• Ackerman, T. P., , and Stokes G. M. , 2003: The Atmospheric Radiation Measurement Program. Phys. Today, 56, 3844.

  • Aumann, H. H., and Coauthors, 2003: AIRS/AMSU/HSB on the Aqua mission: Design, science objectives, data products, and processing systems. IEEE Trans. Geosci. Remote Sens., 41, 253264.

    • Search Google Scholar
    • Export Citation
  • Biggerstaff, M. I., , and Seo E.-K. , 2010: An EOF-based comparison and evaluation of simulated passive microwave signatures to observations over tropical oceans. J. Geophys. Res., 115, D15209, doi:10.1029/2009JD013029.

    • Search Google Scholar
    • Export Citation
  • Clerbaux, C., and Coauthors, 2009: Monitoring of atmospheric composition using the thermal infrared IASI/MetOp sounder. Atmos. Chem. Phys., 9, 60416054.

    • Search Google Scholar
    • Export Citation
  • Clough, S. A., , Shephard M. W. , , Mlawer E. J. , , Delamere J. S. , , Iacono M. J. , , Cady-Pereira K. , , Boukabara S. , , and Brown P. D. , 2005: Atmospheric radiative transfer modeling: A summary of the AER codes. J. Quant. Spectrosc. Radiat. Transfer, 91, 233244.

    • Search Google Scholar
    • Export Citation
  • Divakarla, M. G., , Barnet C. D. , , Goldberg M. D. , , McMillin L. M. , , Maddy E. , , Wolf W. , , Zhou L. , , and Liu X. , 2006: Validation of Atmospheric Infrared Sounder temperature and water vapor retrievals with matched radiosonde measurements and forecasts. J. Geophys. Res., 111, D09S15, doi:10.1029/2005JD006116.

    • Search Google Scholar
    • Export Citation
  • Feltz, W. F., , Smith W. L. , , Knuteson R. O. , , Revercomb H. E. , , Woolf H. M. , , and Howell H. B. , 1998: Meteorological applications of temperature and water vapor retrievals from the ground-based Atmospheric Emitted Radiance Interferometer (AERI). J. Appl. Meteor., 37, 857875.

    • Search Google Scholar
    • Export Citation
  • Feltz, W. F., , Howell H. B. , , Knuteson R. O. , , Woolf H. M. , , and Revercomb H. E. , 2003: Near-continuous profiling of temperature, moisture, and atmospheric stability using the Atmospheric Emitted Radiance Interferometer (AERI). J. Appl. Meteor., 42, 584597.

    • Search Google Scholar
    • Export Citation
  • Feltz, W. F., and Coauthors, 2007: Retrieving temperature and moisture profiles from AERI radiance observations: AERIPROF value-added product technical description. Revision 1, DOE ARM Tech. Rep. DOE/SC-ARM-TR-066.1, 31 pp.

  • Goldberg, M. D., , Qu Y. , , McMillin L. M. , , Wolf W. , , Zhou L. , , and Divakarla M. , 2003: AIRS near-real-time products and algorithms in support of operational numerical weather prediction. IEEE Trans. Geosci. Remote Sens., 41, 379389.

    • Search Google Scholar
    • Export Citation
  • Ha, J.-C., , Lee Y.-H. , , Lee J.-S. , , Lee H.-C. , , and Lee H.-S. , 2008: Development of short range analysis and prediction system. Proc. Ninth WRF Users' Workshop, Boulder, CO, NCAR, P9.8. [Available online at http://www.mmm.ucar.edu/wrf/users/workshops/WS2008/abstracts/p9-08.pdf.]

  • Ha, J.-C., , Lee J.-S. , , Lee Y.-H. & , and Lee H.-C. , 2011: The Korea local reanalysis. Proc. Fifth Korea–Japan–China Joint Conference on Meteorology, Busan, South Korea, Korean Meteorological Society, 163.

  • Hilton, F., and Coauthors, 2012: Hyperspectral earth observation from IASI: Five years of accomplishments. Bull. Amer. Meteor. Soc., 93, 347370.

    • Search Google Scholar
    • Export Citation
  • Huang, H.-L., , and Antonelli P. , 2001: Application of principal component analysis to high-resolution infrared measurement compression and retrieval. J. Appl. Meteor., 40, 365388.

    • Search Google Scholar
    • Export Citation
  • Immler, F., , Dykema J. , , Gardiner T. , , Whiteman D. N. , , Thorne P. W. , , and Vömel H. , 2010: A guide for upper-air reference measurements. Atmos. Meas. Tech. Discuss., 3, 18071842.

    • Search Google Scholar
    • Export Citation
  • Jolliffe, I. T., 1986: Principal Component Analysis. Springer-Verlag, 217 pp.

  • Kidder, S. Q., , and Vonder Haar T. H. , 1995: Satellite Meteorology: An Introduction. Academic Press, 466 pp.

  • Knuteson, R. O., and Coauthors, 2004a: Atmospheric Emitted Radiance Interferometer. Part I: Instrument design. J. Atmos. Oceanic Technol., 21, 17631776.

    • Search Google Scholar
    • Export Citation
  • Knuteson, R. O., and Coauthors, 2004b: Atmospheric Emitted Radiance Interferometer. Part II: Instrument performance. J. Atmos. Oceanic Technol., 21, 17771789.

    • Search Google Scholar
    • Export Citation
  • Koch, S. E., , Feltz W. , , Fabry F. , , Pagowski M. , , Geerts B. , , Bedka K. M. , , Miller D. O. , , and Wilson J. W. , 2008: Turbulent mixing processes in atmospheric bores and solitary waves deduced from profiling systems and numerical simulation. Mon. Wea. Rev., 136, 13731400.

    • Search Google Scholar
    • Export Citation
  • KMA, 2010: Report of global atmosphere watch 2010. Korea Meteorological Administration Rep. 111360000-000084-10, 239 pp.

  • Liljegren, J., , Lesht B. , , Kato S. , , and Clothiaux E. , 2001: Initial evaluation of profiles of temperature, water vapor, and cloud liquid water from a new microwave profiling radiometer. Proc. 11th Atmospheric Radiation Measurement (ARM) Program Science Team Meeting, Atlanta, GA, U.S. Department of Energy. [Available online at http://radiometrics.siteoperations.com/wp-content/uploads/2012/11/MWRP_ARM01.pdf.]

  • Mariani, Z., and Coauthors, 2011: Infrared emission measurements in the Arctic using a new extended-range AERI. Atmos. Meas. Tech. Discuss., 4, 64116448.

    • 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, seagoing infrared spectroradiometer. J. Atmos. Oceanic Technol., 18, 9941013.

    • Search Google Scholar
    • Export Citation
  • Pougatchev, N., and Coauthors, 2009: IASI temperature and water vapor retrievals—Error assessment and validation. Atmos. Chem. Phys., 9, 64536458.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., 2000: Inverse Methods for Atmospheric Sounding: Theory and Practice. World Scientific Publishing Co. Ltd., 240 pp.

  • Rowe, P. M., , Miloshevich L. M. , , Turner D. D. , , and Walden V. P. , 2008: Dry bias in Vaisala RS90 radiosonde humidity profiles over Antarctica. J. Atmos. Oceanic Technol., 25, 15291541.

    • Search Google Scholar
    • Export Citation
  • Schlüssel, P., , Hultberg T. H. , , Phillips P. L. , , August T. , , and Calbet X. , 2005: The operational IASI level 2 processor. Adv. Space Res., 36, 982988.

    • Search Google Scholar
    • Export Citation
  • Smith, W. L., , and Woolf H. M. , 1976: The use of eigenvectors of statistical covariance matrices for interpreting satellite sounding radiometer observations. J. Atmos. Sci., 33, 11271140.

    • Search Google Scholar
    • Export Citation
  • Smith, W. L., , Feltz W. F. , , Knuteson R. O. , , Revercomb H. E. , , Woolf H. M. , , and Howell H. B. , 1999: The retrieval of planetary boundary layer structure using ground-based infrared spectral radiance measurements. J. Atmos. Oceanic Technol., 16, 323333.

    • Search Google Scholar
    • Export Citation
  • Stokes, G. M., , and Schwartz S. E. , 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the Cloud and Radiation Testbed. Bull. Amer. Meteor. Soc., 75, 12011221.

