Paradise Pond Dredging Project - 1st Year_REPORT_2.11.16.pdfPage 1 of 5
PARADISE POND SEDIMENT MANAGEMENT PROTOCOL
Northampton, MA
1st Year Activities Report - 1/21/16
Page 2 of 5
Paradise Pond Sluicing Project – 1st year activities report
Summary: Page 2
Sluicing experience: Page 3 & 4
Sluicing events ‐ Date, gate position, time and flow conditions: Page 4
Conclusion: Page 5
Sediment Redistribution Proposal: Page 5
Baseline monitoring results: Geology Appendix A (12 pages)
Appendix B (34 pages)
Biology Appendix C (19 pages)
Summary: Much data has been collected regarding precipitation events, river flow and turbidity. One significant
storm event during the summer of 2015 was recorded and studied, leading to some interesting conclusions. (see
Appendix A).
Bed load Sediments have been sampled in Paradise Pond and Hulberts Pond, and analyzed for extractable
metals, total Mercury and Grain size. Four downstream reference stations have been established and
crossectional river profiles have been recorded at each reference station. Pond bathymetry has been recorded
and moving bed sediment has been detected at flows above 360 cfs. (see Appendix B).
The River has been tested for “stream health” by evaluating diversity and abundance of invertebrate
communities using several sampling methods at 7 sites in the Mill River. (see Appendix C).
Page 3 of 5
Sluicing Experience: We know from past projects that the river will cut and mobilize bed load sediment under
low flow conditions within the Pond, under full drawdown conditions (16’) and transport the sediment
downstream where it falls out. (see image in the SMP, page 3, from the 1998 full draw down dredging project).
Similarly, we know the same occurs during partial drawdown conditions (of 5’ to 8’ below normal pool) with a
difference ‐ much of the bed load sediment falls out within the Pond, just upstream of the sluice (a.k.a. low
water outlet) with some falling out in the splash pool below the spillway.
River channel in Pond 10/16/12 River channel in Pond 10/17/12
Referencing the two photos above, note significant cutting through bed load sediments in just 24 hours. Most of
the bed load sediments mobilized in the 2012 project fell out of the water column before leaving the Pond in the
channel upstream of the gate. See page 5 of appendix A.
Referencing the two photos below, some bed load sediment did pass through the sluice and accumulated in the
splash pool over the course of the 10 week 2012 partial drawdown. Very little was transported beyond the
splash pool during low flow conditions. The sediment stayed in the splash pool until river flows increased.
(splash pool volume is 1,800 cy).
Sediment sandbar in splash pool, Dec. 2012. Accumulated sediment in splash pool, Dec. 2012.
Page 4 of 5
During the summer of 2015 during the Phase 2 dam repair project we operated the gate many times to control
flow while dam toe repairs were in construction. Most of the time the gate was only partially open, but on 4
occasions the gate was opened to 100%, the 4th time was the day after the project site had been cleared or
equipment (dam toe repair complete).
Sluicing Events:
Date Gate position Time Flow
7/7 Gate open 100% 6:30 AM to 9:10 AM (closed to 25% open). 45 cfs
7/8 Gate open 100% 8:00 AM. Varied that day from 100%, 70% & 25%. 222 cfs
9/14 Gate open 100% 8:00 AM. (Flow had exceed 400cfs night before) 192 cfs
9/30 Gate open 100% 10:10 AM. Closed gate at 9:15 PM – 627 cfs. 3040 cfs
Appendix A reports extensively on the 9/30 event.
Observations: Upstream of Paradise Pond, there are many “run of the river impoundments” that have been in
place for a century or more, where the impoundment is approximately the width of the river (see Cook’s Dam
and the Button Shop Mill dams, in Leeds). The accumulated sediment in these locations has reached an
equilibrium, whereas the amount of sediment brought in is more or less equal to the amount released, up to a
certain grain size. This suggests that the shape and depth of the impoundment will influence annual sediment
accumulation. (see Cook’s Dam & Button Shop Dams photos).
Cook’s Dam Cook’s Dam
Button Shop Dam 1 Button Shop Dam 2
Page 5 of 5
Conclusion: The results of the most recent bathymetric survey suggest that flow velocities need to be far greater
than 200 cfs to move any amount of sediment out of the pond at its current depth. We believe that
redistributing sediment within the Pond, so it is closer to the gate, and shallower in the 100m reach upstream of
the spillway, will improve the probability of releasing sediment.
Sediment Redistribution proposal: We are proposing to proceed with the mechanical relocation of 1000 cy of
sediment from its current location (center of Pond) into the river channel, along the Northern edge of the Pond
under partial drawdown conditions, (approx. 8’) for a period of approximately 2 weeks. The redistribution would
not take place until the channel is sufficiently scoured (at least several days) to create room for each “dose” (1
or 2 cubic yards) of redistributed sediment. This redistribution “experiment”, our first attempt at sediment
redistribution within the pond, will be monitored and recorded as the experiment progresses. (see PPSMP, page
20, E. Experimental Phase Year 2 Activities)
Monitoring will include: aerial photogrammetry at the beginning of the drawdown and after redistribution is
complete (before Pond is refiled), video recording of the work from surface locations, video recording
underwater during active sediment dosing and overhead (droan mounted) cameras. The ADCP will be in the
river stream at various locations to monitor sediment movement while simultaneously video recording each
dosing event. The downstream turbidity sensor will be monitored continuously. A time period of say 10 to 20
minutes will be allowed after the first “dose” of sediment is placed in the river, to determine sediment load
intensity and duration. If the dose does not exceed the permitted limits (see pages 3 & 4 in the Response to
Comments letter from Pare Corp to NHESP dated June, 2014, of Appendix D in the SMP) dosing frequency may
increase. We will measure bathymetry prior to commencing sediment redistribution using a moving boat at 1‐
second intervals using an Innerspace Model 455 Survey Grade Depth Sounder coupled to a Trimble GeoExplorer
XH GPS receiver. After completion of sediment redistribution, with the Pond at normal pool we will measure
Pond bathymetry again using the same method and will report the difference. We are proposing to proceed
with this experiment in summer 2016.
Symposium: We will be hosting the first year symposium to present and discuss our findings on April 8, 2016, at
the Smith College Conference Center overlooking Paradise Pond, 49 College Lane, Northampton, MA. The
symposium will run from 10 AM to 2 PM. The audience will include students, faculty, project regulators and
other interested parties.
