Post on 27-Dec-2021
transcript
Ecography ECOG-03080Bestley, S., Raymond, B., Gales, N. H., Harcourt, R. G., Hindell, M. A., Jonsen, I. D., Nicol, S., Péron, C., Sumner, M. D., Weimerskirch, H., Wotherspoon, S. J. and Cox, M. J. 2017. Predicting krill swarm characteristics important for marine predators foraging off East Antarctica. – Ecography doi: 10.1111/ecog.03080
Supplementary material
Appendix 1
Acoustic data collection and processing
The BROKE-West hydroacoustic data has been described in detail elsewhere (Jarvis et al. 2010) and is publicly
available (Jarvis 2008). The ping rate was 1 Hz and a 1.024 ms pulse duration. Mean vessel survey speed was 10
knots giving a mean along transect inter-ping spacing of 5 m.
The results of the standard sphere calibration exercise (Foote et al. 1987) were applied to the raw acoustic data
(see Jarvis et al. 2010; their Table 2) for calibration parameters. Time varied gain noise correction was carried
out using the procedure described by (De Robertis and Higginbottom 2007). Seabed and near sea surface
acoustic returns were isolated and removed from further analysis. To maintain inter-ping distance, off-survey
effort transect regions were also removed, as were noise spikes and false bottom echoes (grey areas Fig. 3 main
text).
All acoustic data processing was carried out using Echoview v5.4 (Myriax, Hobart, Australia) with the resulting
aggregations exported to the freely available software R version 3.2.0 (R Development Core Team 2015) for
further analysis.
Krill swarms were identified using the dB-difference technique (120 kHz – 38 kHz) and given that three spatial
distinct krill length frequency clusters were identified (Kawaguchi et al. 2010), cluster-specific dB differences and
krill target strengths were used (Table A.1.1). The cluster specific effect krill length ranges are taken from Jarvis
et al. (2010, their Table 4).
Table A1.1: Krill length-frequency cluster specific acoustic data analysis settings. The krill length frequency
clusters were taken from Kawaguchi et al. (2010). Krill were identified using length frequency cluster-specific dB-
difference krill identification ranges. 120 kHz echoes identified as krill were converted to volumetric density
using target strength (units dB re wet mass kg). Biological values (krill volumetric density units g wet mass m-3)
for detection thresholds at – 70 and -80 dB re 1m-1 are provided for each length frequency cluster.
Cluster Effective length range (mm)
120-38 kHz dB difference range (dB re 1 m-1)
Krill 120 kHz target strength (dB re wet mass kg)
Biological detection threshold at Sv = -70 dB re 1 m-1, units: g wet mass)
Biological detection threshold at Sv = -80 dB re 1 m-1, units: g wet mass)
A 20 to 50 4.5 to 14.2 -40.43 1.10 0.11
B 20 to 60 0.4 to 14.2 -39.69 0.93 0.09 C 20 to 60 0.4 to 14.2 -38.95 0.78 0.08
References
De Robertis, A. and Higginbottom, I. 2007. A post-processing technique to estimate the signal-to-noise ratio and remove echosounder background noise. — ICES J. Mar. Sci. 64: 1282-1291.
Foote, K. G. et al. 1987. Calibration of acoustic instruments for fish density estimates: a practical guide. — ICES Coop. Res. Rep. 144: 1-72.
Jarvis, T. 2008, updated 2015. Hydroacoustic data collected onboard the Aurora Australis during voyage 3 of the 2005-2006 season (BROKE-West). Australian Antarctic Data Centre - doi:10.4225/15/57590CA3CFDD5.
Jarvis, T. et al. 2010. Acoustic characterisation of the broad-scale distribution and abundance of Antarctic krill (Euphausia superba) off East Antarctica (30-80°E) in January-March 2006. — Deep Sea Res. Part II 57: 916-933.
Kawaguchi, S. et al. 2010. Krill demography and large-scale distribution in the Western Indian Ocean sector of the Southern Ocean (CCAMLR Division 58.4.2) in Austral summer of 2006. — Deep Sea Res. Part II 57: 934-947.
R Development Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing. — Vienna, Austria. URL http://www.R-project.org/.
Appendix 2
Biophysical variables considered as predictors of krill swarm characteristics
As relationships were not expected to be linear we implemented Generalised Additive Mixed Models (see
Methods). Preliminary analyses investigated potential interactions between continuous biophysical variables, via
full tensor product smooths or tensor product interaction, but found these to be stretching the limits of what
was supportable by the data and ultimately not computable within the complex error structure of the mixed
modelling framework used in the main analysis. Hence, only single terms were included.
The relative spatial pattern of krill swarm densities together with example underway biophysical predictors is
shown schematically in Figure A2.1, and examples of underway predictors and their remotely-sensed analogues
are also shown for reference in Figure A2.2.
Table A2.1. Table of biophysical variables considered as predictors of krill swarm characteristics. These include
variables sourced in situ from underway† (U) BROKE-West research voyage data (Wiley 2006, updated 2014) and
sourced remotely†† (R), primarily sensed via satellite.
