Satellite observations by SABER provide a record of thermal structure during a population of Solar Proton Events (SPEs). Anomalous polar temperature composited from the population is statistically significant over the middle atmosphere and lower thermosphere. Having form similar over the Arctic and Antarctic, it reflects anomalous warming in the mesosphere and lower thermosphere, where anomalous temperature is of order 10–20 K, and anomalous cooling in the stratosphere, where it is a couple of Kelvin. The stratospheric signature is consistent with the radiative impact of reduced ozone during SPEs. As SPEs are prevalent during years near solar max but rare during years near solar min, the thermal response becomes a consideration for interannual changes.
 During solar disturbances, coronal mass ejections emit large quantities of protons and heavy ions. Upon encountering the Earth's magnetosphere, those charged particles are guided by magnetic field lines into the polar caps. There, they precipitate to lower levels, forming a Solar Proton Event (SPE).
 Solar protons have energies of several tens of MeV, enabling them to penetrate into the middle atmosphere. Their absorption at those levels introduces an anomalous energy source, through thermalization and interaction with chemical species. Solar disturbances are prevalent during years neighboring solar max, but rare during years neighboring solar min. Consequently, SPEs represent a decadal influence on the middle atmosphere, one that is related to the decadal variation of solar activity.
 The anomalous proton flux during an SPE enters the photochemistry of HOx, through the formation and neutralization of water cluster ions, and the photochemistry of NOx, through dissociation of molecular nitrogen [e.g., Solomon et al., 1981; Jackman and McPeters, 1985; Rusch et al., 1981]. Both families destroy ozone through catalytic cycles, coupling O3 to solar activity. The HOx produced during an SPE is short lived, having an e-folding time of less than a day. The NOx produced, however, is long lived, at least in the absence of sunlight. Inside polar darkness, this feature enables NOx to be transported downward by the residual mean circulation of the middle atmosphere, the so called Brewer-Dobson circulation [Siskind et al., 1997; Callis and Lambeth, 1998]. Anomalous NOx can then influence lower levels, where ozone is concentrated. In fact, anomalous solar proton flux during an SPE is attended by magnified depletion of ozone [e.g., Thomas et al., 1983; McPeters and Jackman, 1985; Jackman et al., 2001]. During a strong SPE, wherein the proton flux is sharply enhanced, NOx can increase several fold while ozone and its column abundance suffer reductions as large as 50% [Sepålå et al., 2004; Degenstein et al., 2005].
 Inside polar darkness, ozone is generally long lived. Consequently, the anomalous O3 introduced by an individual SPE (negative) can survive for months [ibid]. Consecutive SPEs during an individual winter, likely during years neighboring solar max, can then have an impact that accumulates over the disturbed season. SPEs therefore become a consideration for interannual changes, notably between solar min and solar max.
 The influence of SPEs on composition should be accompanied by an influence on the circulation, one that is reflected in changes of thermal structure. The thermal influence can be especially important inside polar darkness, where SW heating vanishes. Anomalous heating introduced by an SPE then provides virtually the only source of heating. Beyond its direct influence on energetics (eg, through thermalization and photochemistry), an SPE also influences thermal structure indirectly, through radiative transfer. The negative anomaly of ozone introduced by an SPE leads to diminished absorption of UV once sunlight returns to the polar cap. It then introduces anomalous cooling. In the lower stratosphere, anomalous ozone also leads to diminished LW absorption, in the 9.6 μm band of O3, which likewise introduces anomalous cooling.
 The varied mechanisms through which an SPE influences the circulation makes its dynamical impact complex and not well understood. Nonetheless, that impact enters through a single, observed variable: Temperature dictates the circulation through horizontal and vertical gradients. Here, we use recent satellite observations of the middle atmosphere and lower thermosphere to construct a population of SPEs. On it, we perform a statistical analysis to composite the mean temperature anomaly introduced during an SPE, along with its statistical significance.
 The Thermosphere, Ionosphere, Mesosphere, Energetics and Dynamics (TIMED) satellite began operating during the winter of 2001–2002. On board is the Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) instrument [Mlynczak and Russell, 1995; Remsberg et al., 2003]. SABER is a limb-viewing IR radiometer. It observes chemical and thermal structure over the middle atmosphere and lower thermosphere, where SPEs are felt. Below 125 km, SABER observes temperature through 15-μm emission by CO2. At higher altitude, the temperature retrieval relies upon the abundances of CO2 and O, which must be incorporated from other sources. For this reason, our analysis is limited to altitudes below 125 km, where temperature is retrieved directly from observed radiances.
 SABER is in sun-synchronous orbit. It samples thermal structure at two local times. Each provides temperature at about a dozen longitudes. Its coverage in latitude is nearly global. However, it varies with the yaw orientation of the TIMED satellite. For two months, SABER observes thermal structure from 57° latitude in one hemisphere to 87° latitude in the other hemisphere. During this period, only one polar cap is monitored. The TIMED satellite then performs a yaw maneuver, swinging SABER's preferential field of view to the other hemisphere. Thermal structure is subsequently monitored over the opposite polar cap for two months, until the next yaw maneuver reverses the field of view again.