1
Results of the Sluice Experiment
September 30, 2015
Robert M. Newton
Department of Geosciences
An intense rainfall event on September 30, 2015 triggered the highest discharge
event on the Mill River observed in 2015. During this event the sluice gate was opened just
prior to peak flow and remained open for approximately 7.5 hours. Stream discharge and
turbidity were monitored throughout the event together with pond stage. In addition,
discharge and moving bed measurements were made in the pond using an Acoustic
Doppler Current Profiler (ADCP). An estimate of the volume of sediment removed from the
pond just upstream of the dam was determined by comparing pre‐event and post event
pond bathymetry. Suspended sediment flux out of the pond was estimated from turbidity
and discharge data.
The rain gage at the dam station recorded 2.92in of rainfall in an 18hr period from
6pm EST on September 29 to noon on September 30. Other stations further upstream in
the watershed recorded even higher rainfall totals with 4.43in in Haydenville and 4.17in at
the MacLeish Field Station in West Whately (just outside the Mill River watershed). Stream
discharge increased from <10cfs to over 3,100cfs in a little over 6hrs (Figure 1). At the
start of the event the pond was still under a drawdown condition associated with the dam
repair project that was completed on the afternoon of the 29th. The initial pond stage was ‐
3.15ft below the spillway at 8pm EST and the pond reached a filled condition about 9hrs
later.
Analysis of Turbidity
Turbidity throughout the event was measured every 15min at the Lamont Gage
Station using a Campbell Scientific OBS500 Smart Turbidity Meter™ with Clear Sensor
technology™. Turbidity (NTU) was calibrated to suspended sediment load (mg/L) using
data collected during a rain event on July 10, 2015 during which turbidity values ranged
from 1 – 132NTU and discharge ranged from 50 – 1,000cfs. During that event, an ISCO
autosampler was deployed at the Lamont Bridge to collect hourly water samples over the
24 hour period spanning the event. The volume of the water samples was measured and
the weight of suspended sediment was determined gravimetrically from samples filtered
through a 0.45µm membrane filter. Analysis of this data showed a strong correlation (r2 =
0.97) between sediment concentrations (mg/L) and side scatter turbidity (NTU) (Figure 2).
2
Figure 1. Rainfall measured at the pond dam station and discharge measured at the Lamont Bridge
station. The brief depression in discharge at around 2000hrs is associated with the closing
of the sluice gate.
During the September 30 event side scatter turbidity increased from <10NTU to
over 280NTU. Turbidity values over 200 NTU’s occurred over a 4hr period synchronous
with peak flow (Figure 3). Turbidity had reached values in excess of 200NTU’s prior to the
opening of the sluice gate at 9:10am EST. But the spike in turbidity (peak value 298NTU)
that occurred after the gate was opened, likely represents increased sediment load
associated with sluicing sediment through the gate. Towards the end of the event, the lack
of any change in turbidity after closing the gate at 8:15pm EST suggests that, by that time,
the flow had decreased enough such that flow through the open sluice gate was no longer
eroding the pond bottom and adding to the suspended sediment load.
3
Figure 2. Calibration curve relating side scatter turbidity to suspended sediment concentrations.
The total suspended sediment flux can be estimated using the discharge and
turbidity data. For this calculation, suspended sediment (mg/L) was estimated from side
scatter turbidity (NTU) using the July 10 event relationship (Figure 2). Since discharge and
turbidity were measured at different intervals (discharge every 10minutes; turbidity every
15minutes) it was necessary to extrapolate 15minute discharge values from the 10minute
data. Suspended sediment flux was then calculated by multiplying the discharge by the
suspended sediment load by the 15minute time interval after converting cfs to L/sec:
ܵܨ ቀ
ଵହቁൌܵܥቀ
ቁ ൈ ܳቀ
௦ቁ ൈ 900ݏ݁ܿ ൈ
ଵൈଵల
Where:
SF = Suspended sediment flux
SC = Concentration of suspended sediment
Q = Discharge
4
Figure 3. Pond stage and downstream turbidity during Sluicing Experiment.
The suspended sediment flux values represent the mass of sediment leaving the
pond over a 15minute interval. At peak flow, this value reached approximately 34 metric
tons every15minutes (Figure 4). Integrating the 15minute data over the entire hydrograph
yields the total suspended sediment flux that left the pond during the event.
Figure 4. Stream discharge (red) and suspended sediment load (green) during the September 30 event.
5
The total flux for this event was approximately 590 metric tons of which 482 tons
exited during the time the sluice gate was open. If we use the average bulk density of the
bottom sediment (1.2g/cm3) then the total mass flux could be expressed in terms of an
equivalent amount of bottom sediment. In this case, it yields a volume equal to
approximately 490m3 or 650yd3 of sediment.
One of the weaknesses of this analysis is that we did not measure the turbidity of
water entering the pond. Therefore we cannot deduce how much of the 490m3 came from
erosion from the bottom of the pond verses how much was from upstream suspended
sediment moving through the pond.
Bathymetric Analysis
In order to establish the amount of sediment removed from the pond bottom,
bathymetric data was collected immediately after the event and compared to data collected
over the previous summer. Only bathymetric data collected within the 90m reach
upstream of the dam was considered for this analysis (Figure 5).
Figure 5. Bathymetric map showing the bottom elevation of Paradise Pond as determined from bathymetric
surveys conducted during the summer of 2015. Red box indicates area used to evaluate sediment
sluicing during the September 30, 2015 event.
6
Summer 2015 October 2015
Figure 6. Digital elevation Models of pond bathymetry prior to the sluicing experiment (upper left), and
after (upper right). Subtraction of the two maps shows areas of erosion (red) and deposition
(blue) (lower map).
7
Depth data was collected from a moving boat at 1‐second intervals using an
Innerspace Model 455 Survey Grade Depth Sounder coupled to a Trimble GeoExplorer XH
GPS receiver. All GPS positions were differentially corrected using Trimble Pathfinder
Office software and the HAMP CORS station located on the roof of Sabin Reed. Depths were
converted to bottom elevations by subtracting depth values from the pond surface
elevation as determined by stage measurements recorded by the Dam Station datalogger.
The pond staff gage was calibrated to elevation during the summer of 2015 using a Leica
TPS1200 Total Station and campus benchmarks NPW04A and NPW04B.