    • Search Google Scholar
    • Export Citation
  • Szczodrak, M., , Minnett P. J. , , Nalli N. R. , , and Feltz W. F. , 2007: Profiling the lower troposphere over the ocean with infrared hyperspectral measurements of the Marine-Atmosphere Emitted Radiance Interferometer. J. Atmos. Oceanic Technol., 24, 390402.

    • Search Google Scholar
    • Export Citation
  • Tobin, D. C., and Coauthors, 2006: Atmospheric Radiation Measurement site atmospheric state best estimates for Atmospheric Infrared Sounder temperature and water vapor retrieval validation. J. Geophys. Res., 111, D09S14, doi:10.1029/2005JD006103.

    • Search Google Scholar
    • Export Citation
  • Turner, D. D., , Ackerman S. A. , , Baum B. A. , , Revercomb H. E. , , and Yang P. , 2003: Cloud phase determination using ground-based AERI observations at SHEBA. J. Appl. Meteor., 42, 701715.

    • Search Google Scholar
    • Export Citation
  • Turner, D. D., , Knuteson R. O. , , and Revercomb H. E. , 2006: Noise reduction of Atmospheric Emitted Radiance Interferometer (AERI) observations using principal component analysis. J. Atmos. Oceanic Technol., 23, 12231238.

    • Search Google Scholar
    • Export Citation
  • Walden, V. P., , Town M. S. , , Halter B. , , and Storey J. W. V. , 2005: First measurements of the infrared sky brightness at Dome C, Antarctica. Publ. Astron. Soc. Pac., 117, 300308.

    • Search Google Scholar
    • Export Citation
  • Yurganov, L., , McMillan W. , , Wilson C. , , Fischer M. , , and Biraud S. , 2010: Carbon monoxide mixing ratios over Oklahoma between 2002 and 2009 retrieved from Atmospheric Emitted Radiance Interferometer spectra. Atmos. Meas. Tech. Discuss., 3, 12631301.

    • Search Google Scholar
    • Export Citation
  • Zhou, D. K., and Coauthors, 2002: Thermodynamic product retrieval methodology and validation for NAST-I. Appl. Opt., 41, 69576967.

  • Zhou, D. K., , Smith W. L. , , Larar A. M. , , Liu X. , , Taylor J. P. , , Schlüssel P. , , Strow L. L. , , and Mango S. A. , 2009: All weather IASI single field-of-view retrievals: Case study—Validation with JAIVEx data. Atmos. Chem. Phys., 9, 22412255.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) Installed AERI system in the container house. AERI has been operated at (b) Anmyeon-do since June 2010. Field observations were conducted at Anmyeon-do to assess the AERI T/q retrieval performance.

  • View in gallery

    Downwelling radiances from 500 to 3000 cm−1 measured by AERI at 1259 UTC 26 May 2010 (black) and 1809 UTC 23 Mar 2011 (gray). AERIPROF uses 4.3- and 15-μm CO2 bands (light gray bars) for temperature, and 6.6- and 18-μm H2O bands (dark gray bars) for moisture retrieval. Air temperature and amount of precipitable water vapor were 290.8 K and 2.02 cm, respectively, in the May case; and 276.7 K and 0.38 cm, respectively, in the March case.

  • View in gallery

    Comparison of the modified (solid red) and the original (dashed blue) AERI bias spectra. Modified bias spectrum is the mean spectral difference between observed spectrum from the AERI and the simulated spectrum calculated from 25 radiosonde measurements under clear-sky conditions at Anmyeon-do.

  • View in gallery

    Training (a) temperature and (b) mixing ratio profiles, and (c) brightness temperature spectra computed from the (a) and (b) profiles. Radiosonde training profiles (700 profiles) under clear skies from 2009 to 2010 were taken from the Osan site, which is the closest radiosonde station to Anmyeon-do. For each training profile, the brightness temperature spectrum was simulated through LBLRTM version 12.0.

  • View in gallery

    Comparison of temperature (a) RMSE and (b) bias profiles for the optimized (solid black line) and the original (dotted gray line) statistical regressions. Radiosonde measurements (25 samples) on clear-sky days were used for the evaluation during the field experiments at Anmyeon-do.

  • View in gallery

    As in Fig. 5, but for the mixing ratio.

  • View in gallery

    (a) Temperature and (b) moisture profiles of temporally coincident radiosondes at Anmyeon-do (thick solid lines) and Osan (thin solid lines) during the field observation. PC coefficients for Anmyeon-do (dark gray circle), for Osan (dark gray star), and for the training dataset (gray dot) are displayed in the PC space of the training dataset.

  • View in gallery

    (a) RMSE and (b) bias profiles for temperature between radiosonde measurement and the AERI retrieval (solid black lines), the KLAPS profile (dotted gray lines), the AIRS retrieval (dashed gray lines), and the IASI retrieval (solid gray lines). Radiosonde measurements on clear-sky days were obtained from the field experiments at Anmyeon-do in 2010–2011.

  • View in gallery

    As in Fig. 8, but for the mixing ratio.

  • View in gallery

    Comparison of (a) temperature and (b) mixing ratio profiles at (left) 0431 and (right) 1244 UTC 6 Oct 2011. Radiosondes (solid gray lines) are compared with the optimized AERI retrievals (solid black lines), the KLAPS (dashed gray lines), the AIRS (dotted gray lines), and the IASI (dotted gray lines).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 24 24 3
PDF Downloads 17 17 0

Improvement of AERI T/q Retrievals and Their Validation at Anmyeon-Do, South Korea

View More View Less
  • 1 Global Environmental System Division, National Institute of Meteorological Research, Korea Meteorological Administration, Seoul, South Korea
© Get Permissions
Full access

Abstract

An Atmospheric Emitted Radiance Interferometer (AERI), which measures downwelling radiances, has been in operation at Anmyeon-do, South Korea, since June 2010. Temperature and moisture (T/q) profiles with high temporal and vertical resolution can be retrieved from the measured AERI spectrum through the retrieval algorithm AERIPROF. In this work, AERIPROF has been optimized to improve the retrieval performance: 1) a bias spectrum was computed from the coincident radiosondes during the field experiments at Anmyeon-do and 2) regression coefficients were obtained from local radiosondes and associated simulated spectral radiances. An evaluation was performed in the lower troposphere (<700 hPa) with the radiosondes on clear-sky days during the field experiments at Anmyeon-do. The optimized statistical regression results in an improvement of ~0.6 K for temperature and ~0.6 g kg−1 for the mixing ratio on average, in comparison to the original statistical regression. In addition, the optimized AERI T/q retrievals are compared with the satellite [Aqua/Atmospheric Infrared Sounder (AIRS), Meteorological Operation (MetOp)/Infrared Atmospheric Sounding Interferometer (IASI)] T/q retrievals as well as with T/q profiles analyzed from the regional NWP model, the Korea Local Analysis and Prediction System (KLAPS) analysis. The RMS errors of the AERI retrievals are smaller than those of the satellite retrievals (the KLAPS analysis) by ~1.3 K (~0.2 K) for temperature and ~0.3 g kg−1 (~0.2 g kg−1) for the mixing ratio on average. Significant differences could be found between the AERI retrievals with the KLAPS and the satellite retrievals. The local climatic condition seems to be an important factor to bring about this improvement. Considering the training dataset made with spatially distant radiosondes, this is a significant finding. The AERI could bring new information about the lower troposphere.

Current affiliation: Korea Institute of Atmospheric Prediction Systems, Seoul, South Korea.

