Integrated Ocean Skin and Bulk Temperature Measurements Using the Calibrated Infrared In Situ Measurement System (CIRIMS) and Through-Hull Ports

A. T. Jessup Applied Physics Laboratory, University of Washington, Seattle, Washington

Search for other papers by A. T. Jessup in
Current site
Google Scholar
PubMed
Close
and
R. Branch Applied Physics Laboratory, University of Washington, Seattle, Washington

Search for other papers by R. Branch in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The design and performance of a shipboard-integrated system for underway skin and bulk temperature is presented. The system consists of the Calibrated Infrared In situ Measurement System (CIRIMS) and through-hull temperature sensors. The CIRIMS is an autonomous shipboard radiometer system that measures the sea surface skin temperature Tskin for validation of satellite-derived sea surface temperature products. General design considerations for shipboard radiometer systems are discussed and the philosophy behind the CIRIMS design is presented. Unique features of the design include a constant temperature housing to stabilize instrument drift, a two-point dynamic calibration procedure, separate sky- and sea-viewing radiometers for simultaneous measurements, and the ability to use an infrared transparent window for environmental protection. Laboratory testing and field deployments are used to establish an estimated error budget, which includes instrumentation and environmental uncertainties. The combination of this testing and field comparison to the Marine-Atmosphere Emitted Radiance Interferometer (M-AERI) and Infrared SST Autonomous Radiometer (ISAR) instruments indicates that the CIRIMS meets the design goal of ±0.10°C accuracy. Temperature and pressure sensors were installed in custom-designed through-hull ports on the NOAA research vessel (R/V) Ronald H. Brown and the University of Washington R/V Thomas G. Thompson to complement the CIRIMS measurements. The ports allow sensors to be installed while the ship is in water and can accommodate a variety of sensors. The combined system provides the ability to measure near-surface temperature profiles from the skin to a depth of 5 m while underway.

Corresponding author address: Andrew T. Jessup, Applied Physics Laboratory, University of Washington, 1013 NE 40th St., Seattle, WA 98105-6698. Email: jessup@apl.washington.edu

Abstract

The design and performance of a shipboard-integrated system for underway skin and bulk temperature is presented. The system consists of the Calibrated Infrared In situ Measurement System (CIRIMS) and through-hull temperature sensors. The CIRIMS is an autonomous shipboard radiometer system that measures the sea surface skin temperature Tskin for validation of satellite-derived sea surface temperature products. General design considerations for shipboard radiometer systems are discussed and the philosophy behind the CIRIMS design is presented. Unique features of the design include a constant temperature housing to stabilize instrument drift, a two-point dynamic calibration procedure, separate sky- and sea-viewing radiometers for simultaneous measurements, and the ability to use an infrared transparent window for environmental protection. Laboratory testing and field deployments are used to establish an estimated error budget, which includes instrumentation and environmental uncertainties. The combination of this testing and field comparison to the Marine-Atmosphere Emitted Radiance Interferometer (M-AERI) and Infrared SST Autonomous Radiometer (ISAR) instruments indicates that the CIRIMS meets the design goal of ±0.10°C accuracy. Temperature and pressure sensors were installed in custom-designed through-hull ports on the NOAA research vessel (R/V) Ronald H. Brown and the University of Washington R/V Thomas G. Thompson to complement the CIRIMS measurements. The ports allow sensors to be installed while the ship is in water and can accommodate a variety of sensors. The combined system provides the ability to measure near-surface temperature profiles from the skin to a depth of 5 m while underway.

Corresponding author address: Andrew T. Jessup, Applied Physics Laboratory, University of Washington, 1013 NE 40th St., Seattle, WA 98105-6698. Email: jessup@apl.washington.edu

1. Introduction

Validation of sea surface temperature (SST) measurements derived from satellite-based infrared sensors is complicated because the representative depth of the satellite measurements is significantly different from the depth of in situ measurements used for validation. Infrared (IR) techniques have an effective measurement depth of less than 10 μm and provide the surface “skin” temperature Tskin. The most common in situ SST validation measurements use the bulk temperature Tbulk collected from drifting and moored buoys using traditional contact sensors at depths between about 0.5 and 1.5 m. The difference in measurement depth is problematic because the ocean skin is typically cooler (0.1°–0.3°C on average) than the water a millimeter or more below the surface (e.g., Katsaros 1980; Robinson et al. 1984). The situation is further complicated during the day when diurnal warming can lead to the formation of a warm layer with temperature differences of several degrees across the first few meters under low winds (Donlon et al. 2002; Ward et al. 2004).

There is a growing international consensus that SST products derived from satellite-based IR measurements should include Tskin because it is the most relevant to air–sea fluxes and it corresponds directly to the measured radiance. A skin SST product is now a required product of the National Polar-orbiting Operational Environmental Satellite System (NPOESS). To demonstrate the useful accuracy of the satellite-based measurements of Tskin, widespread in situ radiometric measurements will be necessary for validation, especially at wind speeds below 6 m s−1 (Donlon et al. 1999b).

Over the past decade, commercially available IR thermometers have been combined with in situ calibration for validation of a satellite-derived Tskin product (Barton et al. 2004). Optical measurements in the shipboard environment are extremely challenging and the accuracy goal of ±0.1°C set by the community (Kannenberg 1998) is very ambitious. Nonetheless, progress by the community has demonstrated the feasibility of developing an autonomous instrument that could be deployed on ships of opportunity to provide the coverage necessary for global validation of a satellite-derived Tskin product.

Here we report on one such instrument, the Calibrated Infrared In situ Measurement System (CIRIMS), of which three units have been fabricated and deployed a total of 10 times from 1999 to 2006. Through-hull instrument ports have been installed on the National Oceanic and Atmospheric Administration (NOAA) research vessel (R/V) Ronald H. Brown and University of Washington R/V Thomas G. Thompson to complement the CIRIMS Tskin measurements. Data have been collected over a wide range of latitudes and environmental conditions, accumulating a total distance traveled of over 300 000 km (Fig. 1). These data are freely available via a Web page (http://cirims.apl.washington.edu/) and are currently being used for satellite sensor measurement validation. In this paper we outline the CIRIMS design philosophy, report on the field and laboratory testing to establish an error budget, and demonstrate the ability to measure temperature profiles underway with the combination of the CIRIMS and through-hull measurements.

2. Design considerations for shipboard IR radiometric measurement systems

Radiometric determination of Tskin is based on inversion of the equation for the sea surface radiance L(T) measured with an IR radiometer. If the distance to the surface is small enough to neglect atmospheric effects, the radiance measured by a radiometer operating in the wavelength range λ1λλ2 and viewing the sea surface at an incidence angle θ is the sum of the emitted and reflected radiation, given by
i1520-0426-25-4-579-e1
where Lλ,b(λ, T) is the spectral radiance at temperature T given by Planck’s function; R(λ) is the instrument responsivity; and ɛ(λ, θ) and ρ(λ, θ) are the spectral emissivity and reflectivity, respectively. The sky temperature Tsky is an equivalent temperature for the integrated downwelling radiance from the atmosphere at zenith angle θ. Under clear-sky conditions, the sky reflection effect routinely can be as much as 0.5°C and possibly larger in polar regions. Equation (1) can be inverted to solve for Tskin using measurements of both the sea and sky radiance and requires knowledge of the emissivity. In addition to being a function of wavelength and incidence angle, ɛ(λ, θ) becomes a function of surface roughness as θ increases (Hanafin and Minnett 2005; Masuda et al. 1988; Shaw and Marston 2000; Watts et al. 1996). Once an effective emissivity is calculated and L(T) and L(Tsky) are measured, then L(Tskin) is given by
i1520-0426-25-4-579-e2
where ρsea = 1 − ɛsea; Tskin is then calculated by calibrating L(Tskin) using a precision blackbody. During the calibration, a similar reflection correction is applied for the nonunity emissivity of the blackbody.

