• Bates, J. J., Wu X. , and Jackson D. L. , 1996: Interannual variability of upper-troposphere water vapor band brightness temperature. J. Climate, 9 , 427438.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Blumstein, D., Tournier B. , Cayla F. R. , Phulpin T. , Fjortoft R. , Buil C. , and Ponce G. , 2007: In-flight performance of the infrared atmospheric sounding interferometer (IASI) on METOP-A. Atmospheric and Environmental Remote Sensing Data Processing and Utilization III: Readiness for GEOSS, M. D. Goldberg et al., Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 6684), 66840H, doi:10.1117/12.734162.

    • Search Google Scholar
    • Export Citation
  • Cao, C., and Heidinger A. K. , 2002: Intercomparison of the long-wave infrared channels of MODIS and AVHRR/NOAA-16 using simultaneous nadir observations at orbit intersections. Earth Observing Systems VII, W. L. Barnes, Ed., International Society for Optical Engineering (SPIE Proceedings, Vol. 4814), 306–316.

    • Search Google Scholar
    • Export Citation
  • Cao, C., Weinreb M. , and Kaplan S. , 2004a: Verification of the HIRS spectral response functions for more accurate atmospheric sounding. Proc. Conf. on Characterization and Radiometric Calibration for Remote Sensing (CALCON), Logan, UT, Space Dynamics Laboratory of Utah State University, CD-ROM 1.

    • Search Google Scholar
    • Export Citation
  • Cao, C., Weinreb M. , and Xu H. , 2004b: Predicting simultaneous nadir overpasses among polar-orbiting meteorological satellites for the intersatellite calibration of radiometers. J. Atmos. Oceanic Technol., 21 , 537542.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cao, C., Ciren P. , Goldberg M. , and Weng F. , 2005a: Intersatellite calibration of HIRS from 1980 to 2003 using the simultaneous nadir overpass (SNO) method for improved consistency and quality of climate data. Int. TOVS Study Conf., Beijing, China. [Available online at http://cimss.ssec.wisc.edu/itwg/itsc/itsc14/presentations/poster_a/Cao_etal_HIRS_hist_sno_ITSC14.pdf].

    • Search Google Scholar
    • Export Citation
  • Cao, C., Xu H. , Sullivan J. , McMillin L. , Ciren P. , and Hou Y. , 2005b: Intersatellite radiance biases for the High Resolution Infrared Radiation Sounders (HIRS) on board NOAA-15, -16, and -17 from simultaneous nadir observations. J. Atmos. Oceanic Technol., 22 , 381395.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cao, C., Jarva K. , and Ciren P. , 2007: An improved algorithm for the operational calibration of HIRS. J. Atmos. Oceanic Technol., 24 , 169181.

  • Jackson, D. L., and Bates J. J. , 2001: Climate analysis with the 21 yr HIRS pathfinder radiance clear sky radiance data set. Proc. 11th Conf. on Satellite and Oceanography, Madison, WI, University of Wisconsin—Madison, 138–140.

    • Search Google Scholar
    • Export Citation
  • Jackson, J., Wylie D. P. , and Bates J. J. , 2003: The HIRS Pathfinder radiance data set (1979-2001). Preprints, 12th Conf. on Satellite Meteorology and Oceanography, Long Beach, CA, Amer Meteor. Soc., P1.8. [Available online at http://ams.confex.com/ams/pdfpapers/57094.pdf].

    • Search Google Scholar
    • Export Citation
  • Lee, H., and Gruber A. , 2007: Development of the HIRS outgoing longwave radiation climate dataset. J. Atmos. Oceanic Technol., 24 , 20292047.

  • Shi, L., Bates J. J. , and Cao C. , 2008: Scene radiance–dependent intersatellite biases of HIRS longwave channels. J. Atmos. Oceanic Technol., 25 , 22192229.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Soden, B. J., and Bretherton F. P. , 1996: Interpretation of TOVS water vapor radiances in terms of layer-average relative humidities: Method and climatology for the upper, middle, and lower troposphere. J. Geophys. Res., 101 , (D5). 93339343.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Strow, L., Hannon S. , and Tobin D. , 2008: 2008 Proceedings of the Conference on. Characterization and Radiometric Calibration for Remote Sensing (CALCON). Logan, UT, Space Dynamics Laboratory of Utah State University, CD-ROM 1.

    • Search Google Scholar
    • Export Citation
  • Tobin, D., and Coauthors, 2007: 2007 Proceedings of the Conference. on Characterization and Radiometric Calibration for Remote Sensing (CALCON). Logan, UT, Space Dynamics Laboratory, Utah State University, CD-ROM 1.

    • Search Google Scholar
    • Export Citation
  • Wang, L., Cao C. , and Ciren P. , 2007: Assessing NOAA-16 HIRS radiance accuracy using simultaneous nadir overpass observations from AIRS. J. Atmos. Oceanic Technol., 24 , 15461561.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wylie, D. P., and Menzel W. P. , 1999: Eight years of high cloud statistics using HIRS. J. Climate, 12 , 170184.

  • Wylie, D. P., Menzel W. P. , Woolf H. M. , and Strabala K. I. , 1994: Four years of global cirrus cloud statistics using HIRS. J. Climate, 7 , 19721986.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zou, C., Goldberg M. , Cheng Z. , Grody N. , Sullivan J. , Cao C. , and Tarpley D. , 2006: Recalibration of microwave sounding unit for climate studies using simultaneous nadir overpasses. J. Geophys. Res., 111 , D19114. doi:10.1029/2005JD006798.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    Brightness temperature comparison between AIRS and IASI at the SNO on 8 Mar 2007 (red/light = IASI; black/dark = AIRS).

