Validation of ILAS-II version 1.4 O3, HNO3, and temperature data through comparison with ozonesonde, ground-based FTS, and lidar measurements in Alaska

Authors


Abstract

[1] Version 1.4 data from the Improved Limb Atmospheric Spectrometer-II (ILAS-II) were validated through comparison with profile data measured by several instruments at Poker Flat (65.1°N, 147.5°W), Alaska. The height profiles of O3 and HNO3 provided by ILAS-II were compared with those retrieved from spectra measured by a ground-based Fourier transform infrared spectrometer (FTS). The O3 and HNO3 abundances measured by ILAS-II and FTS in the 17–35 km altitude range agreed within the precision of the two measurements. The O3 volume mixing ratios measured by ILAS-II above (below) 20 km were 10% lower (higher) than those from the FTS measurements. The HNO3 values from ILAS-II and FTS agreed to within 10% above 17 km. The O3 profiles obtained from electrochemical concentration cell (ECC) ozonesondes launched from Fairbanks (64.8°N, 147.9°W) were also compared to ILAS-II data. The ILAS-II ozone abundance agreed with the ozonesonde values. Comparison of mesospheric and upper stratospheric temperature data obtained by ILAS-II and Rayleigh lidar indicated that the temperature derived from ILAS-II was significantly higher (lower) than that from lidar at about 60 (40) km.

1. Introduction

[2] The Improved Limb Atmospheric Spectrometer-II (ILAS-II), the successor to ILAS [Sasano, 1999], is a satellite sensor used to measure trace species and aerosols related to the ozone layer. It was onboard the Advanced Earth Observing Satellite-II (ADEOS-II) launched in December 2002 and ceased observations in October 2003. The ILAS-II included several improvements relative to the similar ILAS sensor. Two infrared spectrometers were added for wider spectral coverage. Vertical resolution was improved from 2 km to 1 km. Solar surface scan method was adopted. The characteristics and performance of ILAS-II in orbit are described by Nakajima et al. [2006].

[3] The Alaska Project is an international cooperative research project promoted by the National Institute of Information and Communications Technology (NICT, formerly the Communications Research Laboratory, or CRL), Japan, and the Geophysical Institute, University of Alaska-Fairbanks, USA [Murayama et al., 2002]. The objectives of this project are to develop and demonstrate observational technologies for investigating global environmental changes and abnormalities through comprehensive observation of the middle and upper atmospheres. A set of ground-based atmospheric remote sensing instruments has been installed and operated at the Poker Flat Research Range (PFRR) in Alaska (65.1°N, 147.5°W). The properties observed by both the ILAS-II and the Alaska Project instruments are ozone and other trace gases (observed with millimeter-wave radiometers and Fourier transform infrared spectrometers (FTS)) and aerosols and stratospheric temperature (observed with multiwavelength/Rayleigh lidar).

[4] The ILAS-II targeted both polar regions; it could observe latitudes of 54–71°N and 64–88°S. It made about 14 observations a day, mostly along a latitude circle, for both the Northern and Southern Hemispheres. The observed latitudes varied gradually day to day. Figure 1 shows the latitudes of the measurement locations in the Northern Hemisphere. Since ILAS-II measurements took place around 65°N in March–April and August–September 2003, the NICT participated in cooperative validation experiments. In addition to regular mostly unmanned observations made at PFRR, an observation campaign was carried out at the end of August 2003. During the campaign, multiwavelength/Rayleigh lidars were operated and 7 GPS-equipped ECC ozonesondes were launched once a day from the UAF campus (64.8°N, 147.9°W) so that they were synchronized with the nearby ILAS-II measurements.

Figure 1.

Time series of latitudes of ILAS-II measurements in the Northern Hemisphere. Shaded areas show periods when comparisons were made.

