was composed of silty sand with
68
brick and gravel (urban fill). It was assumed there was no significant variation in hydraulic properties for
69
deep native soil (> 1.52 meter) in the vicinity due to the lack of deep urban soil data. The GI is
70
comprised of four planters numbered #1 to #4 from northwest to southeast, as Fig. 1 and Fig. 2 show.
71
Fig. 3 shows the schematic cross-section view of planter #1. Each planter receives runoff from the road
72
surface from two curb inlets. The planters also receive runoff from the sidewalk through cuts on the
73
planter walls (not visible in the photo).
Ac
ce
pt e
d
ve
rs
io
n
53
74 Fig. 1. Green infrastructure sidewalk planters under investigation (photo date: December 6, 2016)
76
The dimensions and design of the green infrastructure sidewalk planters are provided in Fig. 2 and Fig. 3.
77
The top and bottom drawings of Fig. 2 display the plan view and the longitudinal section views of the
78
design, respectively. Design of planters #1 and #2 are mostly mirrored from that of planters #3 and #4.
79
Each planter has a soil media layer with a depth of 0.61 meter (24 inches) and sits over a rock infiltration
80
bed. Because the sidewalk is sloped, all four planters sit at different elevations with planter #2 the
81
lowest and planter #4 the highest, as shown in Fig. 2. In each of the planters, there is also a domed riser
82
overflow pipe delivering water directly to the infiltration bed if the planter is full. The rim of the domed
83
riser is 0.05 meters (2 inches) above the planter soil for all planters, thus creating an extra storage space
84
before overflowing. Planters #1 and #2 sit in the same infiltration bed, and planters #3 and #4 share
85
another bed. The depth of the rock infiltration beds is about 1 meter (39 inches) as Fig. 2 shows (depth
86
varies because the road/curb surface is not level) and both have the same bottom elevation. A
87
perforated underdrain pipe in the infiltration bed connects planters #1 and #2 (0.5% slope with the
88
lower end in planter #1), and another connects planters #3 and #4 (0.5% slope with the lower end in
89
planter #4). There is another underdrain pipe (0% slope, not perforated) connecting the first set of
90
planters (planters #1 and #2) to the second set of planters (planters #3 and #4). The invert elevation of
91
the zero-slope underdrain pipe joins the invert of the higher end of the perforated pipe in planters #1
92
and #2, but joins the lower end of the perforated pipe in planters #3 and #4. Perforation specifications
93
are not given. Overflow in the planters enters the infiltration beds via the perforated underdrain pipes.
94
None of the underdrain pipes discharge to the combined sewer.
95
The first set of planters and the second set of planters receive street runoff from both directions along
96
the Hellerman Street. PWD estimated the combined drainage area for planters #1 and #2 at 523 square
Ac
ce
pt e
d
ve
rs
io
n
75
97
meters (5,630 square feet), and 536 square meters (5,769 square feet) for planters #3 and #4. If the
98
planters are at full capacity and cannot receive additional runoff, bypassed runoff enters the combined
99
sewer inlet in the middle (between planters #2 and #3) of the GI.
100 101
Fig. 2. Plan view (top) and cross-section view (bottom) of the green infrastructure sidewalk planters
102 Fig. 3. Schematic side view of planter #1
104
In each planter, water depth above soil and above the overflow pipe was collected by HOBO pressure
105
transducers (with built-in loggers, Onset 2018), and soil moisture was collected by Stevens Hydraprobe
106
soil-water senors (Stevens 2018) at a single point with different depths. At planters #1, #2, and #4, soil
107
moisture at 10-cm (3.94 inches) depth was collected. At planter #3, soil moisture at both 10-cm (3.94
108
inches) and 35-cm (13.78 inches) depths was collected. A custom-designed and lab-tested orifice insert
109
was installed on the overflow pipe to facilitate the conversion from water depth above the overflow
110
pipe to flow rate. Water depth In the observation wells and groundwater wells was collected by HOBO
111
pressure transducers (Onset 2018). A weather station comprising a Campbell Scientific CS215 weather
112
probe, a Campbell Scientific LI200X pyranometer, a Campbell Scientific TE525 rain gage (Campbell
113
Scientific 2018), and a Vaisala WXT520 weather probe (Vaisala 2018) was installed on site, collecting air
114
temperature, atmospheric pressure, rainfall depth, relative humidity, solar radiation, wind direction, and
115
wind velocity. Campbell CR800 data loggers (Campbell Scientific 2018) were used to record
116
meteorological data at the weather station and soil moisture data at planters #1 and #3. At planters #2
117
and #4, a low-cost logger utilizing Arduino (Arduino 2018) and Raspberry Pi (Raspberry Pi 2018) were
118
installed to record soil moisture data. Meteorological data (including rainfall) collection started in May
119
2016, and rainfall data collected by PWD’s #17 rain gage (1.8 km, or 1.1 mile, west-northwest from site)
Ac
ce
pt e
d
ve
rs
io
n
103
was used in this study for analyses before that date. Data from the PWD rain gage showed a slightly
121
different temporal rainfall distribution, but still retained very good correlation with the on-site rainfall
122
data. During winter 2016-2017, the PWD data and on-site data had Pearson pair-wise correlation
123
coefficients of 0.95 and 0.85 for storm rainfall depth and storm peak rainfall intensity, respectively.
