A comparison of ground-based hydroxyl airglow temperatures with SABER / TIMED measurements over 23 ◦ N , India

Ground-based observations of OH (6, 2) Meinel band nightglow were carried out at Ranchi (23.3 N, 85.3 E), India, during January–March 2011, December 2011–May 2012 and December 2012–March 2013 using an all-sky imaging system. Near the mesopause, OH temperatures were derived from the OH (6, 2) Meinel band intensity information. A limited comparison of OH temperatures (TOH) with SABER/TIMED measurements in 30 cases was performed by defining almost coincident criterion of ±1.5 latitude–longitude and ±3 min of the ground-based observations. Using SABER OH 1.6 and 2.0 μm volume emission rate profiles as the weighing function, two sets of OHequivalent temperature (T1.6 and T2.0 respectively) were estimated from its kinetic temperature profile for comparison with OH nightglow measurements. Overall, fair agreement existed between ground-based and SABER measurements in the majority of events within the limits of experimental errors. Overall, the mean value of OH-derived temperatures and SABER OH-equivalent temperatures were 197.3± 4.6, 192.0± 10.8 and 192.7± 10.3 K, and the ground-based temperatures were 4–5 K warmer than SABER values. A difference of 8 K or more is noted between two measurements when the peak of the OH emission layer lies in the vicinity of large temperature inversions. A comparison of OH temperatures derived using different sets of Einstein transition probabilities and SABER measurements was also performed; however, OH temperatures derived using Langhoff et al. (1986) transition probabilities were found to compare well.


