Supplemental:
Thermal blankets are used to determine conductive heat flow in marine areas where it is physically
impossible to insert a traditional 3-meter long heat flow probe into the seafloor. Such regions include
carbonate-cemented sediments near hydrocarbon seeps or unsedimented regions of exposed basalt
associated with mid-ocean spreading ridges.
As with any other oceanographic method for data acquisition, thermal blankets have their own
deployment strategies and post-processing routines necessary to obtain the most accurate heat flow
measurements. While thermal blankets have been used previously [Johnson et al., 2010 and references
therein], this supplemental material provides details of an updated version of the data processing methods.
Although the physical configuration of the thermal blankets has remained largely changed since previous
studies, the post processing techniques have been substantially updated.
The core of the thermal blanket consists of urethane foam with a 0.3 mm pore diameter and when
saturated is approximately 98% water by volume. The outer protective shell is made of plastic laminated
fabric with four 1 cm diameter port holes allowing water to saturate the blanket but severely restricts
exchange with outside seawater. Autonomous Antares thermistor/loggers are attached to the fabric shell
with the uncovered tip of the temperature sensor exposed to ambient seawater. The bottom thermistor is
adjacent to the rock surface and detects warming in the small volume of water sealed in the gap between
the overlying blanket and underlying rock. The total volume of this gap is very small (few tens of
milliliters) compared to the blanket volume and assumed to have very little effect on the temperature and
thermal conductivity of the blanket. To guarantee a tight seal with the underlying seafloor, a thin ring of
compliant foam lines the outer perimeter beneath the thermal blanket. To insure thermal isolation of the
underlying thermistor, a heavy motorcycle inner tube filled with CaCl2 solution is attached to the upper
outer edge of the blanket and reinforced with rings of lead wire designed to ensure a solid seal.
Polypropylene line is attached to a buoyant plastic placard to allow identification and manipulation by an
ROV during deployment and recovery.
Pre-Deployment Preparation:
The Antares thermistor probes were calibrated in the laboratory in order to correct for offsets between
instruments by immersion in an ice water bath aboard the ship just prior to deployment. The calibration
time was 2 hours allowing the thermistors to equilibrate with the ice water temperature. The ice water
bath was manually stirred at 15 minute intervals to maintain constant temperature, taking care not to
disturb the thermistors. The offset for each thermistor was calculated from the total mean temperature of
all thermistors during this ice water calibration. This offset was later applied to the corresponding
thermistor during the post-processing. Offsets for all sensors ranged from -0.06°C to +0.05°C and the
detected drift in thermistors over the 2 hour test was within ±0.002°C of the mean. Prior to deployment,
all thermistor internal clocks were synchronized to a single computer clock and were programmed to start
recording at the same time. It is a reasonable assumption that there were no significant time drifts over the
48-hour dive periods.
Thermal Artifacts
Ideal blanket deployments require the top thermistor temperature (measuring seawater) to remain
relatively constant over the entire deployment period. In practice, near-seafloor currents and water
column temperature fluctuations added noise to the thermal gradient measurements. This situation is
especially apparent in stations with very low heat flow where noise from surface temperature variations
propagated downward through the blanket, disrupting the thermal gradient measurement (Supplemental
Figure 1C). Removing the unwanted noise caused by bottom water variations is essential for accurate heat
flow estimations with low heat flow sites, and this noise was minimized during post-processing as
described below.
Due to their high heat capacity and relatively low thermal conductivity, blankets are able to retain a
transient memory of the previous thermal environment (Supplemental Figure 1D). This is particularly
noticeable in thermal blankets deployed on the seafloor directly from the warm surface ocean, but is also
an issue during transfer from a high to low heat flow deployment site. The warmer-than-seawater foam is
not detected by the thermistors during transits until after the blanket was deployed onto the seafloor,
resulting in what appeared to be a heat pulse by the bottom thermistor.
The individual thermal blankets were deployed at numerous stations during an ROV dive. In order to
reduce the effects of retained surface heat on the initial station, all blankets were suspended above the
seafloor for several hours prior to deployment to allow equilibration with the cold near-bottom water
temperatures. This pre-deployment equilibration period considerably reduced the effects of retained
surface heat, but still required post-processing corrections in order to eliminate it completely.
Thermal conductivity analysis:
As with more traditional heat flow instrumentation, the largest source of error is thermal conductivity
[Villinger and Davis, 1987]. The water-saturated foam is normally assumed to have a thermal
conductivity value of non-convecting seawater [Johnson and Hutnak, 1996; Caldwell, 1974]. While the
infinite half space solution to the heat equation (Equation 2, in main article) produced an approximate
estimate of thermal conductivity, it occasionally produced unrealistic values. A likely explanation is at
high heat flow values (>500 mW/m2), the vertical thermal gradient within the blanket can produce porous
convection throughout the interconnected foam cells [Johnson and Hutnak, 1997], altering the effective
thermal conductivity. To overcome this variability, a best-fit line of the measured thermal conductivity
was used instead, which fell at and slightly above the value for non-convecting water thermal
conductivity (Fig. 4).