Corresponding author address: Mi-Lim Ou, National Institute of Meteorological Research, 45 Gisangcheong-gil, Dongjak-gu, Seoul 156-720, South Korea. E-mail: milim@korea.kr

Abstract

An Atmospheric Emitted Radiance Interferometer (AERI), which measures downwelling radiances, has been in operation at Anmyeon-do, South Korea, since June 2010. Temperature and moisture (T/q) profiles with high temporal and vertical resolution can be retrieved from the measured AERI spectrum through the retrieval algorithm AERIPROF. In this work, AERIPROF has been optimized to improve the retrieval performance: 1) a bias spectrum was computed from the coincident radiosondes during the field experiments at Anmyeon-do and 2) regression coefficients were obtained from local radiosondes and associated simulated spectral radiances. An evaluation was performed in the lower troposphere (<700 hPa) with the radiosondes on clear-sky days during the field experiments at Anmyeon-do. The optimized statistical regression results in an improvement of ~0.6 K for temperature and ~0.6 g kg−1 for the mixing ratio on average, in comparison to the original statistical regression. In addition, the optimized AERI T/q retrievals are compared with the satellite [Aqua/Atmospheric Infrared Sounder (AIRS), Meteorological Operation (MetOp)/Infrared Atmospheric Sounding Interferometer (IASI)] T/q retrievals as well as with T/q profiles analyzed from the regional NWP model, the Korea Local Analysis and Prediction System (KLAPS) analysis. The RMS errors of the AERI retrievals are smaller than those of the satellite retrievals (the KLAPS analysis) by ~1.3 K (~0.2 K) for temperature and ~0.3 g kg−1 (~0.2 g kg−1) for the mixing ratio on average. Significant differences could be found between the AERI retrievals with the KLAPS and the satellite retrievals. The local climatic condition seems to be an important factor to bring about this improvement. Considering the training dataset made with spatially distant radiosondes, this is a significant finding. The AERI could bring new information about the lower troposphere.

Current affiliation: Korea Institute of Atmospheric Prediction Systems, Seoul, South Korea.

Corresponding author address: Mi-Lim Ou, National Institute of Meteorological Research, 45 Gisangcheong-gil, Dongjak-gu, Seoul 156-720, South Korea. E-mail: milim@korea.kr

1. Introduction

Radiosonde measurements are important sources of data for climate studies and numerical weather prediction. However, because of high costs and the low temporal resolution of radiosondes, there has been much effort to retrieve temperature and moisture profiles from remote sensing instruments with high-temporal resolution instead of radiosondes. Vertical resolving power of temperature and moisture requires high-spectral resolution measurements. This has led to the development of hyperspectral sensors, which make it possible to acquire profiles with high vertical resolution (Smith et al. 1999; Feltz et al. 2003; Goldberg et al. 2003; Zhou et al. 2009; Hilton et al. 2012).

Satellite-based remote sensors offer a unique way to measure atmospheric structure on a global scale. However, the accuracy of vertical profiles of temperature and moisture is reduced near the earth's surface. Ground-based remote sensing instruments could compensate for the drawback of satellite-based retrievals (Liljegren et al. 2001; Feltz et al. 1998).

The Atmospheric Emitted Radiance Interferometer (AERI) was developed by the Space Science and Engineering Center (SSEC) at the University of Wisconsin–Madison. The AERI system has approximately 5000 spectral channels with 0.5 cm−1 spectral resolution, from which we can retrieve temperature and moisture soundings with high vertical resolution from the surface to 3 km (Smith et al. 1999; Feltz et al. 2003). The AERI is deployed at sites of the Department of Energy's Atmospheric Radiation Measurement (ARM) Program (Stokes and Schwartz 1994; Ackerman and Stokes 2003), in the polar region (Walden et al. 2005; Rowe et al. 2008; Mariani et al. 2011) and on a research vessel (Minnett et al. 2001; Szczodrak et al. 2007).

The AERI system has been installed at Anmyeon-do (36.53°N, 126.33°E), South Korea, and operated since June 2010 to measure downwelling radiances and retrieve atmospheric environmental information, such as temperature and moisture profiles, and a total column concentration of trace gases (Fig. 1a).

Fig. 1.
Fig. 1.

(a) Installed AERI system in the container house. AERI has been operated at (b) Anmyeon-do since June 2010. Field observations were conducted at Anmyeon-do to assess the AERI T/q retrieval performance.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

AERIPROF is the algorithm used for retrieving temperature and moisture (T/q) profiles (Smith et al. 1999; Feltz et al. 2007). Because the algorithm assumes that perturbations of T/q profiles and that their radiances are linearly correlated, it requires an accurate initial (or first guess) profile.

In this paper, we optimized the AERIPROF algorithm to improve the AERI retrievals: 1) a static bias spectrum of the AERI was substituted with that computed from radiosonde observations collected during field experiments and 2) regression coefficients were derived from local radiosonde data for a 2-yr period and associated spectra. However, our statistical regression is significantly different from Feltz et al. (2007). We have retrieved a logarithmic mixing ratio instead of relative humidity. In addition, we made our training dataset with spatially noncoincident radiosondes at a nearby weather station. To assess the sounding ability of the AERI, the optimized AERI retrievals are validated against coincident radiosonde measurements and compared with the satellite-based retrievals from the Aqua/Atmospheric Infrared Sounder (AIRS) and the Meteorological Operation (MetOp)/Infrared Atmospheric Sounding Interferometer (IASI), and with the profiles from the regional NWP, the Korea Local Analysis and Prediction System (KLAPS) analysis, in the lower troposphere (<700 hPa).

2. Instrument and data

a. Site and radiosonde observations

The observation site, Anmyeon-do (36.53°N, 126.33°E), is the sixth largest island in South Korea (87.96 km2) and is located on the west coast of the Chungnam province (Fig. 1b). The annual average temperature for 2010 was 12.3°C, where the lowest (highest) monthly mean temperature was −1.2°C (26.8°C) in January (August). Annual precipitation for 2010 was 1444 mm, and the amount of rainfall in the summer was 765.5 mm (KMA 2010).

Field observations (26–27 May 2010, 3–4 November 2010, 23–24 March 2011, 17–18 May 2011, and 6–7 October 2011) were conducted 5 times by the National Institute of Meteorological Research (NIMR). The primary goal of the field observations was to evaluate the T/q retrieval performance of the AERI.

In situ measurements of vertical profiles of temperature and moisture were obtained by using commercial DFM-06 radiosondes manufactured by Graw in Germany. Typical accuracies of the radiosonde measurements are ±0.2 K for temperature and ±5% for relative humidity (Immler et al. 2010). The radiosondes were launched 4 times per day when the satellites (Aqua/AIRS and MetOp/IASI) passed over the site. A total of 25 radiosonde measurements on clear-sky days were used for the assessment.

b. Atmospheric Emitted Radiance Interferometer at Anmyeon-do

The AERI is a fully automated interferometer that measures the downwelling radiance from 3.3 to 19 μm (520–3020 cm−1) at better than 1 cm−1 resolution. Since the AERI spectrum covers CO2 bands (612–618, 624–660, 674–713, and 2223–2260 cm−1), H2O bands (538–588 and 1250–1350 cm−1), O3 bands (980–1080 cm−1), and the atmospheric window (800–1250 cm−1), it can be used to investigate the vertical distribution of temperature and moisture in the boundary layer (Smith et al. 1999; Feltz et al. 2003), to retrieve cloud properties (Turner et al. 2003) and carbon monoxide (Yurganov et al. 2010), and for other applications (Minnett et al. 2001; Szczodrak et al. 2007; Koch et al. 2008).

The AERI measures interferograms with two detectors (520–1800 and 1800–3020 cm−1). The radiance spectrum is obtained from an interferogram by taking the Fourier transform. The instrument has hot (333 K) and ambient blackbodies for calibration. A typical measurement cycle consists of a 3-min up-looking view to achieve a sufficient signal-to-noise ratio, followed by viewing the two calibration blackbodies for 2 min. Calibrated sky radiances are produced with an absolute calibration accuracy of better than 1% of the ambient radiance about every 8 min. Knuteson et al. (2004a,b) provide more details on the instrument's design, calibration, and validation of radiance.

Figure 2 shows the apodized spectra measured by the AERI on 26 May 2010 (case 1) and on 23 March 2011 (case 2). The air temperature at ground level in case 1 (290.8 K) was larger than in case 2 (276.7 K). Precipitable water vapor (PWV) in case 1 (2.02 cm) was higher than in case 2 (0.38 cm). The CO2 bands are sensitive to a temperature sounding. The radiances in the CO2 bands are higher in case 1 than in case 2. The transparency of the H2O bands and the atmospheric window is sensitive to the column water amount. Especially, the AERI spectrum at the 18-μm band (538–588 cm−1) is sensitive to emission from the wings of absorption lines in the water vapor rotational band (Clough et al. 2005). The transparency of the H2O bands and the atmospheric window bands in case 1 is reduced in comparison to case 2. The two measured spectra show good agreement with the observed temperature and PWV. The AERI retrieval algorithm of temperature and moisture, AERIPROF, is described in chapter 3.

