LABORATORY EVALUATION OF AN IMPROVED
SAMPLING PROCEDURE FOR DISSOLVED OXYGEN
ENVIRONMENTAL PROTECTION AGENCY
NATIONAL ENVIRONMENTAL
RESEARCH CENTER
ALASKA WATER LABORATORY
College, Alaska
-------
LABORATORY EVALUATION OF AN IMPROVED SAMPLING PROCEDURE
FOR DISSOLVED OXYGEN IN STREAMS IN EXTREMELY COLD CLIMATES
by
Frederick B. Lotspeich
and
Eldor W. Schallock
for the
ENVIRONMENTAL PROTECTION AGENCY
ALASKA WATER LABORATORY
COLLEGE, ALASKA
Working Paper No. 15
-------
A Working Paper presents results or investigations
which are to some extent limited or incomplete.
Therefore, conclusions or recommendations--
expressed or implied—are tentative.
-------
INTRODUCTION
Oxygen dissolved in stream waters is essential to most forms of
aquatic life. Under some environmental conditions natural concentra-
tions may be less than optimum. This is especially true for some Alaskan
streams in winter when photosynthetic activity is reduced to insignifi-
cance by thick ice cover and low to zero light intensity caused by
short daylight periods and low solar angles. Even when water temperature
remains at 0°C some oxygen is required for respiration processes by most
aquatic organisms. In Alaska it has been documented that, under the
prolonged ice cover caused by long winters, dissolved oxygen (DO) may
be reduced to less than 2 mg/1 (Frey, 1969; Frey, Mueller, and Berry,
1970). Such low concentrations are well below the level of 5 mg/1
recommended by the Committee for Water Quality Criteria (1968) for
supporting most aquatic life, and any additional oxygen demand caused
by pollutants could be disasterous to aquatic communities. The natural
occurrence of such low concentrations becomes especially relevant when
establishing water quality standards.
Because of this apparent natural occurrence of depressed DO con-
centrations in waters of Interior Alaska, a field study was designed and
completed in the winter of 1970-71 to measure dissolved oxygen in some
representative streams (Schallock and Lotspeich, 1972). Data gathered
by this study verified that levels may range from near zero to saturation
even during a prolonged cold winter. Moreover, because of very cold
temperatures, standard sampling procedures cannot successfully be used
and filling unstoppered sample bottles by immersion with turbulent dis-
placement of air (or other gas) may be required. It was suspected that
at these low concentrations, errors inherent in the method might become
significant.
-------
Because of this suspected error, a new technique of utilizing a
nitrogen-filled sample bottle was designed by Schallock (1972). Compar-
ison of nitrogen-filled bottles with air-filled bottles was made when
measuring low DO during winter field conditions. Results of this work
clearly established that DO concentrations sampled by displacing
nitrogen were consistently and significantly lower than by air displace-
ment. However, these data did not establish, which method was more
accurate. The project described here was designed as a laboratory study
to determine which method is the more accurate and to elucidate on the
causes of differences in determinations between nitrogen and air dis-
placement.
-------
PROCEDURE
A synthetic-rubber coated 55-gallon drum filled with tap water was
placed in a cold room maintained at a temperature of 0°C. A small
stirrer was mounted on the edge of the drum and allowed to run continu-
ously to maintain gentle stirring. The drum was covered with a piece
of clear, rigid plastic to prevent drops from splashing out during
purging and to reduce evaporation. Nitrogen or oxygen gas was bubbled
through the water in the drum to regulate the concentration of dissolved
oxygen at any predetermined level. Gases entered the room through
plastic tubing from cylinders placed outside, and were introduced into
the water at the bottom of the drum through two sintered glass diffusers.
A Yellow Springs Instrument Co. (YSI) DO sensor was used to monitor the
DO of the water during bubbling by nitrogen or oxygen. A YSI DO meter
was mounted outside the cold room to allow it to remain at normal room
temperature during DO measurements.