Important factors to be considered in the design of an autonomous shipboard IR radiometer system and common approaches to addressing them are listed in Table 1. The desire for affordability and reliability has led to combining commercially available radiometers with supplemental calibration techniques in a protective housing. Since the accuracy of affordable radiometers is typically 0.5°C, reliable, long-term accuracy requires in situ calibration.

a. Calibration

A common approach to in situ calibration is to use a two-point method consisting of hot and ambient temperature targets. This approach simplifies the engineering complexity since only one target must be temperature controlled, which is done by heating. The temperature of the hot target is often significantly higher than the Tskin to be measured. The ambient target temperature may also be significantly different than the Tskin if the air is much cooler than the water. This approach therefore relies on both the stability of the instrument response over a long cruise and the characterization of that response using curve fitting. The voltage output of an IR detector is a nonlinear function of scene temperature and is often dependent on the instrument temperature. The calibration curves for the model KT11 radiometer (Heitronics, Germany) used in CIRIMS shown in Fig. 2 illustrate both the nonlinear response and the dependence on instrument temperature. The variation of the response with temperature recommends keeping the instrument at constant temperature to stabilize the calibration. The curves for different instrument temperatures show both an offset and a change in slope, especially at low target temperatures. This behavior suggests that if the difference between the ambient and hot target temperatures is large, the achievable accuracy of a linear interpolation may be affected.

b. Sky reflection correction and sea surface emissivity

The final accuracy of the measured Tskin depends on both the calibration of the instrument and the uncertainty in the sky reflection correction, which depends strongly upon the estimate of emissivity. A narrow bandwidth in the 9–12-μm range minimizes the impact of atmospheric water vapor on the sky correction. Errors associated with the sky correction can be due to either spatial variability in the sky radiance or to the dependence of the emissivity on θ and surface roughness. Under partly cloudy conditions, the downwelling radiance measured from a ship will likely be different than that which is reflected from the spot on the sea surface where the upwelling radiance is measured. The difference is due to the unavoidable separation between the radiometer and the measurement spot and can be exacerbated if the up- and downwelling radiance measurements are sampled at different times.

The emissivity of seawater in the 9–12-μm range is dependent on wavelength and incidence angle; at nadir incidence it varies with λ from roughly 0.98 to 0.99 and becomes a function of surface roughness for θ greater than about 45°. Donlon and Nightingale (2000) discussed errors in skin temperature measurements due to radiometer-viewing geometry (pointing), temporal sampling mismatch between the sea and sky views, and the effect of wind on the sea surface. They found that errors of ±0.25°C can be caused by improper pointing and time mismatch and recommend that pointing errors be less than ±5° and that sea and sky observations be close in time. For incidence (sea viewing) and zenith (sky viewing) angles of 40°, they estimated the worst-case error due to partly cloudy conditions to be 0.2°C for a wind speed of 10 m s−1. Since these errors increase significantly with increasing incidence and zenith angle, they recommend that viewing angles be limited to between 15° and 40°. The geometry necessary to view an undisturbed surface depends primarily on the ship’s wake and the distance from the bow of a suitable mounting location. The superstructure on ferries and cruise ships often provides mounting locations near the bow, so that an undisturbed region can be viewed with a θ = 25° (Robinson et al. 2006). However, on many research vessels, the superstructure and wake characteristics require 40° < θ < 50°.

The dependence of ɛ on wavelength and incidence angle is shown in Fig. 3 for the range of λ and θ used in CIRIMS as tabulated by Shaw and Marston (2000) for zero wind speed. The KT11 responsivity is also shown in Fig. 3 and emphasizes that a secondary advantage of using this bandwidth is that the spectral dependence of the ɛ is relatively weak. The effective emissivity ɛeff(θ) for a radiometer with responsivity R(λ) operating in the wavelength range λ1λλ2 is given by
i1520-0426-25-4-579-e3
The ɛeff(θ) for the three different incidence angles used in the CIRIMS installations described below are listed in Table 2.

c. Radiometer protection

Protection of the radiometer and calibration blackbody is arguably the most challenging aspect of a practical design. A major design goal in the development of CIRIMS was to evaluate the use of an IR transparent window to provide complete protection of the optics and the blackbody during deployments when they are susceptible to spray. The motivation behind the use of a window is to ensure complete protection under all conditions because of the possibility of severe weather and sea conditions during a long deployment. This approach relies on the ability to correct for the effect of the window. The primary concerns regarding the use of a window are wetting, the effect of salt deposits on the transmission, and that the self-emission of the window is a function of ambient temperature.

3. CIRIMS system overview

The CIRIMS was designed as a prototype with the dual purpose of evaluating engineering approaches (such as the IR transparent window) and simultaneously providing operational measurements. It was built to operate autonomously on ships of opportunity for at least six months, withstand harsh weather conditions, and obtain an accuracy goal of ±0.1°C.

Important design factors and common approaches used in IR SST measurement systems are listed in Table 1. The CIRIMS has a unique design to address these issues. Instrument drift is stabilized with an insulated constant temperature housing. A two-point calibration is achieved with dynamic set points of a precision blackbody. Sea and sky measurements are made using separate radiometers. The incidence angle is always kept as close to nadir as possible, which in practice is between 40° and 50°. A radiometer bandwidth of 9.6–11.5 μm is used. An IR transparent window is used when there is risk of contamination by sea spray.

The block diagram in Fig. 4 shows the major system components composing the electronics chassis and sensor housing. The housing is mounted along the side of a ship to view the undisturbed sea surface at an incidence angle between 40° and 50°. The rack-mountable electronics chassis is designed to be inside the ship and can be separated from the sensor housing by up to 30 m. The sensor housing weighs 35 kg and the electronics chassis weighs 23 kg. CIRIMS was designed to be deployed on ocean research vessels and has been deployed on ships ranging in length from 14 to 122 m. The system runs on 120-Vac power and consumes less than 2 kW.

The measurement duty cycle is a 60-min sequence that includes two calibrations separated by 30 min to accommodate the need for the blackbody to change temperature and stabilize. This algorithm is shown in Table 3 and can be changed to have all phases except 5 and 10 viewing the sea without the window when the window is not in use.

a. Sensor housing

The sensor housing (1.0 m × 0.4 m × 0.5 m) contains the sea-viewing radiometer and blackbody, as shown in the photograph and schematic drawing in Fig. 5. The sea-viewing radiometer is mounted on a rotating arm and alternately views the sea surface and the blackbody. Attached to the front of the sensor housing is an insulated external housing that contains the window mechanism. A double-walled tube attached to the bottom face of the view box extends the optical path to approximately 1 m, which provides protection from sea spray when operating without the window. The housing is sealed when the window is in use, and a reservoir of desiccant is used to maintain low humidity to avoid condensation on the blackbody. A GPS module provides time and location and an Iridium satellite-telephone modem is used for sending data to land on a daily basis.

The housing is insulated and maintained at a constant temperature within a range of approximately ±0.10°C about the set temperature (standard deviation 0.10°C) by means of an integrated thermoelectric heater/cooler unit (TEC) and circulation fan. This provides a stable, dry environment for the internal radiometer and the blackbody. The housing temperature is reset on a daily basis to 5°C above the highest air temperature of the previous day. This algorithm is based on minimizing errors in the correction for nonunity emissivity of the blackbody, the use of two different sky- and sea-viewing radiometers, and the window correction.

