Report on Rottnest Ferry SST Validation Tests (5-7 April 2011)
Authors: Ruslan Verein (r.verein@bom.gov.au), Helen Beggs (h.beggs@bom.gov.au), Eric Schulz
(e.schulz@bom.gov.au) and George Paltoglou (g.paltoglou@bom.gov.au), CAWCR, Bureau of
Meteorology, Melbourne, Australia
Hiski Kippo (CSIRO Marine and Atmospheric Research, Perth, Australia)
Date: 15 September 2011
Introduction
The PV SeaFlyte (Call Sign VHW5167) is a fast moving (up to 24 knots) tourist ferry that does regular daytime transects between Hillary’s Boat Harbour and Rottnest Island in Western Australia. The vessel
occasionally also does night-time fishing cruises. As a contribution to the Australian Integrated Marine
Observing System (IMOS), in early 2008 the vessel was equipped with a SeaBird SBE 38 Digital
Oceanographic Thermometer. The SBE 38 has high accuracy (absolute accuracy is better than 0.001 °C (1
mK) and resolution is approximately 0.00025 °C (0.25 mK)) and stability (drift of less than 0.001 °C in sixmonth period) (SeaBird SBE 38 Specification Document). The SBE 38 is installed in the engine cooling
intake in a “dead-end” pipe about 5-10 cm away from the main water flow (Figure 1). Water flow rate
through the engine intake pipe is variable depending on the vessel speed. When the engines are not running
there is no water pumped through and hence there is a threshold of 5 knots for data logging. The depth of the
water intake is 0.1 to 0.5 m depending on vessel speed with faster speed resulting in shallower draft.
Since 30 April 2008, the SeaFlyte SST observations, combined with time of observation, latitudes and
longitudes, have been transmitted daily to the Bureau of Meteorology where they have undergone automatic
quality control (Beggs et al, 2011), averaged over a 1-minute period and uploaded to the Global
Telecommunications System (GTS) as Trackob records and to the IMOS Ocean Portal in daily netCDF files
Data can be downloaded from http://opendap-tpac.arcs.org.au/thredds/dodsC/IMOS/SOOP/SOOPSST/VHW5167_Sea-Flyte/2011/catalog.html. Validation against SST from polar-orbiting satellites (NOAA17, 18 and 19) has shown that there are relatively high warm biases and high standard deviations in the
SeaFlyte SST (at 0.1-0.5 m depth) compared with the satellite SST (Appendix A, Figures A.1 to A.2; Beggs
et al., 2011). The HRPT AVHRR skin SST data from the NOAA satellites were converted to an approximate
foundation (pre-dawn) SST (SSTfnd) by regression against matchups with drifting buoy SST at 20cm to
30cm depths and filtering on NWP 10m wind speeds (Paltoglou et al., 2010). For the matchups, SeaFlyte
SSTdepth data was also converted to foundation SST by rejecting data for winds < 6 m/s during the day and
< 2 m/s at night). The SeaFlyte SSTfnd values were biased warm an average 0.4C (standard deviation of
0.9C, N = 249) during the day and 0.6C (standard deviation of 0.9C, N = 130) during the night compared
with 2 hour matchups with 1 km resolution HRPT AVHRR SSTfnd (quality level 4 and 5) from NOAA-17,
18 and 19 satellites for the period 1 December 2008 to 1 June 2011. These observed warm biases and high
standard deviations were of concern as the daytime matchups between the other nine IMOS ship SSTfnd data
streams included in the study and the NOAA-17, 18 and 19 HRPT AVHRR SSTfnd data for the same period
yielded a mean bias of 0.0C (standard deviation of 0.5C, N = 966) (See Fig A.2). Daytime matchups
between drifting buoy SSTfnd observations and the same satellite SSTfnd data sets over the same region and
period yielded a bias of 0.0C (standard deviation of 0.4C, N = 14146), giving confidence in the use of the
HRPT AVHRR SSTfnd observations as a relatively unbiased and accurate reference SST.
In order to determine possible reasons for the observed warm bias in SeaFlyte SSTs compared with satellite
a trip to
Ocean
foundation SSTs,
Perth was undertaken in April 2011 by the Bureau’s IMOS
Observations Scientist, Ruslan Verein, to inspect the
sensor installation on the vessel and to verify the SBE 38 SST measurements with a reference thermometer.
Figure 1. Photos taken in April 2011 of the location of the SBE 38 Digital Oceanographic Thermometer on
the PV SeaFlyte in a short pipe off the engine cooling intake. The SBE 38 is shown in the centre of the
bottom right-hand-side photo. The thermometer does not protrude into the main water flow.