[5] The Poker Flat FTS, which was installed in September 1999, measures the absorption spectra of molecules in the atmosphere using the sun as a light source [Kasai et al., 2005a, 2005b; Kagawa et al., 2005]. The infrared spectra, which contain absorption features of atmospheric trace species such as O3, HNO3, N2O, NO2, CFCs, HCl, ClONO2, and HF, have been automatically collected and transferred to NICT once a day by the SALMON system (http://salmon.nict.go.jp/) since February 2000.

[6] The trace constituents observed by FTS can be used to validate those observed by satellite measurements. For ILAS, Wood et al. [2002] compared HNO3, CH4, and N2O columns with those observed by FTS at Arrival Heights (77.8°S, 166.6°E). In this paper, comparisons between ILAS-II and FTS on a profile-by-profile basis are presented. Vertical profiles of O3 and HNO3 in terms of the volume mixing ratio (VMR) at certain altitudes were used for comparisons. The abundances of CH4 and N2O are not compared in this paper since the Poker Flat FTS is highly sensitive to tropospheric CH4 and N2O. It is difficult to separately estimate the stratospheric abundances of CH4 and N2O from the tropospheric abundances.

[7] Rayleigh lidar is capable of measuring vertical temperature profiles at altitudes from 30 to 80 km. This is the only system currently available for observing temperature profiles in this altitude range apart from sounding rockets. The instrumentation of the Poker Flat lidar was described in detail by Mizutani et al. [2000]. During the observation campaign, stratospheric and mesospheric temperatures were obtained on 29 August 2003.

[8] This paper is organized as follows. The FTS measurement and analysis are described in section 2. In section 3, we compare the O3 and HNO3 measured by ILAS-II and the Poker Flat FTS. Although both ILAS-II and Poker Flat FTS observe similarly in the infrared region, they are completely different in types of spectrometer, observation frequency, frequency resolution, geometry of the observation, and method of retrieval. The comparisons are made with care, especially regarding FTS's averaging kernels. Section 3 also describes the procedure before comparison. In section 4, we provide details of the observation campaign and compare the O3 and temperature measured by ILAS-II and during the campaign. A summary is given in section 5.

2. O3 and HNO3 Obtained by FTS

2.1. Observation of the Solar Absorption Spectrum

[9] The solar absorption spectra have been recorded with a high spectral resolution FTS, Bruker 120HR. The Poker Flat FTS is equipped with a computer-controlled solar tracker with a hatched weather-tight enclosure, KCl beam splitter, and liquid-nitrogen-cooled InSb and MCT detectors. The system monitors weather conditions; under favorable conditions, the instrument records absorption spectra using six spectral band pass filters every 2 hours. The spectra are recorded at an unapodized resolution of 0.0035 cm−1. Typical solar zenith angles at the time of measurements were 54° in summer and 73° in winter.

2.2. Retrieval

2.2.1. Algorithm

[10] The spectra measured by the FTS were retrieved on the basis of the optimal estimation method (OEM) [Rodgers, 1976, 2000] to derive the vertical profiles of O3 and HNO3 in terms of the VMR at certain altitudes. We used the SFIT2 version 3.7 program to retrieve the solar absorption spectra [Pougatchev et al., 1995, 1996; Rinsland et al., 1998]. The SFIT2 software was developed jointly by NASA Langley Research Center, the University of Denver, the University of Wollongong, and the National Institute of Water and Atmospheric Research, Lauder for retrieving vertical molecular profiles from ground-based infrared solar absorption spectra.

[11] The SFIT2 uses nonlinear retrieval with an iterative Newton scheme. The retrieval scheme for linear OEM can be expressed as

equation image

where y is the vector containing the spectral measurements, x is the vector containing the target gas VMR profile to be retrieved, F(x, b) is the corresponding forward model calculation of the spectrum, b is the model parameters, and xa the a priori parameters. Subscript i is the iteration index. The matrix K = ∂F/∂x contains the differential weighting functions for each of the retrieval parameters. Dy is the contribution function matrix, which is for the OEM given by

equation image

where Sy and Sa are the covariance of the measurement error and the a priori error, respectively.