Variable Source Description and processing steps
Biological production and light
1. Fluorescence U Underway fluorometry data (unit less) is routinely collected in marine science voyages to characterise the in situ chlorophyll a concentrations in surface waters. These data have been calibrated as detailed in Appendix 3. Values used are FLUlme log transformed.
Sea surface chlorophyll-a concentration
R Surface chlorophyll-a concentration (mg m-3) derived from satellite ocean colour. Sourced from the Johnson improved chlorophyll-a estimates using Southern Ocean-specific calibration algorithms applied to the weekly MODIS 9 km resolution product (Johnson et al. 2013). Values are log transformed.
2. Solar radiation U Ship-based measurement of solar radiation (Watts m-2) averaged from the port and starboard data. Values are log transformed.
Solar elevation R Position of the sun (degrees) above or below the horizon, where ±6 – 12° indicates nautical twilight. Calculations based on ship position (latitude, longitude) and time (GMT) via the R library maptools (Bivand and Lewin-Koh 2015), which uses algorithms provided by the National Oceanic & Atmospheric Administration (NOAA).
3. Lunar phase R Phase of the moon, the fractional part of which is 0 for new moon, 0.25 for
first quarter, 0.5 for full moon, and 0.75 for last quarter. Calculations as defined in equation 32.3 of (Meeus 1982) are based on ship position (latitude, longitude) and time (GMT) via the R library oce (Kelley 2014). To aid with modelling as a continuous covariate, values less than 0.028 (n = 80
swarms observed 27-28th Feb) were added to 1.0 (i.e. just beyond a dark moon).
Influence of recent ice history
4. Time since ice melt (wks)
R Time since ice melt (wks) was derived from the daily NSIDC sea ice concentration data, available on a polar stereographic grid at 25km resolution, as processed by the SMMR/SSMI NASA Team. The day of season any pixel was last covered by ice of at least 15% concentration was calculated (day of retreat). This was differenced from the current date and divided by 7d to give a value in weeks. Any pixel for which ice melt was not recorded was given the value zero. The seasonal ice minimum was taken as 15th February.
5. Ice melt rate R Ice melt rate is a spatial gradient calculated from the day of retreat raster (re-projected onto a long-lat grid) for each ice season as described above. To capture negative gradients (i.e. early melt close to the Antarctic continent in the south) a first difference was applied north-to-south. Values are square-root transformed.
Physical ocean habitat characteristics
6. Water temperature
U Ship-based measurement of surface water temperature (°C) from SeaBird SBE3 temperature sensor.
Sea surface temperature (SST)
R NOAA Optimum Interpolation daily sea surface temperature (°C) at 0.25 degree resolution (Reynolds et al. 2008).
7. Salinity U Ship-based measurement of surface water salinity (psu) from a SeaBird SBE21 thermosalinograph. Where possible missing values (n = 2300 of the 21826 records at 1 minute time intervals) were filled via direct linear calibration of the two salinity data streams (OTHER_SAL and TSG_SALIN).
Climatological summer surface salinity
R Salinity (psu) summer climatology at 0m depth (Raymond 2012, updated 2014). Data originally sourced from the World Ocean Atlas (Antonov et al. 2010) and regridded on a 0.1 degree spatial grid -180 to 180E, 80 to 30S (Antarctic).
8. Summer mixed layer depth
U Summer mixed layer (SML) depth (m) estimates derived from ship-based CTD and XCTD casts (Williams 2010). CTD stations occurred on alternate transects (total n = 48 on T7, T9 and T11) with XCTDs filling between (total n = 41 on T8 and T10). SML estimates for a given latitude on each transect were calculated by linear interpolation, excluding near-shelf estimates flagged as suspect (n = 6 and 8, respectively). Values are log-transformed.
Climatological mixed layer depth (February)
R Mixed layer depth (m) monthly climatology for February (Sallee 2013 pers comm). Southern Ocean dataset compiled from the largest number of available ocean observations (225389 profiles from the Argo program and 106682 ship-based and other profiles), available on a 0.5 degree long-lat grid. Values are log-transformed.
9, 10. ADCP current velocities (east-U and north-V)
U Depth-averaged ship-based velocity (cm s-1) vectors in the east-west (U) and north-south (V) direction (see also Figure 1). These have been compiled using both Lowered Acoustic Doppler Current Profiler (LADCP, n = 40 depth-averaged over 90-150m) data collected at CTD stations (Meijers et al. 2010) and underway hull-mounted ADCP (Rosenberg 2006) data (n = 374, depth-
averaged over 89-153m) that have been quality-controlled as detailed in Appendix 3. By transect, velocity estimates were linearly interpolated onto regular latitudes (0.05 degree steps) and a running median (window = 5) applied to smooth the values.
MADT current velocities (east-U and north-V)
R AVISO current velocities (m s-1) in the east-west (U) and north-south (V) direction. Based on the weekly mean absolute dynamic topography (MADT) at 0.25 degree long-lat grid resolution.
