- Email: [email protected]
Interpretation of satellite-derived sea surface temperatures
)
Pergamon
www.elsevier.com/Iocate/asr
A&'. Space Res. Voh 28, No. 1, pp. 165-170, 2001 "C,2001 COSPAR. Published by Elsevier Science Ltd. All rights reserved Printed in Great Britain 0273-1177/01 $20.00 + 0.00 PII: S0273-1177(01)00337-4
INTERPRETATION OF SATELLITE-DERIVED SEA SURFACE TEMPERATURES Ian J. Barton
CSIRO Marine Research, PO Box 1538, Hobart, Tasmania 7001, Australia
ABSTRACT Over the last twenty-five years sea surface temperatures have been available from satellite observations through the application of simple algorithms applied to infrared observations. Algorithm coefficients have been derived from simple regression analyses between surface and space-based observations. Accuracies in the order of 0.6 K have been obtained. However we are now receiving data with improved precision and thus increased care must be exercised in the derivation of algorithm coefficients and the interpretation of the derived temperatures. The most important ancillary data required are estimates of the surface wind speed and future satellites should include such capability if accurate estimates of the mixed layer sea surface temperature are to be obtained. Ship measurements showing the effect of surface wind speed on the vertical structure of near-surface water temperature are presented. © 2001 COSPAR. Published by Elsevier Science Ltd. All rights reserved.
INTRODUCTION Less than two decades ago sea surface temperature (SST) was a simple, unambiguous parameter that could be measured with a standard thermometer or an infrared radiometer. The method of measurement, and the subsequent application of the data, did not provide or require accuracies any better than about one degree Centigrade. Satellite measurements were equated to bulk temperature measurements taken by instruments on ships and buoys and the accuracy was sufficient for oceanographic studies and applications in fisheries, pollution monitoring, and weather prediction. In those early days of satellite measurements with infrared radiometers the major error in satellite-derived SSTs was due to incorrect estimation of atmospheric water vapour absorption. Poor satellite instrument calibration, noisy data sets, and the limits of 8- or 10-bit data digitisation also contributed to errors in the derived SST. However, over the last twenty years several factors have combined to overturn this simple concept of SST. Concern with the possibility of anthropogenic climate change has led to the development of numerical models of the earth's climate system that require accurate measurements of several key parameters including SST. Accurate SST is required for the initialisation and validation of climate models, and for the estimation of surface radiation and energy budgets at the air-sea interface. During 1981 plans were laid to develop a new satellite instrument designed specifically to measure SST with an accuracy that would allow the data to be used in climate research. With the development of this instrument, the Along Track Scanning Radiometer (ATSR), came a realisation that the sea surface temperature was not a simple single parameter, but that the dynamics and radiative balance at the ocean surface meant that usually the radiative temperature measured at the surface was different to the temperature just below the surface (say, 1 mm), and can be different again to the iso-thermal mixed layer
165
166
1. J. B a r t o n
that is well-known to physical oceanographers. Following the launch of the ATSR on the ERS-1 satellite in 1991, and with a more rigorous approach to the SST estimation from the AVHRR instruments on the operational NOAA satellites, the main errors associated with (the interpretation of) satellitederived SST now appear to be more related to the ocean surface rather than atmospheric effects. This change in emphasis has been driven by a better understanding of atmospheric absorption in the infrared as well as an appreciation of the complex nature of the water temperature at the ocean surface. Many times in recent years SST has been touted as one of the major indicators of climate change and variability, and efforts are under way to re-process ground-based and satellite data to gain baseline SST data sets. However, before these data sets can be used in this role, the complex nature of the surface temperature must be fully understood, and these understandings must be incorporated into the data analysis and interpretation. THE NATURE OF SST Sea surface temperature (SST) is no longer an easy parameter to define. To a physical oceanographer interested in ocean currents or the bulk thermal capacity of the ocean, a measurement of the mixed layer temperature at a depth of some metres will suffice. For a fisherman targeting surface species a measurement at a depth of less than a meter may be important, and for researchers interested in air-sea interaction a radiative SST is required. Several attempts have been made to reach some consensus on SST definitions, but all have failed. In this paper three quantities are defined which shall be used in the following discussion. Skin SST (SSST) is the radiative temperature of the ocean surface as measured by an infrared radiometer including a correction for the non-unity emissivity of the sea surface. Physically the SSST is the "black-body" temperature of the ocean surface and is indicative of the temperature some microns below the ocean surface. In reality it is the optical depth (reciprocal of the absorption coefficient of sea water) of electromagnetic penetration at the wavelength of measurement (see McAlister and McLeish, 1970, for further details). Thus the temperature may be different by some mK depending on the wavelength of measurement, but the differences are not significant for air-sea interaction studies and the interpretation of radiometric measurements - whether from a ship, buoy, or satellite. Bulk SST (BSST) is the temperature of the water just beneath the skin layer. This can be considered as the depth where evaporative and thermal radiative effects are not significant and the temperature is determined by shortwave (solar) heating, the temperature of the water below, and the flow of heat from above or below due to both conduction and mixing.