The sky-viewing radiometer is contained in an unheated external housing attached to the side of the sensor housing as shown in Fig. 5. The depth of the sky radiometer housing and the zenith angle were designed so that rain and spray would not reach the lens. The distance from the radiometer lens to the opening is 1 m and the zenith angle of the housing is equal to the incidence angle of the sea-viewing radiometer (between 40° and 50°). While we cannot guarantee that the lens will not get wet, the condition of the lens on the sky-viewing radiometer has been monitored during attended cruises when rain was present and found to be uncontaminated. The sky-viewing radiometer also has a specially modified lens holder that prevents water accumulation on the lens in the unlikely event that water does reach the lens.

b. Radiometers

The sea and sky measurements are made using two model KT11 (Heitronics, Germany) IR radiometers with a spectral bandwidth in the 9.6–11.5-μm range. The manufacturer’s specifications listed in Table 4 include an accuracy of ±1.0°C and temperature drift of 0.03% °C−1, which is consistent with the laboratory calibration curves in Fig. 2. Notable features include nonlinear output (proportional to radiance), small size, and modest cost. The sky-viewing radiometer is mounted in an external housing attached to the sensor housing and is operated without in situ calibration. Although using separate sky- and sea-viewing radiometers introduces some calibration uncertainty (see below), the separation makes it possible to operate the sea-viewing radiometer with and without the IR transparent window with minimal risk of contamination. This provided the ability to evaluate the use of the window (see below).

The rationale for using separate radiometers for the sea and sky measurements is based on the relative magnitude of the sea and sky contributions to L(T) in (1). In general, the sea contribution is nearly two orders of magnitude greater than the sky contribution because the emissivity ɛ(λ, θ) is close to unity and the reflectivity ρ(λ, θ) is equal to 1 − ɛ according to Kirchoff’s law. Since the sky contribution term is so much smaller than the sea contribution term, the accuracy requirement for the sky measurement is less stringent than that for the sea. The error from using two different radiometers was assessed by assuming a maximum calibration difference between the two radiometers and varying both the sea and sky radiance. Laboratory testing has shown that the maximum offset between two radiometers is approximately 10 mV, which corresponds to roughly 2.5°C (see Fig. 2).

The resulting error in Tskin is illustrated in Fig. 6 for an uncertainty in the sky measurement of 2.5°C, which is more than twice the manufacturer’s accuracy specification. In Fig. 6, the skin temperature error is plotted versus Tskin for five values of Tsky ranging from −60° to 30°C. The error was computed as the difference between Tskin calculated with five values of Tsky in (1) and Tskin calculated using those same five values plus 2.5°C. For simplicity, the calculation was done for a constant value of ɛ = 0.985 and a uniform instrument response from 9.6 to 11.5 μm. The error decreases with increasing Tskin and is greatest for cloudy conditions. For reference, typical clear-sky values for Tsky range from −50°C in high latitudes to −20°C in tropical latitudes. The results indicate that the error introduced by foregoing in situ calibration of the sky measurement for clear skies is less than 0.03°C. The error also was estimated using actual data by recomputing Tskin after adding the maximum offset voltage measured in the laboratory (10 mV) to the recorded sky radiometer signal. For a typical 24-h record that included a mixture of sky conditions, the RMS difference in Tskin was 0.015°C, or about half of what was estimated by the calculation illustrated by Fig. 6. We conclude that the value of 0.03°C based on Fig. 6 is a conservative estimate of the maximum calibration error due to the combination of an uncalibrated sky-viewing radiometer and a precisely calibrated sea-viewing radiometer.

c. Calibration blackbody

As noted in section 2a, a common approach for providing a two-point calibration is to use a constant temperature hot target and an ambient (cold) temperature target. Because of the requirement of long deployments and associated uncertainty in the stability of the radiometer transfer function, the CIRIMS design uses a different scheme referred to as a dynamic two-point calibration. In this approach, the hot and cold calibration temperatures follow the temperature to be measured, bracketing it above and below by 2°C. This dynamic interval calibration technique makes linear interpolation possible and ensures consistent accuracy over a wide range of scene temperatures, albeit at additional complexity and cost. The technique is implemented by using a single blackbody calibration target immersed in a precision temperature-controlled bath. The design is similar to the National Institute of Standards and Technology (NIST)-traceable water-bath blackbody described by Fowler (1995) but uses the model 7102 Microbath (Hart Scientific, American Fork, Utah). This portable water bath uses TEC units for both heating and cooling. Although the Microbath has a built-in temperature sensor, a supplemental high-accuracy thermistor model SBE-38 (Seabird, Bellevue, Washington) is used to ensure the desired accuracy. The specifications of the Microbath are given in Table 5.

The blackbody is a custom-designed cylindrical cone made of oxygen-free copper and is shown in Fig. 7 in cut-away view installed in the well of Microbath. The cone is completely immersed and is thermally isolated from the bath by a PVC mounting ring. The ring incorporates the SBE-38 thermistor, which is immersed in the bath fluid directly adjacent to the cone. Factors considered in determining the dimensions of the cone included the radiometer field of view, the effective emissivity based on the ratio of the lengths of the cylinder and the cone (Bedford 1988), and the need to provide adequate stirring space for the Microbath’s magnetic stirrer used to keep the bath temperature uniform (Hart Scientific, B. Cherry 1999, personal communication). The calculated emissivity of the blackbody for a surface emissivity of 0.90 at the intersection with the radiometer field of view is 0.9974, which is approximately 0.13% less than the emissivity at the cone vertex. This dependence of emissivity on spot radius is consistent with the criteria for calibration blackbodies outlined by Donlon et al. (1999a). The surface of the blackbody was painted with black paint (Testor’s #1147), which has an emissivity of 0.95 and is the same paint used in the Fowler (1995) blackbody.

The effective emissivity of the CIRIMS blackbody (CIRIMS-BB) was determined using standard radiometric methods (DeWitt and Richmond 1988) by comparison to a precision reference blackbody using the configuration shown in Fig. 8. The reference (NIST-BB) is a duplicate of the Fowler (1995) blackbody, which has an emissivity of 0.9997 with uncertainty of 0.0003. The KT11 was mounted on a rotating mechanism enclosed in a temperature controlled housing with temperature Tbox = 35°C. The NIST-BB and CIRIMS-BB were set to the same temperatures and stepped from 5° to 35°C. Using this procedure, the emissivity of the CIRIMS-BB was determined to be 0.9986 with an RMS variation of 0.0007. The measured value is slightly larger than the calculated value of 0.9974 given above, which is reasonable since the paint used had a higher emissivity than that in the calculation.

The calibration error, which is the difference between the calibrated temperature from the KT11 and the NIST-BB temperature Tref also was determined using the setup shown in Fig. 8. The dynamic interval calibration method described above was implemented for Tref ranging from 5° to 35°C. The results shown in Fig. 9 demonstrate both the magnitude of the effect of the nonunity emissivity of the CIRIMS-BB as well the effectiveness of the correction. Figure 9 includes measurements made with constant Tbox = 35°C as well as Tbox = Tref. For Tbox = 35°C, the error with and without the emissivity correction is plotted. For Tbox = Tref, only the error without the correction is plotted since in this case the correction for the nonunity emissivity was negligible (as expected). For the case of Tbox = 35°C with the emissivity correction, the mean ± standard deviation and RMS of the calibration error were equal to −0.016° ± 0.010° and 0.018°C, respectively.

d. Window correction algorithm

In the development of the CIRIMS, we put a high priority on the ability to continuously monitor the effect of the window in order to account for changes due to contamination or environmental conditions in an ongoing fashion. The CIRIMS window mechanism design, shown schematically in Fig. 10, allows the effect of the window to be determined by measuring the radiance of a simple flat-plate blackbody that is external to the window. First, the CIRIMS is put in a protected mode by closing a door between the optical path and the outside air (see Fig. 5, left). A two-point, heated, flat-plate blackbody (ɛ = 0.95) is on the back of the door, which protects it from sea spray. The flat-plate blackbody cycles between 40° and 50°C. It takes 3 min to reach the set temperature and then is viewed by the radiometer. This design provides a method to correct for the effect of the window by making measurements of a two-point temperature target while the optics and primary calibration blackbody inside the main housing remain protected.