In Situ Validation Procedure
Over a number of transects during two days (5 and 7 April 2011), the SeaFlyte was slowed at several
positions to allow the deployment of a sea water bucket (MK 3B) to obtain sea water samples at
approximately 0.5m depth. Once the bucket was back on deck, a mercury-in-glass thermometer (AMA, Cat.
No. 00415564, s/n 0138042) was used to measure the sea water temperature. To confirm that the reference
thermometer was valid, it was re-calibrated in June 2011 and the appropriate correction was applied to the
SST data obtained from bucket observations.
Results
Results of the sea water bucket SST comparisons with the SeaFlyte SBE 38 SST for transects on 5 April
2011 are shown in Figure 2 and Table 1, with corresponding results for transects on 7 April 2011 shown in
Figure 3 and Table 2. Conditions during the 5 April transects were calm and sunny with wind speeds of 6
ms-1 and air temperatures ranging from 26C to 31C. Rougher sea conditions were experienced during the 7
April transects where the weather was partly cloudy, with wind speeds of 11 - 12 ms-1 and air temperatures
ranging from 19C to 22C. The difference between the bucket SST measurements taken within a few
minutes of each other was up to 0.2C, and is an indication of the reference thermometer measurement error
and variability in SST over the sampling period. Figures 2 and 3 and Tables 1(a) and 2(a) clearly indicate
that the SBE 38 temperatures increased when the vessel slowed to < 14 knots to take the bucket samples.
Assuming that the SST at 0.1m – 0.5m depth does not vary significantly over the distance that the vessel
slows to collect the bucket sample then Table 1(b) indicates that there was still a warm bias between SBE 38
and bucket sample temperatures of on average 0.3C (standard deviation 0.2C) for average vessel speeds of
22 knots during warm, sunny and calm conditions on 5 April 2011. Counter-intuitively, Figure 3 and Table
2(b) indicate that under cooler, partly cloudy and rougher conditions temperatures on 7 April 2011 the SBE
38 temperatures were on average 0.4C warmer than bucket sample temperatures with a standard deviation
of 0.2C for average vessel speeds of 15 knots (slower than on 5 April due to choppy seas). Although there
may have been a diurnal thermocline present in the upper 1m of the ocean sampled during the 5 April
transects (Figure 2) which may have been responsible for the increase in warm bias between the SBE 38 SST
(0.1 – 0.5m) and the bucket sample at around 02:55 UTC, it is likely that the top 1m of the ocean would have
been well-mixed with little temperature gradient on 7 April. The average bias of 0.3 to 0.4C in the SBE 38
SSTdepth compared with bucket seawater samples observed during 5 and 7 April 2011 is far greater than the
monthly mean bias of 0.01C between the SBE 38 SSTdepth and HRPT AVHRR SSTfnd for April 2011
although comparable to the March 2011 value of 0.20ºC (Figure A.3).
Rottnest Is Ferry data (05.04.2011)
25.0
SBE38
bucket SST
24.8
interpolated SST (SBE38)
SST (° C)
24.6
24.4
24.2
24.0
23.8
2:57
2:54
2:51
2:48
2:45
2:42
2:39
2:36
1:57
1:54
1:51
1:48
1:45
1:42
1:39
1:36
1:33
1:30
1:27
1:24
1:21
1:18
1:15
1:12
1:9
23.6
Time (UTC)
Weather conditions: Calm and sunny,
1:00 UTC: Tair = +25.7, Wind speed = 12 knots (6 m/s)
4:00 UTC: Tair = +31.1, Wind speed = 12 knots (6 m/s)
Figure 2. Measurements of SST on PV SeaFlyte using the SBE 38 sensor located adjacent to the engine
intake water pipe and sea water bucket temperatures for transects on 5 April 2011. The dotted lines
represent SBE 38 temperatures interpolated between those obtained immediately prior to the vessel slowing
to take bucket samples and immediately after it sped up to cruising speed after the samples were taken.