124
To test the performance of the GI in a controlled environment, a simulated runoff test (SRT) was
125
performed at this site by PWD on November 2, 2017 (Fig. 4). Runoff was provided from a street hydrant
126
and the flow rate was throttled and monitored by a Sensus flow meter (White et al. 2016). Fig. 4 shows
127
a SRT for a single planter (Planter #3) with two curb inlets. The flow meter sat on the grate of the
128
upstream curb inlet.
rs
io
n
120
ve
129
Fig. 4. Physical setup of a SRT on the site with flow directions marked (photo date: November 2, 2017)
131
Adjacent to the GI site, PWD has installed and been monitoring three groundwater observation wells,
132
numbered GW1, GW2, and GW3, as Fig. 2 and Fig. 5 show. GW1 is spaced 1.53 meters (60 inches) from
133
the edge of the infiltration bed of planters #1 and #2. The distance between GW1 and GW2 and
134
between GW2 and GW3 are both 1.53 meters (60 inches) as Fig. 2 shows.
Ac
ce
pt e
d
130
135 136
Fig. 5. The three groundwater observation wells and their relative locations from the GI in the field (photo date: December 6,
137
2016).
138
Summary of Available Data
139
The three groundwater wells provided data dated back to July 2014. The periods of record that all three
140
groundwater wells had data is 7/17/2014 - 3/31/2015 and 10/1/2016 - 3/31/2017. The common winter
141
period (October to March) was selected to match seasonality. For the first winter period in 2014-2015,
the sidewalk planters were not well vegetated or maintained as the site had not been transferred to the
143
city from the contractor. After this period, and prior to the 2016-2017 winter, the planters had the
144
upper soil layer replaced and were replanted (due to continuous poor vegetation conditions) in 9/2015
145
and underwent a year of vegetation growth with intensive care from landscapers.
146
GW1 water levels and rainfall depth above 6.35 mm (0.25 inch) free from snow accumulation are
147
plotted in Fig. 6 to demonstrate the trend of groundwater table fluctuation. On several occasions, the
148
data fluctuated rapidly unrelated to rainfall events, as pointed out by red arrows in Fig. 6. Note that
149
“distance to water table” was inverted to intuitively represent the location of water table. Since all
150
fluctuations were observed by all three groundwater wells with similar curve shapes and magnitude,
151
they were unlikely equipment errors. The source of these fluctuations was unclear. Events affected by
152
these fluctuations were excluded from analyses of this study, and a total of 35 events were analyzed as
153
summarized by Table 1.
d
ve
rs
io
n
142
157
ce
156
Fig. 6. Available rainfall and groundwater data in winter 2014-2015
Ac
155
pt e
154
Table 1. Summary of storm events included in the analyses
158 159
Two observations can be drawn from Fig. 6:
160
Firstly, the water table appeared to decline from 2014-2015 to 2016-2017 probably because 2016 was a
161
dry year with an annual precipitation only approximately 75% of that of 2014 and 2015.
162
Secondly, there were significant intra-season variations of the groundwater table. Such variations were
163
more significant in 2014-2015 than 2016-2017 because 2014-2015 had more rainfall. In winter of 2014-
2015, such variations were as large as 0.6 meter (23.6 inches). The overall intra-season trend was rising,
165
which conformed to the observation from other studies (Healy and Cook 2002): the groundwater table
166
rose during the winter due to higher recharge due to less evapotranspiration.
