On the effect of dust aerosols on AIRS and IASI operational level 2 products



[1] Satellite retrievals of environmental parameters can enable a global scale understanding of various atmospheric phenomena. Validation of these retrievals using in situ and other correlative measurements is critical to the success of our accurate interpretation of such phenomena. In this paper, we analyze the effect of dust on AIRS and IASI operational level 2 profiles using data obtained from the 2009–2011 AEROSE campaigns. We find that the presence of dust in the AIRS and IASI measurements biases the cloud-cleared radiances by as much as 4 K. In addition, we find that the temperature and surface temperature retrievals resultant from these cloud-cleared radiances show 2 to 3 K spurious oscillations throughout the troposphere.

1. Introduction

[2] The Atmospheric Infra-Red Sounder (AIRS) and Infra-red Atmospheric Sounding Interferometer (IASI) instruments were designed to provide improved temperature and humidity profiles for numerical weather prediction and long-term climate studies [Aumann et al., 2003; Cayla, 1993]. Both IASI and AIRS have been extensively validated for temperature and water vapor profiles [Divakarla et al., 2006, 2011], and surface temperature [Susskind et al., 2003, 2011; Maddy et al., 2011]. However, the effects of large scale dust aerosols (e.g., aeolian mineral dust from deserts) on measured AIRS and IASI radiances, cloud-cleared radiances (CCRs), and Level 2 (L2) profiles has not been determined.

[3] With an almost complete high-spectral coverage of the 9μm atmospheric window spectrum, AIRS and IASI radiances contain much information about clouds and aerosols, allowing the retrieval of infrared (IR) optical depths τ(ν) of mineral dust [DeSouza-Machado et al., 2006, 2010]. The fact that these instruments are radiatively sensitive to dust warrants an understanding of the effect of dust on L2 profiles and CCRs. Since the L2 atmospheric state retrievals are then derived from the CCRs, it is important to characterize the potential effects on the final L2 retrievals, as these products are what are most easily accessed and used by scientists. For instance, we are aware that some publications correlating dust optical depths and atmospheric thermodynamic profiles using AIRS L2 retrievals have already been published by AIRS product users [see Davidi et al., 2012]. Ongoing research involving atmospheric aerosols includes the suppression of hurricane formation by dust [Amato et al., 2011], as well as, reducing uncertainties in aerosol radiative forcing [DeSouza-Machado et al., 2006]. It is therefore important that current and future users are made aware of potential issues with L2 products of current (and future) sounders such as AIRS and IASI in dusty scenes.

[4] In this paper we analyze operational L2 Version 5 (V5) AIRS retrievals and CCRs as well as Version 2 (V2) IASI retrievals and CCRs obtained from an offline processing system at the National Oceanic and Atmospheric Adminstration (NOAA) National Environmental Data and Information Service (NESDIS). To analyze the effect of dust on these retrievals we compare L2 satellite retrievals to spatially and temporally collocated radio observations (RAOB) and European Centre for Medium Range Weather Forecasts (ECMWF) forecast and analysis surface temperatures as a function of collocated Microtops sunphotometer measurements of aerosol optical depth (AOD). These correlative measurements were obtained on the 2009–2011 AERosols and Ocean Science Expeditions (AEROSE) cruises [Nalli et al., 2011].

[5] The paper is organized as follows. In Section 2, we describe briefly the AIRS and IASI processing systems. The AEROSE correlative measurements are described in Section 3. Statistical comparisons as a function of AOD are presented in Section 4, where it is shown that the presence of dust in both AIRS and IASI measurements biases each instruments CCRs, L2 temperature profiles, and surface temperatures.

2. AIRS and IASI

[6] AIRS was launched into a sun-synchronous polar orbit (1:30 AM/PM local time equatorial crossing time) on board NASA's EOS/Aqua platform on May 4, 2002. AIRS is a cross-track scanning grating spectrometer that measures 2378 channels covering the spectral range 649–1136, 1217–1613, 2181–2665 cm−1. The full widths at half maximum satisfy ν/δν≈ 1200, with the noise equivalent change in temperature (NEDT) ≈ 0.2 K. The AIRS nadir footprint spatial resolution is about 13.5 km. A 3 × 3 array of AIRS footprints, termed the field-of-regard (FOR), is co-aligned with the Advanced Microwave Sounding Unit (AMSU) and together the nine footprint IR and single footprint microwave measurements are used to remove the effects of spatially non-uniform clouds on the AIRS footprints through a technique known as cloud-clearing.