  • View in gallery

    (top) Band 4 radiance ratio between NOAA-6 and -7 [identification number (id) = 11] and NOAA-7 and -8 (id = 10), and (bottom) their spectral response functions (crosses = NOAA-8, triangles = NOAA-7, and pluses = NOAA-6; sample IASI spectral radiance is used as background; data from the sample orbit on 21 Jun 2008).

  • View in gallery

    Channels (top) 4 and (bottom) 8 radiance ratios for successive satellite pairs from TIROS-N (N) to MetOp (computed from selected IASI orbit on 21 Jun 2008). Lable id: 0 = MetOp/N18; 1 = N18/N17; 2 = N17/N16; 3 = N16/N15; 4 = N15/N14; 5 = N14/N12; 6 = N12/N11; 7 = N11/N10; 8 = N10/N9; 9 = N9/N8; 10 = N8/N7; 11 = N7/N6; and 12 = N6/TIROS-N.

  • View in gallery

    Seasonal variations in the band radiance ratio and IASI/HIRS spectral bias bell curves. (right) Note the large contrast between December and June in Antarctica shown inside the rectangular box. Latitude curves (dashed lines) also indicate ascending vs descending of the MetOp/IASI orbit.

  • View in gallery

    Long-term SNO biases between successive NOAA satellite pairs show anomalous oscillation (green) between NOAA-14 and -15 for HIRS band 4.

  • View in gallery

    Relative spectral shift analysis of NOAA-15 and -14 HIRS can reproduce the large bias seen in the SNO data for band 4 (±2 means spectrally shift to a higher/lower wavenumber, respectively).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 160 120 9
PDF Downloads 57 28 2

Spectral Bias Estimation of Historical HIRS Using IASI Observations for Improved Fundamental Climate Data Records

View More View Less
  • 1 NOAA/NESDIS/ORA, Camp Springs, Maryland
  • | 2 Perot Systems Government Services, Fairfax, Virginia
Full access

Abstract

A prerequisite for climate change detection from satellites is that the measurements from a series of historical satellites must be consistent and ideally made traceable to the International System of Units (SI). Unfortunately, this requirement is not met for the 14 High Resolution Infrared Radiation Sounders (HIRS) on the historical NOAA satellites, because the instrument was developed for weather forecasts and lacks accuracy and consistency across satellites. It is well known that for HIRS, differences in the spectral response functions (SRF) between instruments and their prelaunch measurement uncertainties often lead to observations of the atmosphere at different altitudes. As a result of the atmospheric lapse rate, they both can introduce significant intersatellite biases. The SRF-dependent biases are further mixed with other effects such as the diurnal cycle because of observation time differences and orbital drifts, on board calibration, and algorithm issues. In this study, the Infrared Atmospheric Sounding Interferometer (IASI) observations are used to calculate the radiances for the 14 Television Infrared Observation Satellite series N (TIROS-N; to MetOp-A) HIRS instruments in different climate regimes and seasons to separate the SRF-induced intersatellite biases from other factors. It is found that the calculated radiance ratio (a bias indicator) using IASI observations for the HIRS satellite pairs forms bell-shaped curves that vary with the HIRS model and channel as well as climate regimes. This suggests that a bias found in the polar regions at the Simultaneous Nadir Overpass (SNO) cannot be blindly used for bias correction globally; instead, the IASI/HIRS spectral bias bell curves should be used as a complement to more fully address the biases. These bell curves also serve as lookup charts for separating the bias due to true SRF differences from that caused by SRF prelaunch measurement errors to resolve the inconsistency, which sheds new light on reprocessing and reanalysis in generating fundamental climate data records from HIRS.

Corresponding author address: Dr. Changyong Cao, NOAA/NESDIS/STAR, WWB, Rm. 712, 5200 Auth Rd., Camp Springs, MD 20746. Email: changyong.cao@noaa.gov

Abstract

A prerequisite for climate change detection from satellites is that the measurements from a series of historical satellites must be consistent and ideally made traceable to the International System of Units (SI). Unfortunately, this requirement is not met for the 14 High Resolution Infrared Radiation Sounders (HIRS) on the historical NOAA satellites, because the instrument was developed for weather forecasts and lacks accuracy and consistency across satellites. It is well known that for HIRS, differences in the spectral response functions (SRF) between instruments and their prelaunch measurement uncertainties often lead to observations of the atmosphere at different altitudes. As a result of the atmospheric lapse rate, they both can introduce significant intersatellite biases. The SRF-dependent biases are further mixed with other effects such as the diurnal cycle because of observation time differences and orbital drifts, on board calibration, and algorithm issues. In this study, the Infrared Atmospheric Sounding Interferometer (IASI) observations are used to calculate the radiances for the 14 Television Infrared Observation Satellite series N (TIROS-N; to MetOp-A) HIRS instruments in different climate regimes and seasons to separate the SRF-induced intersatellite biases from other factors. It is found that the calculated radiance ratio (a bias indicator) using IASI observations for the HIRS satellite pairs forms bell-shaped curves that vary with the HIRS model and channel as well as climate regimes. This suggests that a bias found in the polar regions at the Simultaneous Nadir Overpass (SNO) cannot be blindly used for bias correction globally; instead, the IASI/HIRS spectral bias bell curves should be used as a complement to more fully address the biases. These bell curves also serve as lookup charts for separating the bias due to true SRF differences from that caused by SRF prelaunch measurement errors to resolve the inconsistency, which sheds new light on reprocessing and reanalysis in generating fundamental climate data records from HIRS.

