June 2025 | Oceanography
71
less for Chl-a. This would slightly overestimate the RWT con
centrations in deeper/colder water, but no RWT was detected
from the bottom-depth Niskin bottle samples.
The map provided in Figure 4a illustrates the GPS paths of the
diving vessel Eastcom used for the deployment. The labeled sites
indicate the profile locations: Stations 1, 2, and 3 in Figure 4b,
4c, and 4d, respectively. The bright colored curves indicate mea
surements while downcasting, whereas the darker dot-dashed
curves indicate upcasting. The results from the Niskin bottle sam
ples are indicated with black asterisks. The bottom-depth Niskin
bottle samples were found to have no detectable RWT across the
entire data set, in agreement with the PIXIE measurements.
The profiles in Figure 4b were captured on August 8, 2023,
two days prior to dye release. No dye was detected while some
Chl-a was detected at Station 1, the location of the DRDC
(Defence Research and Development Canada) Atlantic Acoustic
Calibration Barge. The vertical Chl-a profile measured by the
PIXIE shows qualitative agreement with historical observations
(Giesbrecht and Scrutton, 2018). The Chl-a concentration maxi
mized by 10 m depth and returned to zero/background by 15 m
depth on downcast, though the significant difference in time of
day, season, and year confounds their quantitative comparison.
The upcast profile appears in sharp contrast to the presented
and historical downcast profiles. Because this station achieves
the greatest cast depth of 45 m among the dataset, this discrep
ancy between downcast and upcast may indicate a pressure hys
teresis effect (Shigemitsu et al., 2020) that is uncharacterized and
warrants future investigation.
The profiles in Figure 4c were captured on August 10, 2023,
during the RWT release at Station 2, directly in front of the
Tufts Cove Power Generating Station effluent where the RWT
was released. The RWT channel saturates immediately below
the surface (>82.3 ppb RWT), confined to an apparent stratum
between 1 m and 5 m depths. This indicates a subduction of the
RWT plume that can be confirmed visually in Figure 1, but the
exact mechanism of this stratification is beyond the scope of
this article. The surface-depth Niskin bottle sample recorded an
RWT concentration of 217.8 ppb, clearly in excess of the PIXIE’s
saturation limit. A second, near-surface bottle sample (3.7 m)
recorded a concentration of 10.7 ppb, in good agreement with
the PIXIE’s measurement of 14.5 ppb. The discrepancy between
bottle and PIXIE measurements at this depth could be attributed
to the difference in interrogated volume at this point. The profile
suggests that the bottle sample was taken at the edge of a steep
RWT gradient. The point sampling of the PIXIE’s measurement
is therefore more sensitive to depth than the ~1 m concentration
gradient over which the Niskin bottle averages.
The profiles in Figure 4d were captured on August10, 2023,
three hours later at Station 3, along the anticipated path of the
RWT plume. The RWT channel detected a weaker but cer
tainly present signal (10 ppb) in the first 2 m and returned to
zero by 5 m depth. The surface Niskin bottle sample recorded
an RWT concentration of 15.7 ppb, in modest agreement with
the PIXIE’s measurement. The Chl-a channel shows a simi
lar characteristic to that observed at the previous station, with
no apparent dependence on the presence/absence of the large
(>200 ppb) RWT plume.
To further validate the performance of the PIXIE, the
August 10, 2023, profile at Station 3 can be compared to the near
est RWT transects captured by the ecoCTD, occurring just after
the Niskin bottle samples were collected. See the online supple
mentary materials for a summary of the comparison of the two
sets of profiles along with a waterfall plot (see Figure S2).
CONCLUSIONS
The PIXIE is a low-cost, open-source, multichannel fluoro
meter that demonstrates performance comparable to indus
try standards. It can be assembled at a cost to the end user of
$741.38 USD per channel on average, and alternate configu
rations can be even less expensive. While this cost should not
be compared to the internal cost-per-unit of industrial in situ
fluorometers and the end user must consider the value of the
support and quality assurance industrial devices enjoy, the
PIXIE nevertheless represents an open-source option with sim
ilar performance and a low barrier to entry. The PIXIE’s limit of
detection is 0.01 ppb RWT and 0.02 ppb Chl-a, which is on par
with other in situ fluorometers. The PIXIE was successfully field
deployed and validated as a part of a dye-tracer experiment in
Halifax Harbor. The full availability of the PIXIE’s source files,
from hardware to firmware, allows the end user to customize
the PIXIE as much or as little as desired. The PIXIE makes a
transformative leap in accessibility that can meet the growing
demands for spatio-temporal data from our planet’s waterways,
without sacrificing measurement quality.
POSSIBLE FUTURE DEVELOPMENT
A road map of future work is proposed within the PIXIE
Complete User Guide available on the GitHub project page.
Hysteresis has been identified (Briggs et al., 2011; Cetinić et al.,
2012) as a common problem in fluorometers and similar in situ
devices, and the degrees along which the PIXIE exhibits it should
be studied explicitly. With some modifications to the front end,
the PIXIE could include turbidity and backscattering as poten
tial channel types along with its current fluorometric channels.
Internal temperature sensing can be integrated through firm
ware, and external (in situ) temperature sensing could be per
formed in place of one of the fluorometric channels with only
minor hardware changes. More details toward each of these pro
posed areas of future work can be found on GitHub.
SUPPLEMENTARY MATERIALS
The supplementary materials are available online at https://doi.org/10.5670/
oceanog.2025.309. To access the PIXIE fluorometer files on GitHub, go to:
https://github.com/KylePark0/PIXIE/tree/main.