[12] The averaging kernel matrix A, given by

equation image

provides information on the measurement altitude range with each row of A ideally being a peaked function whose full width at half maximum describes the vertical resolution of the measurement, as the converged solution of equation (1) is expressed using equation (3):

equation image

[Rodgers, 2000], where equation image and xtrue are the retrieved and the “true” profiles, respectively, and I is the unit matrix.

2.2.2. Retrieval Setting

2.2.2.1. O3

[13] The observed spectra were corrected using measured instrument line shape (ILS) functions derived from fits to HBr cell measurements. The ILS functions were used directly in the forward model. An example of the observed spectra is shown by crosses in Figure 2 (top). A microwindow of 3051.29–3051.9 cm−1 was used for retrieving O3. This spectral band was chosen because (1) there is little interference from other molecular absorption lines, (2) the temperature dependence of the intensity of the rotational spectrum is weak, and (3) transitions in this frequency region have sensitivity to O3 in the lower stratosphere. The signal-to-noise ratio (SNR) of the window was set to 200 as was determined from the random noise of the observed spectra.

Figure 2.

(top) Overplot of observed and calculated spectra. Crosses represent the solar spectrum observed with ground-based FTS over Poker Flat at 0910 AKST 15 April 2003. The solid line is the spectrum calculated using SFIT2 version 3.7. (bottom) Residual between observed and calculated spectra.

[14] The SFIT2 version 3.7 software can retrieve vertical profiles of several molecules simultaneously, while interfering molecules in the microwindows are fitted using the a priori profile through parallel translation using the same value for all layers. In this particular case, the O3 profile was retrieved while interfering absorption from atmospheric CH4, CH3D, H2O, and HDO was included as multiplicative scale factors in the calculations.

[15] The spectroscopic parameters, including the transition frequencies and their intensities, were taken from the High-resolution Transmission (HITRAN) 2000 database [Rothman et al., 2003]. Data on H2O, CH4, O2, NO, NO2, and C2H2, which were updated after publication of HITRAN 2000, were also used in the retrieval calculations. The same database was also used in the ILAS-II version 1.4 retrieval.

[16] Meteorological data for input to SFIT2 were prepared as follows. Temperature and pressure data taken from rawinsonde observations made over Fairbanks were used for altitudes below 30 km, UK Meteorological Office (UKMO) data [Swinbank and O'Neill, 1994; Lorenc et al., 2000] were used for 30–50 km altitudes, and CIRA86 data were used for 50–100 km altitudes. Profiles of H2O, which affects the O3 retrieval, were taken from the rawinsonde observations from 0–10 km and connected smoothly to the profile above 10 km based on model calculations made by the Rutherford Appleton Laboratory [Reburn et al., 1998].

[17] The O3 profiles used as a priori data were prepared from statistics from HALOE version 19 data for 60–70°N. The a priori error covariance matrix Sa was prepared on the basis of 10% of the a priori VMR for all layers, which was as large as the average standard deviation of the HALOE data.

[18] The result of spectral fitting is also shown in Figure 2. No systematic spectral residuals were observed.

2.2.2.2. HNO3

[19] Figure 3 shows the same information as Figure 2 for an example of HNO3 retrieval. The HNO3 height profiles were retrieved using all the HNO3 absorption lines in the microwindow range 867.45–869.25 cm−1. Similar spectral bands have been used in many studies [Connor et al., 1998; Rinsland et al., 2000; Hase et al., 2004]. The performance of SFIT2 in retrieving the HNO3 profile was demonstrated by Hase et al. [2004] in the context of comparison with another retrieval code, PROFFIT9.

Figure 3.

Same as Figure 2 but for HNO3. The solar spectrum was measured at 1843 AKST 18 April 2003.

[20] Interfering H2O and OCS spectra in the frequency range were included in the calculations. The a priori HNO3 profiles from 16 to 32 km were made on the basis of MLS version 5.0 in 1996. The Sa for HNO3 were set on the basis of 30% of the a priori VMR for all layers, which was larger than the standard deviations of the monthly MLS data for below 30 km (∼10 hPa). The other parameters were basically the same as those used for the O3 retrieval calculation.