Corresponding author address: Dr. Changyong Cao, NOAA/NESDIS/STAR, WWB, Rm. 712, 5200 Auth Rd., Camp Springs, MD 20746. Email: changyong.cao@noaa.gov

1. Introduction

Consistency in the observations from a series of satellites is a prerequisite for climate change detection. The High Resolution Infrared Radiation Sounder (HIRS) on National Oceanic and Atmospheric Administration (NOAA) satellites (Table 1) provides one of the longest records of satellite observations of the Earth from the surface to the stratosphere, and has been used for a number of climate change detection studies (Soden and Bretherton 1996; Bates et al. 1996; Wylie et al. 1994; Wylie and Menzel 1999; Jackson and Bates 2001; Jackson et al. 2003; Lee and Gruber 2007; Shi et al. 2008). However, there are significant challenges in making the satellite observations from the 14 HIRS instruments consistent for climate change detection. It is well known that for HIRS, the intersatellite radiance biases are significantly affected by the differences in the spectral response functions (SRF) between instruments, which often leads to observations of the atmosphere at different altitudes. The SRF-dependent biases are further mixed with spectral uncertainties due to the lack of stringent requirements in prelaunch instrument SRF characterization. Separating the intersatellite biases due to true spectral response differences from errors in the SRF measurements has become the most challenging and yet unavoidable task for producing reliable fundamental climate data records from HIRS. Since prelaunch calibration becomes an irreversible process for historical satellites once the satellite is launched, in this study we use the Infrared Atmospheric Sounding Interferometer (IASI) (Table 1) observations to calculate the radiances for the 14 Television Infrared Observation Satellite series N (TIROS-N to MetOp-A) HIRS instruments in different climate regimes and seasons, which allows us to separate the true SRF-induced intersatellite biases from SRF prelaunch measurement errors. As it is demonstrated in this study, the results can be used as guidelines in assessing the intersatellite biases for HIRS. When combined with the Simultaneous Nadir Overpass (SNO) methodology, it also helps us identify and ultimately correct the SRF errors to establish intersatellite consistency for climate change detection.

2. Methodology for estimating HIRS spectral biases

a. Quantifying SRF-induced intersatellite spectral biases

In previous studies, it has been demonstrated that several factors can cause observation biases between similar channels of instruments on different satellites. SRF differences and their prelaunch measurement uncertainties are among the major factors for the biases (Cao et al. 2004b, 2005b). These factors can be summarized in the following equation:
i1520-0426-26-7-1378-e1
where
  • β = observation bias between the two instruments,

  • t = observation time difference (including diurnal cycle effect),

  • n = off-nadir effects (both instrument and view path),

  • s = spatial sampling differences, including geolocation, coregistration, alignment, scene uniformity, and sensor modulation transfer functions (MTF; or sidelobe effects for microwave instruments),

  • ɛ = bias in the calibration system (blackbody/diffuser, thermometer, mirror/reflector) and algorithm,

  • l = nonlinearity,

  • υ = SRF difference and uncertainty,

  • o = other factors, including human error and calibration anomaly.

The root cause of β cannot be easily resolved because a combination of these seven factors might be involved. A strategy to solve this problem is to reduce the number of unknown variables to evaluate the effect of specific factors. In this study, the IASI nadir observations from the MetOp-A satellite are used to calculate the HIRS radiances using the HIRS SRF from TIROS-N to MetOp-A. Assuming that the other six variables remain the same, this study focuses on the effect of a single variable—the SRF differences and related uncertainties—which we believe is one of the most important factors in HIRS intersatellite biases. The spectral response uncertainties are typically caused by prelaunch measurement errors, with a smaller possible change or drift postlaunch. Previous studies (Cao et al. 2004a, 2005b) have shown that spectral uncertainty is a major concern for HIRS. However, in general these spectral uncertainties are difficult to quantify without accurate prelaunch measurements. In this study, we first assume that the prelaunch-measured SRFs were correct (as our null hypothesis) and focus on the intersatellite biases due to these (presumably correct) SRF differences based on IASI calculations. Then these biases are compared with actual observed biases at the SNOs to determine the differences. The root cause of these differences is then investigated. We found that prelaunch spectral response measurement uncertainty is the most likely root cause for these differences because the biases would disappear when the SRFs are shifted, while other factors could not explain these biases.

It is known that for the sounding channels of infrared sounders (“channel” and “band” are used interchangeably in this paper), a difference in the SRF means that the channels are probably observing different atmospheric layers at different altitudes, since these channels have different vertical weighting functions. For a perfectly calibrated instrument, the observations from two instruments with slightly different SRFs will have a bias primarily due to the lapse rate of the atmosphere. Since the lapse rate varies globally, this bias also changes with the climate regime, location, and season. It should be noted that since these are not instrument calibration biases, they should be treated spectrally rather than radiometrically, which often is not the case in the current practice of data assimilation in numerical weather predictions. Mixing spectral response–induced biases (i.e., spectral biases) with instrument calibration biases (i.e., radiometric biases) will lead to erroneous analyses. It is generally recognized in scientific literature that different SRFs will cause measurement biases for the infrared sounding channels (Cao et al. 2005b; Wang et al. 2007). However, studies have shown that a prelaunch measurement of the SRFs has uncertainties based on comparisons of measurements by both the vendor and National Institute of Standards and Technology (NIST). As a result, it is difficult to separate the spectral biases due to prelaunch measurement uncertainties from the actual differences in the SRFs. It is possible to use a Radiative Transfer Model (RTM) to simulate the spectral biases. However, in general, the accuracy of the RTM calculations is less than what is desired for instrument calibration. In addition, the RTM models use a limited number of atmospheric profiles, and mostly under clear sky conditions, which makes the simulation less realistic.

