Technical ReportPDF Available

Guidelines for processing RBR CTD profiles


Abstract and Figures

A series of post-processing steps have been developed to improve data quality from RBR loggers when used for vertical profiling. At present, RBR does not provide any official guidance or software for this purpose even though untreated profile data exhibit some classic problems such as salinity spiking. The steps developed and detailed herein apply specifically to the generation of loggers which have a black or gray cylindrical conductivity sensor coil, and a thermistor mounted on the sensor cap. However, some of the processing steps are somewhat generic to any profiler, and are therefore applicable to some degree to the newer generation of RBR instruments that feature a more hydrodynamic conductivity cell and a co-located thermistor. The post-processing steps include 1) correcting for zero-order holds (optional), 2) data despiking (optional), 3) removing atmospheric pressure, 4) low pass filtering, 5) sensor alignment, 6) descent rate thresholding, 7) derived variables, and 8) bin averaging. Most of the parameters in these steps were found by analyzing field data, as opposed to conducting laboratory tank studies or comparing the data to well-characterized reference sensors. We also provide some basic recommendations for obtaining high quality profiles in the field. While the primary focus of the processing concerns conductivity, temperature, and pressure, RBR loggers also interface with external sensors (e.g., fluorescence and dissolved oxygen). We will briefly discuss these sensors with special emphasis on the JFE Alec Co. RINKO III O2 optode.
Content may be subject to copyright.
Guidelines for processing RBR CTD profiles
Mark Halverson, Jen Jackson, Clark Richards, Humfrey Melling, Brian Hunt, Ray Brun-
sting, Mike Dempsey, Germaine Gatien, Andrew Hamilton, Wayne Jacob, Sarah Zim-
Fisheries and Oceans Canada
Institute of Ocean Science
9860 West Saanich Road
Sidney, British Columbia
V8L 4B2
Canadian Technical Report of Hydrography
and Ocean Sciences 314
Canadian Technical Report of Hydrography and Ocean Sciences
Technical reports contain scientific and technical information of a type that represents
a contribution to existing knowledge but which is not normally found in the primary
literature. The subject matter is generally related to programs and interests of the
Oceans and Science sectors of Fisheries and Oceans Canada.
Technical reports may be cited as full publications. The correct citation appears
above the abstract of each report. Each report is abstracted in the data base Aquatic
Sciences and Fisheries Abstracts.
Technical reports are produced regionally but are numbered nationally. Requests for
individual reports will be filled by the issuing establishment listed on the front cover and
title page.
Regional and headquarters establishments of Ocean Science and Surveys ceased pub-
lication of their various report series as of December 1981. A complete listing of these
publications and the last number issued under each title are published in the Canadian
Journal of Fisheries and Aquatic Sciences, Volume 38: Index to Publications 1981. The
current series began with Report Number 1 in January 1982.
Rapport technique canadien sur l’hydrographie et les sciences oc´eaniques
Les rapports techniques contiennent des renseignements scientifiques et techniques
qui constituent une contribution aux connaissances actuelles mais que l’on ne trouve
pas normalement dans les revues scientifiques. Le sujet est g´en´eralement rattace aux
programmes et int´erˆets des secteurs des Oc´eans et des Sciences de Pˆeches et Oc´eans
Les rapports techniques peuvent ˆetre cit´es comme des publications `a part enti`ere. Le
titre exact figure au-dessus du r´esum´e de chaque rapport. Les rapports techniques sont
esum´es dans la base de donn´ees esum´es des sciences aquatiques et halieutiques.
Les rapports techniques sont produits `a l’´echelon r´egional, mais num´erot´es `a l’´echelon
national. Les demandes de rapports seront satisfaites par l’´etablissement auteur dont le
nom figure sur la couverture et la page de titre.
Les ´etablissements de l’ancien secteur des Sciences et Lev´es oc´eaniques dans les r´egions
et `a l’administration centrale ont cess´e de publier leurs diverses s´eries de rapports en
ecembre 1981. Vous trouverez dans l’index des publications du volume 38 du Journal
canadien des sciences halieutiques et aquatiques, la liste de ces publications ainsi que
le dernier num´ero paru dans chaque cat´egorie. La nouvelle s´erie a commenc´e avec la
publication du rapport num´ero 1 en janvier 1982.
Canadian Technical Report of Hydrography and Ocean Sciences 314
Guidelines for processing RBR CTD profiles
Mark Halverson123 , Jen Jackson2, Clark Richards34, Humfrey Melling5, Brian Hunt1 2,
Ray Brunsting2, Mike Dempsey5, Germaine Gatien5, Andrew Hamilton1, Wayne
Jacob2, Sarah Zimmerman5
Fisheries and Oceans Canada
Institute of Ocean Science
9860 West Saanich Road
Sidney, British Columbia
V8L 4B2
1Department of Earth, Ocean, and Atmospheric Sciences, University of British Columbia, Vancouver,
BC, Canada, V6T 1Z4
2Hakai Institute, Heriot Bay, BC, Canada
3RBR Ltd., Ottawa, ON, Canada
4Fisheries and Oceans Canada, Dartmouth, NS, Canada
5Fisheries and Oceans Canada, Sidney, BC, Canada
Her Majesty the Queen in Right of Canada, 2017.
Cat. No. Fs97-18/314E-PDF ISBN 978-0-660-06003-3 ISSN 1488-5417
Correct citation for this publication:
Halverson, M., Jackson, J., Richards, C., Melling, H., Brunsting, R., Dempsey, M., Ga-
tien, G., Hamilton, A., Hunt, B., Jacob, W., and Zimmerman, S. 2017. Guidelines
for processing RBR CTD profiles. Can. Tech. Rep. Hydrogr. Ocean Sci. 314: iv
+ 38 p.
Abstract iv
esum´e iv
1 Introduction 1
1.1 Workshops................................... 1
1.2 Summary of recommended processing steps . . . . . . . . . . . . . . . . . 2
1.3 Structure of this report . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Instruments and sampling locations 3
3 RBR CTD logger principle of operation 3
3.1 Conductivity ................................. 4
3.2 Temperature.................................. 4
3.3 Pressure .................................... 5
3.4 Auxiliarysensors ............................... 5
4 Field use 5
5 Data post-processing 6
5.1 Correct for A2D zero-order hold . . . . . . . . . . . . . . . . . . . . . . . 6
5.2 Datadespiking ................................ 7
5.3 Correction for atmospheric pressure . . . . . . . . . . . . . . . . . . . . . 8
5.4 Lowpasslter................................. 10
5.4.1 Filtertype............................... 11
5.4.2 Dissolvedoxygen ........................... 12
5.5 Sensoralignment ............................... 12
5.5.1 Conductivity and temperature . . . . . . . . . . . . . . . . . . . . 13
5.5.2 Dissolvedoxygen ........................... 13
5.6 Descentrateltering ............................. 15
5.7 Derivedvariables ............................... 16
5.7.1 PracticalSalinity ........................... 16
5.7.2 Dissolved oxygen concentration . . . . . . . . . . . . . . . . . . . 16
5.8 Binaveraging ................................. 17
6 Conclusion 17
Acknowledgements 17
References 36
Appendix 36
A Example processing scheme 37
A series of post-processing steps have been developed to improve data quality
from RBR loggers when used for vertical profiling. At present, RBR does not pro-
vide any official guidance or software for this purpose even though untreated profile
data exhibit some classic problems such as salinity spiking. The steps developed
and detailed herein apply specifically to the generation of loggers which have a
black or gray cylindrical conductivity sensor coil, and a thermistor mounted on
the sensor cap. However, some of the processing steps are somewhat generic to
any profiler, and are therefore applicable to some degree to the newer generation
of RBR instruments that feature a more hydrodynamic conductivity cell and a
co-located thermistor.
The post-processing steps include 1) correcting for zero-order holds (optional),
2) data despiking (optional), 3) removing atmospheric pressure, 4) low pass filter-
ing, 5) sensor alignment, 6) descent rate thresholding, 7) derived variables, and 8)
bin averaging. Most of the parameters in these steps were found by analyzing field
data, as opposed to conducting laboratory tank studies or comparing the data to
well-characterized reference sensors. We also provide some basic recommendations
for obtaining high quality profiles in the field.
While the primary focus of the processing concerns conductivity, temperature,
and pressure, RBR loggers also interface with external sensors (e.g., fluorescence
and dissolved oxygen). We will briefly discuss these sensors with special emphasis
on the JFE Alec Co. RINKO III O2optode.
Plusieurs ´etapes de post-traitement ont ´et´e d´evelopp´ees afin d’accroˆıtre la qua-
lit´e des profils verticaux issus des CTD produites par RBR. A l’heure actuelle,
RBR ne fourni aucune recommandation, ni aucun logiciel permettant de corriger
les probl`emes classiques que l’on trouve dans les donn´ees de profil brut, comme les
pics de salinit´e. Les m´ethodes d´evelopp´ees ici s’appliquent aux mod`eles poss´edant
un senseur cylindrique noir ou gris, ainsi qu’un thermistor mont´e sur le cache du
senseur. Cependant, certaines ´etapes du traitement sont g´en´eriques, et restent ap-
plicables `a n’importe quel profiler, ainsi qu’`a la nouvelle g´en´eration d’instruments
produits par RBR, qui utilisent un capteur de conductivit´e, associ´e `a un thermistor,
ayant une meilleure hydrodynamique.
Les ´etapes de post-traitement incluent 1) la correction des “zero-order holds”
(optionnel), 2) la suppression des pics (optionnel), 3) la suppression de la pression
atmosph´erique, 4) un filtrage par utilisation de filtres passe-bas, 5) l’alignement du
senseur, 6) le filtrage des donn´ees en fonction d’un seuil de vitesse de descente 7) le
calcul des quantit´es d´eriv´ees, et 8) le calcul de la moyenne par cellules. Alors que la
plupart des ´etudes utilisent des donn´ees de laboratoire en r´eservoir, ou comparent
des donn´ees de r´ef´erence capteur, la plupart des param`etres de ces ´etapes ont ici
´et´e valid´ees en analysant des donn´ees de terrain.
Bien que le but premier du traitement concerne la conductivit´e, la temp´erature
et la pression, les CTD RBR s’interfacent ´egalement avec des senseurs externes (e.g.,
fluorescence et oxyg`ene dissous). Nous discuterons bri`evement de ces senseurs, et
particuli`erement de l’optode JFE Alec Co. Rinko III O2.
