ILAS (Improved Limb Atmospheric Spectrometer) /ADEOS data retrieval algorithms

Adv. Space Res. Vol. 21, No. 3, pp. 393-396.1998 Q1998 COSPAR. Published by Elsevier Science Ltd. All rights reserved Printed in Great Britain 0273-I 177/98 $19.00 + 0.00

Pergamon

PII: SO273-1177(97)00919-8

ILAS (IMPROVED LIMB ATMOSPHERIC SPECTROMETER) /ADEOS’ DATA RETRIEVAL ALGORITHMS T. Yokota, M. Suzuki, 0. V. Dubovik and Y. Sasano National Institute for Environmental Studies, Global Environment Division, 16-2 Onogawa, Tsukuba, Ibaraki 305, Japan

ABSTRACT

Data retrieval algorithms (version 1) have been developed for the Improved Limb Atmospheric Spectrometer (ILAS) project. The ILAS sensor is designed to measure atmospheric constituents and parameters with a solar occultation technique. Vertical profiles of temperature, pressure, and aerosol extinction coefficient are determined from visible channels covering the 753 - 784 nm wavelength range for 02 molecular absorption. Profiles of 03, CI&, N20, HN03, H20, NOz, CFC-11, and aerosols are simultaneously obtained from IR channels (6.2 - 11.8 p.m wavelength range) with temperature and pressure profiles as already estimated. The retrieval is realized by vertical inversion with non-linear spectral fitting for each height. A numerical simulation test revealed that each target parameter (pressure, temperature, aerosol, and gas concentrations) can be retrieved by these algorithms to well within required precisions. 01998 COSPAR. Published by Else&r

Science Ltd.

INTRODUCTION Remote sensing from space is a rather effective tool for investigating both the behavior of chemical species involved in stratospheric ozone layer chemistry and mechanisms of ozone change. The Improved Limb Atmospheric Spectrometer (ILAS) aboard the Advanced Earth Observing Satellite (ADEOS) was designed to measure vertical profiles of high latitude stratospheric constituents, pressure, and temperature in a solar occultation mode (Figure 1) (Yokota et al., 1991, and Sasano et al., 1995).

(Zero reference) /’

I-J-..

(Sunset) ILAS is a sensor developed by the Japan Environment Agency. The hardware concept and data retrieval \ algorithms have been developed by the scientists of the ILAS science team. ADEOS is a satellite developed by the National Space Development Agency of Japan Fig. 1. Solar nWllltatinn m (NASDA). ILAS observations will be made 28 times the ILAS mc--_..__ _... each day. Since ADEOS will fly in a sun-synchronous polar orbit, ILAS latitudinal coverage is limited to the high latitude regions of both hemispheres. designed to operate for more than 3 years. ADEOS is scheduled for launch in August 1996. Solar occultation is a very powerful method for measuring minor of the strength of the solar rays used as the light source and the

on)

ray

LAS

is

species in the upper atmosphere, because self-calibration capabilities provided by dark space signals and direct sunlight signals during each sunrise and sunset event. Data inversion algorithms to precisely retrieve target parameters have been developed and tested for ILAS. This paper describes the outline of the operational ILAS data retrieval algorithm (version 1). Here we present a summary of the retrieval precision of the ILAS algorithm estimated by numerical simulation tests. 393

394

T. Yakota er al.

ILAS INSTRUMENT OVERVIEW EbcItonk CIKXJI~Block The ILAS instrument consists of a sun tracking unit, a telescope, an infrared sensor unit, a visible sensor unit, and an electronic circuit unit (Figure 2). The major characteristics of the ILAS instrument (Sasano et al., 1995, and Suzuki et al., 1995) are as follows: (1) grating spectrometers with linear array detectors: 1024 pixel MOS photodiode array for the visible channel and 44 pixel pyro-electric detector for the IR channel, (2) spectral coverage and resolution, Gimbal Assembty visible: 753 - 784 nm with 0.15 nm (FWHM) resolution, IR: 850 - 1610 cm-1 (6.21 - 11.77 urn) Choplnr Motor Asssmbiy with 0.12 urn (FWHM) resolution, (3) IFOV: 2 km Fig. 2. Schematic diagram of the ILAS instrument. vertical x 2 km horizontal for the visible channel and 2 km vertical x 13 km horizontal for the Simulated Transmittance of the ILAS visible channel IR channel, (4) positioning: tracking radiometric center of the sun from the top l.o~;~j of clouds up to 200 km, onboard IFOV position measurement relative to the sun edge with a resolution of 8 arc seconds provided by 1024 pixel linear array detectors, (5) data rate: 12 Hz with 5 17 kbps, (6) weight: 130 kg, (7) power: 78 W (maximum), (8) size: 800(W) x 1630(L) x 550(H) mm.