It was noted during initial removal of DO by nitrogen bubbling
that at high DO levels, nitrogen effectively removed DO in a short time
(within minutes) at small flow rates. As the concentration became lower
(less than about 5 mg/1) removal took more time and higher nitrogen flow
rates. At the lowest DO level (near 0.1 mg/1) strong turbulence was re-
quired to remove oxygen.
Three techniques were used to fill 300 ml BOD bottles: (1) a
siphon similar to that described by Magnuson and Stuntz (1970), (2) a
clear plastic Van Dorn bottle, and (3) immersion of the bottles mouth
side up; removing stoppers after immersion. All sampling was performed
in triplicate. Before sampling, bottles were filled with argon or
3
-------
nitrogen from compressed gas cylinders through a tube held at the
bottom of the bottle for 0.5 minutes. After filling with gas, bottles
were stoppered with ground glass stoppers. In addition, air-filled
bottles were also used. A standard sequence of sampling was maintained:
first the siphon, with least disturbance; then the Van Dorn bottle; and
last immersion. After sampling was completed, six bottles—three filled
with nitrogen and three with ail—were again filled by the siphon method
to determine the amount by which DO was increased owing to disturbance
caused by sample collection. Argon was not used at this step because
preliminary interpretation of data indicated that argon and nitrogen
gave nearly identical results.
After dissolved oxygen sampling was complete, a 500 ml sample was
collected for pH, alkalinity, and conductivity; these measurements were
completed within 10 minutes to eliminate changes caused by increased
temperature. All water removed during sampling was replaced with
similar water from an 8 liter container kept at the same temperature.
Immediately after samples were collected DO was fixed using the
azide modification of the Winkler method (Standard Methods, 1971, pp.477-
481). For very low concentration of DO (<1 mg/1) the titrant was diluted
by a factor of 10 to maintain volumetric accuracy; i.e., the titrant
was approximately 0.0025N instead of 0.025N. Triplicate values for each
sample were arithmetically averaged. Statistical tests for significance
were applied using the t-test for related measures (Bruning and Hintz,
1968, pp. 12-15) which determines the significance of differences between
two correlated means.
-------
RESULTS AND DISCUSSION
Although both nitrogen and argon were used as displaced gases,
results derived from each were nearly identical and differences in
averaged triplicate values between the two were small (usually <0.05
mg/1, only 3 values showing differences of 0.1 mg/1 out of 84 groups)
and may be considered insignificant compared to consistently greater
differences between these gases and air. The t-test (Table 1) showed
that differences in the behavior of these gases were not significant
at the 80% confidence level for the siphon method. Both behaved simi-
larly even though argon is about 1.5 times as dense as nitrogen; these
comparisons lead us to conclude that results when using nitrogen may
be valid and this gas may not be removing significant amounts of DO
from the water during sample collection except at high DO concentrations.
TABLE 1
T-Test for Related Measures to Determine the Significance of Difference
Between Two Correlated Means (Bruning and Hintz, 1968)
Sampling
Method
siphon
nitrogen
filled
immer-
sion
Comparing
N2 vs air
N2 vs Ar
air vs Ar
siphon vs Van Dorn
siphon vs immersion
Van Dorn vs immersion
N£ vs air
N? vs Ar
air vs Ar
tRM
2.44349
0.62988
2.95716
2.58995
1.23951
2.17057
19.19287
1.07853
19.00418
T-Values for a Range
of Confidence Levels
T(0.001
T(0.01)
T(0.02)
T(0.05)
T(0.10)
T(0.20)
) = 4.587
= 3.169
= 2.764
- 2.228
= 1.812
= 0.700
-------
The close similarity of data using nitrogen and argon becomes
apparent when compared with air using the immersion method (Figure 1).
Data for siphon and Van Dorn methods are so similar that they are not
displayed graphically but are recorded in tables to be discussed later.