167
Data Analyses
168
The groundwater table rise across the three groundwater wells was analyzed for storms with rainfall
169
depth above 6.35 mm (0.25 inch). The water table rise from a storm was defined as the difference of
170
the maximum and minimum levels from storm initialization to 12 hours after the storm ends. In general,
171
the change of groundwater elevation reached its peak in 12 hours. Fig. 7 shows the response of the
172
water table for a 72-hour period with both a major (29.7 mm, or 1.17 inch) and a minor (16.5 mm, 0.65
173
inch) event in 11/29/2016 - 12/1/2016.
ve
rs
io
n
164
d
174
Fig. 7. Water table rise at all observation wells and corresponding hourly rainfall intensity on 11/29/2016-12/1/2016
176
The water table rise at all three groundwater wells showed strong Pearson pair-wise correlation to
177
rainfall depth, but weak correlation to either mean rainfall intensity or peak rainfall intensity (Table 2).
178
Fig. 8 graphically shows such correlation for GW2 (3 meters, or 118 inches, from the GI) with a linear
179
regression line added in the subplot of storm depth. Since the water table rise was closely correlated to
180
rainfall depth, rainfall depth was used in following analyses of this study.
181
All rainfall collected by the drainage area of the GI during the time frame was expected to be received by
182
the GI and infiltrated into groundwater (i.e. no loss to the combined sewer inlet) because the highest
183
peak rainfall intensity (11.94 mm/hr, or 0.47 inch/hr) in the same time frame equaled a surface runoff
184
rate of 1.83 liter/sec (0.065 CFS), which was much less than the maximum flow rate (18.33 liter/sec, or
Ac
ce
pt e
175
185
0.647 CFS) that one set of the GI planters can handle before overflowing as measured by the SRT on
186
November 2, 2017.
187
Table 2. Correlation between water table rise and storm attributes
188 Fig. 8. Comparison of water table rise at GW2 with storm attributes from all storms with a linear regression line added for the
190
subplot of storm depth
191
A closer examination of the data revealed several intriguing points:
192
Firstly, the great majority of storms generated a water table rise less than about 0.06 m (2.36 inches) at
193
a distance of 3 m (or 118 inches, the distance between GW2 and edge of the GI) from the GI. The
194
regression line in Fig. 8 shows that water table rise is strongly related to storm depth.
195
Secondly, the responding characteristics of groundwater mounding from storms were possibly different
196
for the two periods. For winter 2014-2015, the water table rise was fairly uniform among all three
197
groundwater wells, but groundwater mounding near the GI (i.e. GW1 > GW2 in water table rise) was
198
statistically significant in winter 2016-2017. In Table 3, water table rise from all three groundwater wells
199
were tested by one-way repeated measures ANOVA first to detect whether at least one well had
200
different characteristics in water table rise. The one-way repeated measured ANOVA was adopted
201
because all measurements were done on the same subjects (i.e. the three groundwater wells) for
202
different conditions (i.e. storms). If the one-way repeated measures ANOVA reported p < 0.05, then a
203
paired Student’s t test was used to compare groundwater well pairs. Note that despite the water table
204
rise at GW1 was statistically higher than that at GW2 in winter 2016-2017, the water table rise of GW2
205
and GW3 was mostly the same at the same time.
Ac
ce
pt e
d
ve
rs
io
n
189
In Table 3, the p-value of one-way repeated measures ANOVA for winter 2014-2015 was very close to
207
0.05 thus had weak statistical power; therefore, Type II error (i.e. false negative) was possible, which
208
means some kind of systemic difference in water elevation among the three groundwater wells in
209
winter 2014-2015 existed but the data cannot detect it. Further studies might be required to strengthen
210
the claim on groundwater mounding characteristics found in winter 2014-2015. Even though
211
groundwater mounding might have existed in both periods, the magnitude of mounding (by comparing
212
the water table rise at GW1 and GW2) near the GI was still much more significant in winter 2016-2017
213
than in winter 2014-2015, clearly indicating an increase in groundwater mounding responses in the
214
latter period.
215
Table 3. Summary of event water table rise from two different time periods
ve
rs
io
n
206
216
Third, groundwater mounding in winter 2016-2017 was further examined by Fig. 9 by two indices: the
218
difference of “distance to water table” between GW1 (1.5 meters, or 59 inches from the GI) and GW2
219
(3.1 meters, or 122 inches from the GI), and between GW2 and GW3 (4.6 meters, or 181 inches from the
220
GI). Positive values indicate GW1 elevation greater than GW2, or GW2 elevation greater than GW3. Fig.