Mixed Layer SST (MLSST) is the temperature of the ocean some metres below the surface, and above the thermocline where the water temperature decreases steadily with increasing depth. Under some conditions the relation between these three SST parameters is well defined. For example, during the night with wind speeds greater than 6 ms -1 the BSST and MLSST are equal due to wind-driven mixing in the upper metres of the ocean and the lack of shortwave (solar) radiative input, and the SSST is 0.2 to 0.4 K lower than the BSST. However, as will be shown later in the body of this paper, there are other cases where the relations are quite complicated and are thus difficult to quantify. The threshold of 6 ms -1 is suggested by Donlon et al. (1999) and is adopted here, although it should be noted that this threshold is somewhat arbitrary and thus open for discussion. Some of the results presented here suggest a threshold less than 4 ms -1. SATELLITE-DERIVED SST The most accurate estimates of SST from space are obtained in cloud-free conditions with infrared radiometers. Measurements are usually taken at several different wavelengths to allow correction of atmospheric effects through differential absorption techniques. Infrared radiometers, whether ship- or satellite-borne, measure the upwelling radiance from the ocean surface and thus, once corrected for
167
Interpretation of Satellite-Derived Surface Tcmperatures
atmospheric effects, give a measure of the SSST. Microwave radiometers have the capability of measuring microwave SSST in cloudy conditions, but the errors are large due to large footprints, a need for atmospheric absorption correction, and the strong dependence of surface emissivity on wind speed and surface roughness. For many years now the AVHRR instruments on the NOAA polar-orbiting satellites have provided the most practical measurement of SST. The Multi-Channel SST (MCSST) and, more recently, the NonLinear SST (NLSST) have provided global estimates of SST with an accuracy of 0.6 K in most regions, but approaching 1.0 K in tropical climates where water vapour absorption effects are most severe. The coefficients for these SST algorithms have been derived from a regression analysis of satellite-derived values with those provided by the drifting buoy network. Therefore this method will account for any absolute calibration errors in the satellite data, and any differences between the SSST and the BSST or MLSST depending on the depth of the buoy measurement. To date no account of wind speed has been included in the derivation of the algorithm coefficients, but diurnal effects are included by the derivation of separate coefficients for day and night algorithms, The ATSR series of instruments (ATSR launched in 1991, ATSR-2 launched in 1995, and the Advanced ATSR to fly on the ENVISAT platform in 2001) has incorporated several design features to allow a more accurate estimate of the SSST. These include, accurate on-board calibration, low noise detectors, 12-bit data digitisation, and a dual view of the ocean surface allowing a more accurate determination of the atmospheric effect. The combined effect of these improvements has provided global SSST estimates with an accuracy of better than 0.3 K. The basic difference with ATSR data analysis is that the algorithm coefficients are derived using a theoretical model of atmospheric absorption allowing a direct measurement of the SSST. Validation of the ATSR SST data products has thus required the use of ship-borne infrared radiometers in specific cases, and comparisons with some BSST data sets after making some assumptions regarding bulk-skin (BSST-SSST) temperature differences.
SSST, BSST, and MLSST DIFFERENCES As mentioned above the relation between the three defined SST parameters is strongly dependent on the local conditions and time of day. The following four cases are each different and the relation between the different SST parameters is discussed. A schematic profile of temperature variation with depth is provided in Figure 1 for each of the cases discussed below. In each case the dotted lines in the figure give the MLSST, while the BSST is the temperature below the skin layer at a depth of 20 to 50 micrometers.
-6 "--" v
E
|
|
|
|
i
i
|
•
•
|
|
]
i
i
i
i
|
|
i
i
-4
c-2 v
o
J
0 2 2(
!
21
22
Typical temperature (°C) Fig. 1. Schematic temperature profiles for the four cases discussed in the text.