From Figure 1 it is quite apparent that DO measured when air is the dis-
placed gas is consistently about 0.3 to 0.5 mg/1 higher than for either
of the other two gases. The t-test was applied to compare argon with
air, nitrogen with air, and nitrogen with argon using the immersion
method. Differences between air and argon or nitrogen were highly sig-
nificant at the 99% confidence level; however, differences between argon
and nitrogen were not significant at the 90% level level. At DO concen-
trations above that normally found in streams without an ice cover this
introduces errors ranging from about 10% at 5 mg/1 to 3% near saturation
at 0.0°C. Below 5 mg/1 errors in the air immersion method increase to
about 300% at the concentrations measured in several unpolluted streams
of Interior Alaska during winter (Frey, 1969, Frey et aj_. 1970). Thus
it becomes extremely important that there be a more accurate sampling
technique for DO at these low concentrations. A final bar graph com-
paring these gases using the siphon method is shown in Figure 1; these
differences are very small at supersaturation. Only the immersion method,
with its attendant turbulence, appears to be inaccurate.
Data comparing sampling methods, using air as the displaced gas, are
given in Figure 2, and show that immersion gives a consistent error at
DO concentrations less than about 9 mg/1. Above this concentration all
three methods produced varying results, near or above saturation the
immersion method gave low values. At low DO concentrations the immersion
method evidently introduces oxygen from the displaced air to the water.
-------
16
1*5-
14-
R-
\C
1 1-
i(-i_
IU
_j y
CD o
2 °
O 7—
2 7
4_
0
CD N NITROGEN
Do AIR j_^
B A ARGON
m
^•W
m
l_l
rrTL
n ,
IN
N o A N a A N
-
Q
^•i
A
i
N
^H
_
a
=
^H
^M
^M
A
•
N
a
=
•••
^H
A
•
MM
•^
^
^
•
^B
•^
•^
••
MM
•M
•
«••
>^
^B
^B
^B
!••
^^
•^
^B
^B
M^
••
^H
MB
••
<^
^M
1^
••
^H
^H
•^
^H
•^
•^
•^
1
=
»
^
=
^
=
=
=
^
=
=
^
=
^
^
=
^
=
=
-
-
^
=
NaA NaA NaA NaA NaA N <^
16
-IR
M
H
Ti.
II
-in
9
q
c;
4
JL
\
0
Figure 1. Comparing dissolved oxygen in tap water (sampled by inversion) using air, nitrogen,
and argon as the displaced gases. One siphon sampling at the highest DO level is included
to show how all gases behave with non-turbulent sampling.
-------
00
IC_
l/l-
II
19-
1 1-
ift-
"7 Q-
_J *>
O o_
2°
v^
O*7_
0
6_
5~
4_
~
2_
i _
ft
gi
CD S SIPHON r
D V VAN
DO
RN
3 1 IMMERSION
0 N NITROGEN
H a AIR
*•!
'^r-R
__n nri
in ri ill FH
f=
^H
^"
•M
1
1 ^M
_
•^
y—
p^
^m
^M
^™
=
^
=
••
.f—
tea
^
i
z
^
^
^
^
•M
mpmm
_
^
•^
^
•=
=
s
=
p
•h—
•
~
^
=
E
7
/
/
/
X1
/
^
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
S
S
s
S
^
s
s
s
s,
s
s
s
s
s
s
s
s
\
s
s
s
s
s
s
s
\
s
\
s
\
>
\
s.
•^d
H6
lu
14
if.
It
1
_IA
"I
ft
SV I SV I SV I SV I SV I SV I SV I SVI SVI SVI
Na
after
Figure 2. Comparing dissolved oxygen in tap water (using air as the displaced gas) using
siphon, Van Dorn, and inmersion sampling procedures. One additional run using nitrogen as
the displaced gas is included to compare sampling methods at high DO concentrations.
-------
This tendency decreases as DO levels approach saturation, where turbu-
lence may remove some DO. The bars for nitrogen in Figure 2 indicate
that immersion actually strips oxygen from the water; although not
graphed, similar results were found for argon. Siphon samples taken
after all techniques were completed are also shown as the final bars
and closely approach DO levels taken with minimum disturbance. This is
interpreted as indicating that little oxygen was introduced or removed
from the container by disturbance from sampling by all three methods.