Fig. 2.
Fig. 2.

Downwelling radiances from 500 to 3000 cm−1 measured by AERI at 1259 UTC 26 May 2010 (black) and 1809 UTC 23 Mar 2011 (gray). AERIPROF uses 4.3- and 15-μm CO2 bands (light gray bars) for temperature, and 6.6- and 18-μm H2O bands (dark gray bars) for moisture retrieval. Air temperature and amount of precipitable water vapor were 290.8 K and 2.02 cm, respectively, in the May case; and 276.7 K and 0.38 cm, respectively, in the March case.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

c. KLAPS analysis data

The KLAPS analysis data were obtained from the NIMR in the Korea Meteorological Administration. The KLAPS analysis product describes the atmospheric state in the region surrounding the Korean Peninsula with a spatial resolution of 5 km and a temporal resolution of 1 h. Vertical profiles are given at 22 levels (1100–50 hPa). Ha et al. (2008) and Ha et al. (2011) offer a detailed description of the KLAPS analysis data. The closest grid data to the location of the AERI system were chosen for an initial guess profile of the AERIPROF algorithm.

d. Satellite-based retrievals (Aqua/AIRS and MetOp/IASI)

The AIRS is a grating spectrometer on the Aqua satellite of the Earth Observing System (EOS). The AIRS covers the infrared bands from 650 to 2700 cm−1 (3.7–15.4 μm) with 2378 spectral channels. It scans with 1.1° footprints at nadir and a ±49.5° swath every 2 s. Global coverage is achieved twice per day (Aumann et al. 2003).

The AIRS level 2 product provides many geophysical parameters, including temperature and moisture profiles, surface temperatures, and trace gases. The required retrieval performance is 1 K in 1-km vertical layers for temperature and 20% in 2-km vertical layers for moisture below 100 hPa (Aumann et al. 2003). Tobin et al. (2006) report that AIRS retrievals for the tropical ocean site satisfy the required accuracy, but those for the midlatitude land site have poor performance, where the RMS error of temperature ranges from 1 to 2 K and of water vapor ranges from 25% to 35%. Divakarla et al. (2006) show that AIRS retrieval accuracies are close to the expected product goal accuracies, and the retrieval accuracy over land is degraded in comparison to over sea. The degradation of the accuracy may arise from the daytime convection, the heterogeneity of the land surface and its emissivity, and the uncertainty of surface pressure over the land (Divakarla et al. 2006). The temperature and moisture products of AIRS are available from the National Aeronautics and Space Administration (NASA) website (http://disc.sci.gsfc.nasa.gov/AIRS).

The IASI is a Fourier transform spectrometer on the MetOp satellite, forming part of the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) Polar System (EPS) since 2006. The IASI has 8461 channels in infrared bands from 645 to 2760 cm−1 (15.5–3.62 μm) at 0.35–0.5 cm−1 spectral resolution. The field of view (3.3° × 3.3°) is composed of 2 × 2 circular pixels, each corresponding to a 12-km-diameter footprint on the ground at nadir. The IASI observes the earth twice a day with an angle of ±48.3° (Clerbaux et al. 2009). The IASI level 2 product provides high-resolution vertical profiles of temperature and humidity, columnar amounts of trace gases, surface temperatures and emissivities, and cloud properties. The required accuracy of temperature and moisture retrievals is 1 K and 10% in the lower atmosphere in clear-sky conditions (Schlüssel et al. 2005). Pougatchev et al. (2009) assessed the retrieval performance, where the temperature RMS error is ~0.6 K between 800 and 300 hPa with an increase to ~2 K at the surface, and the humidity RMS error is ~10% between 800 and 300 hPa. The temperature and moisture products of the IASI are available from the National Oceanic and Atmospheric Administration (NOAA) website (http://www.class.ngdc.noaa.gov/saa/products/welcome).

3. Optimization of AERI T/q retrievals

a. AERI T/q retrievals before optimization

The method of retrieving temperature and moisture profiles is the AERIPROF algorithm developed by the SSEC at the University of Wisconsin–Madison. The retrieval accuracy for temperature is better than 1 K, and the water vapor retrieval accuracy is approximately 5% from the surface to 3 km in the Southern Great Plain (SGP) in the United States (Feltz et al. 2007). The algorithm uses 249 channels in CO2 bands (612–618, 624–660, 674–713, and 2223–2260 cm−1) and 313 channels in H2O bands (538–588 and 1250–1350 cm−1) for retrieving temperature and moisture, respectively (Smith et al. 1999; Feltz et al. 2003).

The AERIPROF algorithm consists of two steps: 1) an initial (first) guess profile is obtained by a linear statistical regression based on local radiosondes and numerical weather prediction (NWP) and 2) physical retrieval is achieved by iteration of radiative transfer calculation to yield a final solution that best fits the radiance observation. Since the retrieval problem is ill posed, it is necessary to have additional information to constrain the retrievals (Rodgers 2000). Therefore, it is essential to increase the accuracy of an initial (or first) guess profile as a proper constraint. To construct a physically reasonable initial guess profile, we used the original statistical regression (Feltz et al. 2007), the KLAPS analysis data, and automated weather station (AWS) data. Before the regression, the original static bias spectrum (Feltz et al. 2007) was subtracted from observed spectra. Then, original regressions were made with the regression coefficients (Feltz et al. 2007), and the KLAPS profiles were blended with the regressions. The “onion peeling” method (Smith et al. 1999) is used for the physical retrieval, where the first-guess profile is first modified at the surface and then changes are made progressively with height in the atmosphere. These adjustments are made to minimize the differences between the observed and the calculated spectrum.

Since the regression coefficients and the bias spectrum are site dependent, we need to modify the original statistical regression and bias spectrum. In this study, we optimized the AERIPROF algorithm by replacing the original static bias spectrum of the AERI with the modified one and deriving new regression coefficients from local radiosonde data for 2 yr and associated spectra.

b. Bias correction

Because of inaccuracies in current knowledge on spectroscopic parameters, instrumentation error, and forward model error, a simulated spectrum is not the same as its observed spectrum. Simulated spectra under clear-sky conditions, however, can be considered as true spectra, since the physics of radiative transfer under a clear sky is well known. Therefore, the difference between a measured and calculated spectrum under clear-sky conditions can be regarded as a bias because of measurement error. A bias correction is a simple and effective method to minimize the difference.

Since the original bias spectrum used for AERIPROF did not work properly at the Anmyeon-do site, a new bias spectrum was calculated by using the coincident radiosonde measurements (25 samples) obtained under clear-sky conditions at Anmyeon-do for the field observation period and the line-by-line radiative transfer model (LBLRTM; Clough et al. 2005) to compute their downwelling spectrum. The mean difference between the observed and simulated spectrum is defined as the modified bias spectrum (Feltz et al. 2007). This static bias spectrum is subtracted from the observed spectrum before regression in AERIPROF. The modified bias spectrum has a wider range of fluctuation in the strong water vapor band (1400–1800 cm−1) and higher radiance in CO2 bands than the original bias spectrum (Fig. 3).

Fig. 3.
Fig. 3.

Comparison of the modified (solid red) and the original (dashed blue) AERI bias spectra. Modified bias spectrum is the mean spectral difference between observed spectrum from the AERI and the simulated spectrum calculated from 25 radiosonde measurements under clear-sky conditions at Anmyeon-do.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

c. Modified statistical regression

The AERIPROF algorithm adopts a statistical regression method to construct a first-guess profile because a statistical regression has two key advantages as an initial guess profile. First, it can reflect local characteristics, since they resemble the statistical structure of the training dataset. Second, it does not require any knowledge of the transmittances or the use of the radiative transfer model, which leads to a low computational cost (Kidder and Vonder Haar 1995).

Given a set of historical radiosondes and their simulated brightness temperatures, the relationship between thermodynamic structures and associated spectra is explained by statistical regression coefficients. In hyperspectral brightness temperatures, however, the large correlation between subsets of channels causes multicollinearity (Huang and Antonelli 2001). If multicollinearities exist, then regression coefficients can lead to unstable and untolerable retrievals (Jolliffe 1986). Though one advantage of AERI is high spectral resolution, it suffers from multicollinearity.