1 Introduction
RBR conductivity-temperature-depth (CTD) profilers provide high quality data straight
from the instrument without post-processing. However, as we will show, applying ad-
ditional processing steps, such as carefully aligning the sensors in time, will improve
the data and reduce artifacts such as salinity spiking. Unlike Seabird Electronics, RBR
does not presently make available any software or guidelines to help with the processing,
although their Matlab software, rsktools, will eventually have this functionality. After
speaking with staff at RBR, and a number of RBR users in the oceanographic commu-
nity, it became clear that there are steps which should be taken to ensure high quality
Staff scientists at the Hakai Institute, a scientific research institution that conducts
long-term research at remote locations on the coastal margin of British Columbia, con-
vened two workshops consisting of experienced RBR and Seabird users to share data
processing procedures and to develop a standard post-processing workflow. Application
of this workflow to data from a properly calibrated RBR profiler will ensure it is of the
highest quality, and therefore suitable for publication in peer-reviewed literature and for
distribution to data archives.
1.1 Workshops
The first workshop was help on November 19, 2015 at the Institute of Ocean Sciences
in Sidney, British Columbia, Canada. The goal of the workshop was to discuss the pro-
cessing procedures used by various organizations and researchers. Participants presented
their RBR processing strategy and experience with the instruments. The outcome was
the formation of working groups, each tasked with a specific objective related to CTD
processing (e.g., sensor alignment, filtering, descent rate). Many of the tasks are similar
to those used in the Seabird processing chain, and thus we include the Seabird terminol-
ogy for reference. The working groups were:
- Collection of field data
- Data conversion
- Atmospheric pressure
- Filter
- Align
- Descent rate / “Loop edit”
- Despiking / “Wild edit”
- Ancillary data (e.g., O2, Turbidity)
- Derived variables
- Technical paper
The second workshop was held on February 1, 2016 at the Institute of Ocean Sciences.
The goal of this workshop was to have each working group present its findings. Following
the presentations, there was an open discussion to determine the next steps. It was
decided that a technical report detailing the steps required to process RBR profiles would
be written. The data collection and processing steps suggested in the two workshops and
in the different working groups are outlined below.
1.2 Summary of recommended processing steps
Here we summarize the processing steps developed by the RBR working group. The
parameters for each step (e.g. conductivity delay) are appropriate for a 6 Hz instrument
descending at about 1 m/s with a RINKO III oxygen sensor in a coastal region. The
steps are arranged in the recommended order in which they should be applied.
1. [Optional] Zero-order hold correction (Section 5.1)
(a) Apply to instruments that have zero-order holds
2. [Optional] Data despiking (Section 5.2)
(a) Identify spikes with median filter and either interpolate over, remove, or re-
place with NaN
3. Correct for atmospheric pressure (Section 5.3)
(a) Subtract atmospheric pressure from measured total pressure. Estimate atmo-
spheric pressure by:
i. Using the total pressure measured in air before or after the cast
ii. For thresholding instruments, use the minimum total pressure at the min-
imum conductivity after the CTD is soaked
4. Filter (Section 5.4)
(a) Low-pass filter temperature and conductivity with a two-way (i.e., zero-phase)
3-point triangular window
5. Align (Section 5.5)
(a) Delay conductivity by 0.33 s (2 scans @ 6 Hz)
(b) Advance oxygen by 3.0 s (18 scans @ 6 Hz)
6. Descent rate filtering / Loop Edit (Section 5.6)
(a) Flag all sensor data when descent rate is less than 0.4 m/s and when acceler-
ation is lower than -0.1 m/s2
7. Derived variables (Section 5.7)
(a) Practical Salinity using filtered and aligned conductivity and temperature.
Ignore RBR’s calculation.
(b) Depth
(c) Oxygen concentration in units of ml/l or µmol/kg from oxygen saturation,
but only after filtering temperature, salinity, and density to match the slow
time response of the oxygen sensor
8. Bin average (Section 5.8)
(a) Average data into depth or pressure bins
(b) In coastal applications bins are typically 1 m wide, and centred on integer
values (i.e., 1 m, 2 m, 3 m, .. . )
(c) In open ocean applications users may wish to use larger bins
1.3 Structure of this report
The document is organized as follows: Section 2 summarizes the instruments and sam-
pling locations used in this report. Section 3 provides a very brief overview on the RBR
conductivity, temperature, and pressure sensors. Section 4 discusses field deployment
best practices. Section 5 describes how the post-processing steps and parameters were
determined. Appendix A includes an example of how the processing steps would be
implemented using the freely available Matlab toolbox RBRproc.
2 Instruments and sampling locations
RBR Limited, established in 1976 in Ottawa, Ontario, designs and builds a variety of
oceanographic sensors including a series of CTD loggers that samples at a sufficient rate
for water column profiling.
The processing recommendations made in this report were developed using RBR
data collected mostly by the Hakai Institute. Hakai owns and routinely profiles with
three RBR loggers: two XRX-620 (S/N 18032, 18066), and one Maestro (S/N 80217). A
fourth instrument, an RBR Concerto (S/N 065679) was on loan from the Pacific Salmon
Foundation in April 2015 for use in the Johnstone Strait program.
Clark Richards (DFO/BIO; formerly RBR) supplied profiles from a 12 Hz Concerto
(rated to 2000 dbar) taken in the Coral Sea. Sarah Zimmerman and Humfrey Melling
(DFO/IOS) provided profiles taken in the Canadian Arctic with four 6 Hz Concertos
rated to 740 dbar (S/N 65578, 65639, 65579, 65636).
3 RBR CTD logger principle of operation
The instruments in this study measure conductivity, temperature, and pressure (CTD)
at either 4, 6, or 12 Hz, and some are fitted with additional sensors to measure biogeo-
chemical parameters. The conductivity cell and the thermistor are physically separated
on these instruments, which means that a parcel of water is sampled at different times
by the conductivity and temperature sensors (Fig. 1). In 2016, RBR introduced a new
generation of the RBR Duo, Concerto, and Maestro instruments. They feature a new
conductivity cell, which was designed so that the thermistor is mounted on the conduc-
tivity cell itself. The hydrodynamics of the new cell facilitate laminar flow, as well. This
updated configuration is expected to change the processing approach, and therefore some
of the recommendations made in this report are not relevant to the newer generation of
instruments (for example, conductivity and temperature alignment). Finally, unlike most
Seabird CTD profilers, RBR profilers are not pumped. This has important implications
for aligning the sensors and for the descent rate.
Most RBR CTDs can operate in two different modes. The first is a manual deploy-
ment, where the user can manually initiate logging before the instrument is placed into
the water. The second is a threshold deployment, where the instrument initiates logging
when a specific channel crosses a threshold defined by the user. This has implications
for methods that might be used to remove atmospheric pressure (Sec. 5.3).
3.1 Conductivity
The conductivity sensor consists of a wire coil housed within a 5-cm diameter ceramic
Delrin cylinder. Its sensing volume, extending far beyond the housing, is 20 to 30 cm in
diameter, however the measurement is weighted more heavily around the centre of the
coils. The conductivity measurement is thus a weighted spatial average of the sensing
volume, which could affect the approach used to filter and align this sensor with the
smaller, but slower, thermistor. Approximately 80% of the measurement is made within
the ceramic tube, with the remaining 20% coming from the volume outside the cell.
Additional details on RBR inductive conductivity cells including calibration can be found
in Shkvorets and Johnson [2010].
The conductivity cell has a response time that depends on the flow rate through the
cell, but it is generally believed to be shorter than the thermistor time constant (Humfrey
Melling estimates 0.3 s). It is important to match the response of the temperature
and conductivity sensors to prevent salinity spiking, which in the case of RBR profilers
means “slowing down” the conductivity sensor. The relatively large sensing volume of
the conductivity sensor, as opposed to “point” measurements made by the thermistor,
means that the conductivity sensor acts somewhat as a spatial filter. Some argue that
this ultimately makes it impossible to truly match the conductivity and temperature
response, however the consensus is that some filtering and alignment considerations can
at least improve the data (i.e., reduce salinity spiking). This will be discussed further in
the alignment and filtering sections.
3.2 Temperature
The temperature sensor is a millimetre-sized thermistor encased within a glass bead,
which is located inside a metal tube that protects the thermistor from varying ambient
pressure. Its sensing volume is a few cubic millimetres. The time constant, defined here
as the e-folding time measured after the thermistor is plunged into a well-stirred bath
(producing a step change in temperature), is typically about 0.6 s. It is possible that the
time constant will be a function of descent speed.
3.3 Pressure
The strain gauge pressure sensor is located either on the sensor end cap or on the side of
the logger. The pressure sensor response time is <0.01 s – effectively zero – meaning the
pressure data do not need to be corrected for alignment or response time. RBR CTDs
measure and report absolute pressure, which is the sum of sea pressure and atmospheric
3.4 Auxiliary sensors
Many RBR profilers are equipped with third-party sensors to measure, for example,
turbidity, chlorophyll-a, photosynthetically active radiation (PAR), and dissolved oxygen.
This report does not address data from these sensors because the processing would likely
differ depending on, for example, the sensor manufacturer (e.g., Seapoint or Wetlabs
chlorophyll-a fluorescence), and their data are not typically combined to derive other
parameters in the same way salinity is calculated from temperature, conductivity, and
pressure. Dissolved oxygen, on the other hand, requires special treatment, especially
when deriving concentration from percent saturation because the results are sensitive to
differences in sensor time constants (Sections 5.4 and 5.5).
4 Field use
Although this report is primarily focused on post-processing, data quality is highly de-
pendent on proper CTD field protocol, therefore, we provide recommendations that will
help ensure the collection of high quality data.
As described earlier, the RBR CTDs use an inductive conductivity sensor. A feature
of this sensor is that it measures the conductivity in a 20 cm diameter sphere of sea
water around the sensor. Conductive objects in this sphere will offset the measured
conductivity. If a guard was installed by RBR or the user to protect the conductivity
cell and thermistor, then the CTD must be calibrated with the guard in place. All other
objects in the 20 cm field may change the calibration. If a bottom line and weight are
used, it should be attached to a frame that is more than 20 cm away from the conductivity
sensor. Such a line should be of non-metallic construction and thin as possible and should
be attached to the bottom of the guard.
Before deploying the CTD, check for objects in the field of the conductivity sensor.
Check the thin platinum wire thermistor for foreign objects. Make sure the clear acrylic
pressure port on the sensor end is free of debris. Remove the cover from the dissolved
oxygen sensor, and the fluorometer if present.