2.2.3. Error Estimation

[21] The total error covariance matrix, S, of the retrieved height profiles of O3 and HNO3 can be expressed as the sum of error covariance matrices [Rodgers, 1990, 2000]:

equation image

where SM, SN, and SB are covariance matrices for the measurement error, smoothing error, and the error due to the uncertainty of the model parameters, respectively. Each covariance matrix can be formulated as

equation image
equation image
equation image

where Sb is the error covariance matrix for the model parameter vector and Kb is the weighting function with respect to each of the model parameters.

[22] Evaluating the errors in a retrieved profile is difficult because there are many complicating factors. In this paper, measurement errors, smoothing errors, and model parameter errors regarding uncertainty of temperature, which are considered as random errors, were estimated with equations (5)(8). Errors from the spectroscopic parameters and effective apodization parameter (EAP), which are considered as systematic errors, are evaluated by a difference between retrievals with perturbed and unperturbed parameters as done by Barret et al. [2002]. We calculated impact of 5% line intensity, 10% of air-broadening coefficient, and 10% of EAP.

[23] Each error was estimated for the columns extending 0–12, 12–18, 18–24, and 24–40 km for O3, and 10–20 and 20–30 km for HNO3. Total errors were then obtained as the root sum square (RSS) of random and systematic errors. Tables 1 and 2 show the estimated total errors of O3 and HNO3 for partial columns, respectively. An example of FTS retrieved profile together with its estimated errors was shown in Figures 4 and 5. Details of the error analysis is beyond the scope of this paper and will appear in the work by A. Kagawa et al. (manuscript in preparation, 2006). Note that the errors of profiles are larger than those of partial columns.

Figure 4.

(left) Example O3 profiles. The solid, long-dashed, short-dashed, and dash-dotted curves show modified ILAS-II, original ILAS-II, FTS retrieved, and FTS a priori profiles, respectively. Error bars indicate the estimated uncertainty in the FTS retrieved profile at each altitude. (right) Averaging kernels of the FTS measurement for each vertical grid level. Solid rectangles indicate elements at the height concerned. Averaging kernels are shown for a height range of 1.3–49 km.

Figure 5.

Same as for Figure 4 but for HNO3.

Table 1. Estimated Total Errors of O3 Partial Columns of 0–12, 12–18, 18–24 and 24–40 km
Height Range, kmTotal Error, %
0–128.4
12–1812.4
18–2411.6
24–4023.9
Total column6.3
Table 2. Same as for Table 1 but for HNO3 Partial Columns of 10–20 and 20–30 km
Height Range, kmTotal Error, %
10–2011.8
20–3018.6
Total column6.7

3. Comparison With O3 and HNO3 From FTS Measurements

3.1. Data

[24] The ILAS-II version 1.4 data were used for the comparison. The retrieval algorithm for the version 1.4 data is described by T. Yokota et al. (manuscript in preparation, 2006). Validation of the O3 data was also done by Sugita et al. [2006] through comparisons with ozonesonde and other satellite measurements. Validations of the HNO3 data through comparison with data from balloon-borne sensors (MIPAS-B2 and MkIV) and based on climatological HNO3 − O3 correlation were done by Irie et al. [2006].

[25] Coincidence criteria for comparison on a profile-by-profile basis between ILAS-II and FTS measurements were set as ±5 degrees in latitude, ±10 degrees in longitude, and ±6 hours in time. Figure 6 shows a scatter diagram between percentage difference in HNO3 stratospheric columns obtained by ILAS-II and FTS and distance of measurement locations. The solid rectangles indicate comparison pairs used in this paper. The distance between the furthest compassion pair was 553 km.

Figure 6.

Scatter diagram between percentage difference D in HNO3 stratospheric columns and distance of measurement locations. Solid rectangles and open circles indicate comparison pairs used and not used in this paper, respectively.