Table 2 presents complete data from tap water and suggests that
a positive error is introduced by sampling with the Van Dorn bottle;
this error increases at high DO. This is interpreted as being caused
by the greater disturbance of Van Dorn sampling compared to the very
quiet siphon. Other parameters measured include pH, alkalinity, and
specific conductance; of these, only pH shifted significantly as DO was
lowered. Stripping with bubbling nitrogen evidently also strips out
carbon dioxide, causing an increase in pH as (#2 is removed. This
removal of dissolved gases by stripping does not simulate stream dynamics
and under natural conditions pH is not necessarily increased at extremely
low DO. In addition to the above observations, this table shows the
close agreement of DO measurement using argon and nitrogen in sampling
bottles by each of the three methods. At low DO concentrations (
-------
TABLE 2
Dissolved Oxygen Sampling Evaluation (Tap Water)
Gas
Dis-
placed
Dissolved Oxygen (mg/1)
Sampling Method
Siphon
Van Dorn
Dunking
Siphon
After
Other Parameters
Temp.
°C
Cond.
umoh/cm
PH
Alk.
mg/1
N2
air
A
N2
air
A
N2
air
A
N2
air
A
N2
air
A
N2
air
A
N2
air
A
N2
air
A
16.47
16.50
16.47
13.28
13.32
13.34
10.11
10.13
10.05
6.96
6.99
6.92
4.95
5.01
4.96
3.85
3.93
3.81
3.15
3.22
3.11
2.00
2.10
1.98
16.40
16.51
16.39
13.32
13.37
13.29
10.13
10.20
10.09
7.01
7.01
6.96
5.05
5.11
4.99
3.96
3.88
3.25
3.23
3.25
2.09
2.21
2.03
15.88
16.37
15.91
12.97
13.30
13.01
9.85
10.18
9.78
6.82
7.26
6.81
4.87
5.40
4.84
3.84
4.21
3.80
3.17
3.54
3.14
2.08
2.63
2.08
16.41
16.44
12.93
10.09
7.03
5.02
5.08
3.93
3.92
3.24
3.25
2.11
2.12
0.8 500 - 78
480 8.05 85
0.3 500 7.85 86
0.3 500 8.10 86.5
1.0 500 8.00 86.0
0.2 500 8.00 85.0
0.3 500 8.30 84.5
0.8 500 8.60 81.0
air
A
aTr
A
&
A
97
03
95
0.93
0.98
0.94
0.11
0.20
0.20
2.05
2.13
2.06
0.97
1.07
1.00
0.23
0.20
0.23
2.03
2.53
2.02
1.04
1.52
1.05
0.27
0.70
0.22
2.05
2.13 0.3 500 8.30 84.5
1.02
1.03
0.20
0.24
0.2 500 8.60 87.0
0.8
500
8.6 81.0
10
-------
t-test (Table 1) was applied to the differences among sampling tech-
niques using nitrogen as the displaced gas. These tests indicate that
differences between the siphon and Van Dorn bottle are insignificant
above the 95% confidence level, for Van Dorn and immersion above the 90%
level, and siphon versus immersion above the 80% level; correlation be-
tween any two methods was consistently above 0.999. This is interpreted
as indicating that all three methods will give reliable DO results when
nitrogen is used as displaced gas.
Data derived in identical sampling procedures are shown in Table 3
for Chena River water collected above 80 river miles above Fairbanks.
Although alkalinity and specific conductance are significantly less
than for the tap water used in the earlier run, results and conclusions
are similar. This suggests that chemical properties of water do not
significantly influence the sampling of DO, at least at the concentra-
tions measured in these two waters.