221
9 shows that groundwater mounding between GW1 and GW2 was prevalent (albeit small) for all storm
222
sizes, but mounding did not spatially extend beyond GW2 for storms less than about 30 mm (1.18 inch).
Ac
ce
pt e
d
217
223 224
Fig. 9. Comparison between height of groundwater mounding and storm rainfall depth based on data from winter 2016-2017
225
Fourth, the water table rise receded quickly after storms. Fig. 10 displays the hourly rainfall depth and
226
water table fluctuation at all three groundwater wells for the storm of 1/18/2015, which was the largest
227
storm analyzed in this study. After the majority the storm passed, the water table receded in
228
approximately 6 hours. A small groundwater mounding is visible in Fig. 10 by comparing the difference
229
in peak water table rise at the three groundwater wells. The data did not show recession of the water
230
table to its original elevation because of the existence of a long-term water table rise either due to a
231
regional groundwater movement or localized infiltration from the nearby park land. Note that the peak
232
of the long-term water table rise was higher than the peak of the short-term change of water table
233
sustained by the GI.
234 Fig. 10. Hourly rainfall depth and groundwater table fluctuation for the storm of 1/18/2015 (total rainfall = 49.5 mm).
236
Discussion
237
Comparison with simulation results
238
The observations from this study was compared to results of the former computer simulation study
239
(Maimone et al. 2011), which modeled water table rise and groundwater mounding due to GI practices
240
in Philadelphia. Maimone et al. (2011) found that most of the simulated water table rise hovers
241
between 0.05-0.1 m (1.97-3.94 inches) with a maximum of 0.15-0.2 m (5.91-7.87 inches) at a distance of
242
3 meters (118 inches) from GI practices, which was confirmed by the observations (Fig. 8) in this study.
243
Machusick et al. (2011) also had similar observations from studying a stormwater infiltration rain garden
244
on the campus of Villanova University (approximately 18 km, or 11 miles, from the site).
245
The highest simulated water table rise was about 0.17-0.28 meter (6.69-11.02 inches, depending on
246
location of the GI of interest) at a distance 1.52 meters (60 inches, equivalent to the distance between
247
GW1 and the GI in this study) from a simulated tree trench (assuming silty sand as the native soil) based
248
on a very large storm of 70.7 mm (2.78 inches) over 45 hours (personal communication with PWD).
249
Although the highest observed water table rise at GW1 in this study was 0.13 meter (5.12 inches, Table
250
3), the associated storm rainfall depth was only 49.5 mm (1.95 inch). Assuming the water table rise was
Ac
ce
pt e
d
ve
rs
io
n
235
proportional to the rainfall depth (Bouwer et al. 1999), the maximum water table rise observed in this
252
study was in the range predicted by Maimone et al. (2011)
253
Trends of water table rise among groundwater wells
254
The different characteristics of water table rise among groundwater wells in the winters of 2014-2015
255
and 2016-2017 can be explained by the principle of superposition, which is applicable to most
256
groundwater hydraulic problems (Reilly et al. 1984). The water table rise near the GI was controlled by
257
two sources: infiltration from the GI, and infiltration from the nearby park space and surrounding areas.
258
The GI is on the northeastern side of the groundwater observation wells, but a baseball field and open
259
space covered with grass which provided significant infiltration to the local water table is located
260
immediately adjacent to the wells to the south, as shown in Fig. 11. In Fig. 11, planters #3 and #4 are
261
covered by tree canopies and are not visible. It is known that infiltrated water creates groundwater
262
mounding (Bouwer et al. 1999), thus the uniform water table rise in winter of 2014-2015 near the GI
263
implied that the GI and the pervious area of the park had about the same influence on rising the local
264
water table. For winter of 2016-2017, the observed localized groundwater mounding indicated that the
265
GI had a higher influence than the park did. There was no evidence that the park open space had any
266
change in soil or vegetation conditions throughout 2014-2017.
io
rs
ve
d
pt e
ce
Ac
267
n
251
268
Fig. 11. Aerial view of planters and the adjacent park pervious area with GI drainage area covered by blue shades (Google 2018)
269
Explanation to change in infiltration of GI planters
270
Even though 2016 was a dry year, there was no statistical difference (p=0.19) in mean event rainfall
271
depth between winter 2014-2015 and winter 2016-2017. Winter 2016-2017 had lower cumulative
272
rainfall depth probably because it had fewer storms (Table 1). Therefore, the most probable
explanation to the GI’s increased infiltration was the influence of vegetation and continued maintenance
274
of the GI planters.