Physical habitat boundaries or edges
11. Bathymetric gradient
R Spatial gradient (radians) of the GEBCO_08 ocean bathymetry (m) data (IOC, IHO and BODC 2003, updated 2008). The 30 arc-second grid was first reprojected to Lambert equal area (centred at 64.5S, 70E) to enable local distances. Gradient was calculated over 8 neighbouring cells using the (Horn 1981) algorithm as implemented in the R library raster (Hijmans 2015). Values are log transformed.
12. Surface water temperature gradient
U Spatial gradient of surface water temperature (|°C| km-1) calculated from the ship-based measurements described above. By transect, temperatures were linearly interpolated onto regular latitudes (0.005 degree steps) and a running median (window = 5) applied to smooth the values. The absolute value of a first difference applied north-to-south was standardized by distance to give the temperature change per kilometre.
SST gradient R Spatial gradient (radians) of gridded SST calculated as described above for bathymetric gradient. Values are square-root transformed.
13, 14. ADCP current velocity gradients (east-U and north-V)
U Spatial gradient calculated from the ADCP velocity vectors (U and V) described above, comprising the absolute value of a first difference applied north-to-south. A running median (window = 5) was applied to smooth the values.
MADT current velocity gradients (east-U and north-V)
R Spatial gradient (radians) of gridded MADT current velocities calculated as described above for bathymetric gradient.
† A full list of the standardised ship instruments and sensors collecting underway data during the 2006 BROKE-West voyage
are archived at the Australian Antarctic Data Centre (http://aadc-
maps.aad.gov.au/aadc/voyages/sensors.cfm?set_code=200506030). All underway voyage data are recorded at 1 minute
time intervals. Individual krill swarms (n = 3097) were assigned the values corresponding to the minute in which the swarm
observation occurred. Or, in the case of spatially derived variables such as SML, ADCP currents and current gradients, and
the spatial gradient of temperature, obtained by direct linear interpolation (using the approx() function in the R library
stats).
†† For data obtained from remotely-sensed raster grids individual krill swarms were assigned the values corresponding to
the pixel in which the krill location occurred, using the extract() function in the R library raster (Hijmans 2015).
Figure A2.1. Schematic spatially depicting post-processed BROKE-West underway data streams used in predictive models. Boxed areas
demarcate data from transects T7-11. Y-axis is latitude (°S) whereas x-axes of boxed areas are rescaled and for relative purposes only. Examples
shown are the response variable krill swarm densities (log[g m-3]) together with the underway biophysical predictors of water temperature (°C),
salinity (psu), fluorometry (log[FLUlme]), and eastward velocity (cm s-1, ADCP and LADCP). Major circulation features overlaid are as in Fig 1 (main
text).
La
titu
de
Longitude
55 E 60 E 65 E 70 E 75 E 80 E 85 E
68
S6
6S
64
S6
2S
xxxxx xxx xx x xxxx xxx x xxxx xxxx xxxxxx xxxxxxxxxxxxxx xxxx xxx xxx xx xx x xxx xxx xx xxx x xxxx xxx xx xxx xxx x xxxx xxx x xx xxx x xxx xxxxx xxxx x xx x xxxx x xx xxx xx xxx x xxx xxx xxxx xxxxx xxxx xxx xx xxxxxxxxx xx xxx xx xx xxx xxx xx xx xx xxxx xx xxx x xx xx xx xxxx x xxx xxxx xx xx xx xxx xx xx xxx xxxx xxxxx xxx xxx xx xxx xx xxxx xxxxxxxxxxxxxx xx xxxxxxxxxxx x