(a) 5/04/2011
Time
01:18
01:19
01:32
01:33
01:44
01:45
02:55
Mean
Stddev
24.524
24.544
24.399
24.335
24.470
24.601
24.633
24.501
0.107
24
24
23.9
23.8
24.1
24
24
23.97
0.095
Difference, °C
0.52
0.54
0.50
0.54
0.37
0.60
0.63
0.529
0.084
Vessel speed, knots
1.9
1.8
1.2
8.5
1.9
1.5
13.0
4.3
-
01:18
01:19
01:32
01:33
01:44
01:45
02:55
Mean
Stddev
24.122
24.105
24.1
24.098
24.351
24.387
24.643
24.258
0.211
24
24
23.9
23.8
24.1
24
24
24.0
0.095
0.11
0.20
0.30
0.25
0.39
0.64
0.29
0.185
22.43
22.0
-
SST, SBE38, °C
SST, reference thermometer, °C
(b) 5/04/2011
Time
SST, SBE38 (interpolated), °C
SST, reference thermometer, °C
Difference, °C
Average vessel speed, knots
0.12
22.57
21.28
21.53
Table 1. PV SeaFlyte SBE 38 SST measurements and vessel speed for 5 April 2011 taken at (a) time of
bucket sample, and (b) interpolated from values measured immediately before and after the vessel slowed to
take the bucket samples. The difference between the SBE 38 SST and the bucket sample temperature is given
in both cases.
Rottnest Is Ferry data (07.04.2011)
25.4
SBE38
bucket SST
interpolated SST (SBE38)
25.2
25.0
24.8
SST (° C)
24.6
24.4
24.2
24.0
23.8
23.6
23.4
8:12
8:10
8:8
8:6
8:4
8:2
7:56
7:54
7:52
7:50
7:48
7:46
7:44
7:42
7:40
7:38
7:36
7:34
7:32
7:30
23.2
Time (UTC)
Weather conditions: Partly cloudy, rough sea,
7:00 UTC: Tair = +21.6, Wind speed = 23 knots (12 m/s)
10:00 UTC: Tair = +19.0, Wind speed = 22 knots (11 m/s)
Figure 3. Measurements of SST on PV SeaFlyte using the SBE 38 sensor located adjacent to the engine
intake water pipe and sea water bucket temperatures for transects on 7 April 2011. The dotted lines
represent SBE 38 temperatures interpolated between those obtained immediately prior to the vessel slowing
to take bucket samples and immediately after it sped up to cruising speed after the samples were taken.
(a) 7/04/2011
Time
SST, SBE38, °C
SST, reference thermometer, °C
07:33
07:35
07:46
07:47
08:01
08:03
08:04
mean
stdev
24.901
24.939
25.134
24.675
24.045
24.773
24.228
24.671
0.395
24.2
24
23.7
23.6
23.5
23.7
23.7
23.77
0.243
0.329
Difference, °C
0.70
0.94
1.43
1.08
0.55
1.07
0.53
0.899
Vessel speed, knots
4.5
5.2
4.6
8.3
10.6
3.1
13.9
7.2
-
07:33
07:35
07:46
07:47
08:01
08:03
08:04
mean
stdev
24.355
24.324
24.278
24.299
23.992
24.026
24.043
24.188
0.159
(b) 7/04/2011
Time
SST, SBE38 (interpolated), °C
SST, reference thermometer, °C
24.2
24
23.7
23.6
23.5
23.7
23.7
23.77
0.243
Difference, °C
0.16
0.32
0.58
0.70
0.49
0.33
0.34
0.417
0.184
14.9
-
Average vessel speed, knots
12.29
14.51
17.84
Table 2. PV SeaFlyte SBE 38 SST measurements and vessel speed for 7 April 2011 taken at (a) time of
bucket sample, and (b) interpolated from values measured immediately before and after the vessel slowed to
take the bucket samples. The difference between the SBE 38 SST and the bucket sample temperature is given
in both cases.
The results for 5 and 7 April 2011 led us to surmise that the SBE 38 temperature warm bias might be
inversely correlated with vessel speed above a 5 knot threshold. Figure 4(a) shows no correlation between
vessel speed and SBE 38 SST minus bucket SST (delta SST) during 5 April, although Figure 4(b) does
indicate a very weak inverse correlation during 7 April (R2 = 0.35).
(a)
Correleation betw een vessel speed
and delta SST (SBE38 - bucket corrected)
0.75
0.7
Delta SST (° C)
0.65
0.6
0.55
0.5
R2 = 0.2512
0.45
0.4
0.35
0.3
0
2
4
6
8
10
12
14
16
18
20
Vessel speed (knots)
(b)
Correlation betw een vessel speed
and delta SST (SBE38 - bucket corrected)
1.6
Delta SST (° C)
1.4
R2 = 0.3493
1.2
1
0.8
0.6
0.4
0.2
0
0
5
10
15
20
Vessel speed (knots)
Figure 4. Difference between PV SeaFlyte SBE 38 and bucket SST measurements (Delta SST) against vessel
speed at the time of each bucket sample for transects on (a) 5 April 2011 and (b) 7 April 2011. The
correlation coefficient between Delta SST and vessel speed (R2) is given in both cases.