The launch of the MetOp-A/IASI hyperspectral sounder provides an excellent opportunity for calibrating HIRS spectrally and quantifying the SRF-induced spectral biases. MetOp-A, successfully launched in October 2006, is the first European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) polar-orbiting satellite, which carries several heritage instruments provided by NOAA, including the Advanced Very High Resolution Radiometer (AVHRR), HIRS, and Advanced Microwave Sounding Unit (AMSU). The IASI provided by EUMETSAT is one of the most advanced instruments, which measures radiation emitted from the surface and atmosphere in the spectral range of 645–2760 cm−1 (i.e., 3.6–15.5 μm) with a high spectral resolution (8461 spectral channels with a sampling interval of 0.25 cm−1, or a spectral resolution of 0.5 cm−1). Therefore, assuming that the IASI has accurate instrument calibration (discussed below), its spectra can be used to evaluate the SRF-induced spectral biases between different models of HIRS and, when combined with the SNO methodology, will allow us to identify SRF measurement errors as demonstrated in section 4 of this study.

b. IASI spectra as a stable on-orbit spectral and radiometric reference standard

On-orbit assessment of infrared radiances relies on stable reference standards. Although infrared instruments have onboard blackbodies for calibration, for most instruments their radiometric traceability to the International System of Units (SI) cannot be guaranteed. This is especially true for broadband instruments, for which spectral and radiometric uncertainties cannot be easily separated. The advance in hyperspectral infrared sounding significantly reduces the spectral uncertainties and also leads to improvements in the radiometric accuracy. The on-orbit agreement of two infrared hyperspectral sounders [such as IASI and the Atmospheric Infrared Sounder (AIRS)] can provide a relatively reliable community reference standard. Both IASI and AIRS provide spectrally resolved radiances in the 3.6–15.5-μm spectral range. Figure 1 shows that IASI and AIRS in general have good agreement at a selected SNO, in addition to several other features including the spectral coverage, and typical scene brightness temperatures. In the shortwave, IASI observations are noisy, due to the low signal at cold temperatures in the polar region where the SNOs occur. The noise is also due to the fact that IASI spectral resolution is finer than that of AIRS. In addition to the sample checks, IASI and AIRS are regularly compared at the SNOs with more than 20 pseudochannels at NOAA/National Environmental Satellite, Data, and Information Service (NESDIS), where it was found that the biases are typically 0.1 K or less. Similar conclusions have been reached in independent studies and presented at conferences and workshops (Blumstein et al. 2007; D. Tobin 2008, personal communication; Strow et al. 2008;). The excellent agreement between IASI and AIRS provides a quasi on-orbit radiometric reference standard for the infrared remote sensing community.

In addition to the SNO comparisons, an aircraft underflight campaign also showed that IASI measurements agree with those of the Scanning High-Resolution Interferometer Sounder (SHIS) at the 0.1-K level (Tobin et al. 2007). Preliminary comparison of the measurements between IASI and the aircraft sensor indicates that IASI has performed well within specifications, maintaining excellent spectral and radiometric calibration accuracy. Based on independent calibration/validation, it is found that IASI is producing accurate hyperspectral radiances that can be used as a reference for other radiometers. This has provided the foundation for evaluating the spectral response–induced biases of HIRS using IASI in the current study.

c. Computing the HIRS radiances from selected IASI orbital data

Excluding HIRS channel 19, which is not fully covered by IASI spectrally, the spectral radiance of HIRS Rh(υ) for a given field of view (FOV) can be calculated using the IASI spectral radiance Ri(υ) with the following equation:
i1520-0426-26-7-1378-e2
where SRF(υ) is the HIRS SRF for a given channel and model, and υ1 and υ2 provide the full wavenumber range (cm−1). Both Ri and Rh have a unit of mW(m2 sr cm−1)−1. Since each of the 14 HIRS instruments and 19 infrared bands has its unique SRF with which each IASI spectrum is convolved, a dataset with a size of 14 × 19 is produced for each orbit of IASI data at the nadir pixel.

The IASI level 1C data are available online in 3-min grandules from the NOAA/NEDSDIS/ the Center for Satellite Applications and Research (STAR) Collaborative Environment (CE) and each grandule has a size of approximately 60 mb. Given the MetOp orbital period of 101 min, one orbit of IASI data takes about 2 GB of storage space. In this study, selected IASI orbital data are used to calculate the HIRS radiances for all models and channels. Four orbits of IASI data, representing four seasons, are selected in this study. It is expected that while the four orbits may not fully represent all atmospheric features, they cover a wide range of climate regimes, and are representative of typical Earth atmospheres. The dates and times of the selected data are listed in Table 2. Note that the 18 March 2008 orbit is primarily an ocean-only orbit that crosses the Atlantic and Pacific Oceans, as well as the Arctic and Antarctic regions, which further reduces the uncertainties associated with land emissivity variations.

