JOUWAL OF GEOPHYSICAL RESEARCH, VOL. ???, UXX, DOI:10.1029/, The impact of ozone field horizontal inhomogeneities on nadir-viewing orbital backscatter UV measurements Mart.in D. X4iiller1’3, Paul Poli2, Joanna Joiner’ M. D. Muller, Leibniz Computing Centre, Barer StraBe 21, D-80333 Munich, Germany (inar- tin. d .m ueller@ web.d e) Paul Poli, Centre National de Recherches M&korologiques, M6tko-France, 42 av. Gustave- Coriolis, 31057 Toulouse Cedex 01, France (paul.poliQmeteo.fr) Joanna Joiner, NASA Goddard Space Flight Center Laboratory for Atmospheres, Greenbelt Road, Greenbelt, MD 20771, USA (joanna.joinerQnasa.gov) ’NASA Goddard Space Flight Center Laboratory for Atmospheres, Greenbelt, USA. 2Centre National de Recherches M&orologiques, M6t&-France, Toulouse, France. 3 N oa~t Leibniz Computing Centre, Munich, Germany. D R A F T October 15, 2005, 9:21pm , D R A F T x-2 MULLER ET AL.: OZONE GRADIENT EFFECTS ON UV SOUNDINGS Abstract. Radiative transfer calculations for nadir-viewing satellites nor- mally assume the atmosphere to be horizontally homogeneous. Yet it has been shown recently that horizontal gradients can lead to significant errors in satel- lite infrared and microwave soundings. We extend the methodology to backscat- ter ultra-violet observations of ozone, and present a first estimate of the ef- fect’s magnitude. The Solar Backscatter Ultra-Violet/a (SBUV/2) instru- ment, a pure nadir sounder, serves as our test bed. Our results indicate that in a vast majority of cases the abovementioned errors can be neglected. How- ever, occurrence of higher errors, particularly at wavelengths longer than 300 nm, coincides with some of the most interesting atmospheric phenomena like tropopause folds and the South polar ozone hole. This leads to a seasonal variation of the magnitude of the effect. Due to the mostly zonal geometry of the ozone distribution, there is also the possibility that biases may be introduced, which is particularly critical if the data are to be assimilated or used to determine trends. The results presented are tested for robustness using different model atmospheres. The influence of horizonal inhomogeneities will be even more pronounced for cross-track sounders and limb viewers, and easier to detect once higher resolution atmopheric models are available. This will be inves- tigated in future studies. D R A F T October 15, 2005, 9:21pm D R A F T MULLER ET AL.: OZONE GRADIENT EFFECTS ON uv SOUNDINGS x-3 1. Introduction It has been known for a while that collocations of ground-based measurements with satellite pixels - which rarely sample the same airmass suffer from errors due to the - horizontal variability of the atmospheric quantity considered [e.g. Lambert et al., 1998; Fioletov et al., 19991. This effect is also relevant for satellite retrievals, and assimilation of retrievals or radiances; Radiative transfer models (RTMs) of the atmosphere commonly assume that the radiances measured by Earth-observing satellites pass through a hori- zontally homogeneous atmosphere. Depending on the variability of the atmospheric pa- rameters or trace gas species observed, this assumption may induce considerable random and/or systematic errors into the calculations. For limb-sounders, this realization has led to the development of 2-dimensional re- trieval algorithms, which deal with entire atmospheric slices at a time. For instance, [ Worden et al., 20041 have found the information content of simulated Thermal Emission Spectrometer CO retrievals to be higher by an order of magnitude for 2-dimensional vs. l-dimensional retrievals. Poli and Joiner [2004] and Poli [2004] report that horizontal gradients are also important for Global Positioning System radio occultation soundings. Recently, investigations have begun regarding the effect of temperature inhomogeneities on scanning nadir infrared and microwave sounders [Joiner and Poli, 20051. While gen- erally negligible, the effect was found to be considerable in some situations. This paper examines the magnitude of ozone inhomogeneities observed by a nadir- viewing instrument merely due to variation of the solar zenith angle (SZA) and solar azimuth angle (SAA). Observation times and locations of the SBUV instrument and an D R A F T October 15, 2005, 9:21pm D R A F T x-4 MULLER ET AL.: OZONE GRADIENT EFFECTS ON uv SOUNDINGS ozone analysis are used to simulate the effects of horizontal inhomogeneities. To this end: analysis and SBUV data are described in Section 2, along with geometrical considerations. Section 3 focuses on the differences between ozone seen along the slant and nadir ray path. Furthermore, simulated radiance differences for SBUV are computed from three different model fields, and the sensitivity towards the model field is investigated. Section 4 discusses the relevance of the effect in practice, and highlights some considerations for the future. 