xxxx xxxxx x x xx xx xx x xxxxx xx x xxxxxx xx x xxxx x xxxx xxx xx x x xxx x xxx xx xxx xx xxx xx xx x xxxxxxx x xx xx xxxx xxx x x xx x xxxx xxx xxxx x x xxx xxx xxxx x xx xxx x xxx xxxxxxxxx xx x xx xx xx xx x x xx x xx xxx xxxx xx xxx xxxx x xxxxxxx xx x xxxxx xx x xx xxxxxx xxxxxx xxx xx xx xxx xxxx x xx xxxx xxx xxxxx x xx xxx xx xxxxx xxxxxx xxx xx xx x xx xxx xxx x xx x xxx xxx xxxxxxx xx xxxx x xxxx xx xxx xx xx xx xxx xxxxx xx xxxxxxx xxxxxxxxx x xxxxxx xxxxxxxxxxxxxxxx xx xxxxx xx xxxxxxxx xx xxxxxxxx xxxxxxxx xxx xxxxx xxx xxxxxxxxxxx xx xxxxxxxx xxxx xxx x xx xx xxx xx xx xxx xx xxx xxxxxx xxxxx xxx xx xxxxxxx xxxxx xxxxxxxxxxxxxxx xxxx xxxxxxxxxxxxxxxxx xxxxxx xxxxxxxxxx xxxx xx xxxxx xx xxx x xxxxxxx xxxxxx xxx xxxx xxx xxx xxxxxx xxx xxx xxxxxx x xxxxxxxxx xxxxxxx xxxxxxxxxxx xxx xxxxxxxxxxxxxxx
xxxxxxx xxx xxxx xx x xxx xxx xxxxxx xxxxxxxxxx xxx
x xxxxxxx xxx xxx xxxxx xxxxxx x xxx xxxxxx
xx xx xxxxx xx xxx xx xx x x
x xxx x x xx xxxx xx x xxxx xxxxx xxx xx x xxxx x xx xx x xx xx xx xx xxx xxx x x xxx xxx xxxxx xx xxx xx xx xxxxxx xx xxx xxx xx xxx xxx xxxxx xxxxxxxxxxxxxx xx xxxxxxx xxxxxxxxxxxx xx xxx xx xx xxxxxxxx xxxx xxxx xx xxx xxxxx xxx xxxxxxxxx xxxxx xxxxxxxxxxxxxxxx xx xxx xx xxxxxxxxxxxxxx xxxxxxxxxx xxxxxxxxxxxxxxxxx xxxx xxx xx xxx xxx xxxxxxx xxxx x xx xxxxxxx x xxxx xxx xx xx xxx x xx xxxx xx xxx xxxxx xxx x xxxxxxxx xx xxxxxxxx xxxx xxxxxxxxxxxxx xx xxxxxxxxxxxx x xxxx xxx xx xx xx xxxx x xx xxx xxx xxxx xxx xxxxx xx xx xx x xxx xxx x xx xxxx xx x x x x
x
x
xxxxxxxx xxx xxxxxx xxx xxx xx
x xx
xx
xxx xxx x
xx x xxxxx xx xx xxxxx xx xxx
x
x xx
x xxx xxxxxx xx xx xxxxx xx
x x xx xxxx xx xx xxxxxx x x xxxx x x xx xxx xx xxx xxxxx x xx xxxxx xxxxxxxx x xxxx xxx
x xx xx xxxxxxxxxxxxxx xx xxxxxx xxxx xxxxxxx xxxx xxxxx xxxxxxxx xx xxx xx x xxxxx xxxxx x xxx xx x xx xxx xxxxxx x xx xxx xxxxxx xxxxxx x xxx xxx xxx xxx xxxxxx xxx xxxx xxx x xxxxxxx x xx x xxx xxxx x xx xx xxx xxxxxx xxxxxxx xxx xx xxx x x xxx x xx xx xx xxxxxx x xx xxxx xxxx xx xx xxxx xx xxx xx
x xxxx x x xxxxx xxxx xx x
xx xx x xxx xx x xx x xxxx xxxxxxx xx x xx xx xx xx xxx xx xxxx xxxx xxxx x xxxxx xxxx xxx xx xxx xxxxxxx xx xx xx xx xx xx x x xxx x xxx xx xxx x xxxx xx xx xx xxxx xxx xxxxx xxxxx xx xxxxx xx x xx xx x xxx xx xx xx xx xx xx xx xxxxxxxxx xx xxxx xx xx xxxx xxx xx xxx x xxx xxxxx xxx xx xx xxx x xxx xxx xx xxx xxx xxx x xx xxx x xx xxxx xxx x xx x x
xxx xxxxxxxxx xxxxx x x xxxx xx x xx xxx xxx xxxxx x xx xx
x x
x
xx
x xx xxx xxx xxx xx
x x xx
xxx
xxx xx
x
x xx
x
xx xxx x
x xx xxx xx xxxxx xx xx xxxxx x xxxxx xxxxxxxx x xx xx xxx xxxx xxxxxx xxxx xxx xxxxx xxxx xxxxx xxxxxxxxxx xx xx xxxx xxx xx xxxxxxxx xxxxx xxxxx xxxxxxxxxx xxx xx xx xxxx xxx xxxx xxxxxxxxxxxx xxxxxxxxxxxx xxxxxxxxxx xxx xxxxxx xx xxx
xxxx xxx xxxxxxx xxxxxxxx xxxx
xx
x x
xxx
xxxx x xx xxx x
x xxxx x xxxx x xx x xxxxx xxxxx xx x xxxx xxxxxxxxx xx xxxxxxxx xxxxxxx xxxx xxxxx xxx x xx x xx xx xx x xx xxxx xxxx xxxx xx x xxxx xxxx xxx xxx xxx x xxxx xx x x xx x xxx xxx x x xx xx xxx x xx xxxx x xx xxx xx xx x xxx xxx x xxxx xxx x xx xxx xx
x
xx xxxxxx xx
xxx xxxxx xxx xxxxx xx
xx xxx xxx xxxxx xx xxx xxxx xxxxx xxxxx xx x x
xx
xx
x
x
x xxx
x
xxxxx x
x
xx x
xx
x
x
xx x
xxx
x
xx
xx xxxxxxxxx xxxxxxx xxxxxxx
xxx xxx xx xxxxxxxx xx xxx xxx x xxxx x
x xxxxx x
xxx xxx
xx xx
xxxxxx
xx xxx
x
x
xx xx
xx
xx
xx
x
xxxx
x
x xxx x xxxx xx x x xxx xxxx xx xx xx xxx xx xxx xxxxx x xxxxxxxx xxx x
x
xx
x x
x
x xxxxxxxx x xxxx x xxxx xxxx xx xxxxx xxxx x xxx x xxxxxxx xx xxx xxx xxxxxx
xxxx xx xx xx
xxx
xx xx xxxxxxx xxxxxxxx xxxxxx
x
xxxx xxxxxxx
xxxx xx xxx xxx x xxx xxx xx xxxx xx
xx xx x xxx xxx
xx xxx xx
xx xxxxx xxxx xxxxxx x xxxxxxxx xx xx xx x xx x xxx xxx xxx x xxx xx
x xx xx x xxx xxx
x
Underway
T (degC)S (psu)
log[FLU_lme]
x
x
Acoustic
Swarm density(log[g/m3])
20 cm/s
ADCP
LADCP
Circulation
ACC
PBGASC
Figure A2.2. Example time-series of underway (gray) biophysical predictors and their remotely-
sensed (black) analogues for (a-b) temperature (°C), (c-d) chlorophyll a (log10[mg m-3]), and eastward
velocity (cm s-1).