23
168
1. J . B a r t o n
Night With no solar heating, and adequate wind-mixing in the upper layers, this is the most predictable situation (see Figure 1, Curve A). BSST and MLSST will be equal and the SSST will be cooler than the other two due to evaporative and radiative cooling, and a heat flux flowing from the ocean to the atmosphere. Donlon et al (1999) report a temperature difference of 0.14 K, but their data plot shows a weak linear dependence on wind speed with a value of 0.20 at 6 ms -1 and 0.14 at 15 ms -l. The paucity of data above 15 ms -l limit any predictions at stronger wind speeds. A similar set of data to that collected by Donlon et a1.(1999) was obtained from the RV FRANKLIN in March 1999 during a cruise in the western Pacific Ocean. The results are shown in Figure 2 tor a seven-day period during the cruise and in Figure 3 for measurements taken under light wind conditions during two days in the latter part of the cruise.
0.5
0.5
oa
v F-O3 03 ._J i
03 03 03
''
'
'
I
0 03 co
'
'
I
'
'
'
I
'
'
'
I
'
'
'
I
'
'
x x
0.0
'
o.o _ x
×
Xx ×
x
~ . ~ m m ~ ~
~
Xx~X
d
-0.5
09 O3 03
-1.0 0
2
4 6 8 10 12 Wind speed (ms')
14
Fig. 2. Night-time skin and mixed layer temperature differences as measured on the RV FRANKLIN during 13-19 March 1999.
-0.5 ~
. x
xx
-1.0f,NI,Gk!,T.,,,,,,,,,,.,,,, 0
2
4 6 8 10 Wind speed (ms")
12
Fig. 3. Night-time skin and mixed layer temperature differences as measured on the RV FRANKLIN during 5-7 April 1999.
Day, Wind Speed > 6 ms "1 The most recent measurements with high quality ship radiometers suggest that, even when there is strong wind mixing in the top metres of the ocean, a diurnal thermocline exists near the surface due to local solar heating (see Figure 1, Curve B). The daytime data for the same seven-day period as above are shown in Figure 4. The results are more spread, but the mean value of absolute temperature differences (MLSST-SSST) at wind speeds above 6 ms -1 is less than during the night. Similar results are shown by Donlon et al. (1999) and Minnett (private communication). Day, Wind Speed < 6 ms "1 It is well known that when the wind speed is less than 2 ms -1 mixing of the upper layers of the ocean does not occur and solar energy warms the near surface layers. Under these conditions the BSST can be some degrees K higher than the MLSST some metres below the surface(see Figure 1, Curve C). Figure 5 shows the SSST-MLSST measurements along with the output of the Licor sensor which measures the relative level of insolation. The near surface heating is evident in the daytime hours. The surface wind speed for the same period is shown in Figure 6. Note how minor wind bursts at minutes 900 (5 ms 4) and 2250 (3 ms l ) disturb the surface heating. It is under these light wind conditions that there is confusion about the definition of the bulk temperature. A diurnal thermocline exists with a temperature that decreases from the BSST just below the skin layer to the MLSST below. The depth of this thermocline increases during the day unless modified by increases in wind speed as shown above. Early in the morning the isothermal mixed layer may be present at depths of less than 1 meter, but with insolation during the day the top of the isothermal mixed layer may be two or more meters below the surface. Figure 7 shows the heating of the near-surface layer for the daytime periods of these two days.
Interpretation of Satellite-Derived Surface Temperatures
169
1.0
1.0
'
O
" "" x " v
o v
Fo3 cO .._1 o9 (/3 cO
t-O9 6O _1
0.0
0.8
0.5
+5 0.6~" o
0.0
0.4 ._~
-0.5 ¢.B O~
×
o f D,A,Y,
...d
-0.5
0.2
-1.0
14
0.0 1000 2000 3000 0 Minutes after 1100GMT, April 5, 1999
Fig. 4. Daytime skin and mixed layer temperature differences as measured on the RV FRANKLIN during 13-19 March 1999.
Fig.5. Skin and mixed layer temperature differences (X) and solar insolation as measured on the RV FRANKLIN during 5-7 April 1999
-1.
,
0
8
2
t t I
,,,
I
, ,
,
I
,,
,1
, ,
,
I,,
4 6 8 10 12 Wind speed (ms")
i l l ~ l l +
ii
1
0.5
i , , l l l + l l l l l z ~ l l l l l ,
(o v
f/)
0.0
_J
o'J
i
v- -0.5
09 ¢/) 09
._ 2 0
i
I
i
I
i
i
I
I
0 1000 2000 3000 Minutes after 1100GMI-, April 5, 1999 Fig. 6. Surface wind speed as measured on the RV FRANKLIN during 5-7 April 1999.
-1.0 0
2
4 6 8 10 Wind speed (ms")
12
Fig. 7. Daytime skin and mixed layer temperature differences as measured on the RV FRANKLIN during 5-7 April 1999.
