Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 For the first time OSIRIS has derived and analyzed mesopause region temperatures at the summer poles. Temperatures have been derived from observations of O2 A-band emission spectra, with a typical estimated accuracy of better than ±3 K at altitudes near 90 km. The current results show that OSIRIS temperatures near the mesopause are usually ∼10 K colder than CO2 temperature results from the SABER instrument on board the TIMED satellite, however the difference between OSIRIS temperatures and SOFIE ice temperatures agrees well with previous model results, and OSIRIS temperatures are consistent with potassium lidar derived atmospheric temperatures. Between 2003–2010, the retrieved OSIRIS polar summer mesopause temperatures are often below 120 K, and on 16 January 2008, at 80°S and 223°E, a temperature of 91 ± 6 K was retrieved at an altitude of 92 km.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 In the upper mesopause region, O2 A-band (b1Σg+ − X3Σg+) 0 − 0 emission at 762 nm is a prominent airglow feature that can be observed in the Earth's limb from a satellite platform during both night and daytime conditions. Excited O2(b1Σg+) molecules are predominantly formed in this region through resonant absorption in the O2 Atmospheric Bands, collisional excitation with O(1D), and a 2-step Barth-type chemical reaction [Bucholtz et al., 1986]. During the day all three of these processes lead to the formation of O2(b1Σg+), however at night there is no solar input for resonant absorption or formation of O(1D) through O2 and O3 photolysis, thus the Barth-type reaction is the primary source of A-band nightglow emission.
 Measurements of O2 A-band emission spectra with the Optical Spectrograph and Infrared Imaging System (OSIRIS) on board the Odin satellite [Llewellyn et al., 2004] have been analyzed to derive temperature profiles in the mesopause region, between altitudes of 85–105 km. Temperature retrieval techniques similar to those used here have been used to derive OH rotational temperature profiles [e.g., Mende et al., 1988; von Savigny et al., 2004; Ortland et al., 1998] have used the High Resolution Doppler Interferometer (HRDI) instrument on board the Upper Atmosphere Research Satellite (UARS) to derive mesospheric temperature profiles from emission intensity measurements of individual rotational lines within the O2 A-band. However, the OSIRIS measurements have now provided an extensive database of mesospheric temperatures derived from observations of the O2 A-band emission spectrum.
 The Odin satellite was launched into a near-circular, near-sun-synchronous orbit at an altitude of ∼620 km with an ascending node at 18:00 LT. As Odin nods in orbit, OSIRIS scans the Earth's limb between tangent heights of approximately 10 and 110 km and the optical spectrograph (OS) observes both scattered sunlight and airglow emission at wavelengths between 275 and 810 nm with a spectral resolution of 0.9 nm. OS observations are dispersed onto a 1353 × 32 pixel (wavelength spectrum × horizontal spatial information) CCD array. The 32 spatial pixels are binned on-chip for an improved signal-to-noise ratio (S/N). This study utilizes data from the 19 pixels across the A-band spectral region that span the wavelength range 759.0–766.7 nm.
 The fine structure of the O2 A-band emission is well understood and has been described by Babcock and Herzberg . O2(b1Σg+) molecules have a radiative lifetime of ∼12 s, and the collisional frequency near the mesopause is on the order of 1 × 104 s−1, allowing O2(b1Σg+) to undergo thousands of collisions before it radiates. Therefore, the O2(b1Σg+) rotational temperature in this region can safely be assumed to be in local thermodynamic equilibrium (LTE).