To alleviate this problem, the AERIPROF algorithm uses principal component regression (PCR). PCR reduces the multicollinearities in hyperspectral channels (Huang and Antonelli 2001) by projecting high-spectral-resolution brightness temperatures on a subset of the principal components (PCs; eigenvectors), which are sufficient to explain most of the variance in the spectra. Because the PCs are linearly uncorrelated, the projected coefficients are independent of each other (Huang and Antonelli 2001; Goldberg et al. 2003). The PCs are generated by decomposing the covariance matrix of the brightness temperatures. The covariance matrix Ω of the spectra are expressed as
eq1
where Δr(k) is the anomalies of brightness temperature of sample k and m is the total number of the samples. The matrices of the principal components U are calculated by singular value decomposition as
eq2
where is a diagonal matrix of eigenvalues, which are ordered from the largest amount of the variance to the lowest. Each eigenvalue represents how much its eigenvector (PC) explains the variance of data. If some PCs with low variance are truncated, then the dimensionality of the data is effectively reduced (Huang and Antonelli 2001) and the signal-to-noise ratio increases by removing uncorrelated random error in measurements (Turner et al. 2006). Reliable regression can be achieved with the projected coefficients on the remaining PCs (Smith and Woolf 1976; Zhou et al. 2002).

The regression coefficients in AERIPROF do not represent the climatology in our region, since the training dataset was chosen from the SGP site (Feltz et al. 2007). Therefore, new regression coefficients were derived from our local radiosondes. The modified regression coefficients for temperature (mixing ratio) were made from the training temperature (mixing ratio) profiles and their brightness temperatures in CO2 (H2O) bands with PCR.

Radiosonde training profiles (700 profiles) under clear sky from 2009 to 2010 were taken from the Osan site (37.10°N, 127.03°E), which is the closest radiosonde station to Anmyeon-do and is located 100 km to the northeast (Figs. 4a,b). A relative humidity threshold scheme is applied to find clear-sky conditions. When the relative humidity was above 80% throughout the entire profile, the radiosondes were filtered out. For each profile, a Beer-apodized spectrum was produced and converted to a brightness temperature spectrum through LBLRTM version 12.0, with the spectral resolution of 0.482 147 22 cm−1 and a maximum optical path difference of 1.037 03 cm of the AERI (Fig. 4c). The profiles of O3, CO2, and other gases were adopted from the U.S. Standard Atmosphere, 1976 for the simulation. In section 3e, the similarity of T/q variance structures between the AERI site (Anmyeon-do) and the radiosonde station (Osan) is examined.

Fig. 4.
Fig. 4.

Training (a) temperature and (b) mixing ratio profiles, and (c) brightness temperature spectra computed from the (a) and (b) profiles. Radiosonde training profiles (700 profiles) under clear skies from 2009 to 2010 were taken from the Osan site, which is the closest radiosonde station to Anmyeon-do. For each training profile, the brightness temperature spectrum was simulated through LBLRTM version 12.0.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

A fundamental question in achieving an optimized regression is to determine the number of appropriate PCs of the training profiles and their spectrum. If the number of PCs is large, then the statistical regression would be affected by measurement noise, whereas if it is small, then the regression would lose the physical information associated with atmospheric structures. A sensitivity test was conducted to find out the optimal numbers of PCs for temperature and moisture retrievals from the AERI measurements. The modified bias spectrum was subtracted from the measured AERI spectrum before each regression. Since retrievals from AERI are valid from the surface to 3 km, the optimal number of PCs is defined as the number that minimizes the averaged RMS error of temperature (or moisture) between 1000 and 700 hPa by using the coincident radiosondes from the field experiments. For the optimal temperature (mixing ratio) retrievals, the numbers of PCs were 12 (22) for the temperature (mixing ratio) profiles and 10 (4) for the brightness temperature in the CO2 (H2O) bands. In section 3d, the modified statistical regressions were compared to the original statistical regression to see the effect of localization on regression.

In our statistical regression process for humidity, there was a significant change in the original algorithm (Feltz et al. 2007). Feltz et al. retrieve relative humidity and then convert it to mixing ratio, whereas we retrieve logarithmic mixing ratio and then recover it to mixing ratio. When relative humidity is retrieved, it could be higher than 100% or lower than 0%. In the original AERIPROF algorithm, the unreasonable values are corrected by setting 0% or 100%. Therefore, the retrieving logarithmic mixing ratio rather than relative humidity is attractive to avoid the unphysical situation. In addition, when the humidity retrievals were compared with radiosondes from the field observations, the retrieving logarithmic mixing ratio shows slight improvements over retrieving relative humidity (Table 1). The RMS errors of the retrieving logarithmic mixing ratio are smaller than retrieving relative humidity at most levels except 880–860 hPa. The averaged RMSE (bias) accuracy of the mixing ratio is reduced from 1.15 (0.36) to 0.92 (0.24) g kg−1, in the lower troposphere (<700 hPa). Significant differences could be found between the retrieving logarithmic mixing ratio and relative humidity from 1000 to 800 hPa at the 90% confidence level (t test is described in section 4).

Table 1.

Retrieving logarithmic mixing ratio was compared to retrieving relative humidity RH for humidity retrieval. Mean μ, standard deviation σ, and t test values were assessed. RMS errors and bias values were calculated by using the radiosondes during the field observations.

Table 1.

Once statistical regression process is completed, regressions are combined with the KLAPS to construct optimized first guess profiles. Then through the physical retrieval process, the final AERI retrievals are obtained. Their performances and the effect of blending with KLAPS are explained in section 4.

d. Evaluation of the optimized statistical retrievals

Two sets of the original and the optimized statistical retrievals were compared to assess the accuracies of temperature and moisture retrievals from the mean bias-corrected AERI spectrum. The evaluation was performed with the radiosonde mesurements (25 samples) on clear-sky days during the field experiments at Anmyeon-do. Figures 5 and 6 show the RMS errors and bias profiles between the radiosondes and the optimized (solid line) and the original (dashed line) regressions. The optimized statistical regressions show better performance than the original statistical regression in the lower troposphere (1000–700 hPa). The averaged RMSE accuracy of the retrieved temperature and mixing ratio has been reduced from 1.75 to 1.15 K and from 1.52 to 0.92 g kg−1, respectively, in the column (1000–700 hPa). The magnitude of the averaged bias of the retrievals has decreased from 1.17 to 0.66 K and from 0.90 to 0.24 g kg−1, respectively.

Fig. 5.
Fig. 5.

Comparison of temperature (a) RMSE and (b) bias profiles for the optimized (solid black line) and the original (dotted gray line) statistical regressions. Radiosonde measurements (25 samples) on clear-sky days were used for the evaluation during the field experiments at Anmyeon-do.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

Fig. 6.
Fig. 6.

As in Fig. 5, but for the mixing ratio.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

e. Evaluation of the training dataset

In statistical regression, the training dataset should represent the local climatology. Even though the radiosonde site (Osan) is 100 km away from the AERI site, the sites are in the westerlies and the clear-sky condition may exclude the situation where two sites would be under different air masses; the result is a climatological correlation between the air masses over the two sites. Even though the T/q variance structure at Anmyeon-do is not the same as that at Osan, a resemblance of two variance structures could exist. The similar variance structure can be retrieved well, while other structures cannot. To demonstrate the similarity of the temperature and moisture structures between the two sites, we calculated the RMS difference between them, and introduced a method based on PC analysis to visualize the analogy of the variance structures.

First, we compared the temperature and moisture profiles at the radiosonde site and the AERI site under clear-sky conditions for the period of the field observations at Anmyeon-do during 2010–2011. The radiosonde measurements at Osan were chosen within ±1.5 h from the radiosondes at Anmyeon-do. The RMS difference for temperature (moisture) varies from ~1.06 K (0.61 g kg−1) to 2.20 K (1.45 g kg−1) between 1000 and 700 hPa (Table 2). The averaged RMS difference was 1.34 K and 1.05 g kg−1, respectively, between 1000 and 700 hPa.

Table 2.

RMS difference (RMSD) and bias values from 1000 to 700 hPa between the radiosonde measurements at Anmyeon-do and those at Osan.

Table 2.

Second, we adapted the PC analysis (Biggerstaff and Seo 2010) to analyze the local climatology at Anmyeon-do across to Osan. Though the radiosonde measurements at Anmyeon-do are somewhat limited, we can examine whether there is similarity between the two sites, and whether the training dataset represents variance structures of the temperature and moisture profiles at Anmyeon-do for the period of the field experiments.