 The temperature record from SABER spans several years following the last solar max. Those years were punctuated by some two dozen solar proton events. The SABER record thus provides thermal structure during a population of SPEs. In contrast to prior analyses, which have concentrated upon individual SPEs, we use this population to composite a statistically-stable estimate of the mean temperature anomaly that is introduced during an SPE and, through statistical significance, to distinguish it from natural variability. (Largely random, natural variability cancels over the population of SPEs, leaving the component of anomalous temperature that varies dependently with solar proton flux.)
 The population is constructed from events that are chronicled at the site http://umbra.nascom.nasa.gov/SEP/seps.html, along with anomalous solar proton flux. Distributed indiscriminately over season, those events must be integrated with sampling of the polar caps by SABER. Its preferential field of view, in tandem with the bi-monthly yaw maneuver of TIMED, yields about a dozen SPEs when the Arctic was under observation (Table 1). All but three are outside of Boreal winter, making sunlight prevalent. SABER's preferential field of view yields another dozen SPEs when the Antarctic was under observation. Most of those are during Austral winter, enabling anomalous NOx to prevail long after individual events. Each population is large enough to construct a stable estimate of anomalous temperature indiscriminately over season, but not separately for individual seasons.
Table 1. Dates of Solar Proton Events When Thermal Structure Over the Polar Caps Was Observeda
Includes the day immediately preceding an individual event and the day immediately following it.
October 25–31, 2003
April 10–14, 2004
November 1–7, 2003
April 16–24, 2002
November 1–5, 2004
May 13–17, 2005
November 6–11, 2004
July 16–26, 2002
November 8–12, 2002
July 24–28, 2004
January 16–19, 2005
July 26–30, 2005
May 27–June 3, 2005
August 13–17, 2002
June 15–19, 2005
August 21–27, 2002
June 17–21, 2002
August 21–25, 2005
July 6–21, 2002
September 6–10, 2002
September 7–11, 2005
December 1–5, 2003
 An individual SPE is bracketed by the time when solar proton flux increases sharply and by the subsequent time when it has decayed by a factor of e−1, a few days later. Using those times to define an event, we consider temperature on the day immediately preceding an individual SPE (Ti) and on the day immediately following it (Tf). Differencing nearly-coincident observations before and after the event,
then measures the anomalous temperature introduced during the SPE. Included in the difference, however, are natural fluctuations. Unrelated to SPEs, those fluctuations vary randomly in space and time.
 To bolster statistical significance, anomalous temperature is collected from all measurements poleward of 60° at each level retrieved by SABER. Fluctuations which are wavelike or which vary randomly over the polar cap then cancel in the mean over measurements, with a commensurate increase in the number of degrees of freedom (dof). Collecting measurements over the Arctic and Antarctic during all solar proton events leads to a similar increase in dof. The result is a population of anomalous polar temperatures ΔT that are contemporaneous with SPEs.
 We wish to establish if the mean temperature anomaly in this population differs significantly from zero, namely, if polar temperature following an SPE, Tf, is consistently different from that preceding the SPE, Ti. The issue is addressed by a Matched Pair t-test [Dowdy and Weardon, 2004]. The matched pair is comprised of nearly-coincident observations of Ti and Tf (eventually averaged over the polar cap), which define the temperature anomaly ΔT. Applied to many such observations, this statistical test, de facto, compares the mean polar temperature before and after an individual SPE.
 The null hypothesis is that the mean temperature anomaly is zero (ie, that Tf does not differ significantly from Ti. The test statistic is then
where N is the number of independent temperature pairs, 2N reflecting the dof, and
accounts for interdependence of temperature before and after an SPE.
 On the days immediately preceding SPEs (Table 1), SABER collects some 28,000 temperature measurements over the polar cap at an individual altitude (±2.5 km). Another ∼28,000 are collected on the days immediately following those SPEs. Thus, in principle, SABER provides a population of some 28,000 matched pairs, which in turn define a respective population of temperature anomalies ΔT. In practice, however, only a fraction of those are independent, because the measurements are made on space and time scales over which temperature is correlated.
 To evaluate the dof, the autocorrelation of ΔT has been calculated along the sequence of SABER observations that are collected along the satellite track. (The sequence of asynoptic measurements is windowed to the days immediately before SPEs (Table 1) and to latitudes poleward of 60°. A companion sequence is formed by windowing to the days immediately following SPEs. The two sequences of SABER measurements are then reordered to represent matching sites on the days immediately before and after SPEs. The resulting time series thus provide a sequence of nearly-coincident matched pairs (Ti, Tf), which define a population of temperature anomalies ΔT.) Plotted in Figure 1 at different altitudes, the autocorrelation measures the number of adjacent observations that are mutually dependent. It is nearly identical at different altitudes. The autocorrelation has the form of a decaying oscillation, reflecting the nearly periodic sampling of the SABER orbit. It decays through an e folding after approximately 400 consecutive observations. Implied is an autocorrelation scale of less than 800 consecutive observations that are mutually dependent (twice the half width of the autocorrelation function). Dividing this scale into the ensemble of ∼28,000 matched pairs then yields more than 70 dof for the overall population of temperature observations.