[26] Each FTS measurement was done ahead of the corresponding ILAS-II measurement since the FTS solar absorption measurements were done in the daytime while the nearby ILAS-II measurements occurred around 2000 AKST (Alaska Standard Time; 9 hours behind UTC) every day during the observation period. We obtained 19 O3 and 23 HNO3 FTS profiles observed between 11 and 30 April (shown by the darkly shaded area in Figure 1) that met the criteria. While the Poker Flat FTS usually operates from February to November every year, in 2003 the spectra obtained before 11 April were not analyzed because the SNR of the spectra was below a preset threshold, and no data were recorded because of a breakdown of the calibration laser in the FTS in August. Note that all the corresponding ILAS-II profiles were obtained to the south of the PFRR site, which could have introduced an apparent bias. The apparent biases, estimated from climatological meridional gradients of O3 and HNO3 from the HALOE data, were 4% and 8% at 17 km and 1% and 2% at 21 km, respectively.

[27] Column O3 and HNO3 were also compared in the context of time series. We used 24 ILAS-II measurements along with 46 O3 and 59 HNO3 FTS measurements that satisfied the above criteria regarding space.

3.2. Preprocessing Before Comparison

[28] When ILAS-II and Poker Flat FTS measurements are compared, care is needed to take into account how the characteristics of profiles differ between the observation systems, particularly the averaging kernels for FTS retrieval [Connor et al., 1994; Rinsland et al., 1998; Rodgers and Connor, 2003]. Thus the following operation similar to equation (4) was applied to the ILAS-II profiles before comparison [Deeter et al., 2004]:

equation image

where equation imageI is a resampled ILAS-II profile at the vertical grid of FTS profiles, AF and xFa are the averaging kernel matrix and a priori profiles of the corresponding FTS retrieval, respectively, and equation imageI is a modified ILAS-II profile with AF and xFa. In this study, equation imageI is obtained by

equation image

where zn is the height of the nth vertical grid of the FTS data, znequation image (zn+equation image) is the middle point between zn−1 and zn (zn and zn+1), xI is the original ILAS-II profile every 1 km, and subscript n attached to vectors denotes their nth component containing VMR at the height of zn. The definite integrals were approximated using the trapezoidal rule.

[29] Figures 4 (left) and 5 (left) show examples of the original and modified ILAS-II profiles, the correlative FTS profile, and the FTS a priori profile. The ILAS-II profile was obtained at (60.42°N, 150.24°W) at 0549 UTC 30 April 2003, while the FTS profiles were retrieved from spectra obtained around 0100 UTC 30 April 2003. Figures 4 (right) and 5 (right) show averaging kernels of the FTS measurements for the heights below 49 km. Since the peaks of the averaging kernels differed from the heights of interest and little information was obtained from the spectra above 37 (31) km or below 11 km for O3 (HNO3), comparisons were made for a height range of 13–35 km (13–29 km).

3.3. O3

[30] Time series of the O3 total columns observed by FTS and ILAS-II are shown in Figure 7, together with those obtained by Total Ozone Mapping Spectrometer (TOMS) observation at the points nearest to PFRR and to the ILAS-II measurements. To calculate the column density from ILAS-II O3 VMR data, we used UKMO assimilated data for the temperature and pressure. The contribution below the ILAS-II minimum height was estimated from the climatological tropospheric O3 profile calculated using ozonesonde flights over Fairbanks in March and April 2001 (obtained from the Climate Monitoring and Diagnostics Laboratory (CMDL) data archive) and was added to the ILAS-II O3 columns. Table 3 shows the obtained average and standard deviations for the O3 columns.

Figure 7.

Time series of O3 total columns obtained by FTS, ILAS-II, and TOMS. Error bars associated with ILAS-II column O3 show the root sum square of uncertainty of ILAS-II column O3 (estimated from VMR uncertainties) and the standard deviation of the tropospheric O3 column as shown in Table 3.