Day/Night, Rain, Wind Speed < 6 ms "1 Under some conditions (usually in the tropics) under light wind conditions rainfall can remain at the ocean surface and provide a cool layer of fresh water at the surface. In this case the BSST (and SSST) can be less than the MLSST (see Figure 1, Curve D). AVHRR ANALYSES Traditional fields of satellite-derived SST have been supplied through the long-term series of AVHRR instruments either as MCSST or as a product of the joint NASA-NOAA Pathfinder Project. Algorithm coefficients have been derived using regression analyses with coincident satellite and in-situ measurements of bulk-SST. Thus the difference between the skin (as measured by the AVHRR) and the bulk (in-situ) temperatures has been inherently included in the AVHRR SST algorithms. Also included in the analysis are situations of diurnal solar heating when the MLSST as measured from a buoy is taken as the SST. Coefficients for AVHRR SST algorithms should be derived using data collected with wind speeds greater than 6 ms -t. The effect of including these anomalous data points has been investigated by Barton (1998) and the results are consistent with model studies. The pathfinder match-up data base offers an opportunity to reprocess the AVHRR data with coefficients that exclude diurnal heating events under light wind conditions. Comparisons between night and day data should identify those cases when heating has occurred. It is now also possible to derive theoretical coefficients and these algorithms can be used to obtain satellite-derived SSST for further investigations of the "skin effect".
170
1. J. Barton
ATSR ANALYSES In contrast to the operational AVHRR analysis, data from the ATSR instruments on the ERS-1 and ERS-2 satellites have been analysed using theoretically derived SST algorithms and thus provide a direct measure of the SSST. Because of this feature precise validation of the ATSR SST product has only been possible with the use of surface- and airborne-based infrared radiometers. Inter-comparisons between ATSR, AVHRR, and surface measurements of SST should be undertaken with care. SST fields provided by the ATSR series of instruments are SSST and will include cool layers due to rain and warm layers due to diurnal heating. When computing average or composite fields of SST it is important to understand the effect of these phenomena. DISCUSSION AND CONCLUSIONS It is evident from the sections above that satellite measurements for estimation of SST must be accompanied by extra information to account for diurnal heating during the day. One technique is to use infrared instruments on both geostationary and orbiting platforms; the former to detect diurnal changes and the latter to provide the required absolute accuracy. Alternatively it may be possible to use satellite instruments to estimate the wind speed. Plans are developing for including scatterometers on operational platforms and this will ease the diurnal heating problem. The international Global Ocean Data Assimilation Experiment (GODAE) is planning to develop a new global SST product at a spatial resolution close to 10 km. Before this can be done it is important that care be taken on how the different data sets are analysed and whether it is SSST, BSST, or MLSST that is included in the analysis, as well as the nature of the final product itself. To obtain a mixed layer temperature it is recommended that only night data are used and a parameterisation for the skin-bulk temperature difference be applied. This makes the assumption that any diurnal heating of the near surface layer will be well-mixed into the top few metres and that the nighttime mixed layer is isothermal (except for the skin layer). Until a robust parameterisation of the skin effect is available the suggestion by Donlon et al. (1999) to use a BSST (or MLSST) measurement for validation of a satellite derived SSST may be the best strategy. Daytime measurements are complicated by the combined effect of the skin layer and the diurnal heating. These measurements should not be used to derive long-term bulk SST data sets but should be used with night measurements to detect any diurnal heating effects. Given that the diurnal heating is very much a transient phenomenon with critical dependencies on both insolation and wind speed, much care is required in the application of daytime measurements. Finally, the direct assimilation of infrared data from satellites into global climate models may be the best method of data analysis in the future. In this way ancillary data will be available to ensure the correct data analysis procedures are adopted. For example, a model may be able to provide an estimate of surface air temperature and humidity, insolation, wind speed, and cloudiness that will ensure that the infrared data product is accurate and reliable. Coupled atmosphere-ocean models should also be able to provide heat flux estimates and define the vertical structure of the upper layers of the ocean.
ACKNOWLEDGMENTS The work undertaken in this paper has received partial financial support from cooperation between CSIRO and Japan's NASDA under the ADEOS-II program. REFERENCES Barton, I.J., Future Directions for Satellite Studies of Skin-Bulk Temperature Differences, Journal of Advanced Marine Science and Technology Society, 4, 197-204, 1998. Donlon, C.J., T.J. Nightingale, T. Sheasby, J. Turner, I.S. Robinson, and W.J. Emery, Implications of the Oceanic Thermal Skin Temperature Deviation at High Wind Speed, Geophys. Res. Lett., 26, 25052508, 1999. McAlister, E.D. and W. McLeish, A Radiometric System for Airborne Measurement of the Total Heat Flow through the Sea, Appl. Opt., 9, 2697-2705, 1970.