3. Radiance ratio and IASI/HIRS spectral bias bell curves

With the selected IASI orbits and the HIRS SRFs as input, the IASI spectral radiances are convolved with the HIRS SRFs to produce the HIRS inband spectral radiance at the nadir pixel for each HIRS channel and model. The radiance ratio between successive HIRS models (such as NOAA-6 and -7, and NOAA-7 and -8) is then calculated to represent the bias or difference between them. The radiance ratios are then plotted as a function of scan line for the orbit. It is known that the atmosphere may have changed since NOAA-6 (e.g., CO2 amount) but it is assumed that the changes impact on the radiance ratio is relatively small, as is found with our preliminary analysis using a line-by-line Radiative Transfer Model (LBLRTM). Another assumption is that the dominant biases are caused by the climate regime and season, not by the daily weather patterns. The effect of clouds is not considered here except that they can be seen as spikes in the radiance ratio. Although many channels are affected by the surface or ground emission, we assume that the surface is not a major factor for the variation in the radiance ratio (or the difference in the biases) between satellites. This is consistent with our analysis of the surface channel (band 8), where the variation of radiance ratios between all pairs of satellites is small.

Figure 2a exemplifies the HIRS channel 4 radiance ratio variation for a summer orbit on 21 June 2008, between NOAA-6/-7, and NOAA-7/-8. It indicates that while the radiance ratio variation between NOAA-6 and -7 is relatively small, the variation is much larger between NOAA-7 and -8 across different latitudes. The latitude curve indicates that in the Arctic and Antarctica, the radiance ratio is between 0.98 and 1.0, while in the tropical regions, this ratio changes to below 0.96 for this June dataset (see discussion on seasonal variation later). Therefore, the HIRS intersatellite radiance ratio between NOAA-7 and -8 varies between ∼0.96 and 1.0, or roughly a total of 4%. The differences in radiance can be converted to differences in brightness temperature using the following equation according to the Planck function:
i1520-0426-26-7-1378-e3
where dT is brightness temperature difference, or delta T, R is radiance, T is blackbody equivalent temperature, υ is wavenumber, c1 is 1.1910427E-05, and c2 is 1.4387752. For example, given the typical brightness temperature of 230 K for band 4 (703 cm−1), a radiance ratio of 0.99 (or difference of 1%) converts to a brightness temperature difference of ∼0.5 K, while a ratio of 0.95 (or difference of 5%) converts to ∼2.6 K. The HIRS band 4 is one of the CO2 channels with a vertical weighting function peaking at 250 hPa. The smaller biases between NOAA-6 and -7 than those between NOAA-7 and -8 can be explained by the SRF differences. Figure 2b shows that there is a smaller difference (<2 cm−1) between NOAA-6 and -7 SRFs than between NOAA-7 and -8 (about 3 cm−1) SRFs. Note that despite the discrepancies in center wavenumber, the SRFs of all models supposedly have met the instrument specifications, which state that the center wavenumber should be 703 ± 1.8 cm−1 for this channel. However, the HIRS specification was written for weather applications and apparently does not necessarily meet the needs for climate applications.

The radiance ratios for both bands 4 and 8 for all successive satellites (NOAA-7/6, NOAA-8/7, NOAA-9/8, …) are plotted in Fig. 3. These IASI/HIRS spectral bias bell curves (named based on the fact that they are “dumbbell” shaped) in Fig. 3a reveal the intersatellite biases for HIRS channel 4 for all satellites. It should be noted that while this channel has large variations in the radiance ratio, some other channels have much smaller variations, notably channels 2, 6, and 8, with maximum radiance ratios less than ∼2% (Fig. 3b). This is because channel 8 is a surface channel that is hardly affected by the lapse rate, and the surface spectral emissivity variation is not a major contributor to the differential radiances, while both channels 2 and 6 are located in relatively flat regions on the spectral radiance curves and therefore are not very sensitive to the small spectral changes. Similar figures have been generated for all HIRS channels. Figures 3a and 3b also reveal a major challenge in using HIRS sounding channels for climate change detection. Since the SRFs are different for each satellite, the intersatellite biases can be large and variable with latitudes.

In addition to the changes with latitude, the radiance ratio also changes with season. Figure 4 shows a clear contrast in the radiance ratio for Antarctica on the winter and summer solstice (indicated in the rectangular box in the right panel). It suggests that a larger lapse rate in Antarctica is observed in June than in December. The latitudes of the observations are indicated on the right axis in these figures. In the tropics, a much smaller seasonal change is observed. Finally, Fig. 3a (similar in Figs. 3b, 4) suggests that the radiance ratios are systematically closer to 1.0 (or smaller biases) for all satellite pairs when clouds are observed, shown as vertical spikes in the bell curves as indicated in the figure.

The implication of the results is that in global analysis, if these intricate effects are not taken into account, the intersatellite biases can be mixed with other factors, which will overshadow the small climate change signal, which is on the order of a fraction of a degree per decade in temperature. As a result, the ability to detect climate change using this approach would be questionable. Therefore, it is recommended that the intersatellite biases be analyzed on a zonal and seasonal basis, as delineated in Figs. 3 and 4, and the intersatellite biases for each zone can be better characterized and corrected based on the lapse rate (Cao et al. 2005b) for the particular zone, and better interpreted based on the seasonal variation.

4. Using the IASI/HIRS spectral bias bell curves to identify HIRS spectral errors and establish intersatellite traceability

The bell curves have several applications in HIRS historical data reprocessing and reanalysis. First, they serve as lookup charts for assessing the theoretical radiance ratio between satellites to estimate the intersatellite biases for a particular climate regime and satellite pair. The maximum radiance ratio serves as a guide in analyzing the historical HIRS intersatellite data at the SNOs. Here we provide an example of identifying major spectral response errors and separating them from the SRF-induced spectral biases. In a previous project funded by NOAA/Environmental Services Data and Information Management (ESDIM), we have quantified the HIRS intersatellite biases at the SNO for all historical satellites, and presented part of the results at conferences and in journal publications (see Cao et al. 2005a,b).