The window effects that need to be determined are the attenuation of the signal, the self-emission of the window, and the reflection from the window. The attenuation is due to the transmissivity of the window being less than unity, and the self-emission is due the emissivity being greater than zero. When the radiance of the external flat-plate blackbody is measured with and without the window, the only difference is the combined effect of the attenuation, self-emission, and reflection from the window. These effects can be separated and quantified because the attenuation dominates but is not a function of temperature, while the self-emission is small and a function of temperature. The effect of reflection is also small but constant since the enclosure is kept at a constant temperature. As outlined below, the slope of a linear regression of the radiance of the flat-plate blackbody measured without the window versus the radiance measured with the window is dominated by the transmissivity. The residual of the regression then is a combination of the self-emission and the reflection. Since the self-emission depends on the temperature of the window but the reflection does not, a regression of the residual against window temperature can be used to separate these two effects. The technique requires that the flat-plate blackbody be stable during the measurements but does not require that the temperature be accurately known. The reason is that the technique is based on the change in the radiance between two temperature settings with and without the window rather than the absolute calibration of the flat-plate blackbody.

When the radiometer views the external flat-plate blackbody without the window in place, the radiance is
i1520-0426-25-4-579-e4
where Lno_window is the radiance without the window in place; and Lbb, ɛbb, and ρbb are the radiance, emissivity, and reflectivity of the flat-plate blackbody; and Lamb is the radiance of the surrounding (ambient) temperature. When the window is in place, the measured radiance is the product of Lno_window and τw, the window transmission coefficient, plus the emission from the window and the reflection by the window of the radiometer housing radiance. Therefore, the radiance from the flat-plate blackbody measured through the window is
i1520-0426-25-4-579-e5
where w denotes the window and the subscript “box stands for the radiometer housing. The CIRIMS window is made of ZnSe with τ = 0.874, ɛ = 0.025, and ρ = 0.101 (clean window, measured in laboratory). The first term on the right-hand side of (5) is the effect of attenuation due to the window. The second term, the window self-emission, is small and varies slowly with the window temperature Twindow, which is measured with an attached thermistor. The third term is also small but constant because the box temperature is fixed.
The effects of attenuation, emission, and reflection are removed using two sequential linear regressions with the result that the radiance corrected for the window is given by
i1520-0426-25-4-579-e6
where a0 and a1 are, respectively, the offset and slope of the regression of Lno_window versus Lwindow; and b0 and b1 are, respectively, the offset and slope of the regression of the residual from the first regression versus Twindow. The slope a1 is roughly equal to τw and the residual is dominated by the second term on the right-hand side in (5) and as such is proportional to Twindow. The slope b1 corresponds to the second term on the right-hand side in (5). The combined offsets from the two regressions correspond to the effect of the third term on the right-hand side in (5). The mean of the residual from the second regression should be negligible and the standard deviation is a measure of the accuracy of the window correction.

The result of these steps are shown in Fig. 11 for data acquired during the Fluxes, Air–Sea Interaction, and Remote Sensing (FAIRS) Experiment, which occurred in 2000 aboard the research platform (R/P) FLIP (Floating Instrument Platform). The regressions are done in terms of the radiometer output voltage since it is proportional to radiance, where the voltages Vwindow and Vno_window correspond to Lwindow and Lno_window, respectively. The slope of the regression of Vwindow versus Vno_window in Fig. 11a is 0.874, which is equal to the window transmission coefficient measured in the laboratory. In the second step, shown in Fig. 11b, a line is fit to the residual of the first regression (first residual) plotted versus the window temperature. The histogram of the residual of this second regression, called the second residual, is shown in Fig. 11c in degrees Celsius. The mean of the second residual corresponds to the constant effect of the window reflection. The standard deviation was 0.04°C and is a measure of the error due to the window.

This error only applies to correction of the window effect when measuring the external flat-plate blackbody, not the sea surface. To evaluate the error when viewing the sea, we made measurements of the sea surface with and without the window during periods when the weather conditions permitted the window to be removed without jeopardizing the optics and internal blackbody. The difference between Tskin without the window and Tskin through the window and corrected using this method during FAIRS showed a mean ± standard deviation of −0.06° ± 0.07°C and an RMS of 0.09°C. The increased standard deviation over the results for the flat-plate blackbody above is likely due to the fact that the measurements with and without the window were made 30 min apart. The increase in the mean difference may be due to inadequately characterizing the window emission because there is evidence that the external flat-plate blackbody causes some additional heating of the window that is not detected by the window thermistor. For purposes of estimating an error budget, the RMS difference of 0.09°C from the FAIRS sea surface results is taken as an overall measure of the error introduced by the window.

4. Measurement uncertainty

The overall measurement uncertainty can be divided into errors due to instrumentation and to environmental factors. Table 6 lists the primary source of both types of errors and an estimate of their magnitude. The primary sources of instrumentation error are the two different radiometers (section 3b), the calibration uncertainty (section 3c), and the IR transparent window correction if it is used (section 3d). Sources of environmental error are changes in emissivity due to surface roughness and variation in local incidence angle produced by ship motion, and uncertainty in the sky correction due to variable sky conditions. Estimates of these individual errors are listed in Table 6a and their cumulative effect for different combinations of sky conditions and window use are summarized in Table 6b. The calibration uncertainty is taken as the RMS error from the laboratory testing described in section 3c and incorporates uncertainty due to the internal blackbody temperature measurement, the stability of the Microbath, the blackbody emissivity, and the radiometer stability. The uncertainties due to using two different radiometers and the use of the window are from the results of sections 3b and 3d, respectively.

The uncertainties due to both the variation in local incidence angle and variable sky conditions were based on analysis of recorded data. For the purposes of the error budget, we consider a change in the effective incidence angle of ±5° relative to the nominal incidence angle of 45°. For the 24-h record with variable sky conditions shown in Fig. 12, Tskin(θ) was computed using the flat surface emissivity for θ = 40°, 45°, and 50°. The middle time series in Fig. 12 shows that changes in the effective emissivity have the greatest effect under clear-sky conditions. The mean difference between Tskin(50) and Tskin(45) was 0.065°C, and the mean difference between Tskin(45) and Tskin(40) was 0.040°C. Since the CIRIMS deployments have used incidence angles ranging from 40° to 50°, we use the average of these two values of 0.053°C as a measure of the error due to changes in effective incidence angle. The effect of sky variability was estimated by computing the standard deviation of the sky radiance σsky for a typical 24-h record that included a range of sky conditions. This value was used as a measure of sky radiance variability and was added to the measured Tsky and then Tskin was recomputed, shown in the bottom time series in Fig. 12. As in the case of changes in the effective emissivity, the sky variability has the most effect under clear-sky conditions. The RMS difference between Tskin and Tskin computed with σsky added to the sky radiance was 0.030°C and is taken as a measure of the uncertainty due to variable sky conditions.

The overall uncertainties are computed as the square root of the sum of the individual uncertainties squared and are summarized in Table 6b. For uniform (clear or cloudy) sky conditions the overall errors are 0.064° and 0.110°C without and with the window, respectively. The overall errors increase for variable sky conditions to 0.081° and 0.121°C without and with the window, respectively. The overall error estimates indicate that the CIRIMS should meet the design goal of ±0.10°C accuracy.

The laboratory comparison to the NIST blackbody (Fig. 9) only quantifies the calibration of the sea-viewing radiometer, not the accuracy of the sky-corrected skin temperature measurement. The complexity of at-sea radiometry is such that comparison with other instruments in the field is perhaps the most practical and convincing method to judge performance. The results of a three-way comparison between the CIRIMS and two other IR systems, which were the Marine-Atmosphere Emitted Radiance Interferometer (M-AERI; Minnett et al. 2001) and the Infrared SST Autonomous Radiometer (ISAR; Donlon et al. 2008), are shown in Table 7. The three instruments were deployed together on the NOAA R/V Ronald H. Brown on a cruise from Miami, Florida, to Santiago, Chile, via the Panama Canal in 2004. The CIRIMS measurements in the comparison were made without the use of the window. The agreement between the ISAR and CIRIMS is remarkably consistent and favorable in comparison to each other and the M-AERI. The mean difference between the ISAR and CIRIMS is effectively zero (less than 0.01°C), and the respective standard deviations between those instruments and the M-AERI are virtually identical at approximately 0.25°C. Both the CIRIMS and the ISAR measured roughly 0.08°C below the M-AERI in the mean. This three-way comparison demonstrates that the mean difference in Tskin between CIRIMS and two other independent instruments is within ±0.10°C. One possible reason that the ISAR and CIRIMS compare better to each other than to the M-AERI is that they are both radiometer-based systems that use a similar approach for the sky reflection correction. The M-AERI is a spectrometer-based system that uses a fundamentally different sky reflection correction method (Minnett et al. 2001). Branch et al. (2008) present a detailed comparison of CIRIMS and M-AERI measurements on several other cruises, including the performance with and without the window. The results of those comparisons are consistent with the three-way comparison summarized above.