Pergamon
www.elsevier.com/Iocate/asr
A&'. Space Res. Voh 28, No. 1, pp. 165-170, 2001 "C,2001 COSPAR. Published by Elsevier Science Ltd. All rights reserved Printed in Great Britain 0273-1177/01 $20.00 + 0.00 PII: S0273-1177(01)00337-4
INTERPRETATION OF SATELLITE-DERIVED SEA SURFACE TEMPERATURES Ian J. Barton
CSIRO Marine Research, PO Box 1538, Hobart, Tasmania 7001, Australia
ABSTRACT Over the last twenty-five years sea surface temperatures have been available from satellite observations through the application of simple algorithms applied to infrared observations. Algorithm coefficients have been derived from simple regression analyses between surface and space-based observations. Accuracies in the order of 0.6 K have been obtained. However we are now receiving data with improved precision and thus increased care must be exercised in the derivation of algorithm coefficients and the interpretation of the derived temperatures. The most important ancillary data required are estimates of the surface wind speed and future satellites should include such capability if accurate estimates of the mixed layer sea surface temperature are to be obtained. Ship measurements showing the effect of surface wind speed on the vertical structure of near-surface water temperature are presented. © 2001 COSPAR. Published by Elsevier Science Ltd. All rights reserved.
INTRODUCTION Less than two decades ago sea surface temperature (SST) was a simple, unambiguous parameter that could be measured with a standard thermometer or an infrared radiometer. The method of measurement, and the subsequent application of the data, did not provide or require accuracies any better than about one degree Centigrade. Satellite measurements were equated to bulk temperature measurements taken by instruments on ships and buoys and the accuracy was sufficient for oceanographic studies and applications in fisheries, pollution monitoring, and weather prediction. In those early days of satellite measurements with infrared radiometers the major error in satellite-derived SSTs was due to incorrect estimation of atmospheric water vapour absorption. Poor satellite instrument calibration, noisy data sets, and the limits of 8- or 10-bit data digitisation also contributed to errors in the derived SST. However, over the last twenty years several factors have combined to overturn this simple concept of SST. Concern with the possibility of anthropogenic climate change has led to the development of numerical models of the earth's climate system that require accurate measurements of several key parameters including SST. Accurate SST is required for the initialisation and validation of climate models, and for the estimation of surface radiation and energy budgets at the air-sea interface. During 1981 plans were laid to develop a new satellite instrument designed specifically to measure SST with an accuracy that would allow the data to be used in climate research. With the development of this instrument, the Along Track Scanning Radiometer (ATSR), came a realisation that the sea surface temperature was not a simple single parameter, but that the dynamics and radiative balance at the ocean surface meant that usually the radiative temperature measured at the surface was different to the temperature just below the surface (say, 1 mm), and can be different again to the iso-thermal mixed layer
165
166
1. J. B a r t o n
that is well-known to physical oceanographers. Following the launch of the ATSR on the ERS-1 satellite in 1991, and with a more rigorous approach to the SST estimation from the AVHRR instruments on the operational NOAA satellites, the main errors associated with (the interpretation of) satellitederived SST now appear to be more related to the ocean surface rather than atmospheric effects. This change in emphasis has been driven by a better understanding of atmospheric absorption in the infrared as well as an appreciation of the complex nature of the water temperature at the ocean surface. Many times in recent years SST has been touted as one of the major indicators of climate change and variability, and efforts are under way to re-process ground-based and satellite data to gain baseline SST data sets. However, before these data sets can be used in this role, the complex nature of the surface temperature must be fully understood, and these understandings must be incorporated into the data analysis and interpretation. THE NATURE OF SST Sea surface temperature (SST) is no longer an easy parameter to define. To a physical oceanographer interested in ocean currents or the bulk thermal capacity of the ocean, a measurement of the mixed layer temperature at a depth of some metres will suffice. For a fisherman targeting surface species a measurement at a depth of less than a meter may be important, and for researchers interested in air-sea interaction a radiative SST is required. Several attempts have been made to reach some consensus on SST definitions, but all have failed. In this paper three quantities are defined which shall be used in the following discussion. Skin SST (SSST) is the radiative temperature of the ocean surface as measured by an infrared radiometer including a correction for the non-unity emissivity of the sea surface. Physically the SSST is the "black-body" temperature of the ocean surface and is indicative of the temperature some microns below the ocean surface. In reality it is the optical depth (reciprocal of the absorption coefficient of sea water) of electromagnetic penetration at the wavelength of measurement (see McAlister and McLeish, 1970, for further details). Thus the temperature may be different by some mK depending on the wavelength of measurement, but the differences are not significant for air-sea interaction studies and the interpretation of radiometric measurements - whether from a ship, buoy, or satellite. Bulk SST (BSST) is the temperature of the water just beneath the skin layer. This can be considered as the depth where evaporative and thermal radiative effects are not significant and the temperature is determined by shortwave (solar) heating, the temperature of the water below, and the flow of heat from above or below due to both conduction and mixing.