One of the major findings was that there are large oscillations in the intersatellite biases between NOAA-14 and -15 satellites for band 4 at the SNOs. As shown in Fig. 5, the biases varied with a clear seasonal oscillation between nearly 0% and 10% from 1999 to 2004 and the maximum biases of up to 10% occurred in the Southern Hemisphere winter, peaking around June. The question is whether this seasonal bias can be explained by the given SRFs of NOAA-14 and -15 HIRS. To address this issue, we performed the following spectral shift analysis.

First we selected an IASI orbit from 21 June 2008 when the largest intersatellite bias occurred (according to the SNO results) and computed the radiance ratio as N14/N15 as shown in the black curve in Fig. 6. The IASI results show that the maximum bias in the polar regions should be small (<2%), which disagrees with the bias of 10% at the HIRS SNOs. We then analyzed the calibration coefficients for both NOAA-14 and -15 HIRS at the SNOs and compared them with prelaunch values to rule out possible onboard calibration errors in the calculations and instrument anomalies during the period analyzed. In addition, we found that the calibration bias for the surface channel band 8 is much smaller (<0.8%), which suggests that the intersatellite biases are not likely caused by onboard blackbody calibration. By ruling out the other possibilities, we believe that prelaunch measurement errors in the SRFs of either NOAA-14 or 15 (or both) are probably the culprit for the bias. To prove this, we performed spectral shift analysis with different combinations of both NOAA-14 and -15 band 4 SRFs and found that when the NOAA-15 SRF is shifted to a smaller value by 4 wavenumbers, the 10% bias between NOAA-14 and -15 band 4 can be reproduced. This suggests that NOAA-15 band 4 SRF could be spectrally deviated (probably due to prelaunch measurement error) from the reported values by 4 wavenumbers (Fig. 6). However, the same result can be produced with a 2-wavenumber shift in both NOAA-15 and NOAA-14 SRFs in opposite directions. This suggests that spectral shift analysis, while it can explain the phenomena, may not be sufficient in underpinning the true SRFs unless a reliable reference is established. Figure 6 also suggests that the biases due to spectral error are not limited to the polar region. In fact, the largest bias could occur in the tropics for this channel. This will have a large impact on both climate and weather applications. The ultimate solution to the spectral error is to perform SI traceable measurements on the HIRS filters or their witness samples. Similar work has been done previously with a demonstrated value of this type of study (Cao et al. 2004a). Another approach is to use IASI as an on-orbit reference, accurately transfer the IASI calibration to HIRS on MetOp, and then use MetOp HIRS to retrospectively calibrate the earlier HIRS [a similar approach was used for MSU recalibration by Zou et al. (2006)]. However, this requires further improvements in the methodology in intercalibrating HIRS and IASI on MetOp, since their FOVs are not collocated and there are significant spatial sampling gaps between them.

The second application of the IASI/HIRS spectral bias bell curves is to identify channels that can be used for calibration transfer between satellites. These are the channels with the smallest orbital variation in the radiance ratio among all satellite pairs. As discussed earlier, channels 2, 6, and 8 have a radiance ratio variation less than 2%. It is known that the surface (window) channel 8 at 11 μm is less affected by atmospheric effects (despite the small water vapor absorption effect), although it is affected by the surface emissivity variations due to different types of landcover. IASI calculations of the HIRS radiances from NOAA-6 to MetOp show that band 8 has the smallest intersatellite biases (less than 0.8% for all successive satellite pairs) among all the infrared channels of HIRS. In addition, the correlation of the radiances between successive satellites for this channel is fairly linear and can be estimated using a linear equation with small residual errors (∼0.04%). Therefore, it is concluded that band 8, when well calibrated radiometrically, can be used to establish the radiometric calibration traceability for all HIRS bands from TIROS-N to MetOp.

A detailed analysis on a selected ocean orbit that passes through the Pacific Ocean, Antarctica, Atlantic Ocean, and Arctica (the 18 March 2008 orbit) is performed to further evaluate the intersatellite biases using IASI, since the surface types under observation are ocean, clouds, sea ice, and snow, for which their emissivities are relatively well known than for other types. It is found that the intersatellite biases are extremely small and predictable between satellites for band 8. This is partly due to the relatively flat spectral emissivity within the spectral region that this channel observes. There are several important implications for this analysis. First, due to the small orbital variation of the intersatellite biases for band 8, an intersatellite bias found in the polar region at the SNOs can be applied to global datasets for this particular channel, especially when the correlation in radiances between satellites is established. This is possible also because HIRS is proven to be stable with little orbital variation in the calibration biases in normal operating conditions because of the fact that the radiation reaching the HIRS detector is dominated by aft optics temperatures, which are very stable (Cao et al. 2007). For example, the variation of the orbital gain for band 8 is typically at 0.009%, and the maximum orbital gain variation is on the order of 0.3% per orbit (e.g., average slope: −0.03769; standard deviation: 0.00003414; maximum − minimum = 0.0001364). On the other hand, a major limitation with band 8 is the large variability in scene temperatures from one FOV to the next, especially in the presence of clouds. In contrast, band 2 observes the stratosphere and is mostly free from cloud effects, and therefore can reduce radiometric (not necessarily spectral) uncertainties in the intersatellite comparisons. The radiance ratio between all successive satellite pairs is less than 1.5% for band 2.