Lower the CTD into the water to initiate the 2 minute “soak”, which allows the
instrument to thermally equilibrate to the water, and for any bubbles which might be
trapped near the sensors to escape or dissolve. Ensure that the water conditions are
sufficient to initiate logging if the instrument is set to threshold mode. A common
approach is to set the pressure threshold to the pressure expected at a depth of 2 m (i.e.,
12 dbar). After 2 minutes, raise the CTD to near the surface, and begin the profile.
RBR CTDs are normally deployed either using a puller or using a mechanical winch.
With a puller, if the rope is not spooled then the CTD will fall freely. Allow the CTD to
free fall smoothly, and avoid rapid acceleration and deceleration. If the CTD is deployed
using a mechanical winch, then lower at a rate of about 1 m/s. Lowering the CTD
at an consistent speed of about 1 m/s is critical to avoid rapid acceleration during the
deployment because an uneven descent rate will negatively impact data quality (Section
5 Data post-processing
5.1 Correct for A2D zero-order hold
The analog-to-digital (A2D) converter on RBR instruments must recalibrate once per
minute. In the time it takes for the calibration to finish, a sample is missed. The
onboard firmware fills this missed scan with the same data measured during the previous
scan, a simple technique called a zero-order hold. Zero-order holds were found in several
Hakai instruments, and they are also known to occur in instruments used by the Pacific
Salmon Foundations’ Marine Survival Project.
A zero-order hold is easily identified by finding where neighbouring samples have the
same value. However, some sensors do not contain holds, which implies that full scans are
not necessarily held. Furthermore, different channels are held on different instruments,
and there are even cases where some sensors are held for more than one point. For
example, on the Pacific Salmon Foundation’s RBR Concerto, S/N 065679, the pressure,
sea pressure, depth, and temperature fields have zero-order holds, while dissolved O2
does not. On this instrument, conductivity and chlorophyll-a have zero-order holds on
the same scan, but also on the previous scan. On the Hakai RBR CTD S/N 080217,
the temperature, chlorophyll-a, PAR, and turbidity fields have zero-order holds, while
conductivity and dissolved O2do not.
It would be trivial to identify and fix the zero order holds if all sensors were held
during a scan. Simply finding where consecutive pressure readings are the same would
be sufficient. As pointed out earlier, in some cases, the hold for temperature and pressure
occurs on the same scan, whereas the hold for conductivity began on the previous scan
(Fig. 2). Thus, not only does each sensor need to be treated independently, the number
of points to fix can also differ.
Before continuing with the following data processing steps, we recommend that users
first test to see if the data has zero-order holds by plotting the time derivative of raw
pressure. This will serve as a reference point for other channels. If the derivative shows a
zero value at regular intervals then we recommend that the user calculate the derivative of
other sensors (e.g., conductivity, temperature, turbidity, oxygen, PAR, and chlorophyll).
The reason for using pressure as a basis for finding zero-order holds in other data is
that pressure increases in a predictable way during the cast, whereas is it possible, albeit
unlikely, for temperature and conductivity to be perfectly uniform with depth (e.g., such
as might be found in the winter surface mixed layer in temperate open ocean regions).
After identifying a hold, the user has a choice on how to fix it. We recommend that
the zero-value hold be replaced either with a filler (e.g., NaN) or with an interpolated
value calculated from the points surrounding the zero-order hold.
Finally, we wish to point out some anomalous behaviour in the pressure record of the
Hakai RBR CTD S/N 080217. Pressure has a zero-order hold, as well as a noticeable
problem on the scan after the hold such that the pressure sensor appears to overshoot the
true value after the hold (Fig. 3). It is possible that other sensors demonstrate analogous
behaviour, but this has not yet been investigated. If it exists, such an error might be
difficult to find because other variables generally do not vary throughout the profile in a
predictable, monotonic fashion. It is not clear at this time if this behaviour is limited to
this instrument.
5.2 Data despiking
Despiking is the general term used to identify and subsequently correct or remove spurious
data points. Spurious data are often “spiky”, meaning that they differ significantly from
neighbouring data. In Seabird’s data processing suite this is referred to as “Wild Edit.”
Median filters are commonly used to find spikes in data. One common algorithm is:
1. Create a smoothed time series with a median filter.
2. Subtract smoothed time series from original.
3. Calculate the standard deviation, s, of this residual.
4. Flag individual values lying outside of a defined tolerance level (e.g., 4s)
In idealized cases, the median filter despiking easily identifies spurious data (Fig. 4).
However, in actual applications it is not always clear whether the algorithm correctly
identifies truly spurious data. For example, consider Fig. 5, which shows a fluorescence
profile taken by Hakai staff on the BC coast on 09 Sep 2014. The 3-pt filter, which was
very effective on our idealized signal, flags data that are clearly bad (note the spike at
30 dbar), but also data which is likely good (7 dbar).
Tests on conductivity show data despiking could be useful for flagging spurious density
features which occur when the CTD momentarily reverses direction during a downcast
(called “loops” by Seabird). However, analyzing for and removing these points is accom-
plished in the descent rate filtering step (Section 5.6). The despiking algorithm also finds
conductivity regions that appear to be density inversions, however, whether inversions
are realistic or not should be diagnosed with salinity and/or density. Despiking could
also be used to remove salinity spikes from improperly aligned and matched conductivity
and temperature data.
Ultimately it is up to the individual researcher to choose whether to apply a despiking
algorithm. Median filtering algorithms are generally inappropriate for smoothly varying
data such as conductivity and temperature (the same is true for Seabird’s algorithm).
Properly aligning, filtering, and descent rate filtering are, in principle, better options to
handle irregularities in conductivity and salinity.
5.3 Correction for atmospheric pressure
RBR CTDs measure and record total pressure, where total pressure is the sum of atmo-
spheric pressure and sea level pressure. Atmospheric pressure must be removed from total
pressure to obtain sea pressure, and sea pressure is required to calculate, for example,
depth and Practical Salinity.
In the absence of independent measurements of atmospheric pressure, a common
approach is to assume that it is 101.325 kPa (10.1325 dbar), which is the globally-
averaged atmospheric pressure at sea level. However, atmospheric tides and synoptic
weather cause sea level pressure to vary naturally by about ±0.6 dbar, and the maximum
difference between the highest and lowest sea level pressures ever recorded is 2 dbar.
These changes are within the expected accuracy for a pressure sensor rated for coastal
applications, which means atmospheric pressure fluctuations could be resolved with the
CTD pressure sensor. The question then becomes: Is the atmospheric pressure reading
from the CTD reasonable? Furthermore, pressure sensors often show hysteresis, which
in this context means that the sensor could produce different values for the same actual
pressure before and after a profile. If hysteresis is a problem, then it should be most
noticeable in near-surface readings. Therefore, it was necessary to check whether the
pressure sensor records consistent atmospheric values before and after the cast.
Data from the Hakai Institute on the BC Central Coast was used to test the viability
of using RBR pressure data itself to estimate atmospheric pressure at sea level (SLP).
The Hakai Institute maintains a meteorological station at sea level in Pruth Bay, Calvert
Island. The instrument is a Campbell Scientific 106, which records barometric pressure
at 5 min intervals with a claimed accuracy of ±0.6 mbar (±6×103dbar). CTD
profiles are collected almost daily from a hydrographic station about 2.5 km from the
meteorological station. An estimate of air pressure from the RBR pressure sensor was
obtained by choosing the minimum pressure for each profile, but only for those times
when the conductivity was near zero to ensure the instrument was at the surface. A
comparison of the two pressure estimates for a number of profiles is shown in Fig. 6. The
comparison shows that the RBR estimate of atmospheric pressure is scattered around
the 1:1 line with less than 0.25 dbar of scatter, with the error presumably caused by
uncertainty in the RBR sensor. The manufacturer’s specified error is 0.35 dbar for the
Hakai instruments (±0.05% of 700 dbar), which is slightly larger than the scatter in the
regression. The positive correlation indicates that the CTD can measure SLP, but the
fluctuations are large enough so that the measurements are generally not different from
10.1325 dbar.
Next we examine the pressure readings made before and after a profile to search for
evidence of hysteresis. If hysteresis exists, then we need to determine whether the pre-
or post-cast pressure reading is the best approximation of air pressure. Data from eight
different instruments in three different regions were used to check for hysteresis. Table 1
summarizes the instruments used in the comparison.
The pre-cast and post-cast near-surface pressure values for the Hakai instruments are
regressed in Fig. 7. The data lie scattered around the 1:1 line with less than 0.1 dbar
of deviation, with a 0.02 dbar bias toward higher pressure before the cast. We note
that the pre- and post-profile pressure readings from the Hakai profiles on the BC coast
were taken while the instruments were submerged but near the water surface instead of
in the air because these instruments do not initiate logging until they exceed a specific
pressure threshold. At any rate, the bias and scatter in this regression is well within
the pressure specification for the instrument, and generally smaller than atmospheric
pressure fluctuations.
The Coral Sea in-air pre- and post-cast values are regressed in Fig. 8. The instrument
used in this comparison was rated to 2000 dbar. The pre-cast pressures were always
higher than the post-cast values by a median value of 0.04 dbar, while the maximum
was 0.1 dbar. As with the comparison made using instruments rated for lower maximum
pressures, it is important to note that the errors are well within the manufacturer’s stated
uncertainty of about 1 dbar (0.05% of 2000 dbar).
In summary, near-surface in-water pressure recordings compare favourably to nearby
measurements made at a meteorological station. The scatter in this comparison falls
within the stated uncertainty in the pressure sensors. This suggests that the CTD pres-
sure sensor provides a reasonable estimate of air pressure, and that this can be used to
remove the air pressure from total pressure. Finally, hysteresis is measurable but in-
significant relative to the stated accuracy of the pressure sensor, meaning either the pre-
or post-cast estimate is equally suitable.
Ultimately the errors caused by using the nominal atmospheric pressure instead of the
local measured atmospheric pressure (whether measured by a CTD or a meteorological
station), translate into small errors in salinity. A 1 dbar error will cause at an error in
Practical Salinity of O(104) PSU, which is much smaller than the manufacturer specifi-
cations. The most important effect of incorrectly accounting for atmospheric pressure is
when trying to accurately measure shallow depths.