Introduction
The mesosphere-lower thermosphere (MLT) region (80-105 km) of the earth's atmosphere is a complex system that is strongly controlled by several physical processes from above and by dynamical processes from below.Its thermal structure is influenced by the absorption of incident solar radiation, auroral heating by currents and particles, solar and secondary energetic particles, incoming cosmic flux, and infrared radiative cooling due to CO 2 .The dynamical forcings (viz.gravity waves, tides and planetary waves) and the anthropogenic changes due to human activity together with the lower atmosphere strongly influence its thermal structure as well (Brasseur and Solomon, 1984;Mlynczak and Solomon, 1993;Mlynczak, 1997;Smith, 2004).Owing to this, knowledge of temperature of the MLT region is one of the crucial parameters in understanding its structure and dynamics.
N. Parihar et al.: A comparison of ground-based hydroxyl airglow temperatures and Telescopes for the Atmosphere (CRISTA) satellite instrument, and Sounding of the Atmosphere by Broadband Emission of Radiation (SABER) on-board the TIMED mission satellite have also contributed immensely to our knowledge of the temperature field of the MLT region (von Savigny et al., 2004;Scheer et al., 2006;Xu et al., 2007;Mulligan and Lowe, 2008;French and Mulligan, 2010;Sheese et al., 2011;García-Comas et al., 2012, 2014;and references cited therein).
Although the ground-based observations provide invaluable information on the local time domain and have tremendous capability for long-term operation with costeffectiveness, they alone are probably not enough to resolve critical issues being globally restricted by landmass distribution.Conversely, the space-borne measurements are unable to provide local information.As such, the coordinated use of similar geophysical datasets is needed to obtain a better understanding of the MLT region (especially to explore seasonal geographical variations and long-term trends).Several investigators have reported the comparative study of the ground-based OH temperatures and satellite-borne measurements.Von Savigny et al. ( 2004) first reported nearglobal satellite-borne measurements of OH (3, 1) Meinel band temperatures performed with the SCIAMACHY instrument.A comparison of these temperatures with the groundbased OH temperatures at Maui, Hawaii (21 • N), Hohenpeißenberg (47 • N) and Wuppertal (51 • N) indicated that the two sets of measurements are in good agreement with each other, with the mean difference being 7.1, 2.6 and 2.7 K respectively.At Wuppertal (51 • N), Oberheide et al. (2006) found OH (3, 1) temperatures to be systematically warmer than the SABER measurement (on average by 7.5 K) during 2003-2005.Using the mesopause region temperature measurements by the CRISTA-1 and CRISTA-2 missions, Scheer et al. (2006) performed such a comparison of groundbased OH temperature measurements at eight sites (spread over 38-63 • N and 32-69 • S).López-González et al. (2007) reported the measurements of OH (6, 2) Meinel band and O 2 (0, 1) atmospheric band temperatures with Spectral Airglow Temperature Imager (SATI) at the Sierra Nevada Observatory (37 • N) and their comparison with SABER observations.These authors noted (i) a similar night-to-night as well as seasonal variation of the temperatures from the two datasets, (ii) SABER temperatures to be colder than SATI measurements by ∼ 5.7 K at 87 km and (iii) SATI temperatures to be colder than SABER measurements by ∼ 2.5 K at 95 km.Mulligan and Lowe (2008) performed a comparison of OH (3, 1) temperatures with ACE-FTS and SABER measurements at three airglow stations -Wuppertal (51 • N), Maynooth (53.2 • N) and Stockholm (59.5 • N).These authors found (i) OH equivalent temperatures derived from ACE-FTS to be in good agreement with SABER observations and (ii) OH temperatures to be warmer than satellite measurements in the 4.5-8.6K range.French and Mulligan (2010) presented an extensive comparison of OH (6, 2) temperature measurements at Davis (68 • S), Antarctica, with Aura-MLS and SABER measurements during 2004-2009and 2002-2009 respectively.These authors observed an annual increasing trend (∼ 0.7 K year −1 ) in warm bias between OH temperatures and SABER measurements and an opposite constant bias of ∼ 10 K between OH temperature and Aura-MLS observations.Sheese et al. (2011) compared O 2 temperature measurements using the ORISIS instrument with SABER and SOFIE values.These authors found ORISIS temperatures to be lower than SABER and SOFIE measurements.Overall, these studies indicate a warm bias between the ground-based OH temperatures and the satellite measurements.Such comparisons serve as a means of identifying biases between the two measurement methods and substantiating their combined use in understanding the MLT thermal structure.
In the present study, a limited comparison of almost coincident measurements of the ground-based OH (6, 2) temperatures and SABER observations was performed at a lowlatitude station in Ranchi (23 • N), India, using a strict spatial and temporal coincidence criterion of ±1.5 • latitudelongitude and ±3 min.As OH temperature represents the weighted temperature of OH emission layer, first the OHequivalent temperatures was estimated from the SABER kinetic temperature profiles and then a comparison was made with ground-based airglow temperatures.Using different sets of Einstein transition probabilities, OH temperatures were derived and a comparison was performed so as to identify the closest matching set of transition probabilities for OH temperature measurements.).The filter wheel unit and CCD detector are thermoelectrically maintained at 25 and −80 • C respectively.Around 825 nm, the quantum efficiency of the CCD detector is 70-80 % and the dark current is less than 0.001 e − pixel −1 s −1 .Parihar and Taori (2015) described this imaging system in detail.Optical filters for monitoring P 1 (2) line and P 1 (4) line had a bandwidth of ∼ 1.1 nm, while that of the 857.0 nm filter was ∼ 2.0 nm.The transparency of these filters is in the range of 77- 86 %.Furthermore, OH emissions in the wavelength range of 705-929 nm were monitored using a 200 nm broadband filter with a transparency of 87 %.More details of the filters used for OH nightglow observations are presented in Table 1.Each emission was monitored for 60 s and the duration of one complete sequence of six filters was 6 min.Such a choice of exposure time for the P 1 (2) line, P 1 (4) line and 857.0 nm filter is based on the experimental set-up of the Mesospheric Temperature Mapper (MTM) described by Taylor et al. (1999).For these observational settings, the signalto-noise ratio was better than 70.Nightglow observations were performed in campaigns (each around 13 days centred on the new moon period and under clear sky conditions) during January 2011-March 2011, December 2011-May 2012 and December 2012-March 2013.Overall, around 120 days of good-quality observations were available for meaningful study.
2.1 Intensity data of P 1 (2) line, P 1 (4) line and the background emission At OH emission height, the field of view of the imager was fairly able to cover a 3 • × 3 • latitude-longitude grid centred on Ranchi.This region of the image has been considered as the region of interest in the present study.For possible contamination with artificial lights of the city around the airglow site, the image data at further lower elevation were avoided.First, in the image data, five sample locations corresponding to Ranchi (named RNC) and to the vertical projections from the OH airglow layer to the geographical locations ±1.5 • latitude-longitude north, south, east and west of Ranchi (named NoRNC, SoRNC, EoRNC and WoRNC respectively) were identified within the 3 • × 3 • geographical grid.Next, the average intensity of a square bin (centred on each sample location and enclosing a circular field of view of ∼ 4 • ) was estimated for an emission feature.Using this intensity and timestamp information of the associated image and repeating this process for the entire dataset of the emission concerned, the time series for each of the sample locations were generated.Using this technique, the time series of the intensity of P 1 (2) line, P 1 (4) line and background emissions for five locations were recorded for further analysis.This process was described in detail in Parihar and Taori (2015).A typical example of such a generated intensity series of P 1 (2)   1.Assuming the temperature dependence of transparency of filters and sensitivity of CCD to be the main sources of the instrumental errors, systematic and random error in the intensity measurements are estimated to be ∼ 8 and 3 % respectively.