Feb 11 Feb 16 Feb 21 Feb 26
-1.5
-1.0
-0.5
0.0
0.5
1.0
d1$gmt
x
(a)T
em
pera
ture
(degC
)
-1.5 -1.0 -0.5 0.0 0.5 1.0x
y
-1.5
-1.0
-0.5
0.0
0.5
1.0
(b)
Feb 11 Feb 16 Feb 21 Feb 26
-1.0
-0.5
0.0
0.5
d1$gmt
x
(c)
Chla
(lo
g10[m
g/m
3])
-1.0 -0.5 0.0 0.5x
y
-1.0
-0.5
0.0
0.5
(d)
Feb 11 Feb 16 Feb 21 Feb 26
-10
0
10
20
d1$gmt
x
(e)
E v
elo
city (
cm
/s)
UnderwayRemote
Time
-10 0 10 20x
y
-10
0
10
20
(f)
Underway
Rem
ote
References
Antonov, J. I. et al. 2010. World Ocean Atlas 2009, Volume 2: Salinity. NOAA Atlas NESDIS 69. — U.S. Government Printing Office. Bivand, R. and Lewin-Koh, N. 2015. maptools: Tools for Reading and Handling Spatial Objects. — R package version 0.8-36. http://CRAN.R-project.org/package=maptools. Hijmans, R. J. 2015. raster: Geographic Data Analysis and Modeling. R package version 2.4-6. http://cran.r-project.org/web/packages/raster/. Horn, B. K. P. 1981. Hill shading and the reflectance map. — Proc. IEEE 69: 14-47. IOC, IHO and BODC. 2003, updated 2008. Centenary Edition of the GEBCO Digital Atlas, published on CD-ROM on behalf of the Intergovernmental Oceanographic Commission and the International Hydrographic Organization as part of the General Bathymetric Chart of the Oceans. — British Oceanographic Data Centre, http://www.gebco.net/data_and_products/gridded_bathymetry_data/gebco_30_second_grid/ Johnson, R. et al. 2013. Three improved satellite chlorophyll algorithms for the Southern Ocean. — J. Geophys. Res. Oceans 118: 3694–3703. Kelley, D. 2014. oce: Analysis of Oceanographic data. — R package version 0.9-14.http://CRAN.R-project.org/package=oce Meeus, J. 1982. Astronomical formulae for calculators. — Willmann-Bell. Meijers, A. J. S. et al. 2010. The circulation and water masses of the Antarctic shelf and continental slope between 30 and 80 degrees E. — Deep Sea Res. Part II 57: 723-737. Raymond, B. 2012, updated 2014. Polar Environmental Data Layers. Australian Antarctic Data Centre - CAASM Metadata. (http://data.aad.gov.au/aadc/metadata/metadata_redirect.cfm?md=/AMD/AU/Polar_Environmental_Data) Reynolds, R. W. et al. 2008. NOAA Optimum Interpolation ¼ Degree Daily Sea Surface Temperature (OISST) Analysis, Version 2. [February 8 -28th 2006]. — NOAA National Climatic Data Center. doi:10.7289/V5SQ8XB5 http://www.ngdc.noaa.gov/docucomp/page?xml=NOAA/NESDIS/NCDC/Geoportal/iso/xml/C00844.xml&view=getDataView&header=none. Rosenberg, M. 2006. BROKE West Survey, Marine Science Cruise AU0603 - Oceanographic Field Measurements and Analysis. http://data.aad.gov.au/aadc/reports/display_report.cfm?report_id=4042. pp. 1-24. Sallee, J. B. 2013 pers comm. Southern Ocean climatological dataset for monthly mixed layer depth. — ftp://ftp.nerc-bas.ac.uk/jbsall/MLfitted_SO.mat Wiley, P. 2006, updated 2014. Aurora Australis Voyage 3 2005-2006 (BROKE-West) Underway Data. Australian Antarctic Data Centre - CAASM Metadata (http://data.aad.gov.au/aadc/metadata/metadata_redirect.cfm?md=/AMD/AU/200506030).