To assess the difference between the variance structures at the two sites, the anomalies of the temperature and moisture profiles at Anmyeon-do and Osan during the field observation were projected onto the first and the second PC (eigenvector) of the training dataset. If the temporally coincident profiles from each site are similar, then the PCs for Anmyeon-do and those for Osan would be close to each other in the scatterplot. This means that the two profiles are affected by similar air masses at a given time. To evaluate the training dataset, the distribution of the projected two PCs for Anmyeon-do is compared to that of the two PCs for the training dataset. By doing so, the variance structures at Anmyeon-do can be evaluated within those at the training dataset. This is an efficient measure of the similarity between Anmyeon-do and Osan with respect to the variation in the temperature and moisture structures.

Figure 7 shows the distribution of the two PCs (gray dot) of the training dataset in its PC space for temperature (Fig. 7a) and mixing ratio (Fig. 7b). The x axis corresponds to the first PC, which explains 90% (72%) of the total variability of the temperature (mixing ratio) profiles in the training dataset. The y axis corresponds to the second PC, which explains 3% (8%) of the total variability of the temperature (mixing ratio) profiles in the training dataset. Physical meaning of the first PC of the temperature (mixing ratio) profiles is related to temperature (mixing ratio) near 1000 hPa. The more the coefficient of the first PC increases, the more the temperature (mixing ratio) near 1000 hPa decreases.

Fig. 7.
Fig. 7.

(a) Temperature and (b) moisture profiles of temporally coincident radiosondes at Anmyeon-do (thick solid lines) and Osan (thin solid lines) during the field observation. PC coefficients for Anmyeon-do (dark gray circle), for Osan (dark gray star), and for the training dataset (gray dot) are displayed in the PC space of the training dataset.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

The dark gray circle (star) shows the distribution of the PC coefficients for Anmyeon-do (Osan). In Fig. 7a, the PCs at points C and D are almost overlapped, and the corresponding temperature profiles also illustrate the analogy between the temperature profiles at the AERI site and the radiosonde station. As for the points A and B, the PC coefficients for the AERI site are only slightly away from those for the radiosonde site, which explains the differences near 1000 hPa in the corresponding temperature profiles. Likewise, in Fig. 7b, the variance structures of the moisture profiles at the AERI site are very similar to those at the radiosonde site. The profiles corresponding to the points A, B, C, and D clearly demonstrate that the patterns of the trapped water vapor layer are well depicted at the two sites.

In terms of overlapping manifolds, most of the variance structures for the AERI site during the field observation seem to match those for the training dataset as well. However, mismatches still exist between the distributions for the AERI site and the training dataset. These mismatches could significantly affect retrievals. This is a limitation of the training dataset. To compensate for the restriction, blending with KLAPS is important. We used KLAPS from 850 to 50 hPa, where the statistical regression begins to increase uncertainties.

4. Retrieval results and validation

The AERI retrievals from the optimized AERIPROF algorithm described in section 3 have been compared with radiosondes, retrievals from the AIRS (the IASI), and the KLAPS model profiles under clear-sky conditions for the period of the field observations at Anmyeon-do during 2010–2011. The number of radiosondes collected throughout the field observations for validation is 25 for the AERI retrievals and the KLAPS, 10 for the AIRS, and 13 for the IASI. The criteria for the temporal and spatial collocation of the data are within ±1 h and ±2° from radiosonde. Since the AERI retrievals are valid from the surface to 3 km, the analysis was performed in the lower troposphere (1000–700 hPa).

Figures 8 and 9 show a comparison of the RMS errors and bias profiles of the temperature and the mixing ratio of the AERI retrievals, the KLAPS, and the retrievals from the AIRS and IASI. The RMS error of the AERI retrievals of temperature is less than 1.1 K throughout the entire column (1000–700 hPa). For the retrieved mixing ratio from the AERI, the RMS error is less than 1 g kg−1 from 1000 to 700 hPa. Compared with the KLAPS, the improvement in accuracies of the temperature profiles is relatively small, but the accuracies of the mixing ratio from the AERI retrievals are much improved in the region (950–850 hPa). The AERI accuracies are as low as 0.5 g kg−1 near 950 hPa, whereas the KLAPS accuracies are as high as 1.1 g kg−1 in the same column. Since the KLAPS is blended with the statistical regressions from 850 to 50 hPa, the improvement mainly originates from the local radiosonde statistics. This explains why local radiosonde statistics are needed for improving retrieval accuracy. The RMSE and bias profiles of AERI show a similar tendency to those of KLAPS from 850 to 700 hPa, which shows that the initial guess profiles are slightly modified though the physical process. For the retrievals from the IASI, the RMS error profile of the temperature ranges from ~1 to ~3 K near the surface and that of the mixing ratio is generally below ~1.0 g kg−1, with the exception of the levels near 850 hPa. The performance of the AIRS retrievals shows similar temperature RMSE accuracies in the column (1000–700 hPa) ranging from ~2 to ~2.5 K near the surface, and mixing ratio RMSE accuracies below ~1.0 g kg−1 in the entire column. Compared with the accuracies of the IASI and the AIRS retrievals, the RMS error of the AERI retrievals is reduced by more than 1 K for temperature and 0.2 g kg−1 for the mixing ratio.

Fig. 8.
Fig. 8.

(a) RMSE and (b) bias profiles for temperature between radiosonde measurement and the AERI retrieval (solid black lines), the KLAPS profile (dotted gray lines), the AIRS retrieval (dashed gray lines), and the IASI retrieval (solid gray lines). Radiosonde measurements on clear-sky days were obtained from the field experiments at Anmyeon-do in 2010–2011.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

Fig. 9.
Fig. 9.

As in Fig. 8, but for the mixing ratio.

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

The RMSE values have been averaged over the 1000–700 hPa range to summarize the evaluation of the accuracies of the retrievals from the AERI, IASI, and AIRS, and KLAPS (Table 3). Compared with the KLAPS, the AERI retrievals result in a slight accuracy improvement of ~0.2 K for temperature and ~0.2 g kg−1 for the mixing ratio. As for the comparison with the satellite-based retrievals, the AERI retrievals show a significant improvement of ~1.3 K for temperature and ~0.3 g kg−1 for the mixing ratio.

Table 3.

Averaged RMSE and bias values from 1000 to 700 hPa for the AERI, the KLAPS, the AIRS, and the IASI.

Table 3.

A statistical hypothesis test, the paired t test, was chosen to determine whether there were significant differences between the AERI retrievals with the following: the KLAPS, the AIRS, and the IASI. The t value is defined as
eq3
where and are the mean and standard deviation of the differences, and n is the number of samples. When n is larger than 30, the t value approaches a normal distribution. The hypothesis of difference between the KLAPS and the AERI retrievals is rejected if | t | < 1.711 at the 90% confidence level with 25 samples. The hypothesis is denied if | t | < 1.833 (1.782) for the AIRS (the IASI). The t statistics are included in Table 4. The AERI retrievals were interpolated to the grids for the KLAPS, the AIRS, and the IASI.
Table 4.

Mean μ, standard deviation σ, and t test values of paired the AERI retrievals and the KLAPS (the AIRS, the IASI) from 1000 to 700 hPa.

Table 4.

Inspection of t values shows that there were significant differences between the KLAPS and the AERI retrievals at the 90% level confidence. For temperature and the mixing ratio, the differences stood out from 1000 to 900 hPa. This is remarkable when we consider that the AERI statistical retrievals were combined with the KLAPS from 850 hPa. It seems that local radiosonde statistics can play a significant role to improve the AERI retrieval performance. For the satellite retrievals, the significant differences for temperature were noticeable near 1000 (986) hPa for the AIRS (the IASI). However, no significant differences for the mixing ratio could be found except the AIRS at 886.7 hPa. Maybe it seems that the number of the samples is not enough to tell the significant differences for the mixing ratio in the IASI. Nevertheless, it should be noted that there is a substantial difference in the temporal availability of the AERI and the IASI.

The radiosonde measurements in Anmyeon-do are used for the calculation of the bias spectrum and the validation of the AERI T/q retrievals. Even though the bias spectrum was calculated from the same radiosonde, the statistical regression coefficients were independently derived with the soundings at a nearby weather station. Furthermore, the AERI T/q retrievals went through a physical retrieval process. Therefore, in this sense, we can consider them as almost independent data and the error statistics are improved.