2. Data and Methods 2.1. Atmospheric model The global ozone field from an SBUV radiance assimilation system [Miiller et al., 2004,2006] based on earlier work by Stajner et al. [2001]; Riish~jgaarde t al. [2000] is used for representing ozone gradients in the atmosphere. The Chemistry and nansport Model (CTM) used for the assimilation is driven by an online version of the GEOS-4 meteoro- logical assimilation [Lin;2 004, and references therein]. This system features a horizontal resolution of 1.25" longitude x 1.0" latitude and generates analyses on 36 pressure levels between 1000 hPa and 0.2 hPa. Since we expect to find significant inhomogeneity effects where strong ozone gradients are present, the periods of 2-4 March 2003 (strong northern hemisphere (NH) ozone gradients) and 16-17 October 2003 (strong south polar ozone gradients) were selected for the various case studies presented. 2.2. SBUV data and geometry The SBUV/2 instrument used here is a 12-channel pure nadir scanner with a 200 km x 200 km footprint, circling the Earth on an ascending sun-synchronous orbit on board the NOAA-16 satellite, with an Equator crossing at 14:30 local time. We use SBUV as a D R A F T October 15, 2005, 9:21pm D R A F T MULLER ET AL.: OZONE GRADIENT EFFECTS ON UV SOUNDINGS x-5 shorthand for SBUV/2 throughout this paper. Depending on atmospheric absorption and scattering properties at the wavelength in question (here between 256 nm and 331 nm, see Table l), not all ozone in a given column will be perceived by the satellite instrument. Figure 1 shows the solar backscatter observation geometry. In order to get an approximate light path for the atmospheric backscattered radiance I, SZA and SAA are calculated for each SBUV observation location and time at the surface. Variation of the SZA with altitude is neglected, and the SAA does not vary with height. The sun-normalized radiances y measured by the SBUV instrument are usually given in “N-values”, whereby With this, the effect of the slanted incidence can be estimated using the linear approxi- mation where x,lant and x d i r are the interpolated analysis fields for slant and nadir path, in- tegrated to yield 21 partial ozone columns on half umkehr layers, as given in the SBUV level 2 retrievals. We can define the Jacobian J of the SBUV RTM for channel i and layer The same approximation is used for the SBUV radiance assimilation [Miiller et ai., 20041. For the relatively small profile differences considered, the RTM linearization errors are negligible. A radiance difference of AN = 1 translates into a 2.33% relative difference between the corresponding intensities I. D R A F T October 15, 2005, 9:21pm D R A F T X-6 MULLER ET AL.: OZONE GRADIENT EFFECTS ON uv SOUNDINGS 2.3. Approximation of the ray path To get an estimate for the average photon path through the atmosphere, the effective single scattering altitude for the incoming solar radiation first has to be estimated. Differ- ent arguments can be made in this regard, but we chose the peak of the single scattering contribution function (CF) for each SBUV channel as shown in Figure 2. A derivation of these functions can be found in [Bhartia et ai., 19961. The function shapes vary strongly with wavelength, solar zenith angle and ozone profile, less so with the atmospheric tem- perature profile (all provided with the SBUV level 2 data). Note that for the purpose of calculating the contribution functions, the nadir ozone profile at each pixel location is obtained by linearly interpolating the analysis, i.e. horizontal homogeneity is assumed. The resulting scattering altitudes for all 11 channels considered are depicted in Figure 3. For extreme SZAs, UV radiation merely glances the atmosphere, but the altitude vari- ability is