[7] IASI was launched into a sun-synchronous polar orbit (1:30 AM/PM local time equatorial crossing time) on board the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) MetOp-A satellite platform in October 2006. IASI is a cross-track scanning Michelson interferometer that measures 8461 channels at 0.25 cm−1 spacing between 645–2760 cm−1 (3.6–15.5μm), in a 2 × 2 array of circular footprints with a nadir spatial resolution of roughly 50 km × 50 km (with a corresponding single footprint spatial resolution at nadir of roughly 12 km). This arrangement of footprints is also co-aligned with the AMSU instrument on Metop-A, thus enabling the removal of the spectral signature of spatially non-uniform clouds on the IASI radiances in a cloud-clearing technique similar to AIRS.

[8] Spectral measurements from AIRS and IASI contain information on the vertical temperature profile, surface parameters (e.g., temperature, emissivity, reflectivity), clouds [Kahn et al., 2008] and aerosols [DeSouza-Machado et al., 2006, 2010; Köhler et al., 2011], and the vertical distribution of tropospheric and stratospheric trace gases such as H2O, CO, CH4, CO2, HNO3, and O3 [Cayla, 1993; Maddy et al., 2008; Xiong et al., 2008; Divakarla et al., 2008; Pittman et al., 2009; Maddy et al., 2009]. The current NOAA IASI and NASA AIRS operational cloud-clearing and physical retrieval methodologies [Susskind et al., 2003; Maddy et al., 2011] uses the same fast eigenvector regression methodology that is described in Goldberg et al. [2003]to provide temperature and moisture geophysical profiles as well as surface parameters using cloudy-sky AIRS or IASI spectral measurements and AMSU microwave sounder brightness temperatures as inputs. These regression output parameters are then matched with climatological trace gas abundances (e.g., O3, N2O, etc.) and used as inputs to an Rapid Transmittance Algorithm (RTA) [Strow et al., 2003] to produce a clear-sky radiance estimate. This clear-sky radiance estimate is then used to extrapolate CCRs from a spatial interpolation of multiple cloudy infrared footprints in the AIRS or IASI FOR collocated to each respective satellite's microwave footprint. Once CCRs are obtained, the AIRS and IASI temperature and various other atmospheric parameter retrievals (e.g., H2O, O3, surface parameters, etc.) algorithms then minimize the difference between the CCRs and a calculation based on the regression solution. The process is iterated several times for temperature and surface parameters (emissivity, surface temperature and reflectivity).

[9] The current IASI operational system is the second operational version of the algorithm. In terms of algorithm steps and constraints the IASI V2 system is an emulation of the AIRS V5 system as applied to MetOp-A IASI and AMSU data.

3. The 2009–2011 AEROSE Correlative Measurements

[10] Since 2004, NOAA has supported a series of trans-Atlantic Aerosols and Ocean Science Expeditions (AEROSE) onboard the NOAA shipRonald H. Brown [Morris et al., 2006; Nalli et al., 2011]. On each ≈30 day AEROSE cruise, atmospheric in situ and remotely sensed marine data are collected over the tropical Atlantic Ocean downwind of Saharan dust and sub-Saharan smoke aerosols. Data relevant to satellite remote sensing validation include: dedicated balloon-borne radiosondes, ozonesondes, and Microtops handheld sunphotometers [Smirnov et al., 2011]. All the 2009–2011 AEROSE RAOBs were obtained from Vaisala RS92-GPS rawinsondes and were launched ≈0.5 hr prior to the expected twice-daily satellite overpasses (roughly 4 total sondes per day). These dedicated RAOB data have been manually inspected for quality, and are not ingested into any numerical forecast models. In addition, AEROSE has an active collaboration with the NASA Goddard Space Flight Center (GSFC) Maritime Aerosol Network (MAN) to apply the MAN methodology for maritime (ship-based) Microtops observations described inSmirnov et al. [2011], including sensor calibration, raw data processing, and quality assurance (QA).

[11] In total, there were 817 Microtops measurements and 260 dedicated RAOB launches acquired on AEROSE campaigns between 2009 and 2011. Figure S1 in the auxiliary material shows meridional and and zonal cross sections of RAOB relative humidities (RH), winds, MAN AOD (circles) and Angstrom exponents, α (triangles) (see figure caption for explanation of the color scheme). Dry Saharan air layers (SAL) were observed between July 28 through July 30, 2009; July 25 through July 27, 2009; and July 31 through August 3, 2011 (see also Figure S2 in the auxiliary material).