The depth data from each survey was imported into ESRI ArcMap software and a 1m
x 1m bottom elevation grid was constructed using a kriging routine to extrapolate
elevations at the 1m grid points. Digital Elevation Models (DEMs) were constructed from
the grids for the two pond bottom surveys and are shown in the upper part of Figure 6.
The post‐experiment DEM was then subtracted from the pre‐experiment DEM to create a
change raster that shows areas of erosion (positive values) and deposition (negative
values)(bottom of Figure 6). This change raster can be integrated to get the net change in
sediment volume over the survey area. In this case the net volume was approximately 50
cubic yards, indicating a small net loss of sediment from the pond due to erosion. Erosion
was concentrated in areas close to the sluice gate and along the eastern bank of the pond
where the channel directed most of the current (Figure 5). This high current area is also
reflected by the high occurrence of sand in cores collected along the eastern side of the
sluice area (Figure 7).
Figure 7, Cores collected in areas subject to higher current velocities have higher concentrations of sand
(red).
8
It should be noted that raster analysis is subject to increasing errors as you
extrapolate away from the actual measured points. The bathymetric surveys were
conducted with the expectation that most of the depth changes would be near the center of
the channel. However, some of the calculated erosion occurred outside of the area of high
intensity depth data, therefore, there is some uncertainty as to the actual amount of erosion
that occurred. In any case, bathymetric analysis shows that opening the sluice gate did not
result in significant erosion of pond bottom sediments during this experiment.
A photogrammetric method for directly determining surface elevations of the pond
bottom was developed during times when pond level was lowered for dam maintenance.
In this method, aerial photographs were collected from a drone flying at an altitude of 50m
and the imagery was processed using Pix4D software to create highly accurate DEM’s.
Figure 8 shows a map of the exposed sand bar just upstream of the island with a 10cm
contour interval.
Figure 8. 10cm contour map of the sand bar just upstream of the island created using photogrammetric
methods. Green areas in lower left and upper right are vegetated surfaces and elevations reflect
surface of the vegetation not the sediment surface.
This method will be used in the future to guide and measure sediment redistribution in the
pond. Prior to redistribution, photogrammetric DEM’s will be constructed and used to
9
determine to what depth the surface needs to be lowered to achieve the desired volume of
sediment to be redistributed using ArcGIS software. A GPS receiver will be used to flag the
area to be excavated using coordinates extracted from ArcGIS. A second photogrammetric
DEM dataset will be collected after redistribution and compared with the pre‐excavation
data to quantitatively measure how much sediment was actually moved.
Figure 9. Hydrograph showing when moving bed tests were made. The highest moving bed velocity
(6.1cm/sec) was measured at a discharge of approximately 1,600cfs.
Moving Bed Measurements
A series of direct measurements of the movement of pond bottom sediment were
conducted during the falling limb of the hydrograph (Figure 9). These measurements were
made from an anchored boat using a Teledyne RD Instruments RiverRay Acoustic Doppler
Current Profiler (ADCP). Each moving bed test was conducted over a 10‐15 minute period
during which the boat was maintained in a fixed position using a 3‐anchor system. Four of
the five tests were conducted in the area just upstream of the dam while the last test was
done near the pond inlet (Figures 10 and 11). Each test determined the velocity and
direction of movement of pond bottom sediment as well as near‐bottom water velocity
(Table 1). All measurements were made well after peak discharge due to boat deployment
safety concerns. The first measurement (site 000) was conducted at a discharge of 1,540cfs
and revealed the highest measured bottom sediment velocity (6.1cm/sec) under a near‐
10
bottom water velocity of 36.7cm/sec. The boat was moved between measurements in an
attempt to determine the area of maximum sediment movement, however, the decrease in
discharge between measurements makes this difficult. Each succeeding measurement
generally showed lower bottom sediment velocity and near bottom velocity except site 004
that was located closest to the open sluice gate and site 001a that was located near the
pond inlet. The pond inlet site was anomalous in that it had the highest near bottom
velocity (55.4cm/sec) and the lowest moving bed velocity (0.46cm/sec). The low bed
velocity reflects the coarse nature of the sediment (coarse sand) at this location. Analysis
of a core collected near this site showed that 95% of the surface sediment was sand size or
greater.
Table 1: Results from ADCP moving bed tests conducted on September 30, 2015.
Site
Discharge
(cfs)
Moving Bed
Velocity (cm/sec)
Near Bottom Water
Velocity (cm/sec)
Time
(EST)
000 1,540 6.04 36.7 14:18
001 1,400 2.41 32.4 14:43
002 1,280 1.19 28.2 15:01
004 1,180 1.22 34.9 15:18
001a 990 0.46 55.4 16:07
The ability of the current to move sediment is primarily a function of the shear
stress imparted by the current together with the particle size of the sediment. A simplified
version of this relationship is described by Hjustlrom’s Curve (Hjulstrom, 1935). This
graph (Figure 12) defines fields of erosion, transport, and deposition as related to current
velocity and grain size. The important aspect of this curve is that it shows that the erosion
curve inclines upward through the finer grain sizes meaning the finer sized material is
more difficult to erode than coarser material. However, once the erosion velocity is
reached and the material is eroded, this finer material will be easily transported.
11
Figure 10. Map showing the results of the moving bed measurements taken during the falling limb of the
hydrograph. The size of the arrows is proportional to the moving bed velocity.. Numbers next to
the arrows show the moving bed velocity in cm/sec.
Figure 11. Near bottom water velocity measured at moving bed sites. Pond inlet near bottom velocity was
highest despite yielding the lowest moving bed velocity.
12
Conclusion
This experiment failed to move significant amounts of bottom sediment out of the
pond. One of the explanations for the lack of significant erosion is that the cohesion of the
fine‐grained sediment, as demonstrated by Hjulstrom’s Curve, prevented the finer sized
material from being eroded. The moving bed detected by the ADCP could be material that
entered the pond as stream bedload and may not represent entrainment of sediment off
the pond bottom. Higher velocities may be required to move this sediment, but this is
unlikely to occur given that they did not occur under this high magnitude event. However,
higher velocities could result from adding more sediment to the sluice area. This would
have the effect of reducing the water depth and therefore the cross sectional area of flow,
forcing an increase in water velocity. Movement of coarser pond sediments from near the
inlet of the pond, to the sluice area, would also provide material that is easier to move
(Figure 12). It may also be possible to increase erosion at the pond bottom by anchoring
small turbulence generating structures that would get the finer material into suspension.
