One of the largest solar proton events in the past thirty years took place on October 28, 2003 and had a significant impact on the Earth's middle atmosphere. The incoming protons produce significant amounts of HOx constituents in the mesosphere and upper stratosphere that lead to ozone depletion. For 11 hours during the solar proton event the OSIRIS instrument on Odin measured high spatial resolution profiles of the oxygen infrared atmospheric band at 1.26 microns which under sunlit conditions can be used as a direct proxy for ozone. Ozone depletion is observed across the southern polar cap for the duration of the observations and extends to latitudes as far north as 45°S. OSIRIS observed significant ozone depletion between 50 and 80 km with a maximum value of 75% around 65 km. The actual maximum depletion could have been even greater as observations ceased while the depletion was still increasing.
 A series of solar flares associated with coronal mass ejections occurred in late October 2003 and caused a major Solar Proton Event (SPE) on October 28. The X-17 class event is the largest solar proton event since 1962. It is well known that incoming high energy protons produce significant amounts of HOx (H, OH, HO2) and NOx (N, NO, NO2) constituents in the polar middle atmosphere that, in turn, directly influence atmospheric ozone density [Swider and Keneshea, 1973; Frederick, 1976; Solomon et al., 1983; Jackman and McPeters, 1985]. The HOx constituents are quite short-lived and lead to large, short-term ozone decreases in the mesosphere and upper stratosphere, typically lasting a few hours or days. The NOx constituents are much longer-lived and can influence the stratospheric ozone density over periods of months to years [Rusch et al., 1981; Jackman et al., 1990; Reid et al., 1991; Jackman and Fleming, 2000]. Several models currently exist which calculate the response of the middle atmosphere to SPEs [Jackman et al., 1995; Krivolutsky et al., 2003; Verronen et al., 2002] and, in particular, Jackman et al.  convincingly modeled the SPE ozone depletion and NOx enhancement measured by the NOAA 14 SBUV/2 and HALOE instruments in July 2000.
 The OSIRIS instrument on the Odin satellite was operational during the SPE of October 2003 although coverage was not as complete as normal due to the harsh operational environment associated with the geomagnetic storm. OSIRIS measured line-of-sight height profiles of the Oxygen InfraRed Atmospheric (OIRA), or O2(a1Δg), band mesospheric airglow emission once every two seconds with a height resolution of 1 km. This emission is an accurate proxy for mesospheric ozone during sunlit and twilight conditions as its dominant sources are related to the photolysis of ozone [Mlynczak et al., 1993]. Consequently the OSIRIS dataset affords a unique opportunity to study the time evolution of mesospheric ozone during the October 2003 SPE with unprecedented height, time and spatial resolution.
2. The Measurements
 The OSIRIS InfraRed Imager (IRI) [Llewellyn et al., 2004] consists of three vertical imagers. An imager simultaneously measures along 100 lines of sight, each separated by approximately 1 km tangent altitude. This study only considers the line-of-sight brightness from the OIRA band measured over a 10 nm passband centred at 1.263 microns. The data from each orbit are used in a tomographic retrieval scheme [Degenstein et al., 2003, 2004; McDade and Llewellyn, 1991, 1993; McDade et al., 1991] to retrieve the two-dimensional OIRA band Volume Emission Rate (VER) profile contained within the Odin orbit plane. The VER profiles are produced at 1 km height resolution every 0.2° along the orbit track. The spatial resolution achieved with this two-dimensional technique is easily adequate for the large scale size features discussed later in this work.
 Odin [Murtagh et al., 2002] is in a dusk-dawn (ascending-descending node), sun-synchronous, 98° inclination, near circular orbit at a height of approximately 600 km above the surface of the Earth. The OSIRIS instrument looks only in the orbit plane so the highest latitude sampled is approximately 82°. This is true in both hemispheres and the latitudinal coverage is constant from orbit to orbit. However, the solar conditions at each point along the orbit vary significantly throughout the year. At the time of the solar proton event the southern hemisphere portion of the Odin orbit was illuminated. The solar conditions for the time of interest were such that at the highest southern latitude, 82°, the solar zenith angle was approximately 84° and evening twilight had just begun.