5. Through-hull temperature and pressure sensors

Two through-hull instrument ports were installed on the sister ships Brown and the Thompson at depths of 2 and 3 m to provide temperature at depths intermediate between the skin and the ship’s standard thermosalinograph depth of 5 m. The ports are located on the starboard side directly above the ship’s existing 5-m intake, as shown in Fig. 13. Each port incorporates a ball valve that allows it to be closed off when the sensor is removed. The sensor housing has O-ring seals that permit it to be installed and removed from inside the ship while it is in the water. The ports were designed and fabricated at the Applied Physics Laboratory and installed when the ships were in dry dock for routine maintenance. The ports were instrumented with two model SBE-39 temperature sensors with integrated pressure transducers (Seabird, Bellevue, Washington). The pressure sensors provide the measurement depth, which changes with ship motion and loading.

a. Vertical temperature profile results

The bulk-skin temperature difference, ΔT = TskinTbulk, is a combination of the effects of diurnal warming and the cool skin effect in the top 10 μm of the water surface. It will depend on the depth of the bulk measurement when there is a near-surface temperature gradient. Figure 14 shows a 3-day time series of Tskin and Tbulk measured at the two through-hull ports and the ship’s 5-m thermosalinograph. The degree of stratification clearly varies with the amount of solar radiation and the wind speed. The greatest stratification occurs on the first day, when the solar radiation is the strongest and the wind speed is below 5 m s−1. On the second day, the stratification begins to develop but is reduced when the wind speed increases and the solar radiation is reduced due to cloudy conditions. A clear separation between the 2- and 3-m measurements appears on the third day, which has somewhat reduced solar radiation and very low wind speed. Except for a few brief periods during strong solar heating, Tskin is always cooler than the shallowest Tbulk. At night the stratification subsides but a cool skin layer remains. An analysis of the degree of stratification under varying conditions is the subject of a forthcoming paper.

Measurements of Tbulk at different depths were made simultaneously with the CIRIMS to examine ΔT as a function of measurement depth over a wide range of conditions. In addition to the through-hull sensors, near-surface temperature was measured by C. Fairall and J. Hare (NOAA/Environmental Technology Laboratory) using the so-called sea snake during several cruises on the Brown in 2001. The sea snake is a thermistor attached to the end of an umbilical towed along the surface that measures temperature at approximately 5-cm depth. The scatterplots in Fig. 15 summarizing all the measurements from the 2001 East Pacific Investigation of Climate (EPIC) cruise show ΔT versus U10 for bulk temperatures at 5, 2, and 0.05 m. At low winds, the 5-m reference erroneously indicates a warm skin, while the 2- and 0.05-m data show a cool skin. The Tskin temperature was always cooler relative to the sea snake temperature and was warmer than the 2-m temperature for only a few low wind cases. This finding itself is not new, but these results demonstrate that an underway measurement at 2 m significantly reduces the occurrence of false warm-skin conditions. Note also that there is considerable scatter at low to moderate winds, even for the shallow-reference data. An examination of the time series during periods shortly after rain suggests that a significant amount of the scatter is due to cold freshwater at the surface. The larger negative ΔT values at high wind speeds were all acquired on one day and could be due to the wind speed dependence of emissivity, as discussed in Branch et al. (2008).

b. Effects of ship motion on bulk temperature measurements

The measurements in Fig. 14 clearly showing near-surface stratification were taken with the ship traveling at approximately 10 kt (5 m s−1). Examination of periods conducive to near-surface stratification when the ship was alternately stopped and steaming indicate that the ship must be underway to provide reliable Tbulk measurements. A clear illustration of the corruption of Tbulk measurements when the ship is stopped is illustrated by Fig. 16, which is a time series from the Brown during which the ship was alternately stopped and steaming at roughly 12 kt (6 m s−1). The time series were taken during a period of high solar insolation and low winds. The time series in Fig. 16a show a 24-h period of wind speed, solar radiation, ship speed, and water temperatures (Tskin, T2m, T3m, and T5m) while Fig. 16b shows the water temperatures during the period from 1000 to 1600 UTC, when the ship abruptly started and stopped. The wind speed and solar radiation in Fig. 16 are comparable to those on the first day in Fig. 15, but the degree of stratification when the ship is stopped in Fig. 16 is significantly less than in Fig. 15. The water temperature time series in Fig. 16 show a well-mixed near-surface layer with a cool skin layer present from 0000 to shortly before 1200 UTC. The stratification was relatively weak while the ship was stationary as the solar radiation increased from roughly 0600 to 1200 UTC. As the ship rapidly got underway shortly after 1200 UTC, all three bulk temperatures were abruptly affected. The most dramatic effect is the decrease of about 0.3°C in T5m, which is from the ship’s intake at the bottom of the hull. The variability in the T5m signal decreases when the ship begins to move, but the variability in both the through-hull signals increases. Furthermore, the difference between T2m and T3m is greater after the ship is underway, showing a degree of stratification that is reasonable based on comparison with the underway example in Fig. 15. When the ship stops at shortly after 1500 UTC, the stratification subsides just as abruptly as it had appeared when the ship got underway.

The behavior of the Tbulk measurements as the ship abruptly starts and stops implies that all bulk measurements, including the ship’s thermosalinograph at a depth of 5 m, are likely to be corrupted when the ship is stopped. As noted above, the character of variability in T5m is different than that of T2m and T3m before and after the ship abruptly starts. When the ship is stationary, waves and currents may cause flow around the hull that could mix near-surface water down below the ship, where T5m is measured. This explanation is consistent with the observation that T5m appears too warm and has high variability prior to the ship getting underway. The increased variability in T2m and T3m while underway is likely due to vertical excursion of the sensors caused by the ship heave. The heave would be expected to affect the through-hull sensors more than the sensor in the ship intake, since the latter is at a depth of 5 m where the temperature gradients are generally less.

If the increased variability in the through-hull temperatures while underway are due to heave, then the vertical excursions should be able to provide higher-resolution profiles of the near-surface temperature structure. The ability to use the vertical motion of a ship underway to profile the near-surface temperature has been demonstrated for bow-mounted sensors (Soloviev and Lukas 1997). The image plot in Fig. 17a illustrates the extent to which the heave can be used to subsample the temperature profile. The image shows temperature using a color scale as a function of time and depth. In this example, the heave amplitude is such that the water column is sampled regularly from roughly 1.5 to 3.5 m. The sampling frequency for the through-hull temperature and pressure data was 0.67 Hz. The image plot was constructed by averaging the data into 0.1-m depth bins and 10-s time bins. Figure 17b shows individual points (black squares) from the temperature profile extracted from the image at the time of the dashed vertical line in Fig. 17a as well the mean values of Tskin, T2m, T3m, and T5m (red). The higher-resolution profile is consistent with the averaged profile and provides additional information about the vertical structure.

The cause of these motion effects cannot be determined definitively without additional information about the true stratification or flow around the hull. Despite any remaining uncertainty in the exact cause of the ship motion effects, the examples included here are significant because, to our knowledge, they are the first published evidence showing the degree to which ship-based bulk temperature can be corrupted while on station.