Note on the temperature dependence of pressure
IOS scientists used four RBR Concertos in the Canadian Arctic Archipelago during the
spring and summer of 2015. Two instruments had similar starting and ending surface
pressures, however the other two had differences larger than the stated pressure accuracy
(Table 1). The air-sea temperature differences were often very large; water temperatures
ranged between -2 and 0C, air temperatures were -20C, and the instruments were
stored indoors at +20C. The pre- and post-cast pressure differences may be due to how
the pressure sensor’s temperature is handled in the pressure calculation.
The stated pressure accuracy is ±0.4 dbar for these four instruments based on
±0.05% accuracy of 740 dbar full scale. Two of the loggers (S/N 65578, 65639) pro-
vided surface pressures from the start and end of the casts within 0.2 dbar (Fig. 9),
however the other two loggers (S/N 65579, 65636) had pressure differences ranging from
0.3 to 1.2 dbar depending on the cast. For the two instruments with larger pressure
differences the ending pressure was always lower than the starting pressure.
A pressure dependence on temperature was clearly seen in the three instruments used
in very cold air temperatures when the instruments were removed from their protective
environment and exposed to cold air before the cast. Interestingly, the direction of the
dependence varied depending on the logger. For S/N 65639, an in-air temperature change
of -15C led to in-air pressure change of -0.13 dbar (+0.1 dbar/10C). For S/N 65636,
a change of -40C led to an in-air pressure change of +1.2 dbar, and for S/N 65579, a
change of -10C led to an in-air pressure change of +0.6 dbar (-0.45 dbar/10C). When
the second two instruments were put into the relatively warmer water, the pressures of
both decreased, consistent with their temperature dependence.
Effects of water freezing around the sensors cannot be ruled out, and a couple of the
instruments described above were only used for 3 and 4 casts.
These results indicate that when working in extreme temperatures, standard practices
of data processing and expected accuracy statements may need to be changed.
The variable dependence of pressure on temperature arises because of how this issue
is handled differently between the model generations. Most instruments using the mu-
sical names (e.g., Concerto), would have pressure compensation via a thermistor in the
pressure sensor, while older instruments either used the external “marine” thermistor or
compensating resistors built into the pressure transducer (e.g., old Concertos or XRX-
620s). The result for the instruments using compensating resistors is that the pressure
calibration is most accurate at the calibration temperature (approximately 20C), but
may deviate some for different temperatures.
5.4 Low pass filter
Low pass filtering conductivity, temperature, and pressure data is used to 1) match sensor
time constants, 2) smooth high frequency noise and, in some instruments, 3) remove
quantization errors in pressure sensors. Each of these will be discussed here. Of course
data from any sensor can be filtered, however in this report we discuss conductivity,
temperature, pressure, and O2.
The primary reason to filter temperature and conductivity is to match their response
times. The rate at which conductivity and temperature sensors respond to impulsive
signals is quantified by a time constant. In the case of RBR instruments, the conductivity
sensor has a faster response, and therefore it needs to be “slowed down” to match the
response time of the temperature sensor. Oxygen sensor time constants are longer than
both temperature and conductivity. Matching the sensor response times of conductivity
and temperature is necessary to reduce salinity spiking, and matching temperature and
salinity to dissolved oxygen percent saturation is necessary to calculate dissolved oxygen
Low pass filters are also applied to CTD data to smooth undesirable high frequency
noise. For example, noise might be introduced into the temperature sensor because of
the hydrodynamic turbulence caused by the conductivity cell.
The pressure sensor on RBR profilers has a fast response time of less than 0.01 s, and
while it is used to calculate practical salinity, it is not necessary to slow down the response
by filtering. However, there are times when it is useful to filter pressure. The first is that
pressure should be filtered when it is used to calculate the descent rate, because taking
the derivative of a time series tends to amplify noise from the raw time series. Filtering
pressure will reduce the noise and produce a smoother descent rate. The second case is
when a pressure sensor with a very deep rating, say 1000 dbar, is used in shallow water.
The precision and accuracy of a pressure sensor depend on its maximum depth rating
(i.e., the full scale reading). For example, a 1000 dbar sensor with an accuracy of 0.05%
translates into a potential error of up to 0.5 dbar. If this is used in a shallow area, the
relative error, in terms of the (potential) bias compared to the profile depth, could be
large. Low-pass filtering can be helpful in this case to remove random errors, but the
bias will still remain.
By way of comparison, Seabird generally recommends filtering pressure for some of
their instruments (e.g., SBE 19+) to remove quantization errors. RBR sensors, while
also using strain gauge sensors, do not appear to suffer from such errors. The difference
is in how the signals are digitized and recorded on the instrument.
Finally, we will need to filter temperature and salinity with a long time constant
filter to match them with the dissolved oxygen sensor response. This is necessary be-
cause oxygen saturation is required to convert from the measurement of relative dissolved
oxygen concentration (i.e., percent saturation) to an absolute concentration, and oxygen
saturation depends on salinity and temperature.
5.4.1 Filter type
In idealized cases, the task of matching time constants and phase responses relies on
recursive filters, which are a type of infinite impulse response (IIR) filter. Recursive
filters have a time history built in to model the response of a sensor to changes in the
environment. Applying a recursive filter to a rapidly changing “step” signal will 1)
smooth the step and 2) change the phase. Schmitt et al. [2005] recommend this type of
filter because it produces the required phase delay, which is to say it lags the signal to
account for a finite response time.
SeaBird uses an IIR filter, however whether it is a Butterworth filter or something else
is unclear. It is a bilinear filter, which is ideal in cases where the sample interval is near
the response time [Schmitt et al., 2005]. In principle, the filter produces the necessary
phase lag, however Seabird’s filter is applied to the data in both directions (in analogy
with filtfilt.m in Matlab). Running the filter in both directions produces two effects:
1) the amplitude response is smoothed more than a single pass with the filter, and 2)
the phase shift is zero. The zero phase shift response is not realistic, however, Seabird
has presumably chosen to address the phase shift issue with the alignment step, whereby
the time series of a particular sensor is advanced or delayed in time. Sensor alignment is
addressed and also recommended by Schmitt et al. [2005].
Oceanographers commonly use moving-average type filters, most of which are consid-
ered Finite Impulse Response (FIR) filters. The most common FIR filter is the moving
average, or boxcar.
An informal survey of the workshop participants showed that users rely on different
filters. For example, Humfrey Melling at IOS recommends using a running 3-point tri-
angular window to smooth a 6 Hz conductivity signal. Clark Richards at BIO tailors
the filter depending on the nature of the T/C data itself. In some cases, it makes sense
to explicitly match time constants and adjust the phase response (e.g., with a recursive
phase-forward filter), while in others it makes sense to simply smooth high-frequency
variations with a simple running mean.
Does the filter type (FIR vs. IIR) matter for the environments sampled by RBR
CTDs? To get some understanding of the difference between FIR and IIR filters, Fig. 10
and Fig. 11 provide a comparison of the IIR single-pole type filter to a FIR 3-pt triangular
window. In both cases the filter was applied in forward and reverse directions to keep the
phase shift at zero (using the Matlab function filtfilt.m). The 1st -order Butterworth
filter had a cutoff frequency of 1 Hz, where the cutoff frequency is the frequency at which
the signal is attenuated by 3 dB.
In this example, the Butterworth cutoff frequency was chosen manually to approxi-
mate the 3-point triangular filter (the cutoff frequency can be fine-tuned to more closely
match the 3-point curve). While there are physical arguments that support the use of
a recursive filter (e.g., the phase delay), it is clear from this simple example that differ-
ence between the filters is minimal if they are applied in both directions. One must still
account for phase delays, which in principle could be achieved by forward filtering only,
however this can be introduced during the sensor alignment correction (Section 5.5).
Therefore, under normal circumstances (e.g., coastal waters or noisy sensors), use
of a simple FIR filter, such as a running mean (perhaps weighted with, for example, a
Hanning or triangular window), applied in the forward and reverse directions, is sufficient.
While the IIR-type filter used by Seabird was designed to emulate the response of a real
sensor, most oceanographers are more familiar with FIR filters. The difference between
the filters is not large enough to warrant using the more sophisticated (yet realistic)
IIR filters. To further ensure that the temperature and conductivity time constants are
closely matched, the same filter should be applied to temperature.
5.4.2 Dissolved oxygen
Finally, in order to calculate O2concentration from O2percent saturation, it is necessary
to compute the O2saturation value, which a function of temperature and salinity. Thus
temperature and salinity should be filtered to match the relatively slow response time of
the dissolved oxygen sensor. The filter time constant or window length will depend on
the time constant of the particular oxygen sensor, but a rough guideline might be found
by considering the lag needed to align the O2sensor to the pressure and temperature
sensors (Section 5.5).
5.5 Sensor alignment
Aligning the sensors on a CTD accounts for two different factors. The first is that each
sensor encounters a water parcel at a different time as the CTD moves through the
water because of physical separation (different location on the instrument). The lag
time depends on the separation and the descent speed because these factors determine
the transit time of a water parcel between the sensors. The second is that the finite
response time of a sensor causes a lag in the data because of the time it takes for it to
equilibrate with its new surroundings. In principle, a well-designed filter could impart
a phase change to account for the finite response time, however the approach usually
taken is to filter in such a way as to keep the phase at zero, and then shift the sensor in
time. The use of sensor alignment to account for response time is especially important
for dissolved oxygen.
5.5.1 Conductivity and temperature
Conductivity and temperature must be aligned to produce an accurate value of practical
salinity. If the sensors are misaligned, then the salinity profile will contain small spikes
because conductivity and temperature were not paired correctly.
For most RBR CTDs, the conductivity sensor encounters a parcel of water before the
temperature sensor because of its physical location on the instrument, and because the
field lines from the conductivity sensor extend some 20 - 30 cm into the water. Thus, it
is necessary to delay the conductivity signal by a small amount. The temperature data
is not shifted in time because the thermistor is physically located near to the pressure
In principle, one could calculate the appropriate delay for the conductivity sensor by
simply dividing the distance between the conductivity sensor and the pressure sensor by
the descent rate. However, the conductivity measurement is a weighted spatial average
of a complicated sensing volume, and as mentioned earlier, we would also like to account
for the response time of the “slower” temperature sensor. Fortunately, we can derive an
empirical lag time by considering the quality of the derived salinity profile. In particular,
we wish to minimize spurious salinity spikes that occur when temperature and conduc-
tivity are misaligned. Salinity spikes generally imply that there is an unstable density
inversion in the water column, and while inversions can occur naturally, they are nearly
always a consequence of sensor misalignment.