275
The GI was built in 2014, but vegetation growth was poor in the first two years. The planters were
276
replanted in September 2015 followed by continuous nurturing by landscapers until summer 2016
277
(personal communications with PWD). Fig. 12 shows the condition of plants in April 2014 (left) and April
278
2017 (right) for planter #3. Vegetation growth was significantly better in 2017. A better vegetation
279
condition is known to have positive effects on soil hydraulic conductivity (Lucas and Greenway 2011).
280
Therefore, it is concluded that the better vegetative condition from replanting/nurturing and better
281
maintenance from PWD in 2016-2017 caused a higher water infiltration rate in the GI, which was
282
responsible for more efficient contribution to the water table, and thus more prominent groundwater
283
mounding.
ve
rs
io
n
273
pt e
d
284
Fig. 12. Comparison of vegetation conditions in 2014 and 2017
286
Conclusion
287
This study examined the fluctuation of the water table in two winters (2014-2015 and 2016-2017)
288
adjacent to GI sidewalk planters in Philadelphia, Pennsylvania. Due to the proximity to a park (Roosevelt
289
Playground), water table rise was the combined effect from the GI and the park. Water table rise from
290
storms was spatially uniform in winter of 2014-2015, which indicated the similarity in contribution to
291
water table rise from the GI and the park pervious area. However, the recharge rate to groundwater
292
caused by the GI became higher than that near the park pervious area in winter of 2016-2017, as the
293
height of groundwater mounding under the GI became statistically significant. It was hypothesized in
294
this study that the improved vegetation conditions and maintenance in 2016-2017 promoted higher
Ac
ce
285
infiltration rates and thus more efficient contribution to the local water table. It was evident that
296
proper maintenance can keep GI practices at their optimal performance.
297
Because the water table rise in this study was the combined effect of both the GI and the park pervious
298
area, it is difficult to isolate the absolute effect of the GI on water table rise at the foundation of
299
adjacent houses. However, the data showed that groundwater mounding was spatially limited, and no
300
mounding was observed 3 meters (118 inches) away from the GI (Fig. 9) except for very large storms. In
301
addition, the rise of water table receded quickly after storms (Fig. 10). Therefore, the effect of
302
groundwater mounding caused by GI practices was very limited in space and time, even in the scenario
303
with higher groundwater recharge rates in 2016-2017. By placing GI practices at least 3 meters (118
304
inches) from houses, the impact on foundations caused by groundwater mounding should be minimal.
305
The data showed significant intra-season variation of water table as high as 0.6 m (23.62 inches), which
306
was far greater than the water table rise caused by local infiltration, whether by GI practices or by
307
adjacent green space. Maimone et al. (2011) studied the effect of city-wide installation of GI practices in
308
Philadelphia by computer modeling, and concluded that a city-wide installation of GI practices can raise
309
groundwater table up to 1.5 meters (59 inches) in certain areas. However, the seasonal (intra-season
310
and/or inter-season) variations of water table were not considered by Maimone et al. in predicting a
311
significant water table rise. The compound effect of these two factors (city-wide GI installation and
312
seasonal variation) and the localized increasing of groundwater recharging rate from higher water table
313
have not been studied in an urban environment. Although such rise of the groundwater table might be
314
a sign of restoration of the urban water cycle to the natural state (Vazquez-Sune et al. 2004), the
315
elevated groundwater table mounding might still be higher than expected adjacent to GI practices in
316
those areas. The elevated mounding would unlikely cause a direct impact to house foundations as the
317
current groundwater table is approximately 4 meters (157 inches) below ground (according to data
318
collected in this study), and the mounding was found to be transient and localized to within 3 meters
Ac
ce
pt e
d
ve
rs
io
n
295
(118 inches) from GI practices. This calls for improved models to simulate such large-scale installations.
320
With improved models, it may be possible to optimize placement of GI practices to minimize the impact
321
caused by a rising water table in those areas.
322
Acknowledgement
323
The Philadelphia Water Department provided data support, access to the site, assistance with installing
324
instrumentation on site, and manpower and equipment during the SRT. This study would not have been
325
possible without this assistance. The contribution from Philadelphia Water Department, particularly Mr.