also higher due to the wider range of possible ozone values in the extratropics [e.g. Hudson et al., 20031. With the direction and penetration depth known, the effective geometric ray path can be calculated using the same method as in [Poli and Joiner, 20041. In this manner, latitude and longitude are obtained for each location where the ray path penetrates one of the 36 analysis levels. The ozone slant profile “seen” by the ray is then calculated by linearly interpolating the model field of each level to these locations. Multiple scattering is not taken int,o account, nor is the finite range of scattering altitudes for each wavelength. Figure 4 shows the variation of penetration depth and direction for those SBUV pixels which exhibit a noticeable difference between the paths. This gives an impression of the D R A F T October 15, 2005, 9:21pm D R A F T MULLER ET AL.: OZONE GRADIENT EFFECTS ON UV SOUNDINGS x-7 horizontal separat,ion between the scattering location at nadir, and the point where the ray enters the model atmosphere (about 60 km altitude). 3. Results In the following, we will compare the differences between the ozone profile seen by the incoming (=slant) and outgoing (=nadir) UV radiation. Obviously, measurements of topof-atmosphere (TOA) backscattered radiation contain contributions from both path directions For the nadir-viewing SBUV, only the incoming solar ray path is affected by horizontal inhomogeneity, which means that the differences are somewhat averaged over. Calculating exactly the contribution from each direction is beyond the scope of this study. As can be seen in Figure 4b, interesting pixels for channel 9 cluster around strong gradients in the total ozone field, because this channel penetrates the atmophere already quite well. Figure 4a is somewhat harder to interpret: Comparing with Figure 2 we note that at high SZA the behaviour of channel 8 is similar the channel 9, while at mid to low latitudes it shows more sensitivity, probably because its secondary CF maximum forms around the region of the subtropical tropopause break. But note also that many of the plotted interesting pixels lie in high latitude regions because due to the satellite orbit geometry their ray paths have a strong zonal component. At 70"s latitude where the ozone hole boundary is located, the model resolution is about 110 km x 50 km (latitude x longitude). Here, gradients along a 200 km longitudinal offset can hence be well resolved. Elsewhere, the ozone gradients have to be much stronger in order to be detectable. Two extreme cases of differences between profiles seen along the nadir and slant paths are depicted in Figure 5. These occurred at the fringe of the southern hemisphere (SH) ozone D R A F T October 15, 2005, 9:21pm D R A F T X - 8 MULLER ET AL.: OZONE GRADIENT EFFECTS ON uv SOUNDINGS hole and at the boundary of an ozone poor region in the NH. The ozone field varies noticeably along the horizontal: Differences of up to 10% are seen in the ozone mixing ratios of slant and nadir paths. This is consistent with the magnitude of slant path related collocation errors reported for zenith-viewing ground instruments as compared to satellites [Lambert et al., 19981. To assess the critical height regions for these differences globally, their variability was calculated for all 2370 SBUV retrieval locations of 16-17 October 2003 (Figure 7). The sunlight penetration depth depends on both SZA and wavelength, and in turn the differ- ences between observed slant and nadir 0 3 p rofiles depend on penetration depth, SZA and ozone variability above the scattering altitude. Therefore two maxima can be identified in Figure 7: At high SZA, the slant path is longest but the penetration depth is shal- low. This leads to a maximum of slant-nadir variability at high altitudes, despite the fact that the ozone distribution is usually quite smooth in the mid-stratosphere and above. A second maximum develops at relatively low SZAs and altitudes for wavelengths reaching the troposphere, mostly in the region around the subtropical tropopause break, where O3 variability is highest. As already noted by Joiner and Poli [2005], the horizontal offsets in the mid- to lower troposphere are too small to cause appreciable ozone differences. It is also interesting to consider that due to consistent