Figure 2 plots, as a function of altitude, the mean temperature anomaly over the Arctic following an SPE. In the mesosphere and thermosphere, temperature is anomalously warm. Anomalous temperature has values of 2–3 K in the mesosphere. However, it increases sharply above the turbopause near 105 km, where it exceeds 20 K. The accompanying uncertainty (shaded) indicates that the warm anomaly over the mesosphere and lower thermosphere differs significantly from zero at the 95% level (ie, the null hypothesis can be rejected with a probability of error of less than 5%). The warm anomaly can be introduced directly, through thermalization of incident protons and attendant photochemistry. As most of the events which contribute to the mean difference over the Arctic are in sunlight (Table 1), the warm anomaly could also be introduced through anomalous SW flux contemporaneous with SPEs. This possibility, however, is not supported by similar structure that is found over the Antarctic, where most of the events are in darkness (see below).
 In the stratosphere, below about 60 km, anomalous temperature reverses sign. Achieving values of 1–2 K, temperature there is anomalously cold. Although weaker, the stratospheric thermal signature is, like that in the mesosphere and thermosphere, statistically significant. The cold anomaly cannot be introduced via thermalization of incident protons. However, it is consistent with reduced ozone during an SPE, which introduces anomalous radiative cooling. (The radiative time scale in illuminated regions is only a couple of days.) The altitudes of cooling, in fact, coincide with those of anomalous O3, which is concentrated below 60 km [Sepålå et al., 2004; Degenstein et al., 2005].
Figure 3 plots the mean temperature anomaly over the Antarctic. It has much the same form as that over the Arctic. A warm anomaly prevails in the thermosphere, where anomalous temperature approaches 15 K, and again in the lower mesosphere. Similarly, a cold anomaly prevails in the stratosphere. Analogous to features over the Arctic, both are significant at the 95% level.
 In contrast to anomalous temperature over the Arctic, the warm anomaly is interrupted near the mesopause, where temperature is anomalously cold. Likewise, the cold anomaly in the stratosphere does not appear until 40 km, somewhat lower than over the Arctic. These distinctions are most likely a reflection of the disparate seasonality of the populations over the Arctic and Antarctic (Table 1). Over the Arctic, observed SPEs are distributed indiscriminately over season. Many coincide with months when the polar cap is illuminated and, thus, can experience anomalous SW heating. However, over the Antarctic, more than half of the SPEs are observed during winter. The polar cap is then shrouded in darkness, eliminating SW heating while rendering anomalous NOx and its impact upon ozone long lived.
 The population of SPEs observed by SABER is large enough to establish a significant thermal perturbation to the circulation of the thermosphere and middle atmosphere. In the lower thermosphere, temperature over the polar cap is anomalously warm by some 10–20 K. In the middle atmosphere, anomalous temperature achieves values of a couple of Kelvin, anomalously warm in the mesosphere and anomalously cold in the stratosphere.
 Anomalous thermal structure over the Arctic has the same general form as that over the Antarctic, even though the two were composited from independent events. Further, the SPEs observed over the Antarctic are concentrated during winter months, when the polar cap receives little sunlight (Table 1). In fact, restricting the population to the dead of winter (Jun–Aug) recovers nearly identical structure (not shown). Anomalous temperature can therefore have only a minor contribution from anomalous SW flux. This leaves as an explanation anomalous proton flux during SPEs, which is guided into the polar caps irrespective of season.
 The overall structure of anomalous temperature reflects a combination of mechanisms through which SPEs influence thermal structure. However, anomalously-cold temperature found in the stratosphere implies the prevalence there of radiative cooling, in response to diminished ozone at those altitudes.
 The cold anomaly in the polar stratosphere could also be introduced through adiabatic cooling, via anomalous upwelling that is driven overhead by anomalous diabatic heating. Regardless of how it is introduced, the thermal anomaly during SPEs invades the middle atmosphere. It should therefore influence the Brewer-Dobson circulation, especially over the winter pole where other sources of heating vanish. Because it is prevalent near solar max but virtually absent near solar min, anomalous heating during SPEs becomes worthy of consideration for interannual changes in the middle atmosphere.
 Although anomalous temperature recovered from the SABER record is statistically significant, it is sensitive to which seasons are included in the population of observed SPEs. This dependence is consistent with the suite of photochemical and radiative mechanisms through which SPEs influence thermal structure. Nevertheless, expanding the population of observed SPEs will enable the structure and seasonality of anomalous temperature to be brought into sharper focus. An expanded population will also enable anomalous temperature to be regressed against anomalous proton flux, quantifying the sensitivity of thermal structure to SPEs. Along with others, these advances will have to await the next solar max, when new observations will hopefully document the thermal response to more solar proton events.
 The authors are grateful for constructive remarks provided during review. This work was supported by NSF grant ATM-0120512.