Figure 12 shows a select depth range of multiple profiles of salinity calculated from a
single profile of conductivity and temperature. Each profile was computed with a different
lag of conductivity relative to temperature. The profile was taken by the Hakai Institute
on September 14, 2015, at station FZH08 on the British Columbia central coast. The
descent rate ranged from 1.8 m/s near the surface to 1.4 m/s at depth. The optimum lag
for conductivity in this example is in the range of -0.167 to -0.334 s (-1 to -2 scans at 6
Hz), where the negative sign indicates a time delay. The optimal lag produces a salinity
profile, which has minimal spiking at steep temperature and conductivity gradients.
5.5.2 Dissolved oxygen
Alignment is particularly important for oxygen sensors because they have a much slower
response time than temperature and conductivity.
To determine the appropriate lag time, we consider data from an JFE-Alec Rinko
III oxygen sensor, which is the sensor used on multiple RBR loggers used by the Hakai
Institute. These are fast response sensors designed specifically for profiling applications.
The Rinko III is an optode-type sensor instrument that measures the changes in phos-
phorescence quenching phase shift of a substrate, caused by changes of oxygen partial
pressure in the water. Oxygen partial pressure is related to in situ oxygen saturation (as
a percentage) by dividing by the partial pressure of oxygen in the atmosphere (21 kPa).
JFE Alec does not provide a recommended advance time; however they JFE Alec
state that the sensor response time is 0.9 s for a 90% change at a temperature of 25C.
This means that the sensor reaches 90% of the true value 0.9 s after being exposed to a
step change in O2. We assume that a lag time should be similar to the response time.
It is worth noting now that this is a gas phase response, which means that the response
time was derived by measuring how long it takes for the sensor to respond to a step
change in gaseous O2.
Two studies have been conducted to characterize the Rinko III response time. Sasano
et al. [2011] conclude that 1 s is a reasonable value, while the more detailed study of Bittig
et al. [2014] find that a reasonable response time based on field data is 4.7 s - substantially
longer than 1 s. The longer response time is thought to be caused by fact that the oxygen
measured at the optode foil must diffuse through a laminar boundary layer, which means
that the response time is a function of the water flow rate past the sensor. Bittig et al.
[2014] also determine that the response time is a function of temperature.
Given the large spread of response times found in these studies, a range of lag times
were applied to a few of the Hakai Institute’s O2profiles to determine an optimum lag.
For simplicity, we seek a single lag value instead of trying to account for variations in
flow rate (i.e. CTD descent rate) and temperature. We follow Seabird’s recommendation
to search for the lag that “collapses” the up and down casts on a T-O2diagram. In other
words, when the lag is set correctly, O2will be paired to the correct temperature (and
therefore pressure and time).
Figure 13 shows the T-O2relationship for a single profile taken on March 30, 2014,
at Hakai station DFO1, for oxygen advance times of 1 s to 4 s in 1 s intervals. It is clear
that an advance of 1 s (6 scans at 6 Hz) is not sufficient to collapse the up and down
casts, whereas an advance of 2 or 3 s brings the O2up and down casts closer together.
Finally, an advance of 4 s over-compensates.
In addition to T-O2plots, O2profiles were plotted for both aligned and unaligned
profiles to gain an intuitive understanding of the O2sensor advance times. Somewhat
counter intuitively, we argue that the optimal advance is not the one that makes the
up and down casts agree perfectly because the hydrodynamic wake generated by the
CTD. The wake, in combination with the sensor location on the instrument, will cause
an inherent mismatch in water properties between up and down casts. For example,
depth profiles of temperature show a roughly 2 m difference between up and down casts.
If we assume the downcast represents the actual temperature profile because the sensor
is encountering undisturbed water, then the up/down mismatch must be caused by the
upcast, which, according to the wake explanation, is because the CTD is measuring
disturbed water caught in its wake. Thus, we seek to delay the O2signal by an amount
that produces a depth difference equal to what is seen in temperature - roughly 2 m.
With the above in mind, a plot of the up and downcast data shows that a 3 s advance
(Fig. 14) causes an up/down cast depth offset of about 2 m.
The 3 s advance found here is longer than JFE Alec’s quoted value and the Sasano
et al. [2011] study, but it is roughly consistent with the findings of Bittig et al. [2014].
The longer time found in our study is presumably due, at least in part, to the much
cooler waters on the BC coast in comparison to the 25C temperature at which the
manufacturer specification was made. The 3 s advance found for the profiles in Fig. 13
was determined by considering a profile made in 7C water. As a quick test to check if the
temperature dependence is important under the range of water temperatures observed
on the BC coast, consider an O2profile from Pruth station on August 10, 2015, in which
the temperature increases from 16C near the surface to 11C at 20 m. The optimum O2
advance for this warm layer was found to be about 2 s, consistent with the expectation
that the response time decreases with increasing temperature.
Finally, we point out that Bittig et al. [2014] found that the flow rate past the sensor
had a significant impact on the time constant. The flow rate dependence was stronger
than the temperature dependence. We don’t address this issue here, however, the flow
rate sensitivity means that users should strive for consist descent rates in the field to
simplify data processing.
5.6 Descent rate filtering
Large seas will cause the CTD descent rate to vary as the boat heaves, particularly when
the CTD is being lowered by a taut line (instead of falling freely). In the case when the
CTD slows, stalls, or reverses, water entrained in the CTD frame or rosette continues
downward and is sampled by the instrument. This is called the wake effect. This water
likely has different properties than its surroundings, which will contaminate the sample.
To examine the impact of inconsistent descent rates on the CTD data, we calculated
the first derivative of density from 117 profiles that were collected at 5 stations (DFO2,
FZH01, HKP01, PRUTH, and QCS01) on the British Columbia central coast from 2012 -
2014. Unless in exceptional circumstances, density increases with depth meaning negative
values could indicate the wake effect. Results suggest that there were density inversions
at all positive descent speeds (i.e. the downcast), which indicates that descent speed
alone is not a good indication of the conditions that create density inversions due to the
wake effect (Fig. 15).
Next, we examined the first derivative of descent speed to get an indication of how
much the CTD is accelerating or decelerating (Fig. 16). Results indicate that density
inversions are normally correlated with times when the CTD decelerates, though, since
about half of the data are associated with some deceleration, deceleration on its own is
not a good metric for diagnosing the wake effect. When the descent speed and deceler-
ation were jointly examined for the density inversions, it was found that 21% - 100% of
the density inversions were associated with a descent speed of less than 0.4 m/s and a
deceleration of less than 0 m/s2, while 1% - 8% of the density inversions were associated
with a descent speed of less than 0.4 m/s and an acceleration less than -0.1 m/s2(not
To estimate the amount of data that would be removed from different loop edit
criteria, the percentage of data that had density inversions greater than 0.01 kg/m3were
calculated. Then the percentage of data that would be removed with a strict (descent
speed less than 0.4 m/s and acceleration less than -0.1 m/s2) and lenient (descent speed
less than 0.4 m/s and acceleration less than 0 m/s2) loop edit criteria was calculated
(Fig. 17). It was found that the number of density inversions ranged from 0.6% - 2.3 %
per station. When the lenient criterion was applied, 0.6% - 2.5 % of the data would be
removed. When the strict loop edit criterion was applied, 0.1% - 0.6 % of the data would
be removed. To minimize data loss while correcting for the wake effect, we recommend
that data where the descent speed is less than 0.4 m/s and acceleration less than -0.1
m/s2are removed.
Low descent rates necessarily occur near the surface just as the CTD begins to descend
after the soak period, and also at the end of the downcast near the bottom. The descent
rate threshold alone does not necessarily flag low descent rate values because of the extra
requirement that the deceleration rate must exceed a threshold. The threshold values
of descent rate and acceleration rate are somewhat conservative in the sense that very
few data points are flagged. However, a researcher may wish to consider implementing a
routine that relaxes the threshold criteria near the bottom and near the surface, especially
if, for example, acceleration rate was not considered.
5.7 Derived variables
In this step standard routines are employed to calculate a range of useful oceanographic
variables such as salinity and depth from measured quantities (typically temperature,
conductivity, and pressure). There are dozens of variables that can be derived from
temperature, conductivity, and pressure, although many are relevant for only a small
number of specialized purposes. Algorithms to compute many of the variables have been
developed and ported to a number of programming languages (TEOS-10 Gibbs Seawater
toolbox). Here we discuss just two quantities, Practical Salinity and dissolved oxygen
5.7.1 Practical Salinity
Practical Salinity is important to mention primarily because users are advised to discard
any salinity variable computed by the RBR logger because it would have been made
before the conductivity and temperature were properly filtered and aligned.
5.7.2 Dissolved oxygen concentration
The fundamental and relevant dissolved oxygen quantity is the actual concentration of
oxygen, which is commonly expressed in units of ml/l or µmol/kg, whereas the Rinko III
dissolved oxygen sensors (and other optodes) measure the percent oxygen saturation. As
discussed in Section 5.5.2, the concentration of oxygen can be calculated by multiplying
the percent oxygen saturation by the saturation level (also called the oxygen solubility).
As with salinity, it is important to note that oxygen concentration, if output by the
optode or RBR logger, is typically an approximate value because dependence of oxygen
solubility on salinity has not been considered. If, by chance, salinity was considered by
the logger firmware or Ruskin, it is still advisable to recalculate oxygen saturation level
because the time constants of the conductivity, temperature, and oxygen sensors must
be matched.
5.8 Bin averaging
Bin averaging is a method that places the sensor measurements onto (often) regular and
standardized depth, pressure, or time values. This reduces high frequency variability at
the expense of resolution. Data can be binned according to pressure, depth, or time.
Bin width and spacing are both (nearly) universally chosen to be 1 m in coastal
waters. Most RBR CTDs sample at a frequency of 6 Hz, and at a descent speed of 1
m/s, bins 1 m in width generally contain the average of 6 samples. Typically, the first
bin is centred on 1 m. Therefore, the first bin contains the average of all data occurring
between depths of 0.5 m to 1.5 m, the 2 m bin contains the average of all data occurring
between depths of 1.5 m and 2.5 m, and so forth until the end of the profile. Note that, in
this particular case, the shallowest bins excludes data from depths shallower than 0.5 m.
