26 / 59
Ref. No. [UMCES] CBL 2016-015
ACT VS16-06
26
Table 4.
Summary of regression statistics for the EXO2 versus reference sample response curves for each
of the nine laboratory trials.
Test ID
R2
y Intercept
Slope
LT15S00
.999
-0.046
.998
LT15S10
.999
-0.123
1.036
LT15S35
.991
-0.133
1.080
LT04S00
1.000
-0.151
1.019
LT04S10
.999
-0.111
1.019
LT04S35
.988
-0.117
1.062
LT30S00
1.000
-0.086
1.022
LT30S10
.999
-0.063
1.043
LT30S35
.996
-0.059
1.094
Results of the 56 day long-term stability and thermal stress challenge for the EXO2 are
shown in figure K. The instrument was maintained in a well circulated tank and oxygen content
manipulated by alternately varying water temperature set point between 15 and 25
o
C several times
per week of deployment. The time series of instrument readings at 15min intervals is plotted
against discrete values for Winkler reference samples (
top panel
) along with the time series of the
difference between instrument and reference measurements (
bottom panel
). The overall mean
difference between measurements was 0.001 (s.d. = 0.326) mg/L for 68 comparisons (out of a
potential total of 75). A low power fault in the EXO occurred on 03/24/15 13:30 and no data were
generated for the final three days of the test. There was no significant trend (slope = -0.0007
mg/L/d) in accuracy over time that would indicate performance drift; however the magnitude of
offset clearly increased after approximately 30 days. It is unknown whether the increased offset
was related to the same component that led to the ultimate power fault.
Results for a sensor response time assessment of the EXO2 are shown in figure L. The top
panel depicts the time series of 10s instrument reads during transfers between adjacent high (9.6
mg/L) and low (2.0 mg/L) DO water baths, maintained commonly at 15
o
C. The bottom panel
(
lower left
) depicts results fit with a 3 parameter exponential decay function: DO
rel
= DO
relMin
+ a
e
-
bt
and indicated τ calculated from fit. Data for low DO to high DO transitions (
lower right
) were
treated similarly but normalized to steady state value in subsequent high DO tank and subsequently
fit with an analogous 3 parameter exponential rise function: DO
rel
= DO
relMin
+ a(1-
e
-bt
) with
indicated τ being directly calculated from fit. The calculated τ
90
was 36 s during high to low
transitions and 26 s for low to high transitions covering a DO range of approximately 8 mg/L at a
constant 15
o
C. However, we acknowledge the EXO2 was incorrectly programmed for a sampling
rate of 10 seconds instead of its highest frequency of 1 s and this may affect the calculated
response rate.