Table 3. Averages and Standard Deviations for O3 Columns Calculated From Ozonesonde Flights Over Fairbanks in March and April 2001
Height Range, kmO3 Column, 1019 cm−2
0–110.152 ± 0.022
0–120.192 ± 0.025
0–130.234 ± 0.029
0–140.284 ± 0.034
0–150.340 ± 0.040
0–160.401 ± 0.045

[31] The absolute value and the seasonal variation of the total column abundance of O3 determined by TOMS satellite measurements over PFRR agreed very well with what was obtained through FTS measurements as already demonstrated by Seki et al. [2002] and Kagawa et al. [2004]. The total column O3 obtained by FTS and ILAS-II measurements also agreed to within at worst 20% and generally about 10% of each other. There were decreases on 13 April and variations during 22–24 April in the FTS total O3 measurements. These features were also seen in the TOMS total O3 measurements and were apparently associated with a synoptic-scale flow pattern. They were not so clearly seen in the ILAS-II data. This reflects the difference in the spatial scale observed by each sensor.

[32] Comparisons in terms of VMR were made in a way similar to that of Sugita et al. [2002]. Figure 8 (left) shows the average O3 profiles measured by ILAS-II and FTS between 11 and 35 km of each other, with the minimum, maximum, and one standard deviation indicated. Figure 8 (right) shows the maxima, minima, medians, 25 and 75th percentiles of the individual percentage difference, D, between the ILAS-II and FTS data. Here, D is defined as

equation image

where V (ILAS-II) and V (val) show the target gas VMRs of the ILAS-II and validation data (in this case, O3 data from FTS), respectively, for each geometric altitude grid. The average RSS errors in the ILAS-II and FTS measurements are also shown in Figure 8 (right) by dashed lines which are symmetrical with respect to the zero line.

Figure 8.

(left) Average profiles of O3 VMR obtained by ILAS-II and FTS measurements. Error bars show one standard deviation at each altitude. Maximum and minimum values are shown as thin solid lines (ILAS-II) and dotted lines (FTS measurements), respectively. (right) Median profiles of the individual percentage differences D between the O3 VMR obtained by ILAS-II and FTS measurements shown by solid rectangles. The maximum and minimum values of D are shown by dash-dotted lines. The 25 and 75th percentiles of D are shown by thin solid lines. Dotted lines symmetrical with respect to the zero line show the average of the individual RSS errors in ILAS-II and FTS measurements.

[33] The general tendency was for the O3 VMR measured by ILAS-II above (below) 20 km to be lower (higher) than that obtained from the FTS measurements. The profile of the median D was almost within the RSS error except at 23 km. This indicates that the O3 VMR in the lower stratosphere measured by ILAS-II agreed with that measured by FTS within the uncertainty levels of the two data sets. The median D was −12% at 23 km and 10% at 17 km.

[34] The maximum D reached around 40% at 17 km. This was due to a significantly smaller VMR being obtained from the FTS measurement than from the ILAS-II one (1.5 versus 2.0 ppmv at 17 km). This small retrieved value might have been due to oscillation of the retrieval of the FTS spectra. Such profiles may be included in the statistics since it would be difficult to objectively exclude them. Nevertheless, the characteristics of the median profile are robust as indicated by the profiles of the 25 and 75th percentiles which are reasonably close to the median profile.

3.4. HNO3

[35] Time series of the HNO3 stratospheric columns observed by FTS and ILAS-II are shown in Figure 9. The FTS stratospheric column was defined as the column amount above 11 km. The column HNO3 obtained by FTS and ILAS-II measurements agreed to within about 10% of each other. The seasonal variation also agreed very well.

Figure 9.

Time series of stratospheric column of HNO3 obtained by FTS and ILAS-II. The FTS stratospheric columns were defined as the column amounts above 11 km.

[36] Figure 10 shows the same information as in Figure 8, except it is for HNO3. The average HNO3 profile measured by FTS agreed very well with the corresponding ILAS-II profile. The D values were less than 10% and within the RSS errors except below 17 km.