In general, this is a positive side effect because near-surface data are often of questionable
Alternative methods for bin averaging include the use of overlapping bins, where the
width of each bin is greater than the bin spacing. One may also choose to window the
data, placing more emphasis on the data occurring near the bin centre. Other possibilities
include methods which produce bin estimates with less bias, which occur in cases when,
for example, the unbinned data is not spaced equally in depth. The LOESS (locally
weighted scatterplot smoothing) approaches could also be applied in this context.
6 Conclusion
Application of the post-processing steps presented in this report will significantly improve
the quality of profiles made by RBR loggers, with a “cleaner” salinity profile being the
most important result. Two different generations of RBR conductivity sensors exist; the
steps here are most relevant for the now discontinued model which has the thermistor
mounted on the pressure case and the cylindrical black or gray conductivity cell.
A correction for the thermal inertia of the conductivity cell was not addressed here
[Lueck, 1990]. The conductivity cell exchanges heat with the ambient water if it moves
through temperature gradients. The heat exchange changes the water temperature of
the sensed water, and thus its conductivity. The resulting impact on salinity can be
significant in regions with strong temperature gradients.
RBR is working to characterize the thermal inertia of its conductivity cells, and as of
late 2016 some progress has been made on deriving a correction formula.
We would like to acknowledge the Hakai Institute and RBR for supporting this work,
and the Institute of Ocean Sciences for hosting the workshops.
Owner/Region CTD Model Maximum Stated Cast Comparison to Difference between Approx. water
& Serial Number pressure rating Accuracy depth met station P pre- and post-cast temperature
[dbar] [dbar] [m] [dbar] [dbar] [C]
Hakai, Coastal BC XRX-620, S/N 18032 700 0.35 75 ±0.25 ±0.05 N/A
XRX-620, S/N 18066 700 0.35 75
Maestro, S/N 80217 1000 0.50 75
RBR, Coral Sea Concerto,S/N 65583 2000 1.00 2000 N/A -0.04 N/A
IOS, Coastal Arctic Concerto,S/N 65578 740 0.37 50 N/A 0.2 -1
Concerto,S/N 65639 740 0.37 50
IOS, Coastal Arctic Concerto,S/N 65579 740 0.37 50 N/A -0.3 to 1.2 -1
Concerto,S/N 65636 740 0.37 50
Table 1: Summary of viability study for the use of CTD pressure to estimate and remove the atmospheric pressure from total
Figure 1: Photo of an RBR concerto logger with the thermistor mounted on the CTD
pressure case, and the old generation of conductivity cell (RBR Ltd).
Figure 2: Time series illustrating a zero-order hold scan (orange dot) for RBR CTD S/N
065679. The yellow and purple dots are the nearest surrounding points. In this example
the black diamond represents the replacement value. Conductivity stays constant for
three consecutive values, including the point before the hold.
Figure 3: Time series showing 6 scans of pressure, conductivity, and temperature for
RBR CTD S/N 080217. The pressure reading following the hold appears to overshoot
its true value. Conductivity does not have a zero-order hold, while temperature does.
Figure 4: The upper panel shows a synthetic test case for the 3-point median filter
despiking algorithm. The signal consists of white noise (σ= 1) centred on 0, 20 single-
point spikes with an offset of ±10, and a short section with a DC offset of -5. Data lying
outside of 3swere flagged.
Figure 5: The panels from left to right are: raw and filtered chlorophyll-a fluorescence,
residual profile (raw - filtered), and the original and “cleaned” profiles. The ±4sthreshold
levels are shown by the dashed lines. In this example, residual values outside of the 4s
threshold are flagged as bad.
Figure 6: Comparison of atmospheric pressure measured at a Hakai meteorological station
to CTD in-water near-surface pressure readings in water before the cast (left panel) and
after the cast (right panel). The solid line represents x=y. The anomalously high CTD
readings were later found to be caused by a faulty sensor.
Figure 7: Comparison of surface pressure data collected after the soak (horizontal axis)
and at the end of a cast (vertical axis) from a station on the British Columbia central
coast. The solid line represents x=y.
Figure 8: Comparison of in-air pressures collected immediately before and after a series
of casts to 2000 dbar in May 2015 off of Eastern Australia in the Coral Sea. Note the
offset of approximately 0.04 dbar, well within the stated accuracy of 1 dbar (0.05% of
2000 dbar). The solid line represents x=y.
Figure 9: Boxplots of pre- and post-cast in-air pressure difference from profiles taken in
summer 2015 on the Canadian Arctic shelf.
Figure 10: Comparison of filter performance for a 6 Hz step change signal. The filters
were applied to the signal twice, once in the forward direction, and once in the reverse
direction, to produce a filtered signal with zero phase shift.
Figure 11: Filter comparison on a time series of 4 s of measured conductivity.
Figure 12: Profiles of Practical Salinity for a range of conductivity lag times. Negative
lags constitute a time delay, while positive values indicate an advance. Conductivity
and temperature profiles are shown for comparison. The CTD descent rate was 1.4 m/s
through this section of the profile.
(a) (b)
(c) (d)
Figure 13: T/O2plots for oxygen sensor advance times of (a) +1 s, (b) 2 s, (c) 3 s, and
(d) 4 s. The profile was obtained on March 30, 2014, at station DFO1.
Figure 14: Section of an O2depth profile obtained on March 30, 2014, at station DFO1
(same profile shown in Fig. 13). O2was advanced by 3 s (18 scans).
Figure 15: Scatter plot of the descent speed versus the first derivative of density for five
different stations on the British Columbia central coast near Calvert Island. For each
station, 1 year of data (which corresponds to 9 - 34 profiles) was examined. The stations
sampled were DFO2, FZH01, HKP01, PRUTH, and QCSOQ (see for a
Figure 16: Scatter plot of the first derivative of descent speed versus the first derivative
of density (δρ) at five stations along the British Columbia central coast. Negative values
of acceleration indicate deceleration and positive values indicate acceleration.
Figure 17: The percentage of raw data where a) the first derivative of density was less
than -0.01 kg/m3, which suggests a density inversion, b) the first derivative of descent
speed was negative (suggesting deceleration) and descent speed was slower than 0.4 m/s,
and c) the first derivative of descent speed was less than -0.1 m/s2and drop speed was
slower than 0.4 m/s. Here only raw data that were taken during the downcast after the
surface soak were included.
Henry C. Bittig, Bjorn Fiedler, Roland Scholz, Gerd Krahmann, and Arne Kortzinger.
Time response of oxygen optodes on profiling platforms and its dependence on flow
speed and temperature. Limnology and Oceanography: Methods, 12:617–636, 2014.
doi: 10.4319/lom.2014.12.617.
Rolf G. Lueck. Thermal inertia of conductivity cells: Theory. Journal of Atmo-
spheric and Oceanic Technology, 7(5):741–755, 1990. doi: 10.1175/1520-0426(1990)
007h0741:TIOCCTi2.0.CO;2. URL
Daisuke Sasano, Masao Ishii, Takashi Midorikawa, Toshiya Nakano, Takayuki Toxkieda,
and Hiroshi Uchida. Testing a new quick response oxygen sensor,“RINKO”. Papers in
Meteorology and Geophysics, 62:63–73, October 2011. doi: 10.2467/mripapers.62.63.
Raymond W. Schmitt, Robert C. Millar, John M. Toole, and W. David Wellwood. A
double-diffusive interface tank for dynamic-response studies. Journal of Marine Re-
search, 63:263–289, 2005.
I. Shkvorets and F. Johnson. Advantages in performance of the RBR conductivity channel
with DelrinTM/ceramic inductive cell. In OCEANS 2010 MTS/IEEE SEATTLE, pages
1–8, Sept 2010. doi: 10.1109/OCEANS.2010.5664276.
A Example processing scheme
A typical implementation of the recommended processing steps would be the following.
The starting point is a single raw RBR profile. The code below is written Matlab, using
functions from the RBRproc Matlab toolbox and the TEOS-10 Gibbs Seawater toolbox.
%% correct for zero-order hold scans (if necessary)
profile = correctHoldRBR(profile,’interp’);
%% remove atmospheric pressure from total pressure
profile = rmPatmRBR(profile);
%% despike Chlorophyll-a and turbidity
profile = despikeRBR(profile,{’Turbidity’,’Chlorophyll’},’median’,3,’NaN’);
%% low pass filter T/C
profile = filterRBR(profile,{’Conductivity’,’Temperature’},3);
%% delay conductivity by 0.3 seconds (2 scans at 6 Hz)
profile = alignRBR(profile,’Conductivity’,-2/6); % negative means delay
%% advance Rinko oxygen by 3 seconds
profile = alignRBR(profile,’DissolvedO2’,3);
%% advance Chl, Turb, and PAR by 0.3 seconds
profile = alignRBR(profile,{’Chlorophyll’,’Turbidity’,’PAR’},0.3);
%% filter for descent rate and acceleration, replace suspect data with NaN
profile = loopRBR(profile,’NaN’);
%% derive some basic quantities
ind = ~strcmp(profile.units,’PSU’);
profile.units = profile.units(ind);
profile = rmfield(profile,’Salinity’); % remove RBR’s calculation
profile.PracticalSalinity = gsw_SP_from_C(profile.Conductivity,...
profile.units(end+1) = {’PSU’};
%% calculate oxygen concentration from percent saturation
filtered = profile; % clone the profile
% smooth SP, T, O2 to match their time constants
fltr = blackman(17)/sum(blackman(17)); % approx 3 sec at 6 Hz
filtered = filterRBR(filtered,{’PracticalSalinity’,’Temperature’,’DissolvedO2’},fltr);
% calculate a bunch of variables needed for saturation value
filtered.AbsoluteSalinity = gsw_SA_from_SP(filtered.PracticalSalinity,...
filtered.units(end+1) = {’g/kg’};
filtered.PotTemperature = gsw_pt0_from_t(filtered.AbsoluteSalinity,...
filtered.units(end+1) = {’deg C’};
filtered.ConTemperature = gsw_CT_from_t(filtered.AbsoluteSalinity,...
filtered.units(end+1) = {’deg C’};
filtered.PotentialDensity = gsw_sigma0(filtered.AbsoluteSalinity, ...
filtered.units(end+1) = {’kg/m^3’};
% calculate saturation value of water with Garcia & Gordon (1990)
filtered.DOsat = gsw_O2sol_SP_pt(filtered.PracticalSalinity,filtered.PotTemperature);
filtered.units(end+1) = {’umol/kg’};
filtered.DOConcentration = 0.01*filtered.DissolvedO2.*filtered.DOsat;
filtered.units(end+1) = {’umol/kg’};
filtered.DOConcentration2 = filtered.DOConcentration.*(filtered.PotentialDensity+1000)./44660;
filtered.units(end+1) = {’ml/l’};
% take the relevant parameters from ’filtered’ and put them back into ’profile’
profile.DOsat = filtered.DOsat;
profile.units(end+1) = {’umol/kg’};
profile.DOConcentration = filtered.DOConcentration;
profile.units(end+1) = {’umol/kg’};
%% trim the soak and deck data, select only downcast
profile = trimRBR(profile);
%% bin by depth to 1 meter intervals
bin = binRBR(profile,’depth’,1);
... where the interval of sensor polling δt = 1/f and some integer n = (1, 2, 3…). The manufacturers of SBE and RBR profilers also suggest the time shift of the measured temperature T1(t) according to Formula (8) in their supplementary materials to the manuals [37,38]. We only note that the efficiency of method (8) is affected by the measure of the difference between the displacement parameter and the phase function |nδt − φT| < 0.5δt. ...