4. Statistical Comparisons of AIRS and IASI to RAOBs

[12] In this section we compare AIRS V5 and IASI V2 operational retrievals to AEROSE RAOBs, Microtops sunphotometer measurements, and ECMWF forecast/analysis surface skin temperatures. For the ship-borne data, we use a matchup time window of ±3.0 hrs and 150 km. For ECMWF comparisons, we select the closest ECMWF forecast or analysis to the satellite observation time and location. Both AIRS and IASI include quality control (QC) fields in the respective output files. Unlike IASI V2 retrievals, which use a single binary flag (i.e., accept/reject) for general retrieval quality, AIRS V5 uses a height dependent tertiary QC (i.e., best, ok, bad) for temperature, and a strict tertiary QC for surface parameters [Susskind et al., 2011]. The AIRS QC model results in different ensembles for temperature as a function of altitude and surface temperature statistics, so determining the cause and effect of poor (or favorable) results for a specific retrieval is difficult. Rather than utilize the different ensembles for the various statistics, we have re-run the AIRS V5 using a QC that emulates the IASI QC. The QC applied is akin to the profile having the best temperature profile QC down to the middle troposphere (e.g., best QC from the top of the atmosphere down to 600 hPa).

[13] Figures 1a (top left) and 1b(top left) show the location of the Microtops sunphotometers measurements of Level 2.0 quality (cloud-screened and QA'd) MAN AOD at 870 nm using the algorithms ofSmirnov et al. [2011]. Angstrom exponents, α are also conveyed in the figures via symbol size, with large α corresponding to a larger symbol and vice versa. We note that the largest MAN AOD are also associated with low α, thus giving credence to the fact that these are coarse mode Saharan dust aerosols.

Figure 1a.

(top left) The locations of the collocated AIRS satellite retrievals, RAOBs, and Microtops sunphotometer AOD measurements from the 2009, 2010, and 2011 AEROSE campaigns. Colors correspond to AOD at 870 nm, while the size of the markers denotes the Angstrom exponent, α. (top right) Mean bias tendency (i.e., bias in each bin minus mean bias over all cases) statistics of the AIRS temperature retrievals minus RAOBs conditioned and color coded as a function of AOD. (bottom left) The standard deviation of the AIRS retrievals conditioned and color coded as a function of AOD. Surface temperature bias tendency and standard deviation statistics relative to the ECMWF forecast/analysis are also shown as the color coded circles. (bottom right) Mean bias tendency statistics of AIRS CCRs minus calculations using the RAOBs as the atmospheric state and ECMWF for skin temperature conditioned and color coded as a function of AOD.

[14] Figures 1a and 1b also show statistical comparisons of AIRS and IASI temperature retrievals to RAOBs and ECMWF model forecast and analysis surface skin temperatures. In addition, in Figures 1a (bottom right) and 1b (bottom right) we show statistics of CCRs minus calculations, which utilize the combined RAOB profile plus ECMWF surface fields for radiative transfer calculations. To facilitate our understanding of the presence of dust on the AIRS and IASI retrievals we have conditioned the mean bias tendency and standard deviation statistics on Microtops sunphotometer AOD at 870 nm in 4 bins. The bias tendency is defined as the mean bias in each AOD bin minus the mean bias over all accepted cases. The 5 AOD bin edges are as follows: 0.00, 0.19, 0.38, 0.56, 0.75. In general, there are anywhere from 27 to approximately 2000 accepted retrievals matched to RAOBS in each AOD bin.

Figure 1b.

As Figure 1a except for IASI RAOB matchups.

[15] Both IASI and AIRS bias tendency statistics show similar vertical oscillations relative to the RAOBs, with a negative peak between 600–700 hPa and slightly positive bias tendency near the surface. This implies that the retrievals are responding to the warm dust layer by warming the temperature profile near the surface and to find a radiative balance in the observations minus calculations (obs-cal), the retrieval cools the temperature profile aloft. These oscillations are much wider in altitude than the expected vertical resolution of the AIRS temperature retrieval as reported inMaddy and Barnet [2008] and IASI vertical resolution should be similar. In addition, we show in the auxiliary material (see Figures S3 and S4) that a large portion of these errors are due to the use of a regression as an a priori.

[16] It is here interesting to note that the standard deviation statistics do not show a large change with AOD. In fact, the statistics seem to improve for the dustiest cases. We expect that this behavior is related to the atmospheric variability in the various marine environments encountered during the AEROSE cruises (e.g., differences in profile variability in the Inter Tropical Convergence Zone (ITCZ) as compared to those in the sub-tropical subsidence regions).

