OMI In-Flight Wavelength Calibration and the Solar Reference Spectrum

The Ozone Monitoring Instrument (OMI) was launched in July 2004 and is one of four instruments onboard NASA's EOS-Aura satellite. OMI is a nadir-viewing UV-VIS spectrometer ranging from 270 to 500 nm, with a spectral resolution of roughly 0.5 nm. OMI obtains daily global coverage at the equator with a nominal sampling at nadir of 13x24 km. This paper discusses the in-flight wavelength calibration and the solar reference spectrum. Wavelength calibration is performed by means of fitting Fraunhofer structure in the radiance and irradiance spectra. It was found that when observing rapidly changing radiance signals, the wavelength scale changed in tune with this. We describe the details of this effect, explain the underlying optical mechanism and show that we can (and do) correct for it with a high degree of accuracy. This effect will be observable in any spectrometer with similar optics as that of OMI. A prerequisite for any in-flight wavelength calibration method that uses Fraunhofer lines in the observed spectra is a good quality high resolution solar reference spectrum. We describe how we calculate such a spectrum, based on combining high resolution ground based data, and medium resolution satellite measurements. 1 The Ozone Monitoring Instrument The ozone monitoring instrument (OMI) is one of four instruments aboard the EOS-Aura satellite and is the result of a collaboration between Dutch and Finnish institutes. Aura was launched successfully on 15 July 2004 and orbits the Earth in a Sun-synchronous orbit at an altitude of about 700 km, passing the equator northward at 13.42 local time. The Aura mission is to observe the Earth’s atmosphere in order to answer questions concerning the possible recovery of the ozone layer, air quality and the changing climate. OMI contributes to these mission objectives in all those fields. OMI is an ultravioletvisible spectrometer with a wide instantaneous field-of-view perpendicular to the flight direction. Using the OMI measurements a number of atmospheric trace gases can be studied, as well as aerosols and clouds. For Ozone, NO2, SO2 and various minor trace gases the total columns are retrieved for small ground pixels (at nadir: 13 km x 24 km). In addition, ozone profile information for the same ground scenes (at 13 km x 48 km resolution) is retrieved. OMI is a successor to instruments like the GOME, TOMS, SBUV and SCIAMACHY. OMI measures the Earth radiance and the Sun irradiance. Most of the retrieval algorithms use the Sun light that is scattered from the Earth and its atmosphere as the main input for their retrievals. Scattered Sun light that enters the instrument is reflected off two telescope mirrors that project an image on the entrance slit of the spectrograph. The spectrograph is divided into two spectral channels: a UV and a VIS channel, that cover the wavelength ranges 270-370 nm and 350-500 nm, respectively. In order to suppress stray light, the UV channel is divided into two sub-channels at about 310 nm: UV1 and UV2. The spectral sampling and resolution are different for the different channels: UV1: 0.33/0.63 nm, UV2: 0.14/0.42 nm, VIS: 0.21/0.63 nm, respectively. The instantaneous field of view in the flight direction is about 1.0 degree, which corresponds to about 10 km on the ground. In nominal operational mode, the information in the across track direction is binned to 60 ground pixels in the UV2 and VIS channels and 30 in the UV1. In the flight direction, images are co-added to restrict the data rate. Typically two to five individual exposures (depending on the expected radiance levels) are co-added. The UV and VIS channels have separate CCD detectors of 780 pixels by 576 pixels, in the spectral and across-track direction, respectively. For one column (wavelength) per CCD detector the individual read-outs are retained. These are the so-called small-pixel column radiances. These smallpixel column radiances are therefore available at a 2-5 times higher frequency than the complete images, and for that reason they allow to see structures on the ground or in the atmosphere with a higher spatial sampling. The small-pixel columns were originally included in the design to study the effect of clouds. They will play a crucial role in the in-flight wavelength assignment, as will be described in this paper. More details on the OMI instrument can be found in [1,2]. 2 Wavelength Calibration with OMI: scene dependence. From scientific sensitivity studies it was determined that the required in-orbit accuracy of the wavelength scale is on the order of 1/100 of a pixel, which is ten times better than can be achieved by using a line lamp spectrum. This accuracy can be reached by using the solar Fraunhofer absorption lines in the Earth spectra. This method has also been employed in previous satellite missions [3]. The basis of the method is a high resolution solar reference spectrum (see next section). This reference spectrum is convolved with the instrument transfer function (spectral slit function) in order to obtain a simulated OMI solar measurement. The wavelength scale of this spectrum, at the instrument's resolution has the same accuracy as that of the original high resolution spectrum, provided the slit function is well known. For OMI the spectral slit functions as a function of viewing angle and wavelength have been accurately calibrated on the ground by use of a purpose-built optical stimulus that utilizes an echelle grating [4]. Thus, we are confident that the accuracy of the wavelength scale of the convolved spectrum is comparable to that of the original solar reference spectrum. The original wavelength scale of a solar measurement is adjusted so that it matches that of the convolved solar reference spectrum at OMI resolution. For Sun light scattered from the Earth and its atmosphere the same method is employed in principle, but in addition to the Fraunhofer lines, the Earth reflected spectra also contain spectral structure originating from absorption and scattering that takes place in the Earth's atmosphere. The most important additional spectral structures are produced by Ozone and by inelastic scattering, or the Ring effect [5,6]. When performing the wavelength calibration of an Earth reflected spectrum, we fit an optimal linear combination of these contributions, using a non-linear solver based on a Levenberg-Marquardt algorithm. The wavelength scale is known to vary with the temperature of the optical bench of OMI. The temperature dependence was studied pre-flight and found to be small, typically 0.01 pixel shift per Kelvin. With a temperature change over an orbit of at most a few tenths of a Kelvin, this results in a change of a few 1/1000 of a pixel. Before launch the wavelength scale was expected to vary mainly as a result of these temperature changes, so the wavelength scale was anticipated to be highly stable inflight. With this in mind, as well as the considerable computational cost of performing wavelength calibration calculations for each individual spectrum and viewing angle, it was decided to calculate the wavelength scale for each spectrum based on the wavelength scale given at a reference temperature (based on the wavelength calibration of a large number of solar spectra obtained at the reference temperature of 264 K) and the correction based on the temperature of the optical bench. Thus, the wavelengths are assigned (predicted) rather than calibrated. In the L1B product the wavelength scale is described by a polynomial rather than given for each sampled point. So, for each channel (UV1, UV2, VIS) and for each row in the measurement, the wavelength for column x is given by