Figure 12. Hjulstroms Curve. From Physical Geology by Steven Earle used under a CC‐BY 4.0 international license.
References
Hjulström, F., 1935, Studies of the morphological activity of rivers as illustrated by the river
Fyris, Bull. Geol. Inst. Univ. Uppsala, v. 25, p. 221–527.
Robert Newton, Maya Domeshek (18’),
Lyn Watts (17’), Marcia Rojas (18’),
Lizzie Sturtevant (18’)
Paradise Pond Sediment
Report
07/22/15
Drone photo of sediment filling pond
Side Scatter Turbidity associated
with the July 10, 2015 rain event.
Water samples collected from this
event were used to correlate the side
scatter turbidity values to suspended
sediment in mg/L
0
200
400
600
800
1,000
1,200
0
20
40
60
80
100
120
140
2000 midnight 0400 0800 noon 1600
Q
(
c
f
s
)
S
i
d
e
S
c
a
t
t
e
r
T
u
r
b
i
d
i
t
y
(
N
T
U
)
July 10, 2015
Calibration curve relating side scatter
turbidity measured at the Lamont
Gage station to suspended sediment
concentrations. Suspended sediment
concentrations determined from
water samples collected at 1 hour
intervals by an ISCO autosampler at
the gage station.
Photo of 0.45µm membrane filters from water samples collected at 1hr intervals during the July 1o event.
Maximum sediment load was approximately 190 mg/L.
Downstream
Monitoring
Stations
-Two benchmarks installed
on the left and right bank
to define reference cross
section
- Staff gauge with pressure
transducer installed to
measure water stage
Reference Sites
-Crossectional profiles were
created using elevations
from a Leica Total Station
and depths from an Acoustic
Doppler Current Profiler
deployed on a Teledyne
Instruments RiverRay
Downstream Bathymetry
Reference Station 1 6/22/15
Stage: 0.52 ft
Discharge: 100cfs
Pond Stage: 3.094 ft
Elevation in ft as a function
of distance across the
channel in ft.
108
110
112
114
116
118
120
122
124
-20 0 20 40 60 80 100 120
Reference Section 1 6/22/15
El
e
v
a
t
i
o
n
(
f
t
)
Distance (ft)
Gradient = 0.0015
Reference Site 2
6/26/15
Stage: 4.41 ft
Discharge: 34. 061 cfs
Pond Stage: 2.948 ft
Elevation in m as a function
of distance across the
channel in m
Reference Site 3
6/30/15
Stage: 7.89 ft
Discharge: 91.627 cfs
Pond Stage: 3.075 ft
Elevation in m as a function
of distance across the
channel in m
Reference Site 4
6/25/15
Stage: 1.035 ft
Pond Stage: 2.962 ft
Elevation in m as a function
of distance across the channel
in m
Measured using a Leica Total Station and
depth sounder
Paradise Pond Bathymetry
Digital elevation model of the pond bottom
showing areas of deposition in yellow and
brown, deeper channel in green and blue
Moving Bed Test Results from moving bed test using RiverRay ADCP
Moving Bed detected:
360-470 cfs
2.56-8.28 ft depth
No Moving Bed
Locations of moving bed tests, green dots
show where in the pond a moving bed was
detected while the red dots indicate no
movement.
Sediment Core Locations
Sediment Core Methods:
-90mm diameter cores
were collected from a
small boat using both a
gravity type Uwitec
Corer and a drive core
method with 76mm
diameter aluminum
tubing
Sediment Core Methods:
-cores were extracted
in the field and
segmented at 2.5 cm
intervals
-samples were dried
overnight in a drying
oven at 50°C
Mercury (Hg)
Methods:
-samples dried at 50°C
-dry samples homogenized using mortar &
pestle
-dried and homogenized samples analyzed
for mercury by thermal decomposition cold
vapor atomic absorption using a Teledyne
Leeman Labs Hydra IIC Mercury Analyzer
Mercury (Hg)
Mercury (Hg)
Metals are often bound to organic
material. In this graph we see the
relationship between loss on ignition
and total mercury in ppb. Loss on
ignition is directly related to the
amount of organic material in these
sediments.
-Lead (Pb)
-Manganese (Mn)
-Zinc (Zn)
-Copper (Cu)
-Chromium (Cr)
-Selenium (Se)
-Arsenic (As)
-Cadmium (Cd)
-Nickel (Ni)
-Phosphorus (P)
Acid Extractable Metals:
Methods:
-extracted using EPA method 3050B:
1-2g of sediment were reacted with
nitric & hydrochloric acid and oxidized
with peroxide
-extracted solution analyzed using
Teledyne Leeman Labs Prodigy
Inductively Coupled Plasma Optical
Emission Spectrometer
Extractable Metals:
Lead (Pb)
Phosphate
Mercury (Hg)
Grain Size
Methods:
-samples were wet sieved through a No. 230
(63 µm) sieve
-sieved samples were dried at 50°C
-percent sand and mud in each sample were
obtained using the recorded weights before
and after drying
Grain Size
Grain Size and Lead
Mercury is associated with coarser material at Hulberts
Pond
Page 1 of 19
Macroinvertebrate and Mussels in the Mill River
The “Before” Samples
Marney Pratt, PhD
Smith College
A common way to asses an impact in a river is to use a Before-After-Control-Impact design (or BACI
design)(Strayer and Smith 2003). A BACI design has at a minimum of one sample before (Before
sample) and after (After sample) an impact for a sample that may be affected by the impact (Impact
sample) as well as for a sample that should not be affected by the impact (Control sample). Ideally, there
will be many samples in each category, but this is not always possible. This report documents results
from the first set of samples in the Before sample for macroinvertebrates and mussels in the Mill River.
The impact we are trying to assess is how the new sediment management protocol of opening the sluice
gate in the Paradise Pond Dam in the Mill River during high flow events will affect the organisms living
in the river.
Part 1: Macroinvertebrate Sampling, Summer 2015
Methods
Data Collection
Aquatic macroinvertebrates were sampled at 5 of the 7 monitoring sites along the Mill River during the
summer of 2015 (Figure 1.1) by kick-sampling using D-nets (LaMotte D-net, 500 micron mesh). Nets
were placed downstream with the opening facing upstream. The area directly upstream of the net in an
area of 0.3m wide x 0.5m long was disturbed by kicking the bottom and picking up and rubbing rocks by
hand in front of the net for 30 seconds to 1 minute. The contents of the net were rinsed into a bucket.
