Comment on “Long‐term variation in the thermosphere: TIMED/GUVI observations” by Y. Zhang and L. J. Paxton

[1] Zhang and Paxton [2011] report on an investigation into the behavior of the dayside ΣO/N2 ratio (for O to N2 vertical column densities down to an N2 depth of 1017 cm−2) between 2002 and 2008. Their results were obtained with dayglow data from the Global Ultraviolet Imager (GUVI) onboard the Thermosphere, Ionosphere Mesosphere, Energetics, and Dynamics (TIMED) satellite. The data of interest come from the 135.6 and LBHS spectral channels spanning the wavelength intervals 133.5–137.7 nm and 141.0–153.0 nm. These regions are dominated by OI 135.6 nm and N2 LBH emission, respectively. After dayside global averaging, a decrease of ΣO/N2 (hereafter, referred to as remotely sensed ΣO/N2 to distinguish it from MSIS-based ΣO/N2 addressed in Zhang and Paxton's reply) is observed with declining solar activity along with variations arising from geomagnetic activity and seasonal effects. The authors attribute the decrease in their remotely sensed ΣO/N2 to a corresponding decrease in the altitude where the N2 column density is 1017 cm−2 (to be referred to as zR) arising from thermospheric cooling. The following quote is from Zhang and Paxton [2011, paragraph 1]: "The O/N2 dependence on Qeuv is due to thermal expansion or contraction that alters the reference height of the fixed N2 column density (1017 cm−2)." The content of the quote is repeated elsewhere in the text and is the main point to be taken away from their work. [2] The suggestion of temperature affecting the remotely sensed ΣO/N2 ratio derived from GUVI 135.6/LBHS is in conflict with the findings reported by Strickland et al. [1995]. There, it was noted that the proper and necessary way to understand O concentration changes from satellite observations of OI 135.6 nm and N2 LBH dayglow from the Earth's disk is not in terms of altitude but in terms of column densities, including total column density. Strickland et al.'s equations 8 and 9 present 135.6/LBH and ΣO/N2 as ratios of integrals involving mixing ratios as functions of total column density. Strickland et al. [1995, Figure 3] demonstrate that thermal effects on density profiles disappear when they are plotted versus total column density. The figure proves that two atmospheres, one cold and the other hot, are indistinguishable versus total column density when both yield the same 135.6/LBH ratio. Conversely, any given observation of 135.6/LBH is associated with an essentially unique atmosphere versus total column density, regardless of the altitude of the reference point. A demonstration of the degree of uniqueness may be seen in Figure 9 of Strickland et al. [1995] which shows a nearly linear relationship with little scatter between the two ratios using 324 unscaled TIGCM atmospheres. [3] We note here that Zhang and Paxton have written a reply to this comment which expands on their conclusion that remotely sensed ΣO/N2 is affected by temperature, which, if true, carries the implication that a generalization of the Strickland et al. algorithm is required to expand it from one temperature invariant lookup table to a series of temperature dependent tables. Figure 3 from Strickland et al. [1995], however, demonstrates that all such tables would be essentially identical. In other words, the claim by Zhang and Paxton that temperature affects remotely sensed ΣO/N2 is dubious since the algorithm they used to deduce the effect does not contain an explicit dependence on temperature. [4] Temperature effects do come into play once a value of remotely sensed ΣO/N2 is obtained. If one wishes to specify an O density profile from this value, it is best to select a model atmosphere with an exospheric temperature thought to be consistent with the time and location of the observations. The estimate of the temperature must be based on additional information, however, since there is no explicit knowledge of temperature available in remotely sensed ΣO/N2. The model O profile can then be scaled by the ratio of the remotely sensed ΣO/N2 to the corresponding value calculated from the model atmosphere. [5] In their reply, Zhang and Paxton state that we have used an "incorrect mental model" in which remotely sensed ΣO/N2 is independent of temperature. It is important to understand, however, that there are two different ideas being discussed by the two sets of authors. This comment addresses the relationship between dayglow (the 135.6/LBH ratio) and ΣO/N2 derived from it. Conversely, Zhang and Paxton address the behavior of ΣO/N2 derived from MSIS atmospheres in complete absence of the remote sensing problem (i.e., no consideration of the constraints placed on ΣO/N2 by 135.6/LBH). The remaining discussion in our comment will first focus on the incorrect conclusion from their paper and at the end will address their reply. [6] We do not dispute that zR changes with exospheric temperature (due to thermal expansion or contraction). In spite of this, a static diffusion model