Appendix 3
Underway survey data processing
1. Fluorometry calibration Underway fluorescence data is routinely collected during research voyages as an inexpensive means
to characterise in situ chlorophyll a concentrations in surface waters. However fluorescence yield
can be highly variable, particularly in relation to the phytoplankton cells immediate light history
(Holm-Hansen et al. 2000). A number of processes can result in “quenching”, where cells exposed to
high light intensity cease to fluoresce, posing significant problems for the reliability of using in situ
fluorescence as a proxy for chlorophyll a concentrations.
A strong diurnal pattern, indicative of quenching, was apparent in the underway fluorescence (FLU)
data collected within the Kerguelen Axis region of the BROKE-West survey (Fig A3.1). However there
are no standardised protocols for calibrating such data. Commonly, a two-step approach involves (a)
calibration of underway FLU with laboratory HPLC measurements, followed by (b) assessment of the
inhibitory effect of solar radiation on the specific fluorescence yield (e.g. Holm-Hansen et al. 2000).
However it is possible that the latter may in fact form a significant source of non-correspondence
between the underway and laboratory measurements.
We investigated two options for calibrating the BROKE-West underway fluorescence data (Wiley
2006, updated 2014) collected within the Kerguelen Axis region (transects 7-11). Firstly, we
modelled FLU in relation to near surface HPLC data collected at CTD stations (Wright et al. 2010)
including underway solar radiation (SOL) directly as a model term (Table A3.1) in a linear mixed
effect model (lme), where the CTD station was included as a random effect with heterogeneous
variance allowed across stations. SHELF was also included as a fixed effect as the underway FLU
appeared to under-represent near-surface chl a at southern stations close to the Antarctic
continental shelf. Linear mixed effect models were fit using the nlme library (Pinheiro et al. 2013) in
the freely available R software (R Development Core Team 2015).
For comparison, we employed a second approach whereby the diurnal signal was stripped via
temporal decomposition (applying the function stl, using a 1440 minute i.e. 24h window, from the R
library stats). The adjusted FLUstl stripped of the periodic component (with amplitude +0.200 and –
0.246, hence comprising only the trend plus remainder component) was then used in the model
fitting stage analogous to that described above (without the SOL term). Log10 transformations were
applied to all continuous variables (FLU, HPLC and SOL) to help linearise the relationships. Models
were fitted to day and night data separately, with preliminary investigation identifying the apparent
threshold value for photo-inhibition of fluorescence at 40 W m-2.
The resulting calibrated FLU time-series show the quenching signal to be in large part removed (Fig
A3.1) and quite good concordance between the two methods. Our results indicate that the
calibrated underway FLU can certainly be used as a proxy indicator to tell us something about in situ
chlorophyll a concentration (Fig A3.2, b-c). Likewise, comparison of the full calibrated FLU time-
series with the corresponding independent remotely-sensed MODIS estimates of chl a (Johnson et
al. 2013) indicate the calibrated underway FLU can be used as a proxy indicator for in situ near-
surface chlorophyll a concentration (Fig A3.3, b-c). However, there does appear to be some remnant
light-related pattern remaining, whereby near-dawn values may be under-estimated (see blue-
purple points in Fig A3.3).
Table A3.1. Algorithms used for estimation of chlorophyll a concentration from underway
fluorescence data. Two approaches, using lme and stl are presented (see text for details). Models
were fitted to day and night data separately.
Model Estimated coefficients*
lme day log10(FLU+0.25) ~ 0.74758 + 1.15661*log10(HPLC) – 0.14789*log10(SOL+1) – 0.30404*SHELF night log10(FLU+0.25) ~ 0.55766 + 0.99821*log10(HPLC) – 0.03242*log10(SOL+1) – 0.27332*SHELF stl day log10(FLUstl+2.66) ~ 0.50665 + 0.18963 *log10(HPLC) – 0. 05579*SHELF night log10(FLUstl+2.66) ~ 0.50182 + 0.18118 *log10(HPLC) – 0. 04954*SHELF
*Here HPLC refers to the near-surface chl a (μg l-1) from CTD station measurements (see Fig A3.2 for details
and also Wright et al. (2010)); FLU refers to the underway fluorometry (unit less); FLUstl refers to the underway
fluorometry (unit less) stripped of the diurnal signal prior to model fitting; SOL refers to ship-based underway
solar radiation (W m-2); and SHELF is a binary (0/1) factor demarcated by the 1500 m isobath for each station.
Note also offsets applied to ensure above-zero values enabling log transformations.
Figure A3.1. Time-series of underway fluorometry data during BROKE-West within the Kerguelen
Axis region (transects 7-11). The raw 1-minute average fluorescence data are coloured by day (gray)
and night (black) where the threshold is set at 40 W m-2 solar radiation. Two fluorometry calibrations
are also shown, the first includes underway solar radiation as a model term in a linear mixed effect
model (lme, magenta), whereas the second uses a temporal decomposition to strip the diurnal signal
prior to model fitting (stl, blue). Vertical dashed lines indicate the start of each of the north-south
transects 7-11, with voyage day referenced to 28-Dec-2003 as day 1.