The current field study shows that in situ thermal conductivities of the sealed blanket foam are relatively
independent of temperature (Supplemental Figure 2). Previous laboratory estimates of thermal
conductivity of the foam under different conditions from the field environment resulted in a thermal
conductivity estimate relation that was more dependent on the applied thermal gradient [Johnson and
Hutnak, 1997]. An estimation of thermal conductivity produced directly from the in situ data should
improve the accuracy of the calculated heat flow. As a byproduct of the post-processing to remove the
impact of bottom water temperature variations, the individual thermal conductivity was produced for each
blanket station deployment. This thermal conductivity was compared to the total change in temperature of
both thermistors over the deployment time, or the difference between the start of deployment temperature
and the estimated bottom thermistor temperature at final equilibrium. For the purpose of this comparison,
thermal conductivity values from thermal blankets that showed signs of residual surface heat, or measured
zero W m-2 or less were removed, as well as records that did not fit the idealized conductive model of a
solid non-convecting slab. This included stations that had a model fit (root mean standard deviation) of
more than ±0.001°C and a curve fit of ±0.0025°C. Anything within these bounding values conformed to
our assumption of the blanket as an ideal non-convecting solid slab and a logarithmic equation was fitted
to the data (Supplemental Figure 2, grey line).
Duplications/Comparison to other Marine Heat Flow methods:
As an experiment-of-opportunity during the same cruise, two thermal blanket deployments, N-0 and J-0,
were placed in Middle Valley, allowing for direct comparison of thermal blanket measurements to
traditional violin bow long (3 meter) thermal probe deployments (Villinger and Davis, 1987; Davis and
Villinger, 1992). During the data acquisition survey presented in this paper, the Alvin short heat flow
probe (1 meter) was also used by ROV Jason II within the area of the thermal blanket deployments
(Supplemental Figure 3). These multiple deployments in approximately the same general area allowed
comparison between the three types of heat flow instrumentation. In this unplanned comparison with
datasets acquired 20 years apart, only one thermal blanket measurement, compared to a nearby Alvin heat
flow probe made on the same cruise, fell on the 1:1 line within measurement error. The comparisons
between the long probe and Alvin heat flow probe appear to have departures from the 1:1 line of the same
scale as the thermal blankets.
Several thermal blankets were placed within visual distance of each other on the 2011 cruise, providing a
test of the precision within the thermal blankets (Supplemental Figure 3). Of the three paired blanket
deployments, two pairs have corresponding Alvin heat flow probes at adjacent sites. Of the third pair of
thermal blankets, both blankets failed to seal to the seafloor. While the two stations with good seals do
not fall within calculated error of the Alvin heat flow probes, they do provide measurements within ~2%
error of each other, a clear example of the thermal blanket’s precision.
One possible reason for the inconsistencies in heat flow measurements include spatial difference ranging
from 100-150 meters and temporal differences of 20 years between the long probe and heat flow stations
placed by Jason. Inconsistencies within individual deployments could be obtained both from non-vertical
deployment (tilting) of the Jason heat flow probe and difficulties accounting for surface residual heat in
the thermal blankets. However, a more recent deliberate comparison between the long OSU heat flow
probe, the short Jason probe and multiple blanket deployments on the Washington margin [Johnson et al,
2013] show general agreement, within 10%, between the three methodologies.
References:
Caldwell, D.R., (1974), Experimental study on the onset of thermohaline convection. J. Fluid
Mech., 64, 347-367.
Davis, E.E., and Villinger, H., (1992), Tectonic and thermal structure of the Middle Valley
sedimented rift, northern Juan de Fuca Ridge. In Davis, E.E., Mottl, M.J., Fisher, A.T., et al.,
Proc. ODP, Init. Repts., 139: College Station, TX (Ocean Drilling Program), 9–41.
doi:10.2973/odp.proc.ir.139.102.1992
Johnson, H. P., and M. Hutnak, (1996), Conductive heat flow measured in unsedimented regions
of the seafloor. Eos, Transactions American Geophysical Union, 77, 321.
Johnson, H. P., and M. Hutnak, (1997) Conductive heat loss in recent eruptions at mid‐ocean
ridges, Geophys. Res. Lett., 24, 3089–3092,
Villinger, H., and E. E. Davis, (1987), A new reduction algorithm for marine heat flow
measurements. Journal of Geophysical Research, 92, 12846.
Davis, E. E., and H. Villinger. (1992), 2. Tectonic and Thermal Structure of the Middle Valley
Sedimented Rift, Northern Juan de Fuca Ridge. Ocean Drilling Program, Initial Reports.
Figure Captions:
Supplemental Figure 1: Examples of difficult blanket deployments. A) The bottom thermistor did not
seal properly with the ground. B) Blanket was deployed over an active diffuse vent. C) Temperature
variability propagating through the blanket. D) Residual heat absorbed on board the ship stored within the
blanket material.
Supplemental Figure 2: Determining the thermal conductivity (k) of the blankets. Plotted thermal
conductivities are from ideal blankets with a good fit with the half-space heat equation. Error bars are
calculated based on how variations of ±0.002°C will affect thermal conductivity estimations. Best fit line
is the logarithmic equation Y = 0.0348*log(x) + 0.7339 used to calculate the final thermal blanket heat
flow values.
Supplemental Figure 3: Comparison of two heat flow blankets (J0 and N0) to two Middle Valley long
heat flow probes (Top Panel; from Stein et al., 1998) and heat flow blankets compared to Alvin short
probe measurements (Middle Panel). For reference purposes, Alvin heat flow probe is compared to the
long probe for Middle Valley (Bottom Panel). Error bars are 20% of the heat flow values and the grey line
is the 1:1 ratio.