6. Conclusions

An autonomous shipboard infrared radiometer system for validation of skin SST has been designed, fabricated, and successfully deployed on a routine basis. The CIRIMS was combined with through-hull sensors to provide an integrated system for measurement of near-surface temperature from the skin to a depth of 5 m. Unique design features of the CIRIMS include a constant temperature housing to stabilize instrument drift, a two-point dynamic calibration procedure, separate up- and sea-viewing radiometers for simultaneous sea and sky measurements, and the ability to use an infrared transparent window for environmental protection. The design philosophy was presented and the uncertainties were estimated for both instrumentation and environmental errors. Instrumentations errors include the calibration procedure, the use of two separate radiometers, and the optional use of the IR transparent window. Environmental errors include the effects of the dependence of emissivity on incidence angle and surface roughness and the effects of variable sky conditions. For uniform (clear or cloudy) sky conditions the overall errors are 0.064° and 0.110°C without and with the window, respectively. The overall errors increase for variable sky conditions to 0.081° and 0.121°C without and with the window, respectively. These results, combined with a three-way field comparison with the ISAR and M-AERI, indicates that the CIRIMS meets the design goal of ±0.10°C accuracy.

Through-the-hull temperature sensors were installed on two ships to complement the CIRIMS measurements. The sensors’ temperature and depth capabilities provide the ability to measure near-surface temperature profiles as the ships were underway. Measurements while the ship was stopped and underway indicate that bulk temperature sensors (including the ship’s thermosalinograph at a depth of 5 m) can be corrupted while on station.

The success of the CIRIMS instrument design has demonstrated the feasibility of making reliable, autonomous Tskin measurements from ships. The knowledge gained during the development of the CIRIMS should be applied to a new, simpler design for autonomous and routine deployment on research and merchant ships. Future and more numerous instruments will continue to improve the validation efforts of satellite-based SST measurements and the study of near-surface thermal boundary layer processes. The temperature profiles measured with the through-the-hull sensors demonstrated the usefulness of combining sensors at multiple depths with surface skin measurements for continuous underway measurements.

Acknowledgments

We gratefully acknowledge the contributions of the Ocean Engineering Department at the APL-UW, which included R. Light, F. Olson, P. Sabin, D. Stearns, M. Welch, and T. Wen. We thank T. Litchendorf (APL-UW) for her able assistance in testing and deployments and G. Wick [NOAA/Earth System Research Laboratory (NOAA/ESRL)] for field assistance and analysis. We thank C. Fairall and J. Hare (NOAA/ESRL) for providing the sea snake data and P. Minnett (University of Miami), C. Donlon (Met Office), and R. M. Reynolds (Brookhaven National Laboratory) for sharing their data for the three-way instrument comparison. Finally, we thank the captains and crews of the NOAA R/V Ronald H. Brown, the University of Washington R/V Thomas G. Thompson, the USCG Polar Sea, and the R/P FLIP. The development and/or deployment of the CIRIMS were funded by NASA, NOPP, and the Applied Physics Laboratory.

REFERENCES

  • Barton, I. J., Minnett P. J. , Maillet K. A. , Donlon C. J. , Hook S. J. , Jessup A. T. , and Nightingale T. J. , 2004: The Miami2001 infrared radiometer calibration and intercomparison. Part II: Shipboard results. J. Atmos. Oceanic Technol., 21 , 268283.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bedford, R. E., 1988: Calculation of effective emissivities of cavity sources of thermal radiation. Theory and Practice of Radiation Thermometry, D. P. DeWitt and G. D. Nutter, Eds., John Wiley and Sons, 653–772.

    • Search Google Scholar
    • Export Citation
  • Branch, R., Jessup A. T. , and Minnett P. J. , 2008: Comparisons of shipboard infrared sea surface skin temperature measurements from the CIRIMS and the M-AERI. J. Atmos. Oceanic Technol., 25 , 598606.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • DeWitt, D., and Richmond J. , 1988: Thermal radiative properties of materials. Theory and Practice of Radiation Thermometry, D. DeWitt and G. Nutter, Eds., Wiley-Interscience, 91–188.

    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., and Nightingale T. J. , 2000: Effect of atmospheric radiance errors in radiometric sea-surface skin temperature measurements. Appl. Opt., 39 , 23872392.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Nightingale T. J. , Fiedler L. , Fisher G. J. , Baldwin D. , and Robinson I. S. , 1999a: The calibration and intercalibration of sea-going infrared radiometer systems using a low cost blackbody cavity. J. Atmos. Oceanic Technol., 16 , 11831197.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Nightingale T. J. , Sheasby T. , Turner J. , and Robinson I. S. , 1999b: Implications of the oceanic thermal skin temperature deviation at high wind speed. Geophys. Res. Lett., 26 , 25052508.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Minnett P. J. , Gentemann C. , Nightingale T. J. , Barton I. J. , Ward B. , and Murray M. J. , 2002: Toward improved validation of satellite sea surface skin temperature measurements for climate research. J. Climate, 15 , 353369.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Robinson I. S. , Wimmer W. , Fisher G. , Edwards R. , and Nightingale T. J. , 2008: An Infrared Sea Surface Temperature Autonomous Radiometer (ISAR) for deployment aboard Volunteer Observing Ships (VOS). J. Atmos. Oceanic Technol., 25 , 93113.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fowler, J. B., 1995: A third generation water bath based blackbody source. J. Res. Natl. Inst. Stand. Technol., 100 , 591599.

  • Hanafin, J. A., and Minnett P. J. , 2005: Measurements of the infrared emissivity of a wind-roughened sea surface. Appl. Opt., 44 , 398411.

  • Kannenberg, R., 1998: IR instrument comparison workshop at the Rosenstiel School of Marine and Atmospheric Science (RSMAS). Earth Obs., 10 , 5154.

    • Search Google Scholar
    • Export Citation
  • Katsaros, K. B., 1980: Aqueous thermal boundary layer. Bound.-Layer Meteor., 18 , 107127.

  • Masuda, K., Takashima T. , and Takayama Y. , 1988: Emissivity of pure and sea waters for the model sea surface in the infrared window regions. Remote Sens. Environ., 24 , 313329.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minnett, P. J., Knuteson R. O. , Best F. A. , Osborne B. J. , Hanafin J. A. , and Brown O. B. , 2001: The Marine-Atmospheric Emitted Radiance Interferometer: A high-accuracy, sea-going infrared spectroradiometer. J. Atmos. Oceanic Technol., 18 , 9941013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Robinson, I. S., Wells N. C. , and Charnock H. , 1984: The sea surface thermal boundary layer and its relevance to the measurement of sea surface temperature by airborne and spaceborn radiometers. Int. J. Remote Sens., 5 , 1946.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Robinson, I. S., Wimmer W. , and Donlon C. J. , 2006: Validation of AATSR sea surface temperature products using the shipborne ISAR Radiometer. Final Rep. ISAR-DEFRA-FREP01, 39 pp.

  • Shaw, J. A., and Marston C. , 2000: Polarized infrared emissivity for a rough water surface. Opt. Express, 7 , 375380.

  • Soloviev, A., and Lukas R. , 1997: Observation of large diurnal warming events in the near-surface layer of the western equatorial Pacific warm pool. Deep-Sea Res. I, 44 , 10551076.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ward, B., Wanninkof R. , Minnett P. J. , and Head M. J. , 2004: SkinDeEP: A profiling instrument for upper-decameter sea surface measurements. J. Atmos. Oceanic Technol., 21 , 207222.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Watts, P. D., Allen M. R. , and Nightingale T. J. , 1996: Wind speed effects on sea surface emission and reflection for the Along Track Scanning Radiometer. J. Atmos. Oceanic Technol., 13 , 126141.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Combined cruise tracks, color-coded for Tskin, for R/V Brown, R/V Thompson, and U.S. Coast Guard (USCG) Polar Sea from 2001 through 2005. The total distance covered is over 300 000 km.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 2.
Fig. 2.