On 6 April 2011 the RV Linnaeus, a small research vessel owned and operated by CSIRO out of Boat
Harbour, was deployed to perform transects between Boat Harbour and Rottnest Island, following the
route as closely as possible of the PV SeaFlyte. The RV Linnaeus is instrumented with an SBE 38 digital
thermometer in the water intake to the thermosalinograph (an SBE 45) at 0.5 m depth. A CTD was
deployed on the cruise to measure SST in the upper 1 m of the upcast in order to compare with the SBE
38 temperatures. The same sea water bucket and thermometer as deployed on the SeaFlyte transects
were deployed during 6 April on the RV Linnaeus in order to validate the Linnaeus SBE 38 SST and also
to compare with the SeaFlyte SBE 38 SST measured while the two vessels were as close as possible to
each other. The wind conditions during the 6 April Linnaeus cruise were light, with winds ranging in the
morning from 3 m/s (08:30 LT) to 7 m/s at 15:00 LT.
Figure 5 shows the SST data from the RV Linnaeus and PV SeaFlyte during the 6 April 2011 transects.
The light green lines in Figure 5 represent the SeaFlyte SBE 38 SST measurements logged while the two
vessels were collocated. The second and third RV Linnaeus CTD SST(0 – 1m) measurements (dark
green points in Figure 5) were within 0.2C of the RV Linnaeus bucket sample SST measurements (gold
lines), giving confidence that the bucket method was valid. The SeaFlyte SBE 38 SST(0.1 - 0.5m)
measurements (light green lines in Figure 5) were biased 0.2C to 0.4C warmer than the RV Linnaeus
SBE 38 SST(0.5m) measurements (red lines in Figure 5).
Figure 5. SST observations from transects of the RV Linnaeus on 6 April 2011. PV
SeaFlyte SBE 38 data (pale green lines) are collocated in time and space with Linnaeus
(where time difference was 5 minutes and distance between boats was 0.001°
lat/lon).
The RV Linnaeus SBE 38 SST measurements were generally ~ 0.1C warmer than the Linnaeus
CTD SSTs, implying that the SeaFlyte SBE 38 SSTs were biased around 0.4C too warm during 6
April. Hence, the results of comparisons between the SeaFlyte SBE 38 SST and the SST
measurements taken on the RV Linnaeus are consistent with the results from the 5 and 7 April
2011 SeaFlyte cruises.
Discussion
Sea surface temperature measurements in the upper metre of waters between Boat Harbour and
Rottnest Island in Western Australia taken on the PV SeaFlyte from engine intake SBE 38
measurements were compared with collocated SST measurements from sea water buckets on
transects on 5 and 7 April 2011 and CTD, SBE 38, SBE 45 and bucket SST measurements at
depths ranging from 1m to 0.5m from RV Linnaeus on 6 April 2011.
What we know from the study:
1. The SBE 38 SST measurements at depths ranging from 0.1m to 0.5m from the engine intake on
the PV SeaFlyte were between 0.2 to 0.4C warmer than those from other in situ SST
measurements taken at depths ranging from 1m to 0.5m.
2. The water temperature measured by the SBE 38 in the engine intake on PV SeaFlyte is
dependent on the vessel speed. Figures 2 and 3 show that the SBE 38 measurements increase as
speed decreases, and also that interpolated SBE 38 – bucket SST increase with decreasing average
vessel speed.
3. The source of this dependence is not clear. There are a number of possibilities:
a.
Engine intake depth is expected to be inversely proportional to vessel speed.
b.
Depth origin of water sampled may be a function of vessel speed as the hull displaces the
water as it moves forward.
c.
Flow rate and water temperature at the sensor in the intake pipe may be a function of
vessel speed, due to stagnation of the water sample and engine room heating.
4. Possibilities a and b require a thermal gradient in the top 1m of the water. This hypothesis is not
well supported by data collected on 5 and 7 April, as we see stronger vessel speed dependency on
the 7th when wind speed was 22kts and no gradient is expected to be present, and a weaker
dependency when stratification is more likely with light winds (12kts) on the 5th.