I jy

!

MEASUREMENT SPECTRA ILAS records spectral intensity as its raw (Level 0) data. These data are transformed to transmittance spectra (Level 1 data). This conversion is performed with 0% and 100% solar spectra measured in outer space. Level 1 data contains information of absorption spectra along the light path between the sun and the satellite. ILAS Level 1 data for the visible and IR channels (Figures 3 and 4) were simulated in consideration of the instrument characteristics such as the instrument function, spectral resolution, signal digitization, cross-talk among IR detector elements, and so on.

@-A-band

molecular absorption)

00. * 0.76

0.71 WAVELENGTH

0.78 (pm)

Fig. 3. Simulated spectra of the ILAS visible channel and the scheme of baseline estimation in the retrieval algorithm.

ILAS IF?Channel

Simulation Spectra

vangent height = 20 km)

DATA RETRIEVAL FLOW Vertical profiles (Level 2 data) of gas mixing ratios, temperature, pressure, and aerosol extinction coefficients are retrieved from Level 1 data. The basic idea (Figure 5) is to fit a theoretically synthesized spectrum to the observed spectra with a non-linear least squares (NLSQ) method. Error bars are estimated and added to each Level 2 data set.

Element Number

( 4 : Window elements for aerosol estimation)

Fig. 4. Simulated spectra of the ILAS IR channel (bottom curve) and each gas transmittance.

ILASIADEOS Data Algorithms

395

For the visible channel, the measured transmittance spectra are first separated into 2 components, one generated by extinction due to aerosol and air molecule scattering and the other by only absorption by oxygen molecules. Then absorption due to molecular oxygen is simulated to generate synthetic ILAS output data for assumed vertical and pressure temperature profiles. These assumed profiles are then iteratively adjusted Profiles of Aerosol Ratbs’(0,. N&3, HNO,, CH,, Profiles of Aerosol (US) (using the Gauss-Newton (IR) Extinction ExtinctionCoefficients N02, H#, (CFC-11)) Coefficients method) to minimize the sum of squares of the differences between the simulated and Fig. 5. ILAS data retrieval scheme (version 1). measured values. With given convergence criterion, the final profiles of pressure and temperature give the retrieved results (Nakajima et aZ., 1993). The ILAS instrument function and all other non-linear factors involved in the measurement are also included in the computation. Vertical structures of pressure and temperature are important for meteorological researches on dynamics and chemical reaction processes in the stratosphere. These are the ILAS output products from the visible channel. Aerosol extinction profiles will be obtained from both the visible and the IR channels. In the visible channel, the contributions of air molecules (Rayleigh scattering) and aerosols (Mie scattering) to the transmittance are estimated from the baseline of the absorption due to oxygen molecules, which is determined from a regression curve fit to the transmittance spectrum. An extinction coefficient profile due to both Rayleigh and Mie components can be derived by vertical inversion. The Rayleigh component can be estimated from air molecule amounts retrieved from the temperature and pressure determination (Mukai e t al., 1994). Aerosol estimation methods for the IR channel, including routines to identify polar stratospheric cloud (PSC) type, will be completed and incorporated into the ILAS version 2 algorithm. A NLSQ procedure (Marquardt method) is used to retrieve gas mixing ratios from the IR channel data (Yokota et al., 1991 and 1993). Unknown parameters are simultaneously retrieved for a tangent layer by the NLSQ method. Vertical inversion is achieved by the onion peeling method. In each step of the Marquardt iteration, analytical rather than numerical derivatives are used for Jacobian calculations, which are precisely and rapidly calculated. One of the great advantages of the ILAS measurement is that precise pressure and temperature profiles, which are required to calculate absorption coefficients for all gas species, are obtained from the visible channel data. A table look-up method is used for a rapid cal-culation of the IR cross-section as a function of pressure and temperature (Yokota, 1994).

