Solar observation of Ozone Mapping and Profiler Suite nadir system during the first 3 years of on-orbit operation

Ozone Mapping and Profiler Suite (OMPS) is a hyperspectral sensor suite that continues the daily global ozone data produced by the Solar Backscatter Ultraviolet Instrument (SBUV/2) and Total Ozone Mapping Spectrometer (TOMS) with new and improved total column and vertical profile ozone data. The OMPS nadir system consists of two sensors, nadir profiler (NP) and nadir mapper (NM), covering spectral range from 250 to 310 nm (NP) and 300 to 380 nm (NM).¹ These two nadir sensors share a nadir viewing telescope with a field of view (FOV) of $∼110deg$. The NM spectrometer captures the full 110 deg FOV cross track to