This process was repeated 20-30 times per site, and all samples from a site were joined into one
composite sample. When the river bottom was softer (sandy or muddy), the net was bumped along the
bottom (to disturb the bottom but to avoid collecting too much sediment) approximately 30 times per
sample. Once the full composite sample was collected, the composite sample was taken back to lab and
sieved down to 50 microns and all macroinvertebrates were sorted out and preserved in 70% ethyl
alcohol. Most macroinvertebrates were identified down to genus or even species whenever possible,
however, some were only identified to family or occasionally just to order when identifications were
particularly difficult (such as for oligochaete worms, water mites, and midge larvae).
We also sampled aquatic macroinvertebrates at all 7 of the monitoring sites during the summer of 2015
using Hester-Dendy (H-D) samplers (WILDCO® Large Square Samplers). The H-D samplers had
fourteen 3" square plates and twenty-four 1"-diamater spacers for a total sampling area of 0.16 m2. The H-
D sampler used is approved for benthic macroinvertebrate collection by the US Geological Survey, the
EPA, and other groups. Three H-D samplers were each attached to a cinder block and placed at each
monitoring site for a total of 21 samplers. H-D samplers were placed in the river at the end of June (June
22-July 2, 2015) and picked up approximately 4 weeks later at the end of July (July 21-27, 2015). All
sites had all three H-D sampler retrieved except for the Diversion Bike Path site (only one was found) and
Hulbert’s Pond (only two were found). When retrieved, the H-D sampler was cut from the cinder block
very carefully and placed immediately in a Whirl-pak bag (1,627 ml capacity, 7-1/2" x 12"). Once in the
lab, the H-D samplers were scraped clean and sieved down to 50 microns and as above, all
macroinvertebrates were preserved in 70% ethyl alcohol. Because these samples contained a lot of
Page 2 of 19
difficult to identify organisms such as midge larvae, the samples are still being identified and the results
from these samples are not presented in this report.
All the above samples from the summer of 2015 represent the Before samples, and all the samples
downstream of the dam are considered potential Impact sites while the site upstream of the dam is
considered a Control site in a BACI design.
Data Analysis
Diversity Measures: There are many ways to look at diversity, and a healthier river is generally more
diverse. One way to compare diversity it to graph rank abundance curves. Species (or the lowest level
the organism was identified to) were ranked from the most abundant (ranked number 1) to the least
abundant. Relative abundance was calculated as the number of that organism found divided by the total
number of all the organisms in the sample times 100. Pairwise Kolmogorov-Smirnov tests were used to
compare the rank-abundance curves.
The Shannon Diversity Index (H’) was used to compare the overall diversity of the different sites:
S
H´ = - Σ (pi )*(ln pi)
i=1
where pi is the proportional abundance of the ith taxa (=proportion of the total sample of each taxa) and S
is the total number of taxa (also called the taxa richness which in this case is the total number of different
taxa noting that most organisms were identified to family or genus.). We also calculated the evenness (J’)
of each site which is a measure of how even the number of organisms in each taxa is (a perfectly even
sample would have J’ = 1 and would mean there were equal number of each organism):
J’ = H’/H’max
where H’max = maximum possible diversity, H’max = ln(S)
Water Quality: Hilsenhoff’s Biotic Index (HBI) was used to calculate the water quality (Table 1.1). This
index is based on the organisms’ tolerance to organic pollution. The tolerance values (ti) used were from
the Biological Monitoring of Surface Waters in New York State (Stream Biomonitoring Unit Staff 2012).
HBI = Σ(xi*ti)/(n), where
xi = number of individuals within a taxon
ti = tolerance value of a taxon
n = total number of organisms in the sample (usually 100)
Table 1.1: Water quality based on the Hilsenhoff’s Biotic Index (HBI) (Hilsenhoff 1987).
HBI Water Quality Degree of Organic Pollution
0.00-3.50 Excellent No apparent organic pollution
3.51-4.50 Very good Possible slight organic pollution
4.51-5.50 Good Some organic pollution
5.51-6.50 Fair Fairly significant organic pollution
6.51-7.50 Fairly poor Significant organic pollution
7.51-8.50 Poor Very significant organic pollution
Page 3 of 19
8.51-10.00 Very poor Severe organic pollution
Figure 1.1: Locations of the 7 sampling sites along the Mill River for the summer of 2015. Kick-samples
using a 500 micron mesh D-net were taken at the 5 sites with the yellow stars. Kick-samples were
collected on the following dates: Upstream – 6/30/15, Lamont Bridge Gage (=Downstream Riffle) –
7/20/15, Diversion Bike Path – 6/22/2015, Mussel – 6/22/2015, Recycle – 6/22/2015. Hester-Dendy
samplers were placed in the river at all 7 sites at the end of June and then picked up approximately 4
weeks at the end of July in 2015.
Page 4 of 19
Figure 1.2: Hester-Dendy (H-D) sampler attached to a cinder block. The cinder block was placed with
the longer axis perpendicular to the river flow so that the H-D sampler was facing upstream.
Figure 1.3: Rank abundance curves for the 5 sites in the Mill River where kick-samples were taken with
D-nets in the summer of 2015. The most abundant organism at a site is given a species rank of 1, the
second most abundant organism is given species rank of 2, and so on. Only the Recycle site had a
significantly different rank abundance curve compared to the other sites (pairwise Kolmogorov-Smirnov
tests, p<0.05).
Interpretation: Most of the sites were fairly similar, where the most abundant organism was between 19-
26% of the sample. The Recycle site stood out as being different, where the most abundant organism (an
isopod) made up 47.5% of the sample.
Page 5 of 19
Table 1.2: Information about the macroinvertebrate samples for the 5 sites in the Mill River where kick-
samples were taken with D-nets in the summer of 2015.
Upstream
Sandy
Downstream
Riffle ‐
Lamont Bike Path
Downstream
Sandy ‐
Mussel Recycle
Shannon Diversity Index (H') 2.66 2.95 2.51 2.68 1.96
Taxa Richness (S) 28 29 20 22 15
Evenness (J') 0.80 0.87 0.84 0.87 0.72
Total Individuals Collected 93 139 74 74 61
S
H´ = - Σ (pi )*(ln pi)
i=1
where pi is the proportional abundance of the ith taxa (=relative abundance of each taxa) and S is the total number of taxa.