... We should note that SBE was one of the first to recommend the users of their profilers to perform dynamic correction of profiling data using a package of programs for primary data processing. Similar proposals for RBR profilers have appeared only in the last few years [38,43,44]. Since the methods of these manufacturers are quite similar, we will limit ourselves here to discussing the primary processing of CTD data by the SBE-Data Processing program, which contains several procedures, executed in sequence [37]. ...
Full-text available
The knowledge of salinity in a specific sea area with high accuracy is required to solve several acoustic and hydrophysical problems on the ocean shelf. Unlike temperature, which can be measured continuously for a long time, with, for example, thermistor strings (thermostrings), salinity values of required accuracy can be obtained only using CTD profiling. This is why methods of estimating salinity from temperature could be helpful. In this paper, the authors propose using the regression method for solving this type of problem and demonstrate the efficiency of this method using examples of temperature measurements from anchored thermostrings. For the correct construction of regressions, the authors analyzed the errors of CTD measurements and suggested a method for the dynamic correction of raw CTD data. From CTD profiling datasets of 12 years (2011–2022), after their dynamic correction, the authors obtained regression polynomial formulas for calculating salinity from temperature and studied data stability in space and time at the hydrophysical test site, located in the shelf zone of the Sea of Japan. The authors consider this method efficient and applicable in solving a variety of acoustic and hydrophysical problems.
... In addition to the advective lag, we also consider the ''response lag,'' which is a constant lag associated with the different response times of the two sensors. Following recommendations from Halverson et al. (2017), the appropriate response lag is determined from T-O 2 curves and the necessary adjustment to ''collapse'' the up and down casts to a mismatch that is equivalent to the mismatch observed in temperature and conductivity profiles (;5 dbar; see Fig. 5e). A lag of 6 scans (0.75 s) is chosen, which is slightly shorter than, but similar to, the 0.9 s response time of the oxygen sensor advertised by the manufacturer. ...
... Although the RBR Concerto 3 was factory calibrated before the cruise, several components were added in proximity to the conductivity cell, namely a plastic guard, a casing, and a weight (see Fig. 1). As the conductivity is measured inductively, any component located close to the conductivity cell would affect the measured ionic current, and thus the conductivity reading (Halverson et al. 2017). An offset of 0.048 32 mS cm 21 was therefore applied to all conductivity observations to compensate for these effects (Fig. 6b). ...
Full-text available
The study of ocean dynamics and bio-physical variability at submeso-scales of O(1 km) and O(1 hour) raises several observational challenges. To address these by underway sampling, we recently developed a towed profiler called the EcoCTD, capable of concurrently measuring both hydrographic and bio-optical properties such as oxygen, chlorophyll fluorescence, and optical backscatter. The EcoCTD presents an attractive alternative to currently used towed platforms due to its light footprint, versatility in the field, and ease of deployment and recovery without cranes or heavy-duty winches. We demonstrate its use for gathering high-quality data at submesoscale spatio-temporal resolution. A dataset of bio-optical and hydrographic properties, collected with the EcoCTD during field trials in 2018, highlights its scientific potential for the study of physical-biological interactions at submeso-scales.
... All CTD data were collected using either a Seabird 19 Plus or RBR XR-4320 or Maestro sensor and all oxygen were collected using an SBE43 or Rinko III sensor. CTD and oxygen sensors were calibrated yearly and data were processed using either Seabird's SeaSoft software or the RBR processing steps recommended by Halverson et al. (2017). In addition, we examine Bute Inlet temperature, salinity and oxygen data collected from 1951 to 2014 by the University of BC and Fisheries and Oceans Canada (see J. M. Jackson et al. (2021)). ...
Full-text available
Arctic outflow winds bring cold air from the continent to the coastline through mountain passes. Using observational data and a 2-D model, we show that a February 2019 outflow event caused the upper 100 m in Bute Inlet, British Columbia (within the traditional territory of the Homalco Nation) to cool up to 1.9°C and gain up to 4.1 mLL−1 of oxygen. The cold, oxygenated water persisted for almost 1 year within the 1,023–1,023.5 kgm−3 isopycnal range (∼50–150 m). Atmospheric (from 1929 to 2022) and oceanographic (from 1951 to 2022) data showed a statistically significant relationship between continental air temperature at Tatlayoko Lake and temperature and oxygen in Bute Inlet. This local mechanism that counters some effects of climate change could create a biological refugium as surrounding waters warm and lose oxygen at a faster rate. The number of outflow events decreased from 1951 to 2018, and increased since.
... CTD stations were in close proximity (0-2 km) to the salmon sampling stations (Fig. S1). An RBR maestro or a SeaBird 19plus V2 CTD was used (Halverson et al. 2017). For the purpose of this study, we present temperature (°C) and salinity data integrated over the upper 10 m of the water column, where salmon were captured. ...
Full-text available
Migrating marine taxa encounter diverse habitats that differ environmentally and in foraging conditions over a range of spatial scales. We examined body (RNA/DNA, length-weight residuals) and nutritional (fatty acid composition) condition of juvenile sockeye salmon ( Oncorhynchus nerka) in British Columbia, while migrating through oceanographically variable waters. Fish were sampled in the stratified northern Strait of Georgia (NSoG); the highly mixed Johnstone Strait (JS); and the transitional zone of Queen Charlotte Strait (QCS). In 2015, body and nutritional condition were high in the NSoG but rapidly declined to reach lowest levels in JS where prey availability was low, before showing signs of compensatory growth in QCS. In 2016, juvenile salmon had significantly lower condition in the NSoG than in 2015, although zooplankton biomass was similar, condition remained low in JS, and no compensatory growth was observed in QCS. We provide evidence that differences in juvenile salmon condition between the two years were due to changes in the food quality available to juvenile fish. We propose that existing hypotheses about fish survival need to be extended to incorporate food quality in addition to quantity to understand changes in fish condition and survival between years.
... The physical environment was characterized weekly with a CTD cast (RBR Maestro or SBE19plus). Temperature, salinity, and pressure were aligned and binned into 1 m intervals prior to analysis (Halverson et al. 2017). A water column stratification index was also calculated from the log 10 -transformed buoyancy frequency calculated from temperature, salinity, and pressure using the swN2 function in the 'oce' R package, and averaged over the upper 30 m. ...
Particulate organic matter (POM) forms the base of the pelagic food web, but is a complex category of material that undergoes substantial changes in quantity and quality across different time scales. As the primary consumers of POM, zooplankton are influenced by these fluctuations, resulting in shifts in the food web pathways that contribute to the production of higher trophic levels. We measured POM fatty acids, which are critical nutritional components in marine food webs, along with a suite of associated biotic and abiotic environmental conditions at a temperate coastal site over four years. Using these data, we investigated the co-occurring patterns of prey quality, quantity, and size, to develop a holistic understanding of how the prey field for zooplankton varies over the seasonal cycle and inter-annually. The seasonal pattern of POM fatty acids corresponded to the succession of phytoplankton taxa, but displayed stronger relationships to size-fractionated Chl a than taxonomic observations. Times with high micro-Chl biomass (spring and fall blooms) had the greatest concentrations of high quality food, but fatty acid levels remained high throughout the summer when Chl a concentrations dropped and the size distribution shifted towards pico-Chl. The 18-carbon polyunsaturated fatty acids 18:3ω3 (α-linolenic acid) and 18:4ω3 (stearidonic acid) increased during this time and were correlated with pico-Chl. In addition, the important nutritional factor for zooplankton and fish, the fatty acid ratio DHA:EPA (22:6ω3/20:5ω3), peaked during the middle of summer separately from the peak in Chl a. These seasonal patterns resulted in tradeoffs among the abundance, size, and nutritional quality of prey for zooplankton. We validated the basal sources of several fatty acids used as trophic markers within food web studies and in particular note their power for resolving size-based trophic connections. Interannual variability, e.g., the occurrence of fall diatom blooms and the timing of community shifts, is also discussed. This work lays the foundation for future studies of the zooplankton community at this location and the incorporation of realistic prey conditions into zooplankton studies.
... ). An RBR maestro or a SeaBird 19plus V2 CTD was used(Halverson et al., 2017). For the 223 purpose of this study, we present temperature ( o C) and salinity data integrated over the upper 10 224 m of the water column, where salmon were captured. ...
Migrating marine taxa encounter diverse habitats that differ environmentally and in foraging conditions over a range of spatial scales. We examined body (RNA/DNA, length-weight residuals) and nutritional condition (fatty acid composition) of juvenile sockeye salmon (Oncorhynchus nerka) in British Columbia, as they migrated through coastal waters that varied oceanographically over tens of kilometers. Fish were sampled in the stratified, productive northern Strait of Georgia (NSoG); the highly mixed, unproductive Johnstone Strait (JS); and the transitional zone of Queen Charlotte Strait (QCS). Body and nutritional condition responded rapidly to changes in prey availability and were lowest in JS with low prey availability, supporting the Tropic Gauntlet Hypothesis, additionally we saw signs of compensatory growth in QCS. Juvenile salmon leaving the SoG in 2016 had significantly lower condition than in 2015, despite higher zooplankton biomass in 2016. We propose that this was due to the higher abundance of low food quality southern zooplankton species in 2016. This study highlights the importance of including food quality as a parameter to understand changes in fish condition and survival between years. Furthermore, small scale variation in oceanographic dynamics impact foraging conditions and need to be considered when assessing early marine survival of juvenile salmon.