OH temperature measurements
Using the well-known ratio approach suggested by Meinel (1950), OH rotational temperatures (hereafter T OH ) were derived from the intensity information of P lines of OH (6, 2) Meinel band.The details of the temperature retrieval from the intensity information of the P 1 (2) and P 1 (4) lines of the OH (6, 2) band are presented elsewhere (Parihar and Mukherjee, 2008;Parihar et al., 2013).Here, the term values given by Kendall and Clark (1979) and the transition probabilities given by Langhoff et al. (1986) were used in deriving the airglow temperatures.The plot of solid circles in Fig. 1 presents an example of such derived temperature on the particular night of 25 December 2011 at the NoRNC location.Using uncertainty information in the intensity measurements, systematic and random errors in the derived T OH are estimated to be 4.2 and 2.0 K respectively.In Fig. 1, the error bars represent this uncertainty in the derived T OH .The southern edge of imaging observations was often affected by artificial lights from surroundings, and T OH of SoRNC was discarded for comparison.Figure 2 presents T OH variation during the night of 10 February 2013 at different sampling locations, wherein the southern edge of images was contaminated.T OH at contaminated SoRNC was found to be systematically lower than that of other sampling locations.The starred discontinuous plot in Fig. 2 presents the variations in averaged T OH over all locations (except the contaminated SoRNC).The corresponding error bar denotes the systematic error of 4.2 K in T OH measurements.

SABER instrument and OH equivalent temperature data
SABER is a 10-channel broadband infrared radiometer on board NASA's TIMED mission satellite that measures earth limb emission (between 1.27 and 16.9 µm) from the lower stratosphere to the lower thermosphere using the limb scanning technique.It was designed to globally explore the energetics, chemistry, dynamics and transport processes of the MLT region on temporal and seasonal timescales.The sounding of the atmosphere by SABER gives vertical scans of limb radiances and their analysis provides information on temperature, pressure, O 3 , H 2 O and CO 2 mixing ratio, the volume emission rates of O 2 ( 1 ) airglow, and OH airglow with vertical resolution of approximately 2 km at the tangent point (see http://saber.gats-inc.com).For example, the kinetic temperatures are retrieved from the CO 2 emission at 15 µm (Mertens et al., 2001).In the present study, the profiles of the kinetic temperature (T k ) and the volume emission rate (VER) of OH emissions (at 1.6 and 2.0 µm) of SABER Version 2.0 Level 2A data were used.

Coincidences of SABER/TIMED overpasses and airglow measurements
In principle, the concurrent measurements from the same location should be considered for such comparisons.It is well known that exact coincidences are not possible, spatial coincidence criterion of ±2.5 • latitude-longitude of Ranchi and temporal coincidence criterion of ±3 min were defined in the present study.All through 13 campaigns of airglow experiments during 2011-2013, nearly 80 such passes were available in a latitude-longitude grid of 5 • × 5 • centred over Ranchi (i.e. in the geographical bin of 20.8-25.8• N and 82.5-87.8• E) and were examined.After defining five sample locations, viz.RNC, NoRNC, SoRNC, EoRNC and WoRNC, in a two-dimensional image, this spatial coincidence is further contracted to a ±1.5 • latitude-longitude bin.This matched the SABER's horizontal resolution of ∼ 300 km along the line of sight (Xu et al., 2006).Complying with this, about 30 coincidences were available for further study.Such a choice of spatial and temporal coincidence is expected to fairly address the limited comparison of two measurements.On 21 nights, such SABER coincidences existed.Sometimes two coincidences in the same day at different locations were also noted.