Figure 10.

Same as for Figure 8 but for HNO3.

[37] The VMRs obtained from FTS measurements at 21–27 km were widely distributed. Especially at 25 km, the maximum D exceeded 50%. This was due to a limited number of outlier values, as for O3 around 15 km. The profile shapes for the 25 and 75th percentiles again showed that the median profile was not affected by these outlier values.

4. Comparison With Data From the Observation Campaign

4.1. Ozonesonde Observation

[38] The observation campaign conducted during 23–30 August 2003 (shown by the lightly shaded area in Figure 1) used En-Sci ECC 2Z-GPS ozonesondes. Each ozonesonde was lifted from the ground to about 35 km by a TOTEX TA-2000 weather balloon. During the flight, data on the ozone concentration, temperature, and humidity were transmitted every 1.5 s which corresponded to a vertical interval of 7–8 m. The horizontal winds were calculated every minute on the basis of the temporal change in location data obtained from a GPS receiver built into the polystyrene case of the ozone sensor. The time interval corresponded to a vertical interval of about 300 m. The O3 VMR accuracy was reported to be within 5% in the stratosphere up to 10 hPa [Komhyr et al., 1995].

[39] The nearby ILAS-II measurements were done at about 0500 UTC every day. We launched ozonesondes at about 0400 UTC every day for the validation since it took about 1 hour for the ozonesondes to reach an altitude of 20 km. These data were included in the ozonesonde statistics of Sugita et al. [2006] which were based on ozonesonde data acquired at many stations.

[40] The ozonesonde campaign also targeted small-scale variations in ozone, both vertically (less than 5 km) and temporally (less than 1 day), in the upper troposphere and lower stratosphere due to atmospheric gravity waves and other features such as tropopause folding events. Launches were made every 3 hours from 1600 UTC 26 August to 0400 UTC 28 August. Altogether, there were 22 ozonesonde and 5 radiosonde measurements. The data with high temporal and vertical resolutions revealed ozone variations due to gravity waves in the stratosphere [Yamamori et al., 2004].

4.2. Comparison With Ozonesonde Ozone

[41] Figure 11 shows a series of vertical profiles of the ozone VMR obtained from nearby ILAS-II measurements and the corresponding ozonesonde launches. Seven ozonesonde profiles were used for comparison. The nearest observation occurred on 30 August, and the distance between the ozonesonde and ILAS-II observation was 137 km. The furthest occurred on 27 August, at a distance of 518 km. The differences in observation times were within 1 hour at a height of 20 km for all observations.

Figure 11.

Series of vertical profiles of O3 VMR observed every day from 24 to 30 August 2003. Each profile is offset by +3 ppmv/day from that observed at 0400 UTC 24 August. Thick curves and diamonds show profiles obtained from synchronously launched ozonesondes and nearby ILAS-II measurements, respectively.

[42] In our observation, ozonesonde O3 VMRs were significantly smaller than the ILAS-II O3 above about 30 km. The difference increased with height. Each profile has a maximum at a height of around 32 km, which is lower than the height of climatological peak of O3 VMR (around 35 km). Thus we speculated that the pump efficiency correction, in order to account for the decrease of flow rate of a gas sampling pump of ECC ozondesonde at ambient air pressures below about 100 hPa, was insufficient above 30 km in our observation, and decided not to use O3 above 30 km for the comparison.

[43] Figure 12 shows the same information as Figure 8, except it was obtained from the ozonesondes rather than from FTS measurements. The ozonesonde data were smoothed using a low-pass filter with a cutoff length of 1 km and resampled every 1 km to fit the vertical grid of ILAS-II data. The ILAS-II ozone VMR around Fairbanks was slightly higher than the ozonesonde ozone VMR. This tendency differed from that reported by Sugita et al. [2006], who showed that the ILAS-II ozone VMR was generally lower than the ozonesonde ozone level for the Northern Hemisphere. The fact that all the ozonesonde observations were made north of the ILAS-II measurement locations (see Figure 1) may explain this opposite tendency. The D values were within the RSS errors, except below 17 km. In the other height regions, the absolute value of D based on the Fairbanks ozonesonde measurements was within the RSS error and considerably smaller than that based on the measurements made at the Northern Hemisphere stations.