Based on the fact that the biases between satellite pairs for band 8 and 2 are very small and predictable with a linear correlation, in theory the calibration links among all satellites can be established for these channels. These two bands can effectively constrain the intersatellite radiometric biases for all other bands, assuming that the blackbody emissivity is spectrally flat (yet to be proven based on laboratory measurements). As a result, any deviation in other bands is likely caused by spectral response differences and related uncertainties. Note that nonlinearity errors are possible but unlikely for HIRS, as demonstrated for band 2 and 8 at both low and high temperatures and also analyzed in previous studies (Cao et al. 2007). Once both the radiometric calibration for these channels and the spectral calibration traceability for the sounding channels are established, a consistent historical HIRS dataset can be established.

5. Conclusions

Spectral differences between HIRS on different satellites and their spectral measurement uncertainties are two major obstacles in creating climate-quality long-term time series from historical data. In this study the spectral radiances from selected IASI orbits are convolved with the spectral response functions of all historical HIRS models and bands to assess the spectrally induced intersatellite biases for HIRS. It was found that the biases are a function of the climate regime related to the latitude and season for most channels. The typical radiance ratios over an orbit form bell-shaped curves that can be used as lookup charts for evaluating the intersatellite biases. The surface channel (band 8), stratosphere channel (band 2), and the low–midtroposphere channel (band 6) have relatively small intersatellite biases and can be used to assess biases due to nonspectral related factors, thus separating the spectral biases based on root cause analysis. However, although the biases at the SNO for these three channels can be applied to global data analysis, the same cannot be performed for other channels because their intersatellite biases are more dependent on the climate regime and season.

The IASI/HIRS spectral bias bell curves are very useful for diagnosing and separating the spectral measurement uncertainties from other effects when combined with the SNO methodology, which will provide the basis for spectral recalibration of historical HIRS instruments. Example analysis of band 4 has identified spectral errors in NOAA-14 and -15. Further studies will allow us to perform accurate spectral calibration for all channels. The bell curves complement the SNO methodology by serving as lookup charts to separate the bias due to true SRF differences from SRF prelaunch measurement errors to resolve the calibration inconsistency across satellites, which will facilitate the reprocessing and reanalysis in generating fundamental climate data records from HIRS.

Acknowledgments

The authors thank Drs. Cheng-zhi Zou, Bob Iacovazzi, Ping Jing, Han Yong, and Mr. Haibin Sun for their critical review and judicious comments. Thanks are extended to Dr. Denis Blumstein of CNES for his help with IASI calibration issues. This study is partially funded by the IPO/IGS and STAR cal/val funds. It also represents a join effort between the World Meteorological Organization (WMO)/Global Space-based Inter-Calibration System (GSICS) and Committee on Earth Observation Satellites/Working Group on Calibration/Validation (CEOS/WGCV) in support of the CEOS climate actions in response to the Global Climate Observation System (GCOS). The manuscript contents are solely the opinions of the authors and do not constitute a statement of policy, decision, or position on behalf of NOAA or the U.S. government.

REFERENCES

  • Bates, J. J., Wu X. , and Jackson D. L. , 1996: Interannual variability of upper-troposphere water vapor band brightness temperature. J. Climate, 9 , 427438.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Blumstein, D., Tournier B. , Cayla F. R. , Phulpin T. , Fjortoft R. , Buil C. , and Ponce G. , 2007: In-flight performance of the infrared atmospheric sounding interferometer (IASI) on METOP-A. Atmospheric and Environmental Remote Sensing Data Processing and Utilization III: Readiness for GEOSS, M. D. Goldberg et al., Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 6684), 66840H, doi:10.1117/12.734162.

    • Search Google Scholar
    • Export Citation
  • Cao, C., and Heidinger A. K. , 2002: Intercomparison of the long-wave infrared channels of MODIS and AVHRR/NOAA-16 using simultaneous nadir observations at orbit intersections. Earth Observing Systems VII, W. L. Barnes, Ed., International Society for Optical Engineering (SPIE Proceedings, Vol. 4814), 306–316.

    • Search Google Scholar
    • Export Citation
  • Cao, C., Weinreb M. , and Kaplan S. , 2004a: Verification of the HIRS spectral response functions for more accurate atmospheric sounding. Proc. Conf. on Characterization and Radiometric Calibration for Remote Sensing (CALCON), Logan, UT, Space Dynamics Laboratory of Utah State University, CD-ROM 1.

    • Search Google Scholar
    • Export Citation
  • Cao, C., Weinreb M. , and Xu H. , 2004b: Predicting simultaneous nadir overpasses among polar-orbiting meteorological satellites for the intersatellite calibration of radiometers. J. Atmos. Oceanic Technol., 21 , 537542.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cao, C., Ciren P. , Goldberg M. , and Weng F. , 2005a: Intersatellite calibration of HIRS from 1980 to 2003 using the simultaneous nadir overpass (SNO) method for improved consistency and quality of climate data. Int. TOVS Study Conf., Beijing, China. [Available online at http://cimss.ssec.wisc.edu/itwg/itsc/itsc14/presentations/poster_a/Cao_etal_HIRS_hist_sno_ITSC14.pdf].

    • Search Google Scholar
    • Export Citation
  • Cao, C., Xu H. , Sullivan J. , McMillin L. , Ciren P. , and Hou Y. , 2005b: Intersatellite radiance biases for the High Resolution Infrared Radiation Sounders (HIRS) on board NOAA-15, -16, and -17 from simultaneous nadir observations. J. Atmos. Oceanic Technol., 22 , 381395.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cao, C., Jarva K. , and Ciren P. , 2007: An improved algorithm for the operational calibration of HIRS. J. Atmos. Oceanic Technol., 24 , 169181.

  • Jackson, D. L., and Bates J. J. , 2001: Climate analysis with the 21 yr HIRS pathfinder radiance clear sky radiance data set. Proc. 11th Conf. on Satellite and Oceanography, Madison, WI, University of Wisconsin—Madison, 138–140.