Samples of the optimized AERI retrievals are compared with the KLAPS and the satellite retrievals on 6 Oct 2011 (Fig. 10). The radiosondes were launched at 0431 (1244) UTC when AIRS (IASI) passed over the Anmyeon-do site. On 6 October 2011, early-morning mist patches occurred and gradually dissipated in the afternoon. The AERI temperature retrievals, in particular, show better performance than the satellite-based retrievals (Fig. 10a). At 0431 UTC, the water vapor distribution was captured only by the AERI retrieval. The water vapor layer trapped below 900 hPa at 1244 UTC was well described in both the AERI retrievals and the KLAPS (Fig. 10b).

Fig. 10.
Fig. 10.

Comparison of (a) temperature and (b) mixing ratio profiles at (left) 0431 and (right) 1244 UTC 6 Oct 2011. Radiosondes (solid gray lines) are compared with the optimized AERI retrievals (solid black lines), the KLAPS (dashed gray lines), the AIRS (dotted gray lines), and the IASI (dotted gray lines).

Citation: Journal of Atmospheric and Oceanic Technology 30, 7; 10.1175/JTECH-D-12-00029.1

5. Summary and discussion

The AERI system has been operational at Anmyeon-do, South Korea, since June 2010. From the measured downwelling radiances, temperature and moisture profiles can be retrieved with the AERIPROF algorithm. Since the statistical regression in the algorithm is site dependent, the original AERIPROF algorithm has been modified to optimize the retrievals. First, a new bias spectrum for the AERI was calculated from the coincident radiosondes and their simulated spectra during the field observation period at Anmyeon-do. Before regression, the modified bias spectrum was subtracted from observed radiances. Second, modified regression coefficients were derived from the local radiosonde statistics. In statistical regression for humidity, the retrieving logarithmic mixing ratio is more accurate than retrieving relative humidity by ~0.23 g kg−1 on average. The accuracy of the optimized statistical regressions for temperature and the mixing ratio has been increased by ~0.6 K and ~0.6 g kg−1, respectively, in the column (1000–700 hPa). Their mean bias has been reduced by ~0.51 K and ~0.66 g kg−1, respectively, in comparison to the original statistical regressions.

To evaluate the sounding ability of the optimized AERI retrievals, radiosondes from field experiments were used for validation under clear-sky days at Anmyeon-do. The optimized AERI retrievals were compared with radiosondes, the KLAPS, and the retrievals from MetOp/IASI and Aqua/AIRS in the lower troposphere (1000–700 hPa). The accuracy of the AERI retrievals shows improvements over the KLAPS and the satellite-based retrievals. The AERI retrievals for temperature and the mixing ratio are more accurate than the KLAPS by ~0.2 K and ~0.2 g kg−1, and the IASI (or the AIRS) by ~1.3 K and ~0.3 g kg−1, respectively, on average.

A series of paired t tests were performed to verify if there were any significant differences between the AERI retrievals with the following: the KLAPS, the AIRS, and the IASI. Significant differences existed between the AERI retrievals and the KLAPS in temperature from 950 to 900 hPa and the mixing ratio from 1000 to 900 hPa. Local climatic conditions seem to be an important factor to bring about this improvement. There were significant differences between the AERI retrievals and the satellite retrievals near 1000 hPa for temperature. A significant difference was found only in the AIRS at 886.7 hPa for the mixing ratio. These results suggest that AERI brought new information in the low troposphere. However, the mixing ratio differences between the AERI retrievals and the IASI were not statistically significant. Perhaps this is caused by the limited number of samples.

Considering the distance of about 100 km between the AERI site and the radiosonde station, these results are remarkable. This implies that the training dataset could be made with spatially distant radiosonde measurements as long as it has similar variance structures. As mentioned in Feltz et al. (2007), to get a good retrieval from the AERI spectrum, its bias spectrum should be corrected and regression coefficients be newly derived. To calculate the coefficients, a sufficient number of radiosonde measurements are needed to represent regional climatology. This could limit the installation locations for AERI. If radiosonde measurements are available, then AERI can be installed anywhere. Szczodrak et al. (2007) used model data as initial guess profiles, but they showed some model data did not catch the thermodynamic structure in the lower troposphere, since the vertical resolution of a model is relatively lower than radiosonde soundings. In this article, we successfully used local radiosonde measurements from an operational weather station, which is 100 km from the AERI system. This means that potential candidates for AERI installation could increase.

To extend the variance structures in the training dataset, we can use data from several nearby weather stations. Huang and Antonelli (2001) used radiosonde measurements in the midlatitudes to get general regression coefficients. In this sense, the present study is to be regarded as a starting point for the development of statistical regression for temperature and moisture in South Korea.

The AERI system installed in South Korea can be used to provide a reliable source of the T/q profile throughout the lower troposphere in clear-sky conditions. The optimized AERI retrievals may provide better representation of the atmospheric state of the lower troposphere than the model and the satellite-based retrievals. In addition, the high-temporal resolution of the fully automated AERI system can compensate for the low-temporal resolution and cumbersome operation of radiosonde measurements.

In the future, we can use other radiosonde measurements in South Korea to extend the variance structure of the training dataset and update the regression coefficients. For better representation of the atmospheric structure, data assimilation of AERI and IASI (or AIRS) could be included. Also, field experiments will be continued to increase confidence in the accuracies of the temperature and moisture profiles from the AERI.

Acknowledgments

This research was carried out as a part of the “Research for the Meteorological and Earthquake Observation Technology and its Application” research project supported by the National Institute of Meteorological Research at the Korea Meteorological Administration. The authors thank Bill Smith, Hyo-Jong Song, David D. Turner, Dong-kyun Kim, and Wayne Feltz for their helpful suggestions.

REFERENCES

  • Ackerman, T. P., , and Stokes G. M. , 2003: The Atmospheric Radiation Measurement Program. Phys. Today, 56, 3844.

  • Aumann, H. H., and Coauthors, 2003: AIRS/AMSU/HSB on the Aqua mission: Design, science objectives, data products, and processing systems. IEEE Trans. Geosci. Remote Sens., 41, 253264.

    • Search Google Scholar
    • Export Citation
  • Biggerstaff, M. I., , and Seo E.-K. , 2010: An EOF-based comparison and evaluation of simulated passive microwave signatures to observations over tropical oceans. J. Geophys. Res., 115, D15209, doi:10.1029/2009JD013029.

    • Search Google Scholar
    • Export Citation
  • Clerbaux, C., and Coauthors, 2009: Monitoring of atmospheric composition using the thermal infrared IASI/MetOp sounder. Atmos. Chem. Phys., 9, 60416054.

    • Search Google Scholar
    • Export Citation
  • Clough, S. A., , Shephard M. W. , , Mlawer E. J. , , Delamere J. S. , , Iacono M. J. , , Cady-Pereira K. , , Boukabara S. , , and Brown P. D. , 2005: Atmospheric radiative transfer modeling: A summary of the AER codes. J. Quant. Spectrosc. Radiat. Transfer, 91, 233244.

    • Search Google Scholar
    • Export Citation
  • Divakarla, M. G., , Barnet C. D. , , Goldberg M. D. , , McMillin L. M. , , Maddy E. , , Wolf W. , , Zhou L. , , and Liu X. , 2006: Validation of Atmospheric Infrared Sounder temperature and water vapor retrievals with matched radiosonde measurements and forecasts. J. Geophys. Res., 111, D09S15, doi:10.1029/2005JD006116.

    • Search Google Scholar
    • Export Citation
  • Feltz, W. F., , Smith W. L. , , Knuteson R. O. , , Revercomb H. E. , , Woolf H. M. , , and Howell H. B. , 1998: Meteorological applications of temperature and water vapor retrievals from the ground-based Atmospheric Emitted Radiance Interferometer (AERI). J. Appl. Meteor., 37, 857875.

    • Search Google Scholar
    • Export Citation
  • Feltz, W. F., , Howell H. B. , , Knuteson R. O. , , Woolf H. M. , , and Revercomb H. E. , 2003: Near-continuous profiling of temperature, moisture, and atmospheric stability using the Atmospheric Emitted Radiance Interferometer (AERI). J. Appl. Meteor., 42, 584597.