Figure 12.

Same as for Figure 8 but for ozonesonde measured O3. The 25 and 75 percentiles of D are not shown.

4.3. Comparison With Lidar Temperature

[44] The ILAS-II temperature and pressure profiles between about 10 and 80 km were retrieved from the O2A-band [Sugita et al., 2004]. Figure 13 shows the temperature profiles from the lidar, ILAS-II, and UKMO stratosphere assimilation on 29 August. Note that the observation times differed between the lidar (observed for 0855–1326 UTC) and ILAS-II (observed at 0529 UTC). At about 40 (60) km, the ILAS-II temperature was more than 10 K lower (higher) than the lidar temperature. These differences were considerably larger than the width of one standard deviation of temperature (shown by the lighter shaded bars) during night. The standard deviations were estimated from a series of temperatures taken every half hour during the nights of 21 August 2001 and 28 August 2002 since the number of temperature profiles for 29 August 2003 was too small to derive the standard deviation. The temperatures were also lower and higher than the UKMO temperatures. These results suggest a significant negative (positive) bias in the ILAS-II version 1.4 temperature measurements at about 40 (60) km.

Figure 13.

Temperature profiles from lidar (open curve), ILAS-II (diamonds), and UKMO (crosses) on 29 August. Solid lines show the error width of measured ILAS-II temperature. Dark and light shaded lines show the error width and one standard deviation, respectively, of the lidar temperature during the night.

5. Summary

[45] ILAS-II version 1.4 data were compared with vertical profiles of O3 and HNO3 retrieved from the solar absorption spectra obtained by the FTS at Poker Flat, Alaska, in April 2003. This comparison was made for lower stratospheric altitudes, where sufficient information from solar absorption spectra could be obtained. The relevant characteristics of FTS retrieved profiles were taken into account.

[46] The data collected indicated that the ILAS-II O3 abundance generally agreed with that obtained from FTS measurement within the level of uncertainty for both data sets. The general tendency was that the O3 VMR measured by ILAS-II above (below) 20 km was 10% lower (higher) than that from the FTS measurements. The HNO3 values from ILAS-II and FTS agreed well. The median difference was smaller than 10%, except below 17 km.

[47] Comparison was also made with the data taken during an observation campaign in August 2003. The ILAS-II ozone abundances agreed with the ozonesonde ones within the level of uncertainty of the ILAS-II data. Although this comparison was made for only one example, mesospheric and upper stratospheric temperature data obtained by the ILAS-II visible sensor and Rayleigh lidar were also compared. The results suggest that the temperatures derived from ILAS-II were significantly lower and higher than those from lidar at about 40 km and 60 km, respectively.

Acknowledgments

[48] The observation campaign in August 2003 was carried out with the participation of Kazuo Shibasaki, Isao Murata, Seiji Kawamura, Kensuke Yoshioka, Fumikazu Taketani, Tomoki Nakayama, Nao Ikeda, and Kazuyuki Miyazaki. We are indebted to William R. Simpson, Kenneth Sassen, and Richard L. Collins for their assistance with the observations at the UAF. The ozonesonde data used to calculate the climatological tropospheric O3 columns were downloaded from the Climate Monitoring and Diagnostics Laboratory Data Archive (http://www.cmdl.noaa.gov/info/ftpdata.html). We are also grateful to Nicholas Jones (University of Wollongong) for his valuable discussions and suggestions on the retrieval procedure. Likewise, we thank Frank Murcray (University of Denver) for his contribution to the FTS measurements. Thanks are also extended to the editor and three anonymous reviewers for their valuable comments. A part of this research work was supported with the Global Environment Research Fund provided by the Ministry of the Environment, Japan.

Ancillary

Advertisement