    • Search Google Scholar
    • Export Citation
  • Jackson, J., Wylie D. P. , and Bates J. J. , 2003: The HIRS Pathfinder radiance data set (1979-2001). Preprints, 12th Conf. on Satellite Meteorology and Oceanography, Long Beach, CA, Amer Meteor. Soc., P1.8. [Available online at http://ams.confex.com/ams/pdfpapers/57094.pdf].

    • Search Google Scholar
    • Export Citation
  • Lee, H., and Gruber A. , 2007: Development of the HIRS outgoing longwave radiation climate dataset. J. Atmos. Oceanic Technol., 24 , 20292047.

  • Shi, L., Bates J. J. , and Cao C. , 2008: Scene radiance–dependent intersatellite biases of HIRS longwave channels. J. Atmos. Oceanic Technol., 25 , 22192229.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Soden, B. J., and Bretherton F. P. , 1996: Interpretation of TOVS water vapor radiances in terms of layer-average relative humidities: Method and climatology for the upper, middle, and lower troposphere. J. Geophys. Res., 101 , (D5). 93339343.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Strow, L., Hannon S. , and Tobin D. , 2008: 2008 Proceedings of the Conference on. Characterization and Radiometric Calibration for Remote Sensing (CALCON). Logan, UT, Space Dynamics Laboratory of Utah State University, CD-ROM 1.

    • Search Google Scholar
    • Export Citation
  • Tobin, D., and Coauthors, 2007: 2007 Proceedings of the Conference. on Characterization and Radiometric Calibration for Remote Sensing (CALCON). Logan, UT, Space Dynamics Laboratory, Utah State University, CD-ROM 1.

    • Search Google Scholar
    • Export Citation
  • Wang, L., Cao C. , and Ciren P. , 2007: Assessing NOAA-16 HIRS radiance accuracy using simultaneous nadir overpass observations from AIRS. J. Atmos. Oceanic Technol., 24 , 15461561.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wylie, D. P., and Menzel W. P. , 1999: Eight years of high cloud statistics using HIRS. J. Climate, 12 , 170184.

  • Wylie, D. P., Menzel W. P. , Woolf H. M. , and Strabala K. I. , 1994: Four years of global cirrus cloud statistics using HIRS. J. Climate, 7 , 19721986.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zou, C., Goldberg M. , Cheng Z. , Grody N. , Sullivan J. , Cao C. , and Tarpley D. , 2006: Recalibration of microwave sounding unit for climate studies using simultaneous nadir overpasses. J. Geophys. Res., 111 , D19114. doi:10.1029/2005JD006798.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

Brightness temperature comparison between AIRS and IASI at the SNO on 8 Mar 2007 (red/light = IASI; black/dark = AIRS).

Citation: Journal of Atmospheric and Oceanic Technology 26, 7; 10.1175/2009JTECHA1235.1

Fig. 2.
Fig. 2.

(top) Band 4 radiance ratio between NOAA-6 and -7 [identification number (id) = 11] and NOAA-7 and -8 (id = 10), and (bottom) their spectral response functions (crosses = NOAA-8, triangles = NOAA-7, and pluses = NOAA-6; sample IASI spectral radiance is used as background; data from the sample orbit on 21 Jun 2008).

Citation: Journal of Atmospheric and Oceanic Technology 26, 7; 10.1175/2009JTECHA1235.1

Fig. 3.
Fig. 3.

Channels (top) 4 and (bottom) 8 radiance ratios for successive satellite pairs from TIROS-N (N) to MetOp (computed from selected IASI orbit on 21 Jun 2008). Lable id: 0 = MetOp/N18; 1 = N18/N17; 2 = N17/N16; 3 = N16/N15; 4 = N15/N14; 5 = N14/N12; 6 = N12/N11; 7 = N11/N10; 8 = N10/N9; 9 = N9/N8; 10 = N8/N7; 11 = N7/N6; and 12 = N6/TIROS-N.

Citation: Journal of Atmospheric and Oceanic Technology 26, 7; 10.1175/2009JTECHA1235.1

Fig. 4.
Fig. 4.

Seasonal variations in the band radiance ratio and IASI/HIRS spectral bias bell curves. (right) Note the large contrast between December and June in Antarctica shown inside the rectangular box. Latitude curves (dashed lines) also indicate ascending vs descending of the MetOp/IASI orbit.

Citation: Journal of Atmospheric and Oceanic Technology 26, 7; 10.1175/2009JTECHA1235.1

Fig. 5.
Fig. 5.

Long-term SNO biases between successive NOAA satellite pairs show anomalous oscillation (green) between NOAA-14 and -15 for HIRS band 4.

Citation: Journal of Atmospheric and Oceanic Technology 26, 7; 10.1175/2009JTECHA1235.1

Fig. 6.
Fig. 6.

Relative spectral shift analysis of NOAA-15 and -14 HIRS can reproduce the large bias seen in the SNO data for band 4 (±2 means spectrally shift to a higher/lower wavenumber, respectively).

Citation: Journal of Atmospheric and Oceanic Technology 26, 7; 10.1175/2009JTECHA1235.1

Table 1.

HIRS and IASI characteristics. Note that HIRS channel 19 is not fully covered by IASI. Also, only 5 out of the 19 HIRS infrared channels are spectrally covered by AIRS. All instruments use onboard blackbody and space view for calibration.

Table 1.
Table 2.

Selected IASI orbits for the spectral bias estimation of HIRS.

Table 2.
Save