(b) Fig. 6. Sample 03 retrieval results: (a) [dashed curve: initial profile], [circle: true profile], [square: retrieved profile], and [horizontal bar: error bar for one sigma], (b) relative estimation error and error bar.

3%

T. Yakota et al.

RETRIEVAL TEST ILAS data retrieval precision for each gas and at each altitude was numerically estimated by applying the IR

retrieval algorithm to theoretically synthesized ILAS signals with realistic noise included. The noise was instrumental random noise, digitization rounding error of the measurement signals, and the error of temperature and pressure estimation from the visible channel. The onion peeling method was used for the vertical inversion. Profiles of 03, CH4, N20, HNOs, H20, NO2, and CFC-11 were simultaneously retrieved for each layer (7 - 73 km altitudes). Extinction due to aerosols was excluded from this test. Figure 6 shows an example of the retrieval results for 03. Fifteen retrieval trials were performed for cases with 3 different sources of random noise and for 5 different pressure and temperature estimation errors. The average of the estimation values < x >, which denotes systematic error from retrieval bias (offset), was calculated. Standard deviation (SD) is calculated as an indicator of retrieval precision. A value defined as E=dm represents an index of the total retrieval error including systematic and, random error. The retrieval test results estimated from the values of E for each gas are roughly summarrzed m Table 1. Table 1.

Estimated

Precision of ILAS Measurements

Altitude (km):

Snecies 03:

10

20

f5%

&3%

30

40

&4% 25%

at Different Altitudes

50 flO%

flO% f3% +15% n.d. n.d. HNo3: n.d. f25% f60% n.d. n.d. N02: 22% &4% &40% n.d. n.d. N20: +5% flO% &50% flOO% f3% CH4: +2% +5% flO% &50% *3% H20: f5% &lo% CFC-11: (n.d.: Not determined due to low spectral signals, or i&e tkk f lO&?estimation

error)

CONCLUSION Here we present the first version of the ILAS retrieval algorithm. The retrieval test results show that the retrieval precision for the target gas mixing ratios are reasonable and within acceptable limits. Both systematic and random retrieval errors were included in these tests. Similar tests for temperature and pressure retrieval from visible channel data were also performed, but are not shown here. Algorithm improvements are planned to, for example, compensate for the effects of aerosol extinction over the whole range of IR channels (e.g. Dubovik et al., and Okamoto et al.).

REFERENCES Dubovik, 0. V., T. Yokota, and Y. Sasano, Improved technique for data inversion and its application to the retrieval algorithm for ADEOWILAS, submitted to Advances in Space Research. Mukai, S., I. Sano, Y. Sasano, M. Suzuki, and T. Yokota, Retrieval algorithms for stratospheric aerosols based on ADEOS/ILAS measurements, IEEE Trans. Geosci. Remote Sensing, 32(5), pp. 1124-l 127 (1994). Nakajima, T., Y. Sasano, and M. Suzuki, Visible remote sensing algorithms for the improved limb atmospheric suectrometer aboard ADEOS Satellite, in Curr. Probl. Atmos. Radiat., A. Deepak Publ., Ed. Sirje Keevallik, pp.383-385 (1993). Okamoto, H., S. Mukai, I. Sano, Y. Sasano, and H. Ishihara, ADEOWLAS aerosol retrieval algorithm via 4 channels, submitted to Advances in Space Research. Sasano, Y., M. Suzuki, T. Yokota, and H. Kauzawa, Improved limb atmospheric spectrometer (ILAS) project: ILAS instrument, performance and validation plan, in Advanced and Next-Generation Satellites, SPIE Proceedings Ser. 2583, pp. 193-204 (1995). Suzuki, M. et al., ILAS, the Improved Atmospheric Spectrometer, on the Advanced Earth Observing Satellite, IEICE Trans. Commun., E78-B, pp. 1560-I 570 (1995). Yokota, T., M. Suzuki, Y. Sasauo, A. Matsuzaki, J. H. Park et al., Improved Limb Atmospheric Spectrometer (ILAS) for Polar Stratospheric Trace Gas Measurement, in Proceedings of IGARSS9Z, 2, IEEE 91CH2971-0, pp. 939-942 (1991). Yokota, T., Operational Data Retrieval Method of the Satellite Sensor ILAS for Polar Stratospheric Ozone Monitoring, in CGER’s Supercomputer Activiry Report 1992, 1, pp. 69-70 (1994).