Calibration curves for KT11 radiometer for housing (case) temperature of 10°, 20°, and 35°C. Here, TBB is the temperature of the reference blackbody and Vout is the voltage output of the radiometer. The mean difference between the output for Tcase of 10° and 35°C was 0.09°C.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 3.
Fig. 3.

Sea surface emissivity as a function of wavelength and the incidence angles used in CIRIMS deployments. Also shown is the response function for the KT11 radiometer used in CIRIMS. The variation of the emissivity is relatively low over the wavelength range of the radiometer.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 4.
Fig. 4.

Block diagram of CIRIMS components in (left) electronics chassis and (right) sensor housing.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 5.
Fig. 5.

CIRIMS sensor housing. (a) Photograph showing deployment along ship railing. (b) Cut-away engineering drawing showing key internal components.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 6.
Fig. 6.

Error due to using a calibrated sea-viewing radiometer and a separate uncalibrated sky-viewing radiometer as a function of Tskin and Tsky assuming an uncertainty of 2.5°C in the sky measurement. The error is greatest for cloudy conditions and cold skin temperatures. For clear-sky conditions, the maximum error is roughly 0.03°C.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 7.
Fig. 7.

Engineering drawing of the cylindrocone blackbody used in the CIRIMS. (Dimensions are in mm)

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 8.
Fig. 8.

Laboratory calibration setup for calibrating KT11 radiometer against laboratory blackbody and to determine emissivity of cylindrocone blackbody used in CIRIMS. The radiometer is enclosed in a temperature-stabilized chamber and alternately rotates to view the reference blackbody on the right and the CIRIMS blackbody on the bottom left.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 9.
Fig. 9.

CIRIMS radiometer calibration error as a function of reference blackbody temperature for different instrument housing (box) temperatures measured using the setup in Fig. 8. The results are shown for a housing temperature fixed at 35°C and set to the temperature of the reference blackbody. Open symbols are the error without correcting for the nonunity emissivity of the CIRIMS blackbody. The error is reduced when the correction is applied (solid symbols).

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 10.
Fig. 10.

Schematic illustration of correction for effect of IR transparent window. (top) The radiance of the flat-plate blackbody is measured at two different temperatures with the window not in place. (bottom) The measurement is repeated with the window in place.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 11.
Fig. 11.

Steps in window correction using the external flat-plate blackbody. (a) Regression of Vwindow vs Vno_window removes attenuation; (b) regression of first residual vs Twindow removes emission (slope) and reflection (offset). The std dev of the second residual is equal to 0.04°C and is a measure error of the window correction.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 12.
Fig. 12.

Time series of (top) Tsky and (middle) Tskin showing variation due to changing the effective incidence angle by ±5° about the measurement angle of 40° and (bottom) adding sky variance σsky to estimate effect of sky variability. The mean difference between Tskin(50) and Tskin(45) was 0.065°C and the mean difference between Tskin(45) and Tskin(40) was 0.040°C. The RMS difference between Tskin and Tskin computed with σsky added to the sky radiance was 0.030°C and is taken as a measure of the uncertainty due to variable sky conditions.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 13.
Fig. 13.

Through-the-hull sensor ports at 2 and 3 m below the water line located in the bow thruster room on the R/V Brown. Identical ports have been installed on the R/V Thompson.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 14.
Fig. 14.

Example of measurements of diurnal warm layer measured from the R/V Thompson over a 3-day period in 2004. The time series are (top) wind speed, (middle) solar radiation, and (bottom) water temperatures: skin temperature, Tskin; bulk temperatures from through-the-hull ports at 2 and 3 m, T2m and T3m, respectively; and bulk temperature from ship’s thermosalinograph, T5m. The degree of stratification depends on both the solar radiation and wind speed.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 15.
Fig. 15.

Scatterplots of all data collected during the EPIC 2001 cruise showing the bulk-skin temperature difference (ΔT = TskinTbulk) vs wind speed U10 computed using bulk temperatures from (left) 5 m, (middle) 2 m, and (right) the sea snake. The erroneously large positive values at low wind speeds using Tbulk at 5 m are due to diurnal warming and emphasize the value of the underway bulk measurements at 2 m.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 16.
Fig. 16.

Time series of wind speed, solar radiation, ship speed, and water temperatures showing the effect of ship speed on bulk temperatures. The ship abruptly goes from being on station to an underway speed of roughly 12 kt shortly after 1200 UTC. Stratification prior to being underway is unreasonably weak based on increasing solar radiation and low wind, while after getting underway the stratification is characteristic of the environmental conditions. Prior to getting underway, the temperature at 5 m shows an unusual variability, which may be due to flow around the ship’s hull. After getting underway, the temperatures at 2 and 3 m show increased variability, which is likely due to the ship’s heave.

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Fig. 17.
Fig. 17.

(left) Temperature map based on vertical excursion of the through-hull sensors (depth bin size: 0.1 m, time bin size: 10 s). (right) Comparison of profiles based on temperature map and averages of Tskin (5 min), T2m, T3m, and T5m (1 min).

Citation: Journal of Atmospheric and Oceanic Technology 25, 4; 10.1175/2007JTECHO479.1

Table 1.

Important design factors and common approaches for shipboard IR systems.

Table 1.
Table 2.

Effective emissivity for CIRIMS installations using different incidence angles.

Table 2.
Table 3.

Measurement algorithm.

Table 3.
Table 4.

Heitronics model KT11 radiometer manufacturer’s specifications.

Table 4.
Table 5.

Specifications for the calibration water bath, Hart Scientific model 7102, and the supplemental SeaBird SBE-38 temperature probe.

Table 5.
Table 6.

Measurement uncertainties for CIRIMS. (a) Instrument and environmental factors and (b) overall errors as a function of sky conditions and use of window.

Table 6.
Table 7.

Comparison of differences between skin temperature measured concurrently by the CIRIMS (TCIRIMS), the M-AERI (TMAERI), and the ISAR (TISAR) during a joint cruise. The comparison used 2577 points acquired over 23 days.

Table 7.
Save
  • Barton, I. J., Minnett P. J. , Maillet K. A. , Donlon C. J. , Hook S. J. , Jessup A. T. , and Nightingale T. J. , 2004: The Miami2001 infrared radiometer calibration and intercomparison. Part II: Shipboard results. J. Atmos. Oceanic Technol., 21 , 268283.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bedford, R. E., 1988: Calculation of effective emissivities of cavity sources of thermal radiation. Theory and Practice of Radiation Thermometry, D. P. DeWitt and G. D. Nutter, Eds., John Wiley and Sons, 653–772.

    • Search Google Scholar
    • Export Citation
  • Branch, R., Jessup A. T. , and Minnett P. J. , 2008: Comparisons of shipboard infrared sea surface skin temperature measurements from the CIRIMS and the M-AERI. J. Atmos. Oceanic Technol., 25 , 598606.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • DeWitt, D., and Richmond J. , 1988: Thermal radiative properties of materials. Theory and Practice of Radiation Thermometry, D. DeWitt and G. Nutter, Eds., Wiley-Interscience, 91–188.

    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., and Nightingale T. J. , 2000: Effect of atmospheric radiance errors in radiometric sea-surface skin temperature measurements. Appl. Opt., 39 , 23872392.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Nightingale T. J. , Fiedler L. , Fisher G. J. , Baldwin D. , and Robinson I. S. , 1999a: The calibration and intercalibration of sea-going infrared radiometer systems using a low cost blackbody cavity. J. Atmos. Oceanic Technol., 16 , 11831197.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Nightingale T. J. , Sheasby T. , Turner J. , and Robinson I. S. , 1999b: Implications of the oceanic thermal skin temperature deviation at high wind speed. Geophys. Res. Lett., 26 , 25052508.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Minnett P. J. , Gentemann C. , Nightingale T. J. , Barton I. J. , Ward B. , and Murray M. J. , 2002: Toward improved validation of satellite sea surface skin temperature measurements for climate research. J. Climate, 15 , 353369.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Donlon, C. J., Robinson I. S. , Wimmer W. , Fisher G. , Edwards R. , and Nightingale T. J. , 2008: An Infrared Sea Surface Temperature Autonomous Radiometer (ISAR) for deployment aboard Volunteer Observing Ships (VOS). J. Atmos. Oceanic Technol., 25 , 93113.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fowler, J. B., 1995: A third generation water bath based blackbody source. J. Res. Natl. Inst. Stand. Technol., 100 , 591599.