S = taxa richness (in our case, this is the total number of different taxa noting that most organisms were identified to family or
genus.)
H’max = maximum possible diversity where H’max = ln(S)
J’ = evenness where J’ = H’/H’max (a perfectly even sample would have J’=1)
Interpretation: The sites had fairly similar diversity, taxa richness, and evenness except for the Recycle
site (which was lower for all of these measures – most likely because the sample was so dominated by
isopods). The most diverse site was the Downstream Riffle site near Lamont Bridge, just downstream of
the dam. Since there was only one sample at each site, no statistics could be used to compare sites.
Future Research: We propose to do more sampling at fewer sites in the future (starting in the summer of
2016) in order to be able to statistically compare them. Riffle sites are the most used types of sites in
assessing water quality because they are relatively easy to sample and usually have the widest diversity.
Thus, we will use the Downstream Riffle site (=Lamont Bridge) as an Impact site and a new Upstream
Riffle site (Figure 2.1) as a Control site as our representative riffle sites. We also believe it is important to
sample from soft sediment run sites, as these may be the most impacted from changes in sediment
management. We propose to use the Downstream Sandy site (=Mussel site) as our Impact site and the
Upstream Sandy site as a Control site as our representative soft sediment run sites. The Downstream
Sandy site was chosen as the representative Impact site because it contains a lot of sensitive organisms
such as mussels, in the Before sample it had a very similar Shannon Diversity Index score compared to
the Upstream Sandy site, it is a place that sediment will likely accumulate if there is any problems with
sediment accumulation, and it is relatively easy to access compared to sites further downstream which
will make it easier to sample repeatedly.
Page 6 of 19
Figure 1.4: Water quality for the 5 sites in the Mill River where kick-samples were taken with D-nets in
the summer of 2015. Hilsenhoff’s Biotic Index (HBI) was used to calculate the water quality. This index
is based on the organisms’ tolerance to organic pollution. The tolerance values used were from the
Biological Monitoring of Surface Waters in New York State (Stream Biomonitoring Unit Staff 2012).
Interpretation: The water quality is actually a little better just downstream of the Paradise Pond Dam
(the Downstream Riffle (=Lamont) and the Diversion Bike Path) than it is just upstream of the dam
(Upstream Sandy). The water quality does go down a little further downstream (the Downstream Sandy
(=Musse ) and Recycle sites), but they are not substantially lower in water quality compared to the
upstream site (Upstream Sandy).
Page 7 of 19
Part 2: Macroinvertebrate Sampling, Fall 2015
The Biodiversity, Ecology, and Conservation Labs (Bio 155 labs) in the Fall of 2015 collected data at the
two riffle sites, where the Upstream Riffle represents the Control sample and the Downstream Riffle
represents the Impact sample (Figure 2.1). The sluice gate was opened once during a large storm at the
end of September in 2015. Thus, these samples can be considered a potential Impact sample for each site.
Note that the Upstream Riffle was not sampled during the summer of 2015, so it unfortunately does not
have a true Before sample. The sampling method used was the same kick-sampling method described in
Part 1 above except the total area sampled was recorded.
Figure 2.1: Locations of the 2 riffle sampling sites along the Mill River for the fall of 2015. Kick-
samples using a 500 micron mesh D-net were taken at each site in the fall of 2015 (10/27/2015 for the
Downstream Riffle, 11/4/2015 for the Upstream Riffle). An additional kick-sample was taken at the
Upstream site in the summer of 2015 (7/20/2015). (Photo from Google Maps)
Upstream Riffle
Downstream
Riffle (Lamont)
Page 8 of 19
Table 2.1: Information about the macroinvertebrate samples for the 2 riffle sites in the Mill River where
kick-samples were taken with D-nets (500 micron mesh) in the summer of 2015 (Downstream Riffle
only) and the fall of 2015 (both sites).
Downstream Riffle Upstream Riffle
Summer 2015 Fall 2015 Fall 2015
Total Individuals Collected 139 98 113
Area sampled NA 7.2 6.9
Density (# of organisms/m2) NA 13.6 16.4
Shannon Diversity Index (H') 2.95 2.69 2.42
Taxa Richness (S) 29.00 23.00 28.00
Evenness (J') 0.87 0.86 0.73
HBI 4.18 4.56 3.73
Water Quality Very Good Good Excellent
%EPT 34% 49% 81%
Density = the number of organisms collected divided by the total area sampled
S
H´ = - Σ (pi )*(ln pi)
i=1
where pi is the proportional abundance of the ith taxa (=relative abundance of each taxa) and S is the total number of taxa.
S = taxa richness (in our case, this is the total number of different taxa noting that most organisms were identified to family or
genus.)
H’max = maximum possible diversity where H’max = ln(S)
J’ = evenness where J’ = H’/H’max (a perfectly even sample would have J’=1)
HBI = Hilsenhoff’s Biotic Index, used to calculate the water quality. This index is based on the
organisms’ tolerance to organic pollution. The tolerance values used were from the Biological
Monitoring of Surface Waters in New York State (2012,
http://www.dec.ny.gov/docs/water_pdf/sbusop12.pdf).
Water Quality is based on the HBI values
%EPT = the percent of Ephemeroptera, Plecoptera, and Tricpotera in the sample
Interpretation: The Downstream Riffle sample was reasonably similar in the summer versus the fall of
2015. While the diversity and taxa richness went down a little, that may be influenced by the smaller
sample: there were 41 fewer individuals collected in the fall, and diversity can be affected by sample size
since a smaller sample has a lower chance of collecting rare organisms. The evenness was similar in the
fall compared to the summer, which indicates that there were similar abundances of each taxa (which
means that the sample is not dominated more by a particular taxa). The Upstream Riffle was less diverse
and less even than the Downstream Riffle, most likely because it was heavily dominated by net-spinner
caddisflies (Hydropsyche sp.), which made up 38% of the sample.
The water quality went down a little from the summer to fall for the Downstream Riffle site, however 4.5
is the border between Very Good and Good so it just barely changed categories going from 4.18 to 4.56.
Interestingly, the %EPT (which is also used as an indicator of water quality – higher percentages means
higher water quality) went up from the summer to the fall for the Downstream Riffle site. While the
water quality appears to be higher at the Upstream Riffle, the diversity and evenness are higher at the
Downstream Riffle. The evidence as whole suggests the water quality is generally quite good and
diversity is relatively high at the Downstream Riffle even after opening the sluice gate one time during the
high flow event at the end of September of 2015.