O3 variation with latitude and systematic patterns in the viewing direction (cf. Figure 4), there are also systematic biases between slant and nadir profiles (Figure 8), even in a global average. Regionally, situations like the ones depicted in Figure 5 may stand for a general trend and thus lead to larger biases. For instance, along the rim of the Antarctic ozone hole, the sun is always seen toward the Equator, and therefore towards higher ozone values, which in turn introduce D R A F T October 15, 2005, 9:21pm D R A F T MULLER ET AL.: OZONE GRADIENT EFFECTS ON UV SOUNDINGS x-9 a high bias in the nadir vs. slant radiances at, altitudes where ozone has been depleted inside the polar vortex. In order to estimate the robustness of the statistics obtained, the experiment was re- peated with the currently operational SBUV v6 profile assimilation system [R iishajgaard et al., 2000; Stajner et al., 20041 and the underlying CTM without any assimilation (Fig- ure 8b). The differences between the ozone fields are much smaller than the differences between SZA categories or wavelengths. It can hence be concluded that the results de- scribed in this paper are also valid for other CTMs and assimilation procedures with similar spatial resolution. To get an idea of the magnitude of the effect as seen by an orbital UV detector, Figure 6 shows the radiance difference for all SBUV pixels in the March case. For comparison, the measurement noise of the SBUV instruments has been estimated to be generally below 0.5% [DeLand et al., 20041. The largest differences for this time period occur at the location of the subtropical jetstreams, as the light travels through the tropopause break and its strong ozone gradients. Again, the st,ructures seen are very similar for the three ozone fields introduced for Figure 8b. Interestingly, while the free-running CTM does yield the smallest slant/nadir differences at almost all latitudes, it seems to produce more scatter around the jetstreams. This may be explained by the fact that strong ozone gradients introduced by transport tend to get smoothed out by assimilating the vertically low resolved SBUV observations with imperfect forecast and observation covariance modeling [ Stajner et al., 2004; Muller et al., 20061. The distributions for the radiance differences feature a very strong central peak around zero, with thin tails that spread beyond the measurement noise (Figure 9). This is consis- D R A F T October 15, 2005, 9:21pm D R A F T x-1 0 MULLER ET AL.: OZONE GRADIENT EFFECTS ON uv SOUNDINGS tent with what Joiner and Poli [2005] found for the AIRS instrument. For the channels between 306 nm and 318 nm as many as 10-20% of the slant-nadir differences are above the SBUV instrument noise level. However, note that in practice this alarming figure is diminished by the fact that the selected ozone hole situation represents a rather extreme case, as can be seen from the seasonal variability of radiance differences shown in Fig- ure 10. Also, while the direct comparison of slant and nadir paths is viable for direct sun ground-based sounders and limb-viewing instruments, the radiance measured by a nadir viewer like SBUV traverses along both slant (incoming) and nadir (backscatter) path. It can be easily shown that with 0 denoting the SZA, is the first-order approximation in sec0 for the measured backscattered radiance. Of course, one has to keep in mind that long slant paths, with corresponding appreciable different ozone fields, often occur at high SZA, where NmWap proaches NslanFt’.u rther- more, most modern earth-viewing satellites are cross-track or along-track scanners, where in most cases both incoming and outgoing path are slanted [cf. Joiner and Poli, 20051. 4. Conclusions In general, the errors made through neglecting the horizontal variation of the ozone field fall within the measurement noise of the SBUV instrument. However, while only a relatively small number of observations would benefit from the explicit treatment of their ray path, these tend to lie near the most interesting structures of the ozone field, e.g. ozone (mini-)holes, the subtropical jetstreams and local phenomena like tropopause folds. Especially for field campaigns at high latitudes, which involve comparisons to satellite in- D R A F T October 15, 2005, 9:21pm D R A F T