... The difference usually could be related to the difference in the investigation volume of the ERT and the probes (Halverson et al., 2017). But, the consistency between the results of water samples conductivity monitoring and resistivity monitoring proves that this difference may not be related to the difference in the investigation volume of the ERT and the probes. ...
Tidally influenced groundwater discharge in tidal flats is an important route by which land-based pollutants enter the sea. Seasonal changes in groundwater flow fields create differences in the process and amount of groundwater discharge in sandy beaches. To explore this subject, we selected Shilaoren Beach in Qingdao as our research area and used electrical resistivity tomography (ERT) and hydrogeological measurement methods to monitor the groundwater discharge process in the same area in both May (rainy season) and November (dry season) in 2019. The study found that there are obvious upper saline plumes (USPs) and freshwater discharge tubes (FDTs) in the sediment layer in both seasons. When sea level falls, the USPs sink and lateral extent decreases due to groundwater discharge, and the FDTs develop from deep below to the beach surface, and divide the USP into multiple sections. During the flood tide, the pore water salinity of USPs increases and lateral and vertical ranges expand rapidly. The groundwater discharge process differs seasonally. In the dry season, the vertical extent of USP is larger around high tide and the distribution range of FDT is more unstable. During the ebb tide, the pore water salinity in USPs and FDTs changes from the moment when seawater starts receding, however, in the dry season, it changes more slowly during the early stage of the ebb tide and more rapidly near the end the ebb tide. The difference of hydraulic driving force is the main factor leading to the difference of distribution range and pore water salinity change of USP and FDT in different seasons. The volume of groundwater discharge also varies with seasons, and the discharge in rainy season is about 1.21 times of that in dry season. In the rainy and dry seasons, the groundwater discharge gradually increases from the land to the sea, and the maximum discharge location is near the low tide line.
... Data collected by the Sea-Bird CTD were processed using the Sea-Bird Scientific data processing software, Seasoft. Data collected by RBR CTDs were processed using the steps outlined in Halverson et al. (2017). All CTD data are available for download on the Hakai Data Portal ( ...
Full-text available
A 4-year (2015–2018) weekly to bi-weekly time series of phytoplankton biomass and composition derived from high-performance liquid chromatography (HPLC) phytoplankton pigments and Chemtax analysis is presented and used to investigate phytoplankton community dynamics at a station in the northern Strait of Georgia (NSoG). Through the time series, blooms were largely dominated by diatoms, which formed the bulk of annual biomass. Spring diatom bloom timing and magnitude varied widely and appears to have been driven by complex interactions of solar radiation, wind, stratification, and grazing. In turn, post-spring diatom blooms were mostly associated with nutrient renewal to the surface layer as suggested by redundancy analysis (RDA), which showed inverse relationships between diatoms and temperature and stratification. A single non-diatom bloom in July 2016, dominated by the silicoflagellate, Dictyocha sp., was the time series maximum biomass and occurred under warm, stratified conditions and a freshening of the surface layer: The Chemtax dictyochophyte group was positively linked to temperature and stratification through RDA. Outside of bloom conditions, diverse communities emerged with prasinophytes and cryptophytes showing persistent contributions and their highest biomass during summer. Uniquely, these groups often persisted through nutrient renewal and drawdown events typically associated with diatom blooms and suggestive of high grazing pressure and nutrient regeneration. The prevalence of these groups through diverse conditions likely precluded statistical links with environmental drivers. This time series is the first of its kind for the NSoG, creates a baseline for future analyses, and highlights the contributions by small species, particularly prasinophytes, to regional phytoplankton communities.
... JS/QCS CTD measurements were obtained using an RBR concerto. CTD data were subsequently processed using Seabird's data processing software for Seabird data and the CTD processing steps outlined in Halverson et al. (2017) for RBR data. Niskin bottles were used to collect water samples at discrete depths in the water column (0, 5, 10, 30 m, and 5 m above the bottom). ...
Temperate coastal marine environments are typified by strong seasonality and highly productive annual spring phytoplankton blooms. However, in areas of strong tidal activity, coastal waters may remain in a state of low productivity year-round due to light limitation induced by deep mixing. Such high-nutrient low-chlorophyll (HNLC) environments can be found in British Columbia, Alaska, Argentina, and the United Kingdom. This study aimed to examine how zooplankton communities are shaped by tidally mixed environments through direct comparison of adjacent stratified and mixed regions on the British Columbia coast. Stations located in five distinct regions of coastal BC were sampled during the most productive months (April to July) over two years. The most seasonally stratified station was characterized by a higher biomass of large zooplankton species, including calanoid copepods and euphausiids. In contrast, the mixed regions had a higher abundance of small (<2 mm) zooplankton species and a prevalence of meroplankton taxa. Zooplankton communities in tidally mixed regions had indicator species that reflected source waters in an adjacent stratified area. However, despite different source waters, tidally mixed regions showed a convergence of community structure pointing to a common community modification process. It may be the result of a combination of advective processes, zooplankton vertical migration behavior, enhanced predation due to close contact with bathymetry and daytime transport to the surface, and a large contribution of meroplankton. In areas with strong tidal exchange, zooplankton communities were more similar than expected based on the physical and chemical characteristics of the water column alone. The mosaic of primary productivity regimes on the BC coast therefore translates into similar spatial scale variation in zooplankton communities, resulting in a spatially heterogenous prey field for zooplankton consumers.
Full-text available
Watersheds of the coastal temperate rainforests of Pacific North America export large amounts of organic carbon (OC) to the coastal ocean. While it has been suggested that terrestrially derived organic matter could subsidize marine food webs and affect ocean biogeochemistry along the coastal margin, little work has been done to quantify and characterize OC across the freshwater to marine continuum. We conducted monthly and targeted rainfall event surveys of dissolved and particulate organic carbon (DOC and POC) quantity and quality (δ13C, dissolved organic matter characterization) across a freshwater to marine salinity gradient between Calvert and Hecate Islands, British Columbia, Canada. Freshwater DOC concentrations (9.97 ± 0.25 mg L−1) far exceeded those in marine waters (1.24 ± 0.03 mg L−1), while POC concentrations were similar across all sites (0.23 ± 0.01 mg L−1). δ13C‐DOC and ‐POC in freshwaters were constant, but varied seasonally at the marine stations with freshwater and marine processes. Rainfall events facilitated the rapid export of terrestrial DOC and POC to coastal waters, altering water quality and potentially subsidizing microbial productivity across marine surface waters. On an annual basis, primary production in marine waters (21–42 Gg C) exceeded total freshwater OC contributions (1.8–2.2 Gg C); however, freshwater exports were more important during the autumn and winter months, when rainfall was highest and primary production was limited by shorter days and deep turbulent mixing. Our results highlight the importance of storms for connecting the coastal temperate rainforest with surface coastal waters, especially during the summer when connectivity between the freshwater and marine ecosystems is otherwise low.
Full-text available
We evaluated a new quick response dissolved oxygen (DO) sensor “RINKO”, currently under development by JFE Advantech Co., Ltd., to determine its in situ response and applicability to hydrographic and hydrochemical observations in the western North Pacific and in the Sea of Japan. The response time of the RINKO sensor to ambient DO changes was demonstrated at about 1 s in these in situ experiments. RINKO data were calibrated to temperature, pressure, and bottle sample DO data obtained by conventional Winkler method, yielding a precision of < ± 1 μmol kg−1, except in the upper layers of the ocean where vertical DO variations are large. Although the sensor was subject to instrumental drift during the period of a cruise (equivalent to < 6 μmol kg−1 DO) and to pressure hysteresis between downcasts and upcasts (< 4 μmol kg−1 DO), the RINKO sensor is capable of observing continuous vertical DO profiles with a precision that is sufficient for most practical purposes. The RINKO sensor will contribute to an improved understanding of DO variability, providing an expanded DO database with enhanced spatial and temporal resolution.
Conference Paper
Full-text available
This paper demonstrates several key advantages of the inductive conductivity sensor. In calibration the sensor has a linear response and may be calibrated with direct traceability to primary standards without assumptions about the salinity scale; one calibration can be used for a wide range of salinities and temperature compensation can be directly measured independently. In the field it has demonstrably superior passive exchange of measurand within the sensor and this is confirmed by the comparative TS plots when used simultaneously with an electrode based sensor.
Full-text available
A large tank capable of long-term maintenance of a sharp temperature-salinity interface has been developed and applied to measurements of the dynamical response of oceanographic sensors. A two-layer salt-stratified system is heated from below and cooled from above to provide two convectively mixed layers with a thin double-diffusive interface separating them. A temperature jump exceeding 10°C can be maintained over 1-2 cm (a vertical temperature gradient of order 103°C/m) for several weeks. A variable speed-lowering system allows testing of the dynamic response of conductivity and temperature sensors in full-size oceanographic instruments. An acoustic echo sounder and shadowgraph system provide nondisruptive monitoring of the interface and layer microstructure. Tests of several sensor systems show how data from the facility is used to determine sensor response times using several fitting techniques and the speed dependence of thermometer time constants is illustrated. The linearity of the conductivity-temperature relationship across the interface is proposed as a figure of merit for design of lag-correction filters to accurately match temperature and conductivity sensors for the computation of salinity. The effects of finite interface thickness, slow sensor sampling rates and the thermal mass of the conductivity cell are treated. Sensor response characterization is especially important for autonomous instruments where data processing and compression must be performed in-situ, but is also helpful in the development of new sensors and in assuring accurate salinity records from traditional wire-lowered and towed systems.
The time response behavior of Aanderaa optodes model 3830, 4330, and 4330F, as well as a Sea-Bird SBE63 optode and a JFE Alec Co. Rinko dissolved oxygen sensor was analyzed both in the laboratory and in the field. The main factor for the time response is the dynamic regime, i.e., the water flow around the sensor that influences the boundary layer’s dynamics. Response times can be drastically reduced if the sensors are pumped. Laboratory experiments under different dynamic conditions showed a close to linear relation between response time and temperature. Application of a diffusion model including a stagnant boundary layer revealed that molecular diffusion determines the temperature behavior, and that the boundary layer thickness was temperature independent. Moreover, field experiments matched the laboratory findings, with the profiling speed and mode of attachment being of prime importance. The time response was characterized for typical deployments on shipboard CTDs, gliders, and floats, and tools are presented to predict the response time as well as to quantify the effect on the data for a given water mass profile. Finally, the problem of inverse filtering optode data to recover some of the information lost by their time response is addressed.