326
Stephen White and Mr. Chris Bergerson, is noted and highly appreciated.
327
This publication was developed under Assistance Agreement No. 83555601 awarded by the U.S.
328
Environmental Protection Agency to Villanova University. It has not been formally reviewed by EPA.
329
The views expressed in this document are solely those of Villanova University and do not necessarily
330
reflect those of the Agency. EPA does not endorse any products or commercial services mentioned in
331
this publication.
332
References
333
Arduino. (2018). “What is Arduino?” (August 21,
334
2018).
335
Bouwer, H., Back, J.T., and Oliver, J.M. (1999). “Predicting infiltration and ground-water mounds for
336
artificial recharge.” J. Hydrol Eng., 4(4), 350-357.
337
Campbell Scientific. (2018). “Campbell Scientific.” (June 15, 2018).
338
Civic Federation. (2007). “Managing urban stormwater with green infrastructure: case studies of five
339
U.S. local governments.”
Ac
ce
pt e
d
ve
rs
io
n
319
(December 14, 2017).
342
Endreny, T., and Collins, V. (2008). “Implications of bioretention basin spatial arrangements on
343
stormwater recharge and groundwater mounding.” Ecol. Eng., 35, 670-677.
344
Google. (2018). “Google Earth.” (June 18, 2018).
345
Healy, R.W., and Cook, P.G. (2002). “Using groundwater levels to estimate recharge.” Hydrogeology J.,
346
10, 91-109.
347
Lucas, W.C., and Greenway, M. (2011). “Hydraulic response and nitrogen retention in bioretention
348
mesocosms with regulated outlets: Part I – hydraulic response.” Water Environ. Res., 83, 692-702.
349
Machusick, M., Welker, A., and Traver, R. (2011). “Groundwater mounding at a storm-water infiltration
350
BMP.” J. Irrig. Drain. Eng., 137(3), 154-160.
351
Maimone, M., O’Rourke, D.E., Knighton, J.O., and Thomas, C.P. (2011). “Potential impacts of extensive
352
stormwater infiltration in Philadelphia.” Environmental Engineer: Applied Research & Practice., 14, 1-3.
353
National Oceanic and Atmospheric Administration. (2017). “Climate Data Online”.
354
(December 14, 2017).
355
Onset. (2018). “Onset.” (June 15, 2018).
356
Peel, M.C., Finlayson, B.L., and McMahon, T.A. (2007). “Updated world map of the Koppen-Geiger
357
climate classification.” Hydrology and Earth System Sciences, 11(5), 1633-1644.
358
Philadelphia Water Department. (2017). “Green City, Clean Waters.”
359
(December 14, 2017).
Ac
ce
pt e
d
ve
rs
io
n
340
Raspberry Pi. (2018). “Raspberry Pi.” (August 21, 2018).
362
Reilly, T.E., Franke, O.L., and Bennett, G.D. (1984). The principle of superposition and its application in
363
ground-water hydraulics. United States Geological Survey, Reston, Virginia, USA.
364
Stephens, D.B., Miller, M., Moore, S.J., Umstot, T., and Salvato, D.J. (2012). “Decentralized groundwater
365
recharge systems using roofwater and stormwater runoff.” J. Am. Water Resour. As., 48(1), 134-144.
366
Stevens. (2018). “Stevens.” (June 15, 2018).
367
Thomas, B.F. and Vogel, R.M. “Impact of storm water recharge practices on Boston groundwater
368
elevations.” J. Hydrol. Eng., 17(8), 923-932.
369
Tu, M.-C. and Smith, P. (2018). “Modelling pollutant buildup and washoff parameters for SWMM based
370
on land use in a semiarid urban watershed.” Water Air Soil Poll., 229: 121. doi:10.1007/s11270-018-
371
3777-2.
372
Tu, M.-C., Smith, P., and Filippi, A.M. (2018). “Hybrid forward-selection method-based water-quality
373
estimation via combining Landsat TM, ETM+, and OLI/TIRS images and ancillary environmental data.”
374
PLoS ONE, 13(7): e0201255. doi: 10.1371/journal.pone.0201255.
375
Tu, M.-C. and Traver, R.G. (2018). “Clogging impacts on distribution pipe delivery of street runoff to an
376
infiltration Bed.” Water, 10(8): 1045. doi: 10.3390/w10081045.