 The retrieval algorithm used in this study is described in detail by Sheese et al. . Each temperature profile is derived from a single OS scan. Within a single scan in the mesosphere – lower thermosphere (MLT) region, measurements have an exposure time of 1 s, and the vertical separation between the start of two consecutive measurements is ∼1.5 km. The typical horizontal separation between the start and end of an MLT scan is ∼250 km. For a single OS scan, the radiance profiles of each of the 19 pixels in the A-band region are interpolated onto a 1-km grid and separately inverted using a tomographic-like algorithm [Degenstein et al., 2003] to produce A-band volume emission rate (VER) spectra profiles. The inverted VER spectrum at each altitude is fit to modeled emission spectra, and the temperature that yields the best fit between the modeled and the inverted spectrum is the retrieved temperature. At observation tangent heights below 90 km, absorption along the line-of-sight in the A-band region is non-negligible, and each pixel inversion requires the calculation of absorption coefficients, as detailed by Degenstein et al. . With use of developed A-band emission and absorption intensity models, the pixel absorption coefficients are calculated line-by-line with the assumption that the rotational lines are only Doppler broadened, and assume a priori temperature and O2 density, [O2], profiles obtained from a locally run NRL-MSISE-00 model [Picone et al., 2002]. Since the absorption coefficients depend on the temperature profile, the retrieval algorithm is iterative, with the absorption coefficients recalculated at the beginning of each iteration. The retrieval is considered to have converged on a temperature profile when the change in temperature between iterations at each altitude level is less than 0.1%, and retrievals typically converge in less than 5 iterations. The a priori temperature profile, 200 K at all altitudes, has no influence on the final retrieved temperature profile. For instance, retrievals starting with an MSIS a priori temperature profile yield retrieved temperatures that differ from the 200 K a priori retrieval results by less than 0.1% (at all altitude levels). The same is true when a 0 K a priori profile is used.
 The random error due to instrument noise for each OSIRIS pixel in the A-band region is typically on the order of 0.1%, which is negligible in the retrieval of temperature profiles. The retrieval is insensitive to OSIRIS absolute calibration uncertainties, as only relative intensities are required. The most significant systematic uncertainties arise due to uncertainties in the background correction for upwelling radiation and off-axis signal, uncertainty in the MSIS [O2] profile, and uncertainty in the OS wavelength registration. Total estimated systematic uncertainties are less than ±6 K at 100 km, less than ±3 K at 90 km, and less than ±8 K at 85 km.
Sheese et al.  described a method of estimating the total error of a retrieval by the determining the retrieval “S-value”. The S-value is simply the sum of the square difference between the inverted spectrum and the modeled spectrum, both normalized to a pixel intensity of 1000; equal weight is given to each pixel in the S-value calculation. Simulated retrievals determine the error in temperature that reproduces the measured S-value. Maximum total errors are typically less than ±10 K near 100 km, less than ±5 K near 90 km, and less than ±8 K near 85 km.
 All OSIRIS results and other mission datasets discussed in this section, unless stated otherwise, are from the summer high latitudes during the polar mesospheric cloud (PMC) season. In the Northern hemisphere (NH) all data are from June and July, and in the Southern hemisphere (SH) all data are from December and January and each SH summer season is designated by the later year in the season (e.g. SH Dec/Jan 2008/2009 is referred to as SH 2009).
 Latitudinal cross-sections of the OSIRIS 2003–2010 mean summer season temperature profiles in the SH and NH are plotted in Figures 1a and 1b. The NH mean exhibits a minimum temperature of 124 K at 80°N at an altitude of ∼89 km, while the SH mean exhibits a minimum temperature of 129 K at 80°S at an altitude of ∼90 km. The mean minimum temperature and mean height of the temperature minimum both decrease with latitude. The SH mean mesopause height (not shown) was relatively constant, ∼89.5 km, during 2003–2006. The mesopause height increased to ∼91 km in 2008 and dropped back down to ∼89.5 km in 2010. There was much less variation in the NH, with a minimum mean mesopause height of 88.3 km in 2007 and a maximum mean height of 89.0 km in 2008. Figure 1d shows the yearly mean mesopause temperatures in both hemispheres from 2003–2010. The mean OSIRIS mesopause temperatures are colder than the majority of results published by other satellite missions, and slightly warmer than the potassium lidar results of Höffner and Lübken ; the interhemispheric difference in mesopause temperatures can still be examined. The results of two separate model simulation studies, discussed by Siskind et al.  and Lübken and Berger , predict that the summer SH upper mesosphere should be warmer than the summer NH by 3–8 K, and ∼5 K, respectively. Siskind et al. , using a 2-D chemical-dynamical model, concluded that this interhemispheric difference primarily arises because of hemispheric asymmetries in the mean summer zonal winds in the lower stratosphere, which ultimately lower gravity-wave drag and increase temperatures in the SH upper mesosphere. Lübken and Berger , using a 3-D GCM model, also concluded that dynamical asymmetries in the lower atmosphere were ultimately the leading cause of this difference at the mesopause. The OSIRIS data from 2003–2010, shown in Figures 1c and 1d, is consistent with the model predictions, as mean SH upper mesospheric temperatures are ∼2–8 K warmer than those in the NH. However, in the 2010 seasons the mean SH mesopause temperature was slightly colder than the NH value by < 1 K. The cause of the larger decrease in SH mesopause temperatures than those in the NH over the past 8 years is a topic for a later study.
3.1. Comparisons With Other Instruments
 Two often-cited missions that measure CO2 temperatures in the mesosphere are the Sounding of the Atmosphere using Broadband Radiometry (SABER) instrument onboard the Thermosphere-Ionosphere-Mesosphere Energetics and Dynamics (TIMED) satellite [Mlynczak, 1997] and the Solar Occultation for Ice Experiment (SOFIE) instrument onboard the Aeronomy of Ice in the Mesosphere (AIM) satellite [Russell et al., 2009].
Russell et al.  compared SABER and SOFIE mesospheric CO2 temperatures for the NH 2007 and 2008 seasons and concluded that there is a bias between the datasets in the mesopause region, where SOFIE temperatures are warmer than those of SABER. Hervig and Gordley  discuss this bias as being on the order of 5–10 K. Figure 2 shows the statistical results of a comparison between all 242 coincident OSIRIS and SABER temperature profiles during the NH 2007–2008 seasons, as well as all 332 coincident profiles in the SH 2007–2008 seasons. The SABER data (V1.07) was downloaded directly from the SABER website, http://saber.gats-inc.com, and the coincidence criteria were set to 2° in latitude, 5° in longitude, and 1 h in time. At altitudes below 92 km, OSIRIS results are systematically 5–14 K colder than SABER; above 92 km, OSIRIS temperatures are on average 10–15 K colder. The large standard deviations in temperature difference, up to 28 K, is not unexpected, as such lax coincidence criteria, in an extremely dynamic region, lead to comparisons of regions that are far from common-volume. These results indicate that OSIRIS temperatures are also 10–25 K colder than SOFIE CO2 temperatures.
 Although there is a large bias between OSIRIS A-band and SABER/SOFIE CO2 temperatures, there is very good agreement between the OSIRIS A-band temperatures and SOFIE ice temperatures. Hervig and Gordley [2010, Figure 9] present a mean July 2008 SOFIE ice temperature profile for latitudes between 67° and 71°N. The mean SOFIE ice temperature is ∼134 K at all altitudes between 85 and 89 km, whereas the mean July 2008 OSIRIS temperature for the same region, Figure 3a, increases from a minimum of 130 K at 86 km to 133 K at 89 km. Hence, OSIRIS observes the mean kinetic temperature in this region as 1–4 K colder than the mean SOPHIE ice temperatures. Espy and Jutt  have modeled equilibrium ice particle temperatures in the mesosphere, taking into account radiative heating due to both solar and terrestrial sources and ice particle radiative and collisional cooling. They concluded that ice temperatures in the summer polar mesopause region should be 3–8 K warmer than the local kinetic temperature. However, Siskind et al.  determined that ice particles within an ice layer can radiatively heat the ambient atmosphere by 2–6 K. This would imply that ice particle temperatures would typically be anywhere from 6 K colder to 3 K warmer than the ambient temperature.
 A time series of OSIRIS 3-day average mesopause temperatures at SOFIE latitudes in 2008 and a similar time series of 3-day average minimum SOFIE ice temperatures is shown in Figure 3b. The mean OSIRIS mesopause temperatures are 7 K colder to 4 K warmer than the mean minimum SOFIE ice temperatures, and on average are 3 K colder. This difference between ice and ambient temperatures is consistent with the combined model results of Espy and Jutt  and Siskind et al. . Also shown in Figure 3b are 3-year mean mesopause temperatures derived from potassium lidar measurements over Spitsbergen (78°N, 15°E) between 2001 and 2003, obtained from Höffner and Lübken . The potassium lidar climatology exhibits similarly low temperature values, with slightly lower temperatures than OSIRIS near solstice. Lower potassium lidar temperatures are observed since they are measured at a higher latitude (78°N). Although OSIRIS and the lidar measurements are in good agreement in the mesopause region, the two datasets quickly diverge at heights above the mesopause. The potassium lidar temperatures are approximately 40 K lower than OSIRIS near 100 km. Since SABER and OSIRIS mean vertical temperature gradients are very similar and do not match the K lidar gradient, the divergence is most likely due to inherent differences in the measurement techniques. Two possible explanations are that the limb measurements are averaging out lower temperatures along the lines-of-sight, or that the ground measurements are losing signal above the mesopause, or a combination of both.
3.2. The Coldest Place on Earth
 OSIRIS frequently retrieves mesopause temperatures that are below 120 K, and even temperatures below 100 K are observed on occasion. Figure 4a shows the number of times a retrieved mesopause temperature was recorded in the January and June 2008 data. Over 12% of retrieved mesopause temperatures in January, and over 58% in June, were below 120 K. In June there were 12 retrieved profiles with mesopause temperatures at or below 100 K, and in January there was a retrieved profile with a mesopause temperature of 91 K at an altitude of 92 km, Figure 4b. The fit between the retrieved VER spectrum at 92 km and the modeled emission spectrum at 91 K is shown in Figure 4c, with a reference 110 K modeled spectrum. The comparison between the retrieved and modeled 91 K spectrum yields an S-value of 78, which corresponds to a total uncertainty of less than ±6 K, whereas the S-value is over 900 (±20 K) when comparing with the 110 K spectrum.
4. Discussion and Conclusions
 At and just below the mesopause, OSIRIS A-band temperatures are typically on the order of 15–25 K below the frost point temperature of ∼145 K, and only a small percentage of retrieved mesopause temperatures each season are above this threshold. OSIRIS A-band temperatures are more consistent with the ubiquitous nature of polar mesospheric clouds in the summer high-latitudes than the warmer CO2 temperatures, and are in very good agreement with SOFIE ice temperatures, given the equilibrium ice temperature results of Espy and Jutt .
 OSIRIS A-band temperatures are systematically lower than SABER CO2 temperatures by 5–15 K and lower than SOFIE CO2 temperatures by up to ∼25 K, and at mesopause heights these biases are outside the range of the combined instruments' estimated systematic errors. As well, further study has shown that modeled A-band emission spectra that assume the warmer retrieved CO2 temperature profiles are inconsistent with observed OSIRIS A-band spectra. An investigation into the source of these biases is needed.
 The mesopause has long been known as the coldest region of the atmosphere. Schmidlin  reported a falling sphere temperature of ∼95 K at an altitude of 92 km over Kiruna, Sweden (68°N, 21°E), and Lübken et al.  have presented multiple potassium lidar derived temperature profiles with mesopause temperatures below 90 K at altitudes near 90 km. OSIRIS is now demonstrating that these cold mesopause temperatures are not rare occurrences. A few times each summer, OSIRIS retrieves mesopause temperatures below 100 K, and on 16 January 2008, at 80°S, a temperature profile was retrieved with a mesopause temperature of 91 ± 6 K.
 The authors thank the anonymous referees for their valuable comments. Odin is a Swedish-led satellite project funded jointly by Sweden (SNSB), Canada (CSA), France (CNES), and Finland (Tekes), with support by the third-party mission programme of the European Space Agency (ESA).
 The Editor thanks two anonymous reviewers for their assistance evaluating this manuscript.