Page 9 of 19
Note on odonates found so far
In all the kick-samples from the summer and the fall in 2015 as well as the Hester-Dendy samples from
the summer of 2015 analyzed to date, a total of 6 larvae from the Order Odonata have been found.
Table 2.2: All the Odonata found in all the macroinvertebrate samples taken in 2015.
Species (# found) Location Substrate type Date Type of sample
Aeshna eremita (1) Upstream Sandy Soft Sediment, run 6/29/15 Kick-sample
Aeshna eremita (1) Downstream Riffle
(Lamont) Hard substrate, riffle 7/27/15 Hester-Dendy
Hagenius brevistylus (1) Downstream Riffle
(Lamont) Hard substrate, riffle 10/27/15 Kick-sample
Lanthus parvulus (1) Downstream Riffle
(Lamont) Hard substrate, riffle 10/27/15 Kick-sample
Boyeria vinosa (2) Upstream Riffle Hard substrate, riffle 11/4/15 Kick-sample
Interpretation: There was a small number of odonates found in the samples taken in 2015. Thus, they
were not particularly abundant in the Mill River in the locations we sampled and the times we sampled.
They seem a little more common at the riffle sites and are more often found by kick-sampling. None of
the 4 state-listed odonates historically found in the Mill River were found in our samples (Boyeria
grafiana – Special Concern, Gomphus ventricosus - Threatened, Gomphus abbreviatus - Special Concern,
and Neurocordulia yamaskansis - Special Concern).
Page 10 of 19
Part 3: Mussel Sampling, Summer and Fall 2015
Study Sites
Freshwater mussels are often used as indicator species in aquatic environments because they can be very
sensitive to impacts. For this study, the Impact sample was taken from the same Downstream Sandy
(=Mussel) site used in the Macroinvertebrate sampling in Part 1 above (Figure 3.1a). Because no mussels
were found upstream of Paradise Pond in the Mill River, we had to use another similar river in the area as
the Control sample. The Manhan River is a similar river in the area, and a site downstream of a dam
where no change in sediment management is being made was used as the Control sample (Figure 3.1b).
Figure 3.1: Sampling locations (yellow stars) on the (a) Mill River in Northampton, MA and (b) Manhan
River in Easthampton, MA. The Manhan River serves as a control and the Mill River serves as the
impact site. The summer samples were taken before the sluice gate was opened (samples taken on
7/14/2015 for the Mill River and 7/17/15 for the Manhan River), while the fall samples were taken after
work was done on the Paradise Pond Dam and the sluice gate was opened once at the end of September in
the Mill River (samples taken on 10/20/2015 for the Manhan River and 10/26/15 for the Mill River).
(a) Mill River (b) Manhan River
Page 11 of 19
Species and Size Class Data Collection 1994-2015:
The Smith College Ecology class taught by Dr. Steven Tilley sampled the mussels in the Mill River site
almost every fall for 16 years (1994-2010). Unfortunately, they did not take any measures of sampling
effort, but the data can still give us an idea of the size class distribution of Elliptio complanata (the most
abundant mussel consistently found in their samples) as well as the proportion of Lampsilis radiata (the
second most abundant mussel found in their samples) found. They collected mussels by sticking their
fingers into the sediment and moving them around in the sediment (noodling) and pulling out any mussels
they found. Then they sorted the mussels by species and size class using the number of annual growth
rings to put them into size classes (Figure 3.2). Because it is harder to count growth rings on larger
mussels, they used a size class of 5+ rings to indicate larger mussels. This method was repeated by
Steven Tilley, Marney Pratt, and Molly Peak on May 20, 2015.
Figure 3.2: Annual growth rings in freshwater mussel shells can be used as an age approximation for
mussels. Due to the difficulty in seeing the growth rings in older mussels, and the chance that false rings
can show up, it is not a totally accurate measure of age. However it can still be useful to tell old from
young mussels in general. (photo from Nedeau, McCollough, Swartz 2000)
Page 12 of 19
Figure 3.3: Percentage of the total sample that was Lampsilis radiata in the Mill River. All samples were
collected in the Mill River at the site indicated in Figure 3.1a. Samples from 1994-2010 were collected by
the Smith College Ecology class taught by Steven Tilley (now Professor Emeritus) during the Fall
semester, while the 2015 sample was collected on 5/20/2015 by Steve Tilley, Molly Peek, and Marney
Pratt. Mussels were sampled by “noodling” through the sediment and no attempt was made to keep track
of sampling effort (number of people sampling per time).
Interpretation: The percentage of L. radiata in the Mill River has steadily gone down through the years.
It is not known why, but it is suspected that it has to do with a decrease in reproduction (McLain, 2006).
These data indicate that there has been a consistent decrease in at least one species on freshwater mussel
at this site that is most likely unrelated to how Smith College is planning to manage the sediment in
Paradise Pond.
Page 13 of 19
Figure 3.4: Frequency of Elliptio complanata mussels in different size classes over time where the x-axis
is the frequency (expressed as a proportion of the total sample (n)) and the y-axis is the number of annual
growth rings (5+ represents all the shells with 5 or more growth rings). All samples were collected in the
Mill River at the site indicated in Figure 3.1a. Samples from 1994-2010 were collected by the Smith
College Ecology class taught by Steven Tilley (now Professor Emeritus) during the Fall semester, while
the 2015 sample was collected on 5/20/2015 by Steven Tilley, Molly Peek, and Marney Pratt. Mussels
were sampled by “noodling” through the sediment and no attempt was made to keep track of sampling
effort (number of people sampling per time).
Interpretation: While counting annual growth rings on freshwater mussels is problematic for precise
measurement of mussel age, and no attempt was made to determine population density, this dataset of
over 15 years is still useful at showing how the size class structure of mussels has changed at this site over
time. When the population is mature, it is dominated by larger mussels with more annual growth rings.
At some point between the fall of 2001 and 2004, the population went from being dominated by larger,
older mussels to being dominated by small, young mussels. The reason for this is unknown. Then it is
clear that from 2004-2010 the young mussels that recruited steadily grew larger. By 2015, the population
of E. complanata is again dominated by larger, older mussels.
Page 14 of 19
Species, Size Class, and Population Density, Summer and Fall 2015
Mussels were sampled in the Mill River in the summer (7/14/2015) and the fall (10/26/2015) and in the
Manhan River in the summer (7/17/2015) and the fall (10/20/2015) of 2015. A grid was set up by
stretching one measuring tape across the river and one along the edge of the river running parallel to the
river. We used a systematic sampling method based on methods described in Strayer and Smith (2003).