377
Tzoulas, K., Korpela, K., Venn, S., Yli-Pelkonen, V., Kazmierczak, A., Niemela, J., and James, P. (2007).
378
“Promoting ecosystem and human health in urban areas using Green Infrastructure: A literature
379
review.” Landscape Urban Plan., 81(3), 167-178.
380
Urbonas, B. and Stahre, P. (1993). Stormwater Best Management Practices and Detention for Water
381
Quality, Drainage, and CSO Management. PTR Prentice-Hall: Englewood Cliffs, NJ.
Ac
ce
pt e
d
ve
rs
io
n
361
Vaisala. (2018). “Vaisala.” (June 15, 2018).
383
Vazquez-Sune, E., Sanchez-Vila, X., Carrera, J. (2004). “Introductory review of specific factors influencing
384
urban groundwater, an emerging branch of hydrogeology, with reference to Barcelona, Spain.”
385
Hydrogeology Journal, 13, 522-533.
386
White, S., Krechmer, T., Heffernan, T., Manna, N., Mannarino, E., Bergerson, C., Olson, M., and Cruz, J.
387
(2016). “Green infrastructure performance model in the real world: Modeling natural and simulated
388
runoff events.” Proc., International Low Impact Development Conference 2016, Portland, ME, 163-172.
n
382
Ac
ce
pt e
d
ve
rs
io
389
Table 1. Summary of storm events included in the analyses
Mean 2014-2015
Count
rainfall depth (mm, in)
Mean
Mean
event
peak
intensity
intensity1
(mm/hr,
(mm/hr,
in/hr)
in/hr)
Mean
Mean
event
peak
intensity
intensity1
(mm/hr,
(mm/hr,
in/hr)
in/hr)
Mean 2016-2017
Count
rainfall depth (mm, in)
5
15.0 (0.59)
1.7 (0.067)
5.1 (0.201)
October
3
14.7 (0.58)
1.9 (0.075)
4.8 (0.189)
November
5
20.9 (0.82)
1.3 (0.051)
5.1 (0.201)
November
3
17.7 (0.70)
1.1 (0.043)
5.9 (0.232)
December
3
21.8 (0.86)
1.5 (0.059)
4.3 (0.169)
December
4
14.1 (0.56)
1.7 (0.067)
4.0 (0.157)
January
3
30.4 (1.20)
2.3 (0.091)
5.8 (0.228)
January
2
17.8 (0.70)
1.1 (0.043)
6.2 (0.244)
February
0
-
-
-
February
2
9.3 (0.37)
2.1 (0.083)
5.3 (0.209)
March
3
17.6 (0.69)
1.4 (0.055)
3.6 (0.142)
March
Overall
19
20.5 (0.81)
1.6 (0.063)
4.8 (0.189)
5-min. peak intensity in mm/hr and in./hr.
pt e
391
Ac
ce
392
io
rs 2
24.5 (0.96)
1.2 (0.047)
5.2 (0.205)
16.0 (0.63)
1.5 (0.059)
5.1 (0.201)
Overall
d
1
n
October
ve
390
16
393
Table 2. Correlation between water table rise and storm attributes
Event rainfall depth
Mean intensity
Peak intensity
Water table rise @ GW1
0.82
0.38
0.41
Water table rise @ GW2
0.80
0.31
0.36
Water table rise @ GW3
0.78
0.34
0.35
394
Ac
ce
pt e
d
ve
rs
io
n
395
GW2
GW3 1
Std. dev.
mean rise
of rise (m,
(m, inch)
inch)
0.030, 1.18
0.027, 1.06
0.032, 1.26
0.029, 2.54
Max rise
(m, inch)
p1
ce
GW1
Event
pt e
Winter of 2014-2015 (n=19)
d
ve
rs
io
n
Table 3. Summary of event water table rise from two different time periods
Ac
396
0.024, 0.94
0.022, 0.87
0.125, 4.92 0.108, 4.25 0.101, 3.98
0.06
Winter of 2016-2017 (n=16)
Event
Std. dev.
mean rise
of rise (m,
(m, inch)
inch)
0.026, 1.02
0.016, 0.63
0.018, 0.71
0.018, 0.71
One-way repeated measures ANOVA; 2 Paired Student’s t test
0.014, 0.55
0.012, 0.47
Max rise (m, inch)
p1
0.062,