  • Hanafin, J. A., and Minnett P. J. , 2005: Measurements of the infrared emissivity of a wind-roughened sea surface. Appl. Opt., 44 , 398411.

  • Kannenberg, R., 1998: IR instrument comparison workshop at the Rosenstiel School of Marine and Atmospheric Science (RSMAS). Earth Obs., 10 , 5154.

    • Search Google Scholar
    • Export Citation
  • Katsaros, K. B., 1980: Aqueous thermal boundary layer. Bound.-Layer Meteor., 18 , 107127.

  • Masuda, K., Takashima T. , and Takayama Y. , 1988: Emissivity of pure and sea waters for the model sea surface in the infrared window regions. Remote Sens. Environ., 24 , 313329.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Minnett, P. J., Knuteson R. O. , Best F. A. , Osborne B. J. , Hanafin J. A. , and Brown O. B. , 2001: The Marine-Atmospheric Emitted Radiance Interferometer: A high-accuracy, sea-going infrared spectroradiometer. J. Atmos. Oceanic Technol., 18 , 9941013.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Robinson, I. S., Wells N. C. , and Charnock H. , 1984: The sea surface thermal boundary layer and its relevance to the measurement of sea surface temperature by airborne and spaceborn radiometers. Int. J. Remote Sens., 5 , 1946.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Robinson, I. S., Wimmer W. , and Donlon C. J. , 2006: Validation of AATSR sea surface temperature products using the shipborne ISAR Radiometer. Final Rep. ISAR-DEFRA-FREP01, 39 pp.

  • Shaw, J. A., and Marston C. , 2000: Polarized infrared emissivity for a rough water surface. Opt. Express, 7 , 375380.

  • Soloviev, A., and Lukas R. , 1997: Observation of large diurnal warming events in the near-surface layer of the western equatorial Pacific warm pool. Deep-Sea Res. I, 44 , 10551076.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Ward, B., Wanninkof R. , Minnett P. J. , and Head M. J. , 2004: SkinDeEP: A profiling instrument for upper-decameter sea surface measurements. J. Atmos. Oceanic Technol., 21 , 207222.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Watts, P. D., Allen M. R. , and Nightingale T. J. , 1996: Wind speed effects on sea surface emission and reflection for the Along Track Scanning Radiometer. J. Atmos. Oceanic Technol., 13 , 126141.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Combined cruise tracks, color-coded for Tskin, for R/V Brown, R/V Thompson, and U.S. Coast Guard (USCG) Polar Sea from 2001 through 2005. The total distance covered is over 300 000 km.

  • Fig. 2.

    Calibration curves for KT11 radiometer for housing (case) temperature of 10°, 20°, and 35°C. Here, TBB is the temperature of the reference blackbody and Vout is the voltage output of the radiometer. The mean difference between the output for Tcase of 10° and 35°C was 0.09°C.

  • Fig. 3.

    Sea surface emissivity as a function of wavelength and the incidence angles used in CIRIMS deployments. Also shown is the response function for the KT11 radiometer used in CIRIMS. The variation of the emissivity is relatively low over the wavelength range of the radiometer.

  • Fig. 4.

    Block diagram of CIRIMS components in (left) electronics chassis and (right) sensor housing.

  • Fig. 5.

    CIRIMS sensor housing. (a) Photograph showing deployment along ship railing. (b) Cut-away engineering drawing showing key internal components.

  • Fig. 6.

    Error due to using a calibrated sea-viewing radiometer and a separate uncalibrated sky-viewing radiometer as a function of Tskin and Tsky assuming an uncertainty of 2.5°C in the sky measurement. The error is greatest for cloudy conditions and cold skin temperatures. For clear-sky conditions, the maximum error is roughly 0.03°C.

  • Fig. 7.

    Engineering drawing of the cylindrocone blackbody used in the CIRIMS. (Dimensions are in mm)

  • Fig. 8.

    Laboratory calibration setup for calibrating KT11 radiometer against laboratory blackbody and to determine emissivity of cylindrocone blackbody used in CIRIMS. The radiometer is enclosed in a temperature-stabilized chamber and alternately rotates to view the reference blackbody on the right and the CIRIMS blackbody on the bottom left.

  • Fig. 9.

    CIRIMS radiometer calibration error as a function of reference blackbody temperature for different instrument housing (box) temperatures measured using the setup in Fig. 8. The results are shown for a housing temperature fixed at 35°C and set to the temperature of the reference blackbody. Open symbols are the error without correcting for the nonunity emissivity of the CIRIMS blackbody. The error is reduced when the correction is applied (solid symbols).

  • Fig. 10.

    Schematic illustration of correction for effect of IR transparent window. (top) The radiance of the flat-plate blackbody is measured at two different temperatures with the window not in place. (bottom) The measurement is repeated with the window in place.

  • Fig. 11.

    Steps in window correction using the external flat-plate blackbody. (a) Regression of Vwindow vs Vno_window removes attenuation; (b) regression of first residual vs Twindow removes emission (slope) and reflection (offset). The std dev of the second residual is equal to 0.04°C and is a measure error of the window correction.

  • Fig. 12.

    Time series of (top) Tsky and (middle) Tskin showing variation due to changing the effective incidence angle by ±5° about the measurement angle of 40° and (bottom) adding sky variance σsky to estimate effect of sky variability. The mean difference between Tskin(50) and Tskin(45) was 0.065°C and the mean difference between Tskin(45) and Tskin(40) was 0.040°C. The RMS difference between Tskin and Tskin computed with σsky added to the sky radiance was 0.030°C and is taken as a measure of the uncertainty due to variable sky conditions.

  • Fig. 13.

    Through-the-hull sensor ports at 2 and 3 m below the water line located in the bow thruster room on the R/V Brown. Identical ports have been installed on the R/V Thompson.

  • Fig. 14.

    Example of measurements of diurnal warm layer measured from the R/V Thompson over a 3-day period in 2004. The time series are (top) wind speed, (middle) solar radiation, and (bottom) water temperatures: skin temperature, Tskin; bulk temperatures from through-the-hull ports at 2 and 3 m, T2m and T3m, respectively; and bulk temperature from ship’s thermosalinograph, T5m. The degree of stratification depends on both the solar radiation and wind speed.

  • Fig. 15.

    Scatterplots of all data collected during the EPIC 2001 cruise showing the bulk-skin temperature difference (ΔT = TskinTbulk) vs wind speed U10 computed using bulk temperatures from (left) 5 m, (middle) 2 m, and (right) the sea snake. The erroneously large positive values at low wind speeds using Tbulk at 5 m are due to diurnal warming and emphasize the value of the underway bulk measurements at 2 m.

  • Fig. 16.

    Time series of wind speed, solar radiation, ship speed, and water temperatures showing the effect of ship speed on bulk temperatures. The ship abruptly goes from being on station to an underway speed of roughly 12 kt shortly after 1200 UTC. Stratification prior to being underway is unreasonably weak based on increasing solar radiation and low wind, while after getting underway the stratification is characteristic of the environmental conditions. Prior to getting underway, the temperature at 5 m shows an unusual variability, which may be due to flow around the ship’s hull. After getting underway, the temperatures at 2 and 3 m show increased variability, which is likely due to the ship’s heave.

  • Fig. 17.

    (left) Temperature map based on vertical excursion of the through-hull sensors (depth bin size: 0.1 m, time bin size: 10 s). (right) Comparison of profiles based on temperature map and averages of Tskin (5 min), T2m, T3m, and T5m (1 min).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 427 78 23
PDF Downloads 256 46 6