In its lower reaches, Chena River receives domestic waste waters
from primary treatment plants from Fort Wainwright and Fairbanks. This
limited treatment results in a pollution load many times higher than
that present in reaches upstream. To test the possible effects of this
pollution on sampling methods and avoid possible criticism that BOD
from pollution might influence DO mechanisms in our laboratory procedure,
100 gallons of river water from below Fairbanks was brought into the
laboratory and treated in the same manner as that described earlier.
Data generated using this water are recorded in Table 4. Despite the
distance separating these sampling stations (about 80 to 100 river miles)
with two major tributaries entering between these stations and the fact
11
-------
TABLE 3
Dissolved Oxygen Sampling Evaluation (Upper Chena River)
Gas
Dis-
placed
Dissolved Oxygen (mg/i;
Sampling Method
Siphon
Van Dorn
Dunking
Siphon
After
Other Parameters
Temp.
°C
Cond.
umoh/cm
PH
Alk.
mg/1
air
A
A
N2
atr
A
air
A
N2
atr
A
N2
atr
A
air
A
2?r
A
air
A
17.62
17.70
17.63
14.00
14.04
13.96
12.50
12.55
12.46
10.09
10.12
10.07
8.19
8.26
8.17
5.81
5.84
5.77
3.56
3.60
3.54
2.11
2.13
2.10
0.16
0.23
0.17
17.62
17.59
17.57
13.96
14.03
14.93
12.48
12.52
12.56
10.04
10.14
10.02
8.20
8.29
8.16
5.83
5.91
5.80
3.63
3.72
3.59
2.16
2.24
2.18
0.24
0.37
0.24
17.11
17.64
17.08
13.53
14.02
13.48
12.10
12.56
12.07
9.80
10.22
9.76
7.97
8.50
7.98
5.72
6.20
5.68
3.57
4.03
3.52
2.14
2.63
2.13
0.36
0.72
0.26
17.57
17.67
13.92
13.91
12.39
12.46
10.04
10.10
8.23
8.26
5.80
5.91
3.61
3.70
2.18
2.21
0.26
0.33
0.6 200 7.80 57.0
0.6 200 7.80 57
0.5 200 7.70 56.5
1.2 210 7.75 57.0
0.5 210 7.90 56.5
0.4 200 7.85 57.0
0.9 210 7.90 57.0
0.9 200 8.15 57.0
1.0 200 8.40 57.0
12
-------
TABLE 4
Laboratory Study of DO (Lower Chena)
Gas
Dis-
placed
Dissolved Oxygen (mg/1 )
Sampling Method
Siphon
Van Dorn
Dunking
Siphon
After
Other Parameters
Temp.
°C
Cond.
umoh/cm
PH
Alk.
mq/1
N2
air
A
No
a?r
A
N2
air
A
N2
air
A
N2
air
A
N2
air
A
No
air
A
N2
air
A
13.37
13.50
13.36
16.89
16.97
16.89
10.69
10.79
10.71
6.80
6.89
6.81
5.04
5.13
5.05
2.87
2.96
2.85
1.19
1.27
1.15
0.43
0.47
0.39
13.38
13.44
13.37
16.90
16.99
16.84
10.73
10.79
10.72
6.83
6.93
6.85
5.11
5.17
5.14
92
09
99
38
33
48
12.99
13.46
13.09
16.38
16.83
16.32
10.44
10.52
10.37
6.70
6.62
6.66
4.
5,
99
46
4.96
,89
,45
,90
,26
,73
,30
0.50
0.70
0.56
0.60
1.12
0.56
13.30
13.41
16.77
16.85
10.70
10.78
6.85
6.89
5.08
5.16
2.96
3.08
1.28
1.35
0.51
0.63
0.2
0.1
0.2
0.1
1.1
1.1
0.5
1.2
200
200
190
180
185
7.8 72.5
7.8 71
7.8 71
7.7 69
190 8.0 69
8.0 69
185 8.3 69
185 8.4 69
13
-------
that one site is below a seriously polluting source, data in Table 4
closely resembles that in Table 3. There are no significant differences
in the behavior of oxygen in these sampling methods between the two
waters.