with fixed volume densities at a fixed lower boundary in the thermosphere [e.g., Walker, 1965; Jacchia, 1977] produces insignificant changes in either 135.6/LBH or ΣO/N2 as the exospheric temperature changes (for further discussion of this behavior, see Strickland et al. [2004, paragraph 8]). To be consistent with the findings in the 1995 study, a decrease in remotely sensed ΣO/N2 must arise from a decrease in 135.6/LBHS which in turn must reflect a decrease in atmospheric O relative to N2. [7] The choice of 1017 cm−2 is made because it offers the best (most unique) relationship to the 135.6/LBH ratio. An N2 column density of this value should always be used as the base of the column for computing the O column density and ΣO/N2 when relating radiance observations to models. It is not a variable of the modeling process. [8] Zhang and Paxton reach their conclusion that ΣO/N2 dependence on Qeuv is due to thermal expansion or contraction that alters zR by showing a correlation between ΣO/N2 and zR using the MSIS-86 model [Hedin, 1987]. This is presented in Figure 6 of Zhang and Paxton [2011] versus the solar activity proxy F10.7. We do not dispute this correlation, but their discussion is now uncoupled from remotely sensed ΣO/N2 and in turn is not relevant to their observed decrease in GUVI-based ΣO/N2. As already noted above, any decrease must arise from associated 135.6/LBH decreases, which, in turn, reflect a drop in overall concentration of O relative to N2 in the atmosphere being observed. MSIS itself exhibits this latter behavior with declining solar activity. Smith et al. [2010] report on O variations in the upper mesosphere, also using TIMED data but from the SABER (Sounding of the Atmosphere using Broadband Emission Radiometry) instrument. Their observing period is from 2002 to 2010, similar to that considered by Zhang and Paxton (2002 to 2008). In Figure 10 of Smith et al. [2010], dayside O at 94 km is observed to decrease with decreasing solar activity starting with SABER observations in 2002. We quote the following words from paragraph 67 of Smith et al.'s [2010] paper: "As the solar UV flux declines, there is reduced photolytic production of O and H, reduced temperature …, and reduced magnitude of molecular diffusion. All of these could contribute to a solar cycle variation in O." [9] The more likely explanation for the changes seen by Zhang and Paxton in their remotely sensed ΣO/N2 is in O production (by photodissociation), transport (diffusion and convection), and recombination chemistry and not "thermal expansion or contraction that alters the reference height of the fixed N2 column density." The key to a given remotely sensed ΣO/N2 value is the O density in the vicinity of its peak (∼95 km), which, for a given profile shape, determines the magnitude of this profile. Peak density is sensitive to the above processes as well as to heating and associated upwelling at nearby higher altitudes. Upwelling on the dayside increases with solar activity leading to decreases in dayside ΣO/N2. This is more than countered by increased O production if remotely sensed ΣO/N2 increases with solar activity, as reported by Zhang and Paxton. In support of this conclusion we note that the ∼20% decrease in averaged daytime O from 2002 to 2010 reported by Smith et al. is within the uncertainty of the ∼30% decrease in GUVI ΣO/N2 reported by Zhang and Paxton over the same time period. [10] As noted above, the Zhang-Paxton reply is uncoupled from the remote sensing problem in that it addresses a series of MSIS runs with no consideration of the relationship between MSIS ΣO/N2 and 135.6/LBH. We take exception, as we did in section 2, to statements that "the change in the N2 reference height is the dominant source for the observed O/N2 change" if such statements are meant to refer to remotely sensed ΣO/N2. Their findings regarding the behavior of MSIS [O]/[N2] (volume density ratio) and MSIS ΣO/N2 versus F10.7 (81-day average and previous-day values as model inputs) and altitude may be of general interest but are of no consequence to the remote sensing problem since there are no constraints which a measured value of 135.6/LBH can place on ΣO/N2. To illustrate this point, we note that in their reply Zhang and Paxton claim "the variation in [O]/[N2] (16–25%) at reference heights and above contributes a small portion to the variation in ΣO/N2 (80%). The major contribution to the ΣO/N2 variation has to come from the changes in N2 density profile and therefore the N2 reference height." Yet they show that the exospheric temperature varied from ∼1400 K to ∼700 K from 2002 to 2008 [Zhang and Paxton, 2011, Figure 5] whereas the remotely sensed ΣO/N2 varied by ∼30%. The fact that the 80% variation predicted by MSIS ΣO/N2 is considerably higher than the ∼30% variation observed in remotely sensed ΣO/N2 is not surprising, since a comparison between these independent quantities is not meaningful. [11] Robert Lysak thanks the reviewer for his or her assistance in evaluating this paper.