OH-equivalent temperatures from SABER kinetic temperature profiles
As hydroxyl emissions emanate from an extended altitude regime of ∼ 8-10 km width centred around 87 km (Baker et al., 2007;Nikoukar et al., 2007), an approach concerned with OH-equivalent temperature from the SABER kinetic temperature profiles was adopted by several investigators for fair comparison of ground-based and SABER measurements (Oberheide et al., 2006;López-González et al., 2007; Mulli-   while OH (5, 3) and (4, 2) band emissions are mainly collected at 1.6 µm (Baker et al., 2007).As OH emissions from different υ peak at different heights (von Savigny, 2015; Noll et al., 2016), the difference in OH 2.0 and 1.6 µm emission peaks is observed.Generally, the 2.0 µm emission peak lies above that of the 1.6 µm channel by ∼ 1.6 km at about 89 ± 2 km, thereby resulting in different values of two OHequivalent temperatures.In about 20 % of the events, the difference was about 1 K or less and 4 K or less in about 53 % of the events.Sometimes differences as high as 11-13 K were also noted.Figure 5 presents the scatter plot of the separation of 1.6 and 2.0 µm VER peaks and the difference of two OH-equivalent temperatures.Note that two data values (marked in red) show large temperature differences even for small separation of two emission peaks, and are thus considered as outliers.Barring two exceptions, a linear fit was applied to the scatter plot and indicates a marginal linear dependence of the temperature difference of T 1.6 and T 2.0 on the separation of their peaks with a Pearson correlation coefficient of ∼ 0.49.Also, the FWHM thickness of both profiles was generally 8 ± 2 km.OH-equivalent temperature estimated using a SABER kinetic temperature profile strongly depends upon the choice of weighing function used.French and Mulligan (2010) estimated OH-equivalent temperatures using different weighing functions and found a maximum difference of 3 K among the different selections considered.Herein, OH-equivalent temperatures were also determined using another type of weighing function -a Gaussian fitted to SABER's OH VER profiles (based on French and Mulligan, 2010); however, the disparity between the ground-based and satellite measurements was found to increase further.Hence, T 1.6 and T 2.0 determined using OH VER weighting was con-  2 summarizes coincidental SABER and T OH measurements along with the difference observed between two measurements.The difference can mostly be seen to lie in the range of 0-15 K.A difference as high as 32 K is also noted on 22 May 2012.
Sometimes the horizontal field of view of the SABER instrument subsumed two or more of the sampling locations; however, agreement improved between two measurements in general as SABER approached Ranchi.In Fig. 7b, the encircled SABER overpass is closer to RNC than the other one and is observed to be in good agreement with T OH .Clearly T OH can be seen to match better with T 2.0 than with T 1.6 .In about 37 % of cases, fair agreement between T OH and T 2.0 can be noted within the limits of experimental uncertainty of about 5.6 K (the combined systematic error of T OH and SABER measurements), and indicates that the matching of two measurements is within an acceptable range of their combined systematic error.A difference of 7-12 K between T OH and T 2.0 was observed in a large number of remaining coincidences.Scheer et al. (2006) reported the difference of 0.2-11.5K between the ground-based T OH and CRISTA temperatures.As a trial, OH-equivalent temperatures were calculated using an intermediate weighing function, viz. the average of OH 1.6 and 2.0 µm VER profiles, and compared it with T OH measurements.However, the difference between two measurements was found to increase further.Some of these cases are of interest in the context of a report on SABER kinetic temperature errors by García-Comas et al. (2008).In the presence of large vertical gradient in temperature, which may be produced by tidal influence, inversion layers or other phenomenon, the temperatures derived using non-LTE (local thermodynamic equilibrium) algorithms are highly sensitive to uncertainty in collisional rates.These errors will be more dominant in the region of a large temperature gradient.García-Comas et al. (2008) estimated the maximum error of ±8 K in SABER temperature at around 90 km, especially near the crest and trough of the temperature inversion, for a typical case in such a situation.A similar scenario was noted in several events reported herein, and Fig. 10 presents a few examples of such cases.On 21 February 2012, a strong temperature inversion of more than 35 K was seen in SABER kinetic temperature profiles in the 83-88 km region.Near the vicinity of this inversion crest (about 88 km), a peak of OH 2.0 µVER can be noted.As such, errors in estimated OH-equivalent temperature are expected for the reasons pointed out by García-Comas et al. (2008).This possibly resulted in the large difference of 8-16 K between the two measurements for this coincidence.This difference lessened as the separation between the crest of inversion and OH peak increased (as can be seen for 17 March 2012).Generally, T OH was found to be lower than SABER OH-equivalent temperature whenever the temperature inversion existed and the OH emission peak lay in close proximity to its crest.In such cases, the OH layer generally lay around or below 89 km.A similar discrepancy between two measurements was noted during the presence of strong tides as well.
Overall, the mean values of T OH , T 1.6 and T 2.0 were 197.3 ± 4.6, 192.0 ± 10.8 and 192.7 ± 10.3 K respectively.The ground-based temperatures were 4-5 K higher than SABER values, similar to earlier reports (von Savigny et al., 2004;Oberheide et al., 2006;Scheer et al., 2006;López-González et al., 2007;Mulligan and Lowe, 2008).As T OH derived from Meinel band line intensities strongly depend on the choice of transition probabilities (French et al., 2000), T OH were also derived using the transition probabilities given by Mies (1974), Turnbull and Lowe (1989) and Goldman (1998); however, the mean difference between T OH and SABER measurements increased from 3 to 15 K with their use.The T OH derived using Mies (1974) and Goldman (1998) were in general 8-9 K warmer than SABER measurements.A difference of about 15 K was noted between T OH derived using transition probabilities given by Turnbull and Lowe (1989) and SABER measurements.