Figure A3.2. (a) Raw and (b-c) calibrated underway fluorometry (n=4306) plotted in relation to total
chl a (μg l-1) determined via HPLC analysis of Niskin bottle samples taken at 44 CTD stations (72-99
and 103-118) along the north-south legs of transects 7, 9 and 11 (Wright et al. 2010). The calibration
utilised the HPLC data from the shallowest depth sampled (range 4-14m) at each CTD station. The 1-
minute averaged underway fluorescence data was included for the shortest period of either (i) the
duration of the CTD station time (start to end), or (ii) a 2 h period straddling the CTD station time
mid-point. Diagonal line indicates 1:1 relationship for reference. Data are coloured as in Fig A3.1.
Figure A3.3. Southern Ocean MODIS estimates of chl a (mg m-3) (Johnson et al. 2013) plotted in
relation to (a) raw and (b-c) calibrated BROKE-West underway fluorometry (n=27642) measurements
within the Kerguelen Axis region (transects 7-11). Data are coloured according to normalised time of
day (see key, panel c).
-1.0 -0.5 0.0 0.5
-1.0
-0.5
0.0
0.5
raw
uw ay$flu.log[ff]
log10(u
way$chla
RJb[ff])
log10(M
OD
IS)
log10(flu)
(a)
-1.0 -0.5 0.0 0.5
-1.0
-0.5
0.0
0.5
lme
uw ay$F_lme
log10(u
way$chla
RJb)
log10(flu)
(b)
-1.0 -0.5 0.0 0.5
-1.0
-0.5
0.0
0.5
stl
uw ay$F_stl2
log10(u
way$chla
RJb)
log10(flu)
(c)
Time of day
daw n (0.5)
midday (1)dusk (1.5)
midnight (2)
2. ADCP processing
The errors associated with the underway ADCP data obtained from the hull-mounted ADCP on the
Aurora Australis are well documented (Rosenberg 2006). In general, these data are contaminated by
the ships motion when the ship accelerates i.e. changes direction or speed. The most reliable data
are collected when the ship is "on station" (≤ 0.35 m/s) or travelling at speeds ≤ 1 m/s. Underway an
erroneous vertical ADCP current shear occurs, most significantly above bin 10 (i.e. bin centres 21 –
93m), likely due to acoustic ringing from an air/water interface inside the transducer sea-chest.
Best quality underway ADCP data should be obtained when the ship is steaming in a straight line at a
suitable constant speed, at depths below those in which the vertical shear is observed. Here, we
utilise the 30 minute ensemble ADCP data available1 along transects 7-11 representing subsurface
flows at around 100m depth (i.e. U and V components separately averaged across depth bins 10-17
[covers the depth range 89 – 153 m] using those data with an initial quality flag of 5 or above; n =
492). It should be noted that for the major currents these ADCP velocities will likely be conservative
relative to the surface flows, but there may be additional subsurface influences such as those related
Ekman drift or subsurface eddies (Meijers et al. 2010, Williams et al. 2010). Whilst krill swarms at
and below 89m represent only 20% of the swarms recorded in this region of the survey, it is likely
through diurnal vertical migrations that a substantially greater portion of the krill biomass will move
through these depth layers.
To help interpret whether the observed variability in current flows was likely to be real or an artefact
relating to the ship’s motion, further quality control involved cross-referencing the underway ADCP
velocities with the original 1 second ship track GPS data. Mean and variance statistics for ship speed
and direction were calculated for each 30 min ensemble period (Fig A3.4). For direction these were
calculated via circular statistics using the R library CircStats (Lund and Agostinelli 2012). These data
clearly show the low-speed “on station” periods and also the high-speed periods of directed transit.
We applied threshold criteria such that ADCP data were utilised where (i) mean ship speed was ≤ 1
m/s, OR (ii) at higher mean ship speeds if the variance in both speed AND direction were low (≤ 1
and < 0.2 , respectively). This comprises 76% of the available underway ADCP data (n = 374). These
ADCP data are mapped within the context of the regional ocean circulation in Fig 1 of the main text.
Figure A3.4. Mean (_mn) and variance (_var) statistics for ship speed (s) (m s-1) and direction (d)
(true degrees) during each 30 minute ADCP ensemble period calculated from the raw 1 second ship
track GPS data.
Acknowledgements
Rob Johnson provided useful information contributing to the fluorometry calibration approaches
adopted here. Mark Rosenberg and Andrew Meijers provided helpful suggestions contributing to the
ADCP quality control approaches adopted here. MR also provided the raw 1 second ship track GPS
data. A full list of the standardised ship instruments and sensors collecting underway data are
archived at the Australian Antarctic Data Centre (http://aadc-
maps.aad.gov.au/aadc/voyages/sensors.cfm?set_code=200506030). The underway ADCP current
velocity data from BROKE-West are publicly available at:
http://gcmd.nasa.gov/KeywordSearch/Metadata.do?Portal=amd_au&MetadataView=Full&Metadat
aType=0&KeywordPath=&OrigMetadataNode=AADC&EntryId=BROKE-West_ADCP
References
Holm-Hansen, O. et al. 2000. Reliability of estimating chlorophyll a concentrations in Antarctic waters by measurement of in situ chlorophyll a fluorescence. — Mar. Ecol. Prog. Ser. 196: 103-110.