    • Search Google Scholar
    • Export Citation
  • Feltz, W. F., and Coauthors, 2007: Retrieving temperature and moisture profiles from AERI radiance observations: AERIPROF value-added product technical description. Revision 1, DOE ARM Tech. Rep. DOE/SC-ARM-TR-066.1, 31 pp.

  • Goldberg, M. D., , Qu Y. , , McMillin L. M. , , Wolf W. , , Zhou L. , , and Divakarla M. , 2003: AIRS near-real-time products and algorithms in support of operational numerical weather prediction. IEEE Trans. Geosci. Remote Sens., 41, 379389.

    • Search Google Scholar
    • Export Citation
  • Ha, J.-C., , Lee Y.-H. , , Lee J.-S. , , Lee H.-C. , , and Lee H.-S. , 2008: Development of short range analysis and prediction system. Proc. Ninth WRF Users' Workshop, Boulder, CO, NCAR, P9.8. [Available online at http://www.mmm.ucar.edu/wrf/users/workshops/WS2008/abstracts/p9-08.pdf.]

  • Ha, J.-C., , Lee J.-S. , , Lee Y.-H. & , and Lee H.-C. , 2011: The Korea local reanalysis. Proc. Fifth Korea–Japan–China Joint Conference on Meteorology, Busan, South Korea, Korean Meteorological Society, 163.

  • Hilton, F., and Coauthors, 2012: Hyperspectral earth observation from IASI: Five years of accomplishments. Bull. Amer. Meteor. Soc., 93, 347370.

    • Search Google Scholar
    • Export Citation
  • Huang, H.-L., , and Antonelli P. , 2001: Application of principal component analysis to high-resolution infrared measurement compression and retrieval. J. Appl. Meteor., 40, 365388.

    • Search Google Scholar
    • Export Citation
  • Immler, F., , Dykema J. , , Gardiner T. , , Whiteman D. N. , , Thorne P. W. , , and Vömel H. , 2010: A guide for upper-air reference measurements. Atmos. Meas. Tech. Discuss., 3, 18071842.

    • Search Google Scholar
    • Export Citation
  • Jolliffe, I. T., 1986: Principal Component Analysis. Springer-Verlag, 217 pp.

  • Kidder, S. Q., , and Vonder Haar T. H. , 1995: Satellite Meteorology: An Introduction. Academic Press, 466 pp.

  • Knuteson, R. O., and Coauthors, 2004a: Atmospheric Emitted Radiance Interferometer. Part I: Instrument design. J. Atmos. Oceanic Technol., 21, 17631776.

    • Search Google Scholar
    • Export Citation
  • Knuteson, R. O., and Coauthors, 2004b: Atmospheric Emitted Radiance Interferometer. Part II: Instrument performance. J. Atmos. Oceanic Technol., 21, 17771789.

    • Search Google Scholar
    • Export Citation
  • Koch, S. E., , Feltz W. , , Fabry F. , , Pagowski M. , , Geerts B. , , Bedka K. M. , , Miller D. O. , , and Wilson J. W. , 2008: Turbulent mixing processes in atmospheric bores and solitary waves deduced from profiling systems and numerical simulation. Mon. Wea. Rev., 136, 13731400.

    • Search Google Scholar
    • Export Citation
  • KMA, 2010: Report of global atmosphere watch 2010. Korea Meteorological Administration Rep. 111360000-000084-10, 239 pp.

  • Liljegren, J., , Lesht B. , , Kato S. , , and Clothiaux E. , 2001: Initial evaluation of profiles of temperature, water vapor, and cloud liquid water from a new microwave profiling radiometer. Proc. 11th Atmospheric Radiation Measurement (ARM) Program Science Team Meeting, Atlanta, GA, U.S. Department of Energy. [Available online at http://radiometrics.siteoperations.com/wp-content/uploads/2012/11/MWRP_ARM01.pdf.]

  • Mariani, Z., and Coauthors, 2011: Infrared emission measurements in the Arctic using a new extended-range AERI. Atmos. Meas. Tech. Discuss., 4, 64116448.

    • 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, seagoing infrared spectroradiometer. J. Atmos. Oceanic Technol., 18, 9941013.

    • Search Google Scholar
    • Export Citation
  • Pougatchev, N., and Coauthors, 2009: IASI temperature and water vapor retrievals—Error assessment and validation. Atmos. Chem. Phys., 9, 64536458.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., 2000: Inverse Methods for Atmospheric Sounding: Theory and Practice. World Scientific Publishing Co. Ltd., 240 pp.

  • Rowe, P. M., , Miloshevich L. M. , , Turner D. D. , , and Walden V. P. , 2008: Dry bias in Vaisala RS90 radiosonde humidity profiles over Antarctica. J. Atmos. Oceanic Technol., 25, 15291541.

    • Search Google Scholar
    • Export Citation
  • Schlüssel, P., , Hultberg T. H. , , Phillips P. L. , , August T. , , and Calbet X. , 2005: The operational IASI level 2 processor. Adv. Space Res., 36, 982988.

    • Search Google Scholar
    • Export Citation
  • Smith, W. L., , and Woolf H. M. , 1976: The use of eigenvectors of statistical covariance matrices for interpreting satellite sounding radiometer observations. J. Atmos. Sci., 33, 11271140.

    • Search Google Scholar
    • Export Citation
  • Smith, W. L., , Feltz W. F. , , Knuteson R. O. , , Revercomb H. E. , , Woolf H. M. , , and Howell H. B. , 1999: The retrieval of planetary boundary layer structure using ground-based infrared spectral radiance measurements. J. Atmos. Oceanic Technol., 16, 323333.

    • Search Google Scholar
    • Export Citation
  • Stokes, G. M., , and Schwartz S. E. , 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the Cloud and Radiation Testbed. Bull. Amer. Meteor. Soc., 75, 12011221.

    • Search Google Scholar
    • Export Citation
  • Szczodrak, M., , Minnett P. J. , , Nalli N. R. , , and Feltz W. F. , 2007: Profiling the lower troposphere over the ocean with infrared hyperspectral measurements of the Marine-Atmosphere Emitted Radiance Interferometer. J. Atmos. Oceanic Technol., 24, 390402.

    • Search Google Scholar
    • Export Citation
  • Tobin, D. C., and Coauthors, 2006: Atmospheric Radiation Measurement site atmospheric state best estimates for Atmospheric Infrared Sounder temperature and water vapor retrieval validation. J. Geophys. Res., 111, D09S14, doi:10.1029/2005JD006103.

    • Search Google Scholar
    • Export Citation
  • Turner, D. D., , Ackerman S. A. , , Baum B. A. , , Revercomb H. E. , , and Yang P. , 2003: Cloud phase determination using ground-based AERI observations at SHEBA. J. Appl. Meteor., 42, 701715.

    • Search Google Scholar
    • Export Citation
  • Turner, D. D., , Knuteson R. O. , , and Revercomb H. E. , 2006: Noise reduction of Atmospheric Emitted Radiance Interferometer (AERI) observations using principal component analysis. J. Atmos. Oceanic Technol., 23, 12231238.

    • Search Google Scholar
    • Export Citation
  • Walden, V. P., , Town M. S. , , Halter B. , , and Storey J. W. V. , 2005: First measurements of the infrared sky brightness at Dome C, Antarctica. Publ. Astron. Soc. Pac., 117, 300308.

    • Search Google Scholar
    • Export Citation
  • Yurganov, L., , McMillan W. , , Wilson C. , , Fischer M. , , and Biraud S. , 2010: Carbon monoxide mixing ratios over Oklahoma between 2002 and 2009 retrieved from Atmospheric Emitted Radiance Interferometer spectra. Atmos. Meas. Tech. Discuss., 3, 12631301.

    • Search Google Scholar
    • Export Citation
  • Zhou, D. K., and Coauthors, 2002: Thermodynamic product retrieval methodology and validation for NAST-I. Appl. Opt., 41, 69576967.

  • Zhou, D. K., , Smith W. L. , , Larar A. M. , , Liu X. , , Taylor J. P. , , Schlüssel P. , , Strow L. L. , , and Mango S. A. , 2009: All weather IASI single field-of-view retrievals: Case study—Validation with JAIVEx data. Atmos. Chem. Phys., 9, 22412255.

    • Search Google Scholar
    • Export Citation
Save