Conclusions
A limited comparison of the ground-based measurements of hydroxyl temperatures around the mesopause region were made with the OH-equivalent temperatures retrieved from SABER on-board TIMED observations of 30 coincidences.The results of comparison are very encouraging in the sense that the ground-based temperatures derived using the transition probabilities given by Langhoff et al. (1986) are in good agreement with the satellite retrievals, within the limits of experimental errors.Similar to earlier reports (von Savigny  et al., 2004;Oberheide et al., 2006;López-González et al., 2007;Mulligan and Lowe, 2008), OH ground-based temperatures are found to be warmer than SABER measurements by 4-5 K on average in this present study.This study also indicates that the difference between two measurements is large (8 K or more) in cases when OH layer lay in the vicinity of large temperature inversions.T OH was found to be closer to T 2.0 in comparison to T 1.6 .In the future, efforts will be made to perform such a study with a longer dataset of ground-based OH temperature measurements.
The OH temperature data used in this study are available upon request from Navin Parihar (email: navindeparihar@gmail.com).

Figure 1 .
Figure 1.A typical example of the intensity series generated using the image data and airglow temperatures over a geographic spot situated 1.5 • latitude north of Ranchi (named NoRNC) on 25 December 2011.
line and P 1 (4) line on 25 December 2011 at location NoRNC (i.e. at 1.5 • latitude north of Ranchi) is shown in Fig.

Figure 2 .
Figure 2. Nocturnal variation of T OH at different sampling locations on 10 February 2013.

Figure 3 .
Figure 3.An illustration of estimation of OH-equivalent temperatures from SABER temperature profiles using OH 1.6 and 2.0 µm VER profiles as the weighing function.(SABER measurements reference: Orbit 49113 Event 38 of 1 January 2011).The left panel presents OH 1.6 and 2.0 µm VER profiles in the blue and red curves respectively.The blue and red broken curves in the right panel symbolize the weighing functions defined using OH 1.6 and 2.0 µm VER profiles respectively.SABER temperature profile is shown by a solid curve in right panel.

Figure 5 .
Figure 5. Scatter plot showing the dependence of the difference of T 1.6 and T 2.0 OH-equivalent temperatures on the separation of 1.6 and 2.0 µm VER peaks.

Figure 6 .
Figure 6.A comparison of ground-based OH temperature and SABER measurements on 15 May 2012.The variation of T OH at RNC, NoRNC, EoRNC, WoRNC, SoRNC and overall mean (excluding the contaminated SoRNC) are presented by "×", " ", "•", "•", " " and " " respectively.The uncertainty in T OH measurements is shown by error bars across mean T OH values.Solid circles in blue and red represent SABER OH-equivalent temperatures T 1.6 and T 2.0 respectively.

Figure 8 .
Figure 8. Plots showing limits of T OH variation in a 3 • × 3 • latitude-longitude grid over Ranchi for coincidental events and comparison with T 1.6 and T 2.0 measurements.

Figure 9 .
Figure 9. Histograms showing the frequency of coincidental cases against the observed difference of ground-based T OH measurements and SABER OH-equivalent temperatures (T 1.6 and T 2.0 ) at steps of 4 K.

Figure 10 .
Figure 10.Typical examples of SABER kinetic temperature profiles marked by strong inversions and nearly co-located OH layer.In each plot, the temperature measurements are shown by solid curves, while the broken curve symbolizes the OH 2.0 µVER over FWHM.

Table 1 .
Characteristics of optical filters used for OH nightglow observations at Ranchi.

Table 2 .
List of airglow-SABER coincidences, corresponding two OH-equivalent temperatures, viz.T