5. Comparisons between the SeaFlyte SSTs measured since 1 December 2008 and skin SST
observations from infrared radiometers on three satellites, converted to SST at drifting buoy
depths, indicate a widely varying warm bias, peaking during July 2009 to March 2010 and
October 2010 to January 2011 (1.2 to 2.3C) and falling to less than 0.5C during January 2009 to
May 2009, April to May 2010 and February 2011 to July 2011 (Figure A.3). There were very
high biases observed during winter months, unlikely to be linked to thermal stratification of the
water column, and some smaller (although still significant) biases observed during summer (eg.
January to March 2009).
6. The monthly varying bias could be due to:
a. mean bias in vessel speed as a function of weather conditions resulting in rougher
conditions, or
b. sporadic choking of the flow to the “dead end” pipe housing the SBE 38 sensor due to
seaweed.
Recommendations
The cause of the PV SeaFlyte SBE 38 bias is unclear. Moving the SBE 38 to a location with
consistent flow and good thermal insulation from the engine room environment appears to be the
simplest way of establishing if the hypothesis outlined in 3c above is the cause of the observed
warm bias.
References
Beggs H., R. Verein, G. Paltoglou, H. Kippo and M. Underwood (2011) Enhancing ship of opportunity
sea surface temperature observations in the Australian region, Journal of Operational Oceanography,
(under internal review).
Paltoglou, G., H. Beggs and L. Majewski (2010) New Australian High Resolution AVHRR SST Products
from the Integrated Marine Observing System, In: Extended Abstracts of the 15th Australasian Remote
Sensing and Photogrammetry Conference, Alice Springs, 13-17 September, 2010.
http://imos.org.au/srsdoc.html
SeaBird SBE 38 specification document: http://www.seabird.com/products/spec_sheets/38data.htm
SeaFlyte IMOS netCDF format data files: http://opendaptpac.arcs.org.au/thredds/dodsC/IMOS/SOOP/SOOP-SST/VHW5167_Sea-Flyte/2011/catalog.html
APPENDIX A. Satellite SST versus PV SeaFlyte SBE 38 SST Matchup Statistics
Figure A.1. Matchup statistics for IMOS HRPT AVHRR SSTfnd (from NOAA-17, 18 and 19)
minus SeaFlyte SBE 38 SSTfnd for the period 30 April 2008 to 10 September 2011.
Observations are matched if within 2 hours and within the same 1 km x 1 km satellite pixel.
See Paltoglou et al. (2010) for further details of the comparison method. The top left-hand
figure shows GAMSSA analysis SSTfnd minus SeaFlyte SSTfnd versus HRPT AVHRR
SSTfnd (from NOAA-17, 18 and 19) minus SeaFlyte SSTfnd (in K). The red stars represent
AVHRR SSTs for quality level 5 (best), yellow squares represent quality level 4 and green
diamonds quality level 3. The same symbols are used in the right-hand figure which shows the
same HRPT AVHRR SSTfnd minus SeaFlyte SSTfnd matchup values against date for 1
December 2008 to 1 June 2011. The bottom figure shows HRPT AVHRR SSTfnd minus
SeaFlyte SSTfnd plotted at the matchup location. The NOAA satellite overpass times were
approximately 10 am and 10 pm (NOAA-17) and 2 am and 2 pm (NOAA-18 and 19).