Mixed Layer SST (MLSST) is the temperature of the ocean some metres below the surface, and above the thermocline where the water temperature decreases steadily with increasing depth. Under some conditions the relation between these three SST parameters is well defined. For example, during the night with wind speeds greater than 6 ms -1 the BSST and MLSST are equal due to wind-driven mixing in the upper metres of the ocean and the lack of shortwave (solar) radiative input, and the SSST is 0.2 to 0.4 K lower than the BSST. However, as will be shown later in the body of this paper, there are other cases where the relations are quite complicated and are thus difficult to quantify. The threshold of 6 ms -1 is suggested by Donlon et al. (1999) and is adopted here, although it should be noted that this threshold is somewhat arbitrary and thus open for discussion. Some of the results presented here suggest a threshold less than 4 ms -1. SATELLITE-DERIVED SST The most accurate estimates of SST from space are obtained in cloud-free conditions with infrared radiometers. Measurements are usually taken at several different wavelengths to allow correction of atmospheric effects through differential absorption techniques. Infrared radiometers, whether ship- or satellite-borne, measure the upwelling radiance from the ocean surface and thus, once corrected for
167
Interpretation of Satellite-Derived Surface Tcmperatures
atmospheric effects, give a measure of the SSST. Microwave radiometers have the capability of measuring microwave SSST in cloudy conditions, but the errors are large due to large footprints, a need for atmospheric absorption correction, and the strong dependence of surface emissivity on wind speed and surface roughness. For many years now the AVHRR instruments on the NOAA polar-orbiting satellites have provided the most practical measurement of SST. The Multi-Channel SST (MCSST) and, more recently, the NonLinear SST (NLSST) have provided global estimates of SST with an accuracy of 0.6 K in most regions, but approaching 1.0 K in tropical climates where water vapour absorption effects are most severe. The coefficients for these SST algorithms have been derived from a regression analysis of satellite-derived values with those provided by the drifting buoy network. Therefore this method will account for any absolute calibration errors in the satellite data, and any differences between the SSST and the BSST or MLSST depending on the depth of the buoy measurement. To date no account of wind speed has been included in the derivation of the algorithm coefficients, but diurnal effects are included by the derivation of separate coefficients for day and night algorithms, The ATSR series of instruments (ATSR launched in 1991, ATSR-2 launched in 1995, and the Advanced ATSR to fly on the ENVISAT platform in 2001) has incorporated several design features to allow a more accurate estimate of the SSST. These include, accurate on-board calibration, low noise detectors, 12-bit data digitisation, and a dual view of the ocean surface allowing a more accurate determination of the atmospheric effect. The combined effect of these improvements has provided global SSST estimates with an accuracy of better than 0.3 K. The basic difference with ATSR data analysis is that the algorithm coefficients are derived using a theoretical model of atmospheric absorption allowing a direct measurement of the SSST. Validation of the ATSR SST data products has thus required the use of ship-borne infrared radiometers in specific cases, and comparisons with some BSST data sets after making some assumptions regarding bulk-skin (BSST-SSST) temperature differences.
SSST, BSST, and MLSST DIFFERENCES As mentioned above the relation between the three defined SST parameters is strongly dependent on the local conditions and time of day. The following four cases are each different and the relation between the different SST parameters is discussed. A schematic profile of temperature variation with depth is provided in Figure 1 for each of the cases discussed below. In each case the dotted lines in the figure give the MLSST, while the BSST is the temperature below the skin layer at a depth of 20 to 50 micrometers.
-6 "--" v
E
|
|
|
|
i
i
|
•
•
|
|
]
i
i
i
i
|
|
i
i
-4
c-2 v
o
J
0 2 2(
!
21
22
Typical temperature (°C) Fig. 1. Schematic temperature profiles for the four cases discussed in the text.
23
168
1. J . B a r t o n
Night With no solar heating, and adequate wind-mixing in the upper layers, this is the most predictable situation (see Figure 1, Curve A). BSST and MLSST will be equal and the SSST will be cooler than the other two due to evaporative and radiative cooling, and a heat flux flowing from the ocean to the atmosphere. Donlon et al (1999) report a temperature difference of 0.14 K, but their data plot shows a weak linear dependence on wind speed with a value of 0.20 at 6 ms -1 and 0.14 at 15 ms -l. The paucity of data above 15 ms -l limit any predictions at stronger wind speeds. A similar set of data to that collected by Donlon et a1.(1999) was obtained from the RV FRANKLIN in March 1999 during a cruise in the western Pacific Ocean. The results are shown in Figure 2 tor a seven-day period during the cruise and in Figure 3 for measurements taken under light wind conditions during two days in the latter part of the cruise.