[17] Surface temperature bias tendency and standard deviation statistics relative to the ECMWF forecast/analysis are also shown as the color coded circles near 1000 hPa. The bias tendency of the surface temperatures from AIRS generally fall within 1 K, while the bias tendency of the surface temperatures from IASI show a stronger dependence on MAN AOD. It has been shown [e.g., Maddy et al., 2011] that IASI V2 surface temperatures are biased cold in cloudier scenes. It is likely that the presence of dust causes further complications in the IASI cloud-clearing algorithm; hence, the larger dependence on AOD.

[18] The conditional mean differences between the AIRS and IASI retrieval algorithms CCRs and RTA calculations using the RAOB temperature and moisture profiles, ECMWF forecast analysis surface temperature, and water emissivity from [Masuda et al., 1988] as input to the radiative transfer model are shown in of Figures 1a (bottom right) and 1b (bottom right). For each AOD bin we have removed the mean bias over all cases so that the bias tendency becomes clearer. The shape of the bias tendency of the dustiest cases (e.g., third and forth bins) matches well with the shape expected from dust contaminated IR radiances [e.g., see DeSouza-Machado et al., 2006, Figure 5]. We also note that the spectral shape of the biases in the CCRs are similar in shape to those noted in Nalli et al. [2006, Figures 9 (bottom) and 10 (bottom)] for dusty scenes.

5. Conclusions

[19] We have shown that current v5 AIRS and IASI temperatures and surface temperature retrievals can have large errors in the presence of dust aerosol, and may not accurately represent the atmospheric state. In particular, we have shown that AIRS and IASI temperature retrievals are biased warm by several degrees near the surface and cold by several degrees in the lower-middle troposphere (e.g., 500–700 hPa) relative to high quality dedicated RAOBs. In addition, we have found IASI V2 surface temperature tends to be biased cold relative to ECMWF for the dustiest scenes. Using high quality RTA calculations we have shown that these uncertainties are due to dust contamination in each instrument's CCRs. We expect that this is due to the fact that dust tends to be spatially uniform, a situation where cloud-clearing algorithms perform poorly.

[20] While we have not explicitly looked at other types of aerosols (e.g., smoke), we expect that L2 profile retrievals will be less impacted in presence of smoke because smoke particles are smaller and their radiative effect on IR satellite measurements is smaller [e.g., see Köhler et al., 2011]; such investigation may be the subject of future research. In addition, we expect similar biases to show up in AIRS and IASI L2 products obtained in the presence of dust from other regions, such as E. Asia during the springtime Asian dust season [e.g., Nalli and Reynolds, 2006]. In general, one would expect the spectral signature of mineral dust to be dominated by silicate features. However, a number of papers [e.g., see Turner, 2008; DeSouza-Machado et al., 2006] have demonstrated that dust from different geographic regions has different spectral signatures based upon the mineral composition of the desert surface and crust from which the airborne dust originates. Care needs to be taken therefore, to ensure that the channels used in a retrieval avoid regions where spectral signatures of one or more dust species feature prominently to avoid problems in the temperature and humidity retrievals of the lowest atmospheric layers.

[21] We plan a follow up to this letter showing that combined aerosol plus geophysical state retrievals can enable accurate sounding in dust laden aerosol regions, which should ultimately strengthen conclusions drawn from comparisons of L2 products with other atmospheric parameters such as aerosol optical depths. An analysis of H2O and O3 retrievals in the presence of dust will be the subject of future research.


[22] This research was supported in part by NASA Research Announcement (NRA) NNH09ZDA001N, Research Opportunities in Space and Earth Science (ROSES-2009), Program Element A.41: The Science of Terra and Aqua, the NOAA Joint Polar Satellite System (JPSS)Office (NJO), the GOES-R Algorithm Working Group, and the STAR Satellite Meteorology and Climatology Division (SMCD) (M. D. Goldberg, SMCD Division Chief). AEROSE is supported by the NOAA Educational Partnership Program grant NA17AE1625, NOAA grant NA17AE1623 to establish the NOAA Center for Atmospheric Science (NCAS) at Howard University; we also wish to acknowledge A. Flores and M. Oyola (Howard University), and A. Smirnov (NASA/GSFC), for their assistance with AEROSE Microtops measurements. The views, opinions, and findings contained in this paper are those of the authors and should not be construed as an official National Oceanic and Atmospheric Administration, National Aeronautics and Space Administration or U.S. Government position, policy, or decision.