Night-time
Daytime
Number of 2 hour Matches
1.5
Drifting
Buoys
All IMOS
Ships
Pacific Sun
Iron Yandi
Stadacona
Highland
Chief
Portland
Spirit of
Tasmania II
SeaFlyte
Iron Yandi
Stadacona
Highland
Chief
Portland
Spirit of
Tasmania II
SeaFlyte
Iron Yandi
Stadacona
Highland
Chief
Portland
Spirit of
Tasmania II
SeaFlyte
Aurora
Australis
L'Astrolabe
Drifting
Buoys
All IMOS
Ships
Pacific Sun
Iron Yandi
Stadacona
Portland
Highland
Chief
0
Spirit of
Tasmania II
0
SeaFlyte
0.5
Aurora
Australis
0.5
Southern
Surveyor
1
Non-IMOS
ships
1
L'Astrolabe
Drifting
Buoys
1.5
Southern
Surveyor
Drifting
Buoys
2
Kelvin
2
Non-IMOS
ships
L'Astrolabe
Stand. Dev. ( AVHRR SST - In Situ SST)
Stand. Dev. ( AVHRR SST - In Situ SST)
Kelvin
All IMOS
Ships
-1.5
Pacific Sun
-1.5
All IMOS
Ships
-1
Pacific Sun
-1
Southern
Surveyor
-0.5
Non-IMOS
ships
Kelvin
Drifting
Buoys
All IMOS
Ships
Pacific Sun
Iron Yandi
Stadacona
Highland
Chief
Portland
Spirit of
Tasmania II
SeaFlyte
Aurora
Australis
L'Astrolabe
0
Southern
Surveyor
0
Non-IMOS
ships
Kelvin
0.5
Aurora
Australis
Mean (AVHRR SST - In Situ SST)
Mean (AVHRR SST - In Situ SST)
0.5
-0.5
Aurora
Australis
L'Astrolabe
Non-IMOS
ships
Drifting
Buoys
All IMOS
Ships
Pacific Sun
Iron Yandi
1
Highland
Chief
1
Stadacona
10
Portland
10
SeaFlyte
100
Spirit of
Tasmania II
100
Aurora
Australis
1000
L'Astrolabe
1000
Southern
Surveyor
10000
Non-IMOS
ships
10000
Southern
Surveyor
Number of 2 hour Matches
Figure A.2. Total number of night-time (left) and daytime (right) matchups (top), mean bias
(middle) and standard deviation (bottom) of 1 km resolution HRPT AVHRR L2P SSTfnd from
NOAA-17 (blue), NOAA-18 (orange) and NOAA-19 (green) satellites minus collocated
observations of SSTfnd from IMOS and non-IMOS ships or drifting buoys over the region
60ºE – 170ºW, 70ºS – 20ºN for the period 1 December 2008 to 1 June 2011. The HRPT
AVHRR L2P SST data was derived from brightness temperatures regressed against drifting
buoys for quality level > 3. Observations were considered matched if measured within  2
hours and within the same 1.1 km pixel. Matchups were rejected if the absolute difference
exceeded 5ºC and daytime 10m NWP winds were < 6 ms-1 or night-time NWP 10 m winds
were < 2 ms-1. Note that “All IMOS Ships” corresponds to all the ships shown expect PV
SeaFlyte. Refer to Paltoglou et al. (2010) for further details.
Number Matchups AVHRR SST vs SeaFlyte SST
Degrees Celsius
1000
100
10
1
Jul-11
Jun-11
May-11
Apr-11
Mar-11
Feb-11
Jan-11
Dec-10
Nov-10
Oct-10
Sep-10
Aug-10
Jul-10
Jun-10
May-10
Apr-10
Mar-10
Feb-10
Jan-10
Dec-09
Nov-09
Oct-09
Sep-09
Aug-09
Jul-09
Jun-09
May-09
Apr-09
Mar-09
Feb-09
Jan-09
Month
Mean (AVHRR SST minus Rottnest Ferry SST)
Degrees Celsius
0
-0.5
-1
-1.5
-2
-2.5
Jul-11
Jun-11
May-11
Apr-11
Mar-11
Feb-11
Jan-11
Dec-10
Nov-10
Oct-10
Sep-10
Aug-10
Jul-10
Jun-10
May-10
Apr-10
Mar-10
Feb-10
Jan-10
Dec-09
Nov-09
Oct-09
Sep-09
Aug-09
Jul-09
Jun-09
May-09
Apr-09
Mar-09
Feb-09
Jan-09
Month
Std Dev (AVHRR SST minus SeaFlyte SST)
Degrees Celsius
2.5
2
1.5
1
0.5
0
Jul-11
Jun-11
May-11
Apr-11
Mar-11
Feb-11
Jan-11
Dec-10
Nov-10
Oct-10
Sep-10
Aug-10
Jul-10
Jun-10
May-10
Apr-10
Mar-10
Feb-10
Jan-10
Dec-09
Nov-09
Oct-09
Sep-09
Aug-09
Jul-09
Jun-09
May-09
Apr-09
Mar-09
Feb-09
Jan-09
Month
Figure A.3. Monthly number of matchups (top), mean bias (middle) and standard deviation
(bottom) of 1 km HRPT AVHRR SSTdepth from NOAA-17, -18 and -19 satellites minus
SeaFlyte SBE 38 SSTdepth for 1 January 2009 to 31 July 2011. Observations are matched if
within  1 hour and within the same 1.1 km pixel. The HRPT AVHRR L2P SST data was
derived from brightness temperatures regressed against drifting buoys. Matchups were selected
where the quality level of the HRPT AVHRR SST values in the L2P files were either 3, 4 or 5.
No filtering was performed on the matchups relating to absolute temperature difference or wind
speeds.