0.5
0.5
oa
v F-O3 03 ._J i
03 03 03
''
'
'
I
0 03 co
'
'
I
'
'
'
I
'
'
'
I
'
'
'
I
'
'
x x
0.0
'
o.o _ x
×
Xx ×
x
~ . ~ m m ~ ~
~
Xx~X
d
-0.5
09 O3 03
-1.0 0
2
4 6 8 10 12 Wind speed (ms')
14
Fig. 2. Night-time skin and mixed layer temperature differences as measured on the RV FRANKLIN during 13-19 March 1999.
-0.5 ~
. x
xx
-1.0f,NI,Gk!,T.,,,,,,,,,,.,,,, 0
2
4 6 8 10 Wind speed (ms")
12
Fig. 3. Night-time skin and mixed layer temperature differences as measured on the RV FRANKLIN during 5-7 April 1999.
Day, Wind Speed > 6 ms "1 The most recent measurements with high quality ship radiometers suggest that, even when there is strong wind mixing in the top metres of the ocean, a diurnal thermocline exists near the surface due to local solar heating (see Figure 1, Curve B). The daytime data for the same seven-day period as above are shown in Figure 4. The results are more spread, but the mean value of absolute temperature differences (MLSST-SSST) at wind speeds above 6 ms -1 is less than during the night. Similar results are shown by Donlon et al. (1999) and Minnett (private communication). Day, Wind Speed < 6 ms "1 It is well known that when the wind speed is less than 2 ms -1 mixing of the upper layers of the ocean does not occur and solar energy warms the near surface layers. Under these conditions the BSST can be some degrees K higher than the MLSST some metres below the surface(see Figure 1, Curve C). Figure 5 shows the SSST-MLSST measurements along with the output of the Licor sensor which measures the relative level of insolation. The near surface heating is evident in the daytime hours. The surface wind speed for the same period is shown in Figure 6. Note how minor wind bursts at minutes 900 (5 ms 4) and 2250 (3 ms l ) disturb the surface heating. It is under these light wind conditions that there is confusion about the definition of the bulk temperature. A diurnal thermocline exists with a temperature that decreases from the BSST just below the skin layer to the MLSST below. The depth of this thermocline increases during the day unless modified by increases in wind speed as shown above. Early in the morning the isothermal mixed layer may be present at depths of less than 1 meter, but with insolation during the day the top of the isothermal mixed layer may be two or more meters below the surface. Figure 7 shows the heating of the near-surface layer for the daytime periods of these two days.
Interpretation of Satellite-Derived Surface Temperatures
169
1.0
1.0
'
O
" "" x " v
o v
Fo3 cO .._1 o9 (/3 cO
t-O9 6O _1
0.0
0.8
0.5
+5 0.6~" o
0.0
0.4 ._~
-0.5 ¢.B O~
×
o f D,A,Y,
...d
-0.5
0.2
-1.0
14
0.0 1000 2000 3000 0 Minutes after 1100GMT, April 5, 1999
Fig. 4. Daytime skin and mixed layer temperature differences as measured on the RV FRANKLIN during 13-19 March 1999.
Fig.5. Skin and mixed layer temperature differences (X) and solar insolation as measured on the RV FRANKLIN during 5-7 April 1999
-1.
,
0
8
2
t t I
,,,
I
, ,
,
I
,,
,1
, ,
,
I,,
4 6 8 10 12 Wind speed (ms")
i l l ~ l l +
ii
1
0.5
i , , l l l + l l l l l z ~ l l l l l ,
(o v
f/)
0.0
_J
o'J
i
v- -0.5
09 ¢/) 09
._ 2 0
i
I
i
I
i
i
I
I
0 1000 2000 3000 Minutes after 1100GMI-, April 5, 1999 Fig. 6. Surface wind speed as measured on the RV FRANKLIN during 5-7 April 1999.
-1.0 0
2
4 6 8 10 Wind speed (ms")
12
Fig. 7. Daytime skin and mixed layer temperature differences as measured on the RV FRANKLIN during 5-7 April 1999.
