Oceanography | Vol. 38, No. 2
70
To compensate measurements for the temperature of the
sample volume, the slopes of each fit line are plotted, and a best-
fit line is calculated. From the parameters of this fit, the all-cause
temperature exponent can be approximated. The PIXIE achieved
an approximate temperature exponent of –0.019°C–1 for RWT,
consistent with previous results (Park et al, 2023). For Chl-a, the
approximate temperature exponent obtained was –0.008°C–1,
which is consistent with equivalent temperature parameters for
Chl-a from the literature (Watras et al., 2017). The coefficient of
determination for these curves exceeded 0.99.
The standard deviation of each 0 ppb (blank) data point was
calculated. Applying the calibration curves to these blank stan
dard deviations, the limit of detection was taken as three times
the worst-case standard deviation across blanks. The limits of
detection were found to be 0.01 ppb for RWT and 0.02 ppb for
Chl-a. The upper limit of the PIXIE’s RWT detection range was
found to be 58.9 ppb at 5°C, whereas it was 87.7 ppb at 20°C.
For Chl-a, the upper limit of the detection range was found to
be 78.8 ppb at 5°C, whereas it was 90.2 ppb at 20°C. Further
details about these results can be found in the PIXIE Complete
User Guide on at the GitHub link. Bubbles were observed during
two of the 40 ppb Chl-a calibration trials, resulting in anoma
lously high standard deviations in those data sets as the bubble
periodically stirred through the sample volume. The inset plot of
Figure 3c illustrates 1-sigma error bars on the 40 ppb, 5°C data
point. This contrasts with the 10-sigma error bars on the 30 ppb,
5°C RWT data point illustrated in the inset plot of Figure 3a.
To directly compare the cross-sensitivity of the two chan
nels to the opposite fluorophore, the PIXIE’s RWT channel mea
sured a standard of Chl-a and vice-versa. A complete descrip
tion of the comparison and its results (Figure S1) is available in
the online supplementary materials.
Field Deployment
Figure 4 depicts the first deployment of the PIXIE in August 2023
during the dye tracer release experiment. Three key stations’
worth of collected data are presented. Pre- and post-release data
are provided to illustrate the PIXIE’s multichannel functional
ity. The pre-release station was chosen based on depth and on
the availability of historical Chl-a profile data (Giesbrecht and
Scrutton, 2018), whereas the post-release stations were chosen
for their strong near-surface RWT concentrations as sampled by
the Niskin bottle, so that the measured profile and bottle samples
can be meaningfully compared. The PIXIE data were downsam
pled to 1 sample per second to parallel the calibration results and
to align with the logger’s depth sensor data. With default gain
settings, the device saturation limits at 18°C are 82.3 ppb and
88.5 ppb for RWT and Chl-a, respectively. The channels’ data
are calibrated assuming a uniform temperature of 18°C. This
temperature is chosen to match the recorded in situ tempera
ture of the surface Niskin bottle samples, rounded to the near
est degree. The August surface bottle samples fell within a range
of 16° to 20°C, so the theoretical error in measurement is no
more than 5.1% (Smart and Laidlaw, 1976) for RWT and even
8/8/2023 16:14
b
8/10/2023 14:21
c
8/10/2023 16:54
d
FIGURE 4. (a) GPS paths of Eastcom on August 8, 2023 (violet path) and August 10 (indigo path). Stations 1, 2, and 3 indicate locations of the vertical
profiles in (b), (c), and (d), respectively. Inset: Station 2, magnified, indicates Tufts Cove Power Generation Station, the release site of the RWT. Red text in
the following indicates UTC time of profile start. (b) August 8 (afternoon) pre-release profiles show zero RWT response (top) and a Chl-a profile (bottom).
Solid curves indicate downcast, and dot-dashed curves indicate upcast. (c) August 10 (morning) post-release profiles stationed next to the RWT outflow;
RWT channel saturates near the surface, decently agrees with Niskin bottle sample (black asterisks). (d) August 10 (afternoon) post-release profiles were
made further along the anticipated path of the dye plume. Note modest underestimation of Niskin bottle RWT sample.