In sampling by the siphon and Van Dorn bottle, the recommended
procedure is to allow 1 to 2 volumes of water to overflow by displacing
water from the bottom of the sample bottle. This procedure is usually
impossible in winter because of the intense cold that may freeze samples
in less than one minute. To evaluate the error that might result from
not following this procedure, one and two volumes of water were dis-
placed using the siphon method and tap water. Applying the t-test to
this procedure showed that differences between filling and overflowing
one and two volumes were not significant at the 90% confidence level
for either air or nitrogen. For nitrogen, the differences were not
significant at the 80% confidence level. From the test of these data
it may be concluded that any error introduced by not overflowing (which
is impossible by the immersion procedure) is insignificant.
14
-------
CONCLUSIONS
Dissolved oxygen concentrations measured in water samples at 0°C
temperature, collected by immersion with air as the displaced gas, are
significantly higher at low concentrations than when another gas is
displaced. This leads to the conclusion that at very cold temperatures,
where immersion must be used, nitrogen or other gas-filled sample bottles
should be used. At DO levels near saturation and above, both nitrogen
and argon give low values, evidently causing by stripping of easily
removed oxygen at these concentrations.
Quiet displacement of any gas is preferable to the turbulence
caused by immersion. Thus, if the siphon technique could be adapted
to very cold temperatures, thick ice, and varying water levels, it would
be the preferred method. However, results using nitrogen-filled bottles
with the immersion sampling method are comparable to those derived by
alternate recommended procedures which may not be feasible under extreme
winter cold. The immersion method is easily adapted to a wide range
of transportation modes; it has been used with a pickup truck when the
temperature was -44°C and fixed wing airplane at temperatures as low
as -30°C. It is doubtful if other methods would prove as satisfactory
under these conditions. Finally, some disturbance of the water surface
during sampling does not appear to significantly increase the oxygen
content of the water, even at very low DO concentrations.
Results acquired during the course of this investigation appear
independent of sample source. One, tap water, is a treated ground water
and the other two were from the Chena River, one from about 75 river
miles above Fairbanks—unpolluted—and the other about five miles down-
stream from Fairbanks and strongly polluted. This suggests that, within
15
-------
the range of conditions encountered on these waters, dissolved oxygen
can be measured using the same techniques without special precautionary
measures.
Of the three chemical parameters measured only pH appeared to
undergo a significant change. This change only occurred at very low
DO concentrations where pH increases about 0.5 units. This increase
may be the effect of removing COp by Np scrubbing and is not believed
to be directly related to DO concentrations at these low concentrations.
Elimination of the overflow portion of recommended procedures
(extremely difficult under our climatic conditions) does not introduce
a serious inaccuracy with this procedure and can be neglected in evalu-
ating field data acquired under winter conditions.
16
-------
REFERENCES
Anonymous, 1971. Standard Methods for the Examination of Water and Waste-
Water, 12th edition, 874 pp.
Bruning, James L. and Kintz, B.L. 1968. Computational Handbook of
Statistics. Scott, Foresman and Co. 369 pp.
Frey, Paul J., 1969. Ecological Changes in the Chena River. USDI, FWQA,
Alaska Water Laboratory, College, Alaska.
Frey, Paul J., Mueller, Ernst W., and Berry, Edward C., 1970. The Chena
River; A Study of a Subarctic Stream. USDI, FWQA, Alaska Water Labora-
tory, College, Alaska.
Magnuson, John J., and Stuntz, Warren E., 1970. "A Siphon Water Sampler
for use Through the Ice." Limnology and Oceanography, 15:156-158.
National Technical Advisory Committee, 1968. Water Quality Criteria.
USDI, FWPCA, Washington, D.C.
Schallock, Eldor W., and Lotspeich, Frederick B., 1972. "Dissolved Oxygen
Under Late Winter Ice Cover in Alaskan Streams." In preparation.
17
-------
|