Day/Night, Rain, Wind Speed < 6 ms "1 Under some conditions (usually in the tropics) under light wind conditions rainfall can remain at the ocean surface and provide a cool layer of fresh water at the surface. In this case the BSST (and SSST) can be less than the MLSST (see Figure 1, Curve D). AVHRR ANALYSES Traditional fields of satellite-derived SST have been supplied through the long-term series of AVHRR instruments either as MCSST or as a product of the joint NASA-NOAA Pathfinder Project. Algorithm coefficients have been derived using regression analyses with coincident satellite and in-situ measurements of bulk-SST. Thus the difference between the skin (as measured by the AVHRR) and the bulk (in-situ) temperatures has been inherently included in the AVHRR SST algorithms. Also included in the analysis are situations of diurnal solar heating when the MLSST as measured from a buoy is taken as the SST. Coefficients for AVHRR SST algorithms should be derived using data collected with wind speeds greater than 6 ms -t. The effect of including these anomalous data points has been investigated by Barton (1998) and the results are consistent with model studies. The pathfinder match-up data base offers an opportunity to reprocess the AVHRR data with coefficients that exclude diurnal heating events under light wind conditions. Comparisons between night and day data should identify those cases when heating has occurred. It is now also possible to derive theoretical coefficients and these algorithms can be used to obtain satellite-derived SSST for further investigations of the "skin effect".
170
1. J. Barton
ATSR ANALYSES In contrast to the operational AVHRR analysis, data from the ATSR instruments on the ERS-1 and ERS-2 satellites have been analysed using theoretically derived SST algorithms and thus provide a direct measure of the SSST. Because of this feature precise validation of the ATSR SST product has only been possible with the use of surface- and airborne-based infrared radiometers. Inter-comparisons between ATSR, AVHRR, and surface measurements of SST should be undertaken with care. SST fields provided by the ATSR series of instruments are SSST and will include cool layers due to rain and warm layers due to diurnal heating. When computing average or composite fields of SST it is important to understand the effect of these phenomena. DISCUSSION AND CONCLUSIONS It is evident from the sections above that satellite measurements for estimation of SST must be accompanied by extra information to account for diurnal heating during the day. One technique is to use infrared instruments on both geostationary and orbiting platforms; the former to detect diurnal changes and the latter to provide the required absolute accuracy. Alternatively it may be possible to use satellite instruments to estimate the wind speed. Plans are developing for including scatterometers on operational platforms and this will ease the diurnal heating problem. The international Global Ocean Data Assimilation Experiment (GODAE) is planning to develop a new global SST product at a spatial resolution close to 10 km. Before this can be done it is important that care be taken on how the different data sets are analysed and whether it is SSST, BSST, or MLSST that is included in the analysis, as well as the nature of the final product itself. To obtain a mixed layer temperature it is recommended that only night data are used and a parameterisation for the skin-bulk temperature difference be applied. This makes the assumption that any diurnal heating of the near surface layer will be well-mixed into the top few metres and that the nighttime mixed layer is isothermal (except for the skin layer). Until a robust parameterisation of the skin effect is available the suggestion by Donlon et al. (1999) to use a BSST (or MLSST) measurement for validation of a satellite derived SSST may be the best strategy. Daytime measurements are complicated by the combined effect of the skin layer and the diurnal heating. These measurements should not be used to derive long-term bulk SST data sets but should be used with night measurements to detect any diurnal heating effects. Given that the diurnal heating is very much a transient phenomenon with critical dependencies on both insolation and wind speed, much care is required in the application of daytime measurements. Finally, the direct assimilation of infrared data from satellites into global climate models may be the best method of data analysis in the future. In this way ancillary data will be available to ensure the correct data analysis procedures are adopted. For example, a model may be able to provide an estimate of surface air temperature and humidity, insolation, wind speed, and cloudiness that will ensure that the infrared data product is accurate and reliable. Coupled atmosphere-ocean models should also be able to provide heat flux estimates and define the vertical structure of the upper layers of the ocean.
ACKNOWLEDGMENTS The work undertaken in this paper has received partial financial support from cooperation between CSIRO and Japan's NASDA under the ADEOS-II program. REFERENCES Barton, I.J., Future Directions for Satellite Studies of Skin-Bulk Temperature Differences, Journal of Advanced Marine Science and Technology Society, 4, 197-204, 1998. Donlon, C.J., T.J. Nightingale, T. Sheasby, J. Turner, I.S. Robinson, and W.J. Emery, Implications of the Oceanic Thermal Skin Temperature Deviation at High Wind Speed, Geophys. Res. Lett., 26, 25052508, 1999. McAlister, E.D. and W. McLeish, A Radiometric System for Airborne Measurement of the Total Heat Flow through the Sea, Appl. Opt., 9, 2697-2705, 1970.