We chose x (for the tape measure across the river) and y (for the tape measure parallel with the river)
coordinates of 3-5 random starts. 1-5 samples were taken starting at each random start, and the distance
in meters along the y-axis from each random start was determined using the methods described in Strayer
and Smith (2003). At each sampling location, a 0.5 m x 0.5 m quadrat was used to sample the mussels.
We first used a clear-bottomed view bucket to observe any mussels on the surface, and then dug up the
sediment down to 15 cm deep. The sediment was sifted through a sieve with 4.67 mm mesh. For each
mussel found, we identified it to species, noted if it was found on or below the surface, measured the shell
length to the 0.1mm, and estimated the number of annual growth rings.
Figure 3.5: Proportion of all freshwater mussels found (including mussels that were alive and dead) in the
summer and fall of 2015 in the Mill and Manhan Rivers. The three species of mussels found include the
Eastern elliptio Elliptio complanata, Eastern lampmussel Lampsilis radiata, and Eastern pondmussel
Ligumia nasuta.
Interpretation: E. complanata is the most abundant species of mussel found at both sites in both seasons.
L. radiata is spotty and fairly rare, while L. nasuta (which is listed as a species of Special Concern in
Massachusettes) was only found in the Mill River in the fall sample.
Page 15 of 19
In order to put our data in a larger context, below we include data on freshwater mussels collected by
David McLain in the Mill River:
Figure 3.6: Percentage of difference species of mussels in the Mill River from sampling done by David
McLain (McLain 2006). The 1998-pre sample was taken before dredging of Paradise Pond, while 1998-
post was taken after the dredging.
Interpretation: Just like our samples in 2015, these data show that E. complanata is the dominant mussel
found in the Mill River. These data also support our result that the percentage of L. radiata has been
decreasing over time. According to these data, the percentage of L. nasuta was fairly stable ranging from
0.5-1.3% of the sample, whereas we found 0% in the summer and 13% in the fall in 2015. Mussels can
be very patchy in distribution, and we sampled a fairly small area very intensively (sifting through the
sediment to a 15cm depth and measuring the size and counting the growth rings of every mussel found).
Thus, our sampling protocol may not be the best way to sample rare species (our protocol was designed to
look at population density and size distributions).
Page 16 of 19
Figure 3.7: Frequency distributions of Elliptio complanata shell lengths in the Mill River (a, b) and
Manhan River (c, d) sampled in the summer (a, c) and fall (b, d) of 2015. See Table 3.1 for more
information on the median size and total sample size.
Interpretation: There appears to be two peaks in the size distribution of E. complanata at the Mill River.
This suggests there was a year or more of low recruitment and/or high mortality at this site. The Mill
River has a greater mix of sizes (with many large mussels), while the Manhan River is more dominated by
smaller mussels.
Page 17 of 19
Figure 3.8: Frequency distributions of Elliptio complanata mussels with annual growth rings from 1 to
5+ (5+ includes mussels with 5 or more growth rings) in the Mill River (a, b) and Manhan River (c, d)
sampled in the summer (a, c) and fall (b, d) of 2015. See Table 3.1 for more information on the total
sample size.
Interpretation: There appears to be more really young mussels in the summer for both sites. This may
be because the little mussels are growing into larger mussels through the summer into the fall. It is also
possible that the smaller mussels have higher mortality rates.
Page 18 of 19
Table 3.1: Information about the mussel samples for Elliptio complanata (includes mussels that were
alive as well as dead) in the Mill and Manhan Rivers in the summer and fall of 2015.
Interpretation: For E. complanata mussels, the Mill River had greater density and larger size of mussels
compared to the Manhan River in both seasons. The density and size of mussels decreased from the
summer to the fall for both sites. The density decreased 2-fold for the Mill River, but decreased 11-fold
for the Manhan River. It is not known why the density of mussels decreased so much more at the Manhan
River, but numerous animal tracks (most likely raccoon) were seen all around the Manhan River site
during the fall sampling. It is suspected that predation may play a role in the decrease in density from the
summer to the fall, and predation may be stronger at the Manhan River site. From the summer to the fall,
the median length decreased by 1.3-fold and 1.1-fold for the Mill and Manhan Rivers respectively. The
decrease was similar between sites, and may also be due to predation if predators generally choose larger
mussels. The reason the mussels at the Mill River site are so much larger than the Manhan River (a
difference of 23.3 cm in median shell length in the summer) may also be due to greater predation at the
Manhan River site. It is also worth noting that a very small percentage of mussels were found at the
surface (7-25%), and the percentage of mussels at the surface decreased from summer to fall for both sites
and the percentage went down more of the Mill River (decreased 3.4-fold) than the Manhan River
(decreased 1.8-fold). This small percentage of mussels at the surface is important to note because a
sample taken from just seeing what is visible from the surface would vastly underestimate the population.
Overall Conclusion for Part 3: So far, there is no evidence of a large negative impact of using the sluice
gate in the Paradise Pond Dam on the freshwater mussels in the Mill River. While the sluice gate was
opened once at the end of September in 2015 during a large storm, there does not seem to be any
indication of higher mortality rates for the mussels in the Mill River (Impact sample) compared to the
Manhan River (Control sample). In fact, the density of mussels went down much more in the Manhan
River than in the Mill River from July to October in 2015 (Table 3.1), and there were more Eastern
pondmussels (Ligumia nasuta) in the Mill River in October than in July in 2015 (Figure 3.5), which is
important because this is a species of Special Concern in Massachusetts.
Page 19 of 19
Literature Cited
Hilsenhoff, W.L. 1987. An improved biotic index of organic stream pollution. Great Lakes Entomol.
20:31-39.
McLain, D. 2006. Effects of upstream, dredging on a mussel bed in the Mill River at Arcadia Wildlife
Santuary.
Strayer, D. L., and D. R. Smith. 2003. A guide to sampling freshwater mussel populations. American
Fisheries Society, Monograph 8, Bethesda, Maryland.
Stream Biomonitoring Unit Staff. 2012. Standard Operating Procedure: Biological Monitoring of Surface
Waters in New York State. New York State Department of Environmental Conservation Division of
Water http://www.dec.ny.gov/docs/water_pdf/sbusop12.pdf