Johnson, R. et al. 2013. Three improved satellite chlorophyll algorithms for the Southern Ocean. — J. Geophys. Res. Oceans 118: 3694–3703.
Lund, U. and Agostinelli, C. 2012. CircStats: Circular Statistics, from "Topics in circular Statistics" (2001). R package version 0.2-4. http://CRAN.R-project.org/package=CircStats.
Meijers, A. J. S. et al. 2010. The circulation and water masses of the Antarctic shelf and continental slope between 30 and 80 degrees E. — Deep Sea Res. Part II 57: 723-737.
Pinheiro, J. et al. 2013. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-108.
Rosenberg, M. 2006. BROKE West Survey, Marine Science Cruise AU0603 - Oceanographic Field Measurements and Analysis. http://data.aad.gov.au/aadc/reports/display_report.cfm?report_id=4042. pp. 1-24.
R Development Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computing. — Vienna, Austria. http://www.R-project.org/.
Wiley, P. 2006, updated 2014. Aurora Australis Voyage 3 2005-2006 (BROKE-West) Underway Data. Australian Antarctic Data Centre - CAASM Metadata (http://data.aad.gov.au/aadc/metadata/metadata_redirect.cfm?md=/AMD/AU/200506030).
Williams, G. D. et al. 2010. Surface oceanography of BROKE-West, along the Antarctic margin of the south-west Indian Ocean (30-80 degrees E). — Deep Sea Res. Part II 57: 738-757.
Wright, S. W. et al. 2010. Phytoplankton community structure and stocks in the Southern Ocean (30–80°E) determined by CHEMTAX analysis of HPLC pigment signatures. — Deep Sea Res. Part II 57: 758-778.
Appendix 4
Satellite tracking data
Table A4.1. Further details on the species for which tagging data was available in or near to the Broke-West study region during midsummer (January and
February). Raw Argos locations were filtered using the state-space model of (Jonsen, Flemming et al. 2005) (see Methods) prior to model predictions being
made along individual tracks (Fig. 7, main text).
Year Species, release location
Data date range
No. of Individuals and relevant life stage information
Tag type No. track locations*
Data status and custodian
References
2001 Adelie penguins, Béchervaise Island, Antarctica (62.81°E, 67.58°S)
1/1 to 15/2
22 guard and crèche
Argos PTT 2741 Public, J. Clarke
(Emmerson et al. 1999, updated 2014)
2004 Antarctic fur seals, Spit Bay, Heard Island (73.75°E, 53.10°S)
8/1 to 29/2
15 males
Kiwisat PTT (incl. 4 SMRU SRDLs)
4053 Public, N. Gales
(Gales et al. 2004, updated 2015, Frydman and Gales 2007)
2006 White chinned petrels, Cañon des Sourcils Noirs colony, Kerguelen Island (70.26°E, 49.61°S)
19/1 to 18/2
10 incubation and chick-rearing
Argos PTT (incl. 5 solar PTTs)
1900 Private, H. Weimerskirch
(Péron et al. 2010)
2007 Emperor penguins, Auster rookery, Antarctica (64.03°E, 67.38°S)
1/1 to 28/2
10 fledglings
Argos PTT 527 Public, B. Wienecke
(Wienecke and Robertson 2000, updated 2015, Wienecke et al. 2010)
*indicates state-space filtered locations.
References
Emmerson, L. et al. 1999, updated 2014. Satellite Tracking of Adelie Penguins Around Mawson Station. Antarctica Australian Antarctic Data Centre - CAASM Metadata (https://data.aad.gov.au/metadata/records/Tracking_BI). Frydman, S. and Gales, N. 2007. HeardMap: Tracking marine vertebrate populations in near real time. Deep Sea Res. Part II 54: 384-391. Gales, N. et al. 2004, updated 2015. Animal Tracking at Heard Island 2003/2004 - ARGOS data. Australian Antarctic Data Centre - doi:10.4225/15/55C9581205236. Jonsen, I. D. et al. 2005. Robust state-space modeling of animal movement data. Ecology 86: 2874-2880. Péron, C. et al. 2010. Seasonal variation in oceanographic habitat and behaviour of white-chinned petrels Procellaria aequinoctialis from Kerguelen Island. Mar. Ecol. Prog. Ser. 416: 267-284. Wienecke, B. et al. 2010. Maiden journey of fledgling emperor penguins from the Mawson Coast, East Antarctica. Mar. Ecol. Prog. Ser. 410: 269-282. Wienecke, B. and Robertson, G. 2000, updated 2015. Foraging ecology of emperor penguins in summer and potential overlap with fisheries. Australian Antarctic Data Centre - CAASM Metadata (https://data.aad.gov.au/metadata/records/ASAC_1252).