 On October 28, 2003 the IRI began normal operations approximately 6 hrs after the SPE began. Prior to this, Odin was operating in a special mode and no data immediately preceding the SPE are available. OSIRIS was forced to shut down after 11 hours, or 7 orbits, due to the severe geomagnetic conditions. Two days later the IRI resumed normal operation and after a comparison with data collected the previous year it was seen that all the visible effects from the SPE had disappeared. For analysis we have seven orbits of solar storm data collected between October 28, 2003 17:14:07 UT and October 29, 2003 6:06:05 UT and a baseline orbit collected two days later starting on October 31, 2003 19:13:22 UT.
 Two examples of the VER profiles used in this study are shown in Figure 1. The upper panel illustrates the retrieved VER profile for the last orbit of OSIRIS operation during the SPE while the lower panel is the VER profile for the first orbit of OSIRIS operation after the SPE. Excluding the region of obvious depletion centred at 270° along the satellite track both profiles are visually identical except for minor dynamical perturbations. We use the lower panel as the baseline orbit for the remainder of this work because it is typical of all profiles measured at this time of year. All further comparisons and ratios are made with respect to the orbit shown in Figure 1b.
 The VER profile presented in Figure 1a clearly identifies a large depletion in the OIRA band VER centred at 270° along the satellite track. The along track extent of this region is variable as a function of altitude but has a width of approximately 60° corresponding to the time the IRI measures inside the polar cap region. Depletion is observed at all altitudes between 50 km and 80 km. Figure 1a is typical of all 7 VER retrievals during the SPE and it is only the spatial extent and the depth of the depletion region that vary with time.
 Figure 2 shows VER cross sections from the SPE and similar altitude cross sections from the baseline orbit. The cross sectional data for Figure 2a is taken at 65 km altitude while that of Figure 2b is from 55 km altitude. At both altitudes the depleted region gets deeper as the event progresses and moves in latitude due to the geographic offset of the auroral oval.
 While depletion of the OIRA band VER is very interesting, it is much more useful to examine the fractional ozone depletion that can be derived from the OIRA band emissions. In between 50 km and 80 km and under sunlit conditions the OIRA band VER is primarily due to photolysis of ozone but also has contributions from several other sources [Mlynczak et al., 1993]. In order to convert the fractional decrease of OIRA band emissions to a fractional decrease in ozone the relative contributions from ozone photolysis and the additional sources must be modelled for the normal, baseline state. Because the fractional contributions depend on the ozone profile, this was derived from the OIRA measurements with a steady-state photochemical retrieval similar to Rusch et al.  and Thomas et al. . The assumption of steady-state is reasonable but will introduce small errors as the region of interest had just entered evening twilight. A simple one-dimensional time dependent photochemical model based on Mlynczak et al.  was then used along with the retrieved ozone profile to determine the relative contributions to the baseline OIRA band VER profile from ozone photolysis and other sources. If it is assumed that the absolute production of the OIRA band by mechanisms other than those related to the photodissociation of ozone is the same during and after the SPE then the fractional decrease in ozone is easily derived from the fractional decrease in OIRA band VER. It should be noted that production of OIRA band emissions through energy transfer from O(1D) formed in the dissociative recombination of O2+ suggests that the derived ozone depletion could be underestimated.
 To determine the actual errors associated with the measurements of the fractional change in ozone is not simple as the dominant errors are systematic errors contained within the forward model and the ozone retrieval. The random errors associated with the OIRA band VER profile are very small at the altitudes of ozone depletion and because only a fractional change is required any systematic errors associated with absolute calibration disappear. At the altitudes in question where approximately 80% of the OIRA band volume emissions are a result of photodissociation of ozone it is reasonable to estimate the overall error in fractional ozone depletion is less than 10%.
 Figure 3 shows the time evolution of the altitude dependent ozone depletion at 82°S. The depletion is seen from 50 km to almost 80 km and is maximum around 65 km. The two-dimensional plot is a collage of the seven orbits collected during the storm period where the vertical profiles are separated by 1:36:30 in UT but are made at the same local time and latitude. The cross sections displayed in Figure 3 show a steady decrease in ozone with time that is layered with a maximum around 65 km.
 Figure 4 shows a polar plot of the spatial extent of the ozone depletion region along with its temporal evolution. Since orbits intersect each other at different times the storm data are separated into two polar plots, one for data collected on the ascending tracks of the orbits and the other for the descending tracks. The orbit tracks are indicated on both plots and the results between the tracks are produced with a cubic spline interpolation. Although these plots are a useful visual aid, care must be exercised as they blur the spatial and temporal components of the effects of the SPE. For each latitude time increases with longitude in a counter clockwise sense.
 The destruction of mesospheric ozone by the SPE is readily evident in the OSIRIS data set and provides a beautiful example of the phenomenon. The physical processes responsible for the destruction are thought to be well known: the SPE causes the production of water cluster ions that undergo dissociative recombination to form OH and H that, in turn, catalytically destroy ozone [Solomon et al., 1983]. The reaction rates are relatively rapid and the ozone depletion only persists for a few hours after the end of the SPE.
 The hemispherical, polar plots shown in Figure 4 are quite remarkable in that they are devoid of significant structure within the polar cap. Some of this smoothness is due to the spline interpolation of the coarse longitudinal sampling but this is certainly not the case for latitudinal slices along the orbit track which have an intrinsic sampling resolution of 0.2°. Clearly the proton precipitation was relatively uniform across the entire polar cap region and produced similar atmospheric responses at all locations. The lower boundary of the proton precipitation is readily identified from Figure 4 as the transitional region from significant ozone depletion to minor depletion and is seen to extend equatorward to 45–50°S.
 The OSIRIS data, due to operational reasons, are only available during the peak hours of the storm but nonetheless clearly show the time evolution of the mesospheric ozone in response to the SPE. The OSIRIS data begin approximately 6 hours after the start of the SPE event as recorded by the GOES 11 Proton Flux measurements (http://solar.sec.noaa.gov). By this time the mesospheric ozone density at 65 km (∼0.17 hPa) had already decreased to 50% of its normal value in response to the SPE and continued to decrease for the next 11 hours of OSIRIS observations. When OSIRIS was turned off, the ozone had depleted to 75% of normal values at 65 km and showed no sign of abatement. If the downward trend continued at its observed rate then the ozone density at 65 km would have reached zero by October 29, 2003 18:00 UT. By October 31, when OSIRIS was next powered on, the mesospheric ozone had completely recovered from the entire effect of the SPE as expected.
 The altitude of maximum ozone depletion is significantly below the altitude measured and modelled by Jackman et al.  for the SPE of July 2000. In that event HALOE observed a maximum depletion over 70% at altitudes centered on 0.05 hPa (73 km) while OSIRIS observed a maximum depletion of 75% at 65 km in the SPE of October 2003. The 8 km difference is well within the resolution of both instruments and must be assumed to result from the different response of the atmosphere to differing proton spectra. However, it is not obvious from a simple inspection of the GOES 11, 5 minute, proton data that one of these SPE is dominantly stronger than the other. A proper resolution of the issue would require accurate modelling of the mesospheric ion and neutral chemistry for the October SPE which is beyond the scope of this study.
 The OSIRIS instrument on Odin measured two dimensional OIRA band VER profiles in the Southern Hemisphere for 11 hours during the October 2003 SPE with a height resolution of 1 km and a separation of 0.2° along the orbit track. A uniform depletion of the VER was observed across the entire polar cap region throughout this period. This region extended as far north as 45°S. We conclude there was a corresponding depletion of ozone in the same region as this emission is a direct proxy for ozone under sunlit conditions.
 We have quantified the amount of ozone depletion using a simple photochemical model and referencing baseline data collected two days after the storm. Ozone depletion was observed from 50 to 80 km altitude with maximum depletion occurring around 65 km. The ozone depletion at 65 km was 50% at the start of OSIRIS observations, approximately 6 hours after the SPE onset, and steadily increased to 75% when OSIRIS was turned off 11 hours later. There was no indication that the depletion had stabilized and may have continued to increase after observations ceased.
 The observations appear to be in qualitative agreement with previous modelling studies as several models have predicted mesospheric ozone depletions greater than 70%. However, the height of maximum depletion observed by OSIRIS on October 28 2003 was 65 km while model predictions and measurements for previous storms generally place this height closer to 73 km (0.05 hPa).
 This work was supported by the Canadian Space Agency and the Natural Sciences and Engineering Research Council (Canada). Odin is a Swedish-led satellite project funded jointly by Sweden (SNSB), Canada (CSA), France (CNES) and Finland (Tekes).
- 2003), Volume emission rate tomography from a satellite platform, Appl. Opt., 42, 1441–1450.
- 2004), Tomographic retrieval of the oxygen infrared atmospheric band with the OSIRIS infrared imager, Can. J. Phys., 82, 501–515.
- 1976), Solar corpuscular emission and neutral chemistry in the Earth's middle atmosphere, J. Geophys. Res., 81, 3179–3186.
- 2000), Influence of extremely large solar proton events in a changing stratosphere, J. Geophys. Res., 105, 11,659–11,670.
- 1985), The response of ozone to solar proton events during solar cycle 21: A theoretical interpretation, J. Geophys. Res., 90, 7955–7966.
- 1990), Effect of solar proton events on the middle atmosphere during the past two solar cycles as computed using a two-dimensional model, J. Geophys. Res., 95, 7417–7428.
- 1995), Two-dimensional and three-dimensional model simulations, measurements, and interpretation of the influence of the October 1989 solar proton events on the middle atmosphere, J. Geophys. Res., 100, 11,641–11,660.
- 2001), Northern Hemisphere atmospheric effects due to the July 2000 solar proton event, Geophys. Res. Lett., 28, 2883–2886.
- 2003), Proton activity of the sun during 23rd solar maximum and its response in ozonosphere of the Earth, Adv. Space Res., 31, 2151–2156.
- 2004), The OSIRIS Instrument on the Odin Spacecraft, Can. J. Phys., 82, 411–422.
- 1991), Inversion techniques for recovering two-dimensional distributions of auroral emission rates from tomographic rocket photometer measurements, Can. J. Phys., 69, 1059–1068.
- 1993), Satellite limb tomography: Methods for recovering structured emission rates in the mesospheric airglow layer, Can. J. Phys., 71, 552–563.
- 1991), A rocket tomography measurement of the N2+ 3914A emission in an auroral arc, Planet. Space Sci., 39, 895–906.
- 1993), An updated model for O2(a1Dg) concentrations in the mesosphere and lower thermosphere and implications for remote sensing of ozone, J. Geophys. Res., 98, 18,639–18,648.
- 2002), An overview of the Odin atmospheric mission, Can. J. Phys., 80, 309–319.
- 1991), Response of the middle atmosphere to the solar proton events of August–December 1989, Geophys. Res. Lett., 18, 1019–1022.
- 1981), The effect of particle precipitation events on the neutral and ion chemistry of the middle atmosphere, 1, Odd nitrogen, Planet. Space Sci., 29, 767–774.
- 1983), Ozone densities in the lower mesosphere measured by a limb scanning ultraviolet spectrometer, Geophys. Res. Lett., 10, 241–244.
- 1983), Mesospheric ozone depletion during the solar proton event of July 13, 1982: 2. Comparison between theory and measurements, Geophys. Res. Lett., 10, 257–260.
- 1973), Decrease of ozone and atomic oxygen in the lower mesosphere during a PCA event, Planet. Space Sci., 21, 1969–1973.
- 1983), Ozone density distribution in the mesosphere (50–90 km) Measured by the SME Limb Scanning Near Infrared Spectrometer, Geophys. Res. Lett., 10, 244–248.
- 2002), Modelling the effects of the October 1989 solar proton event on mesospheric odd nitrogen using a detailed ion and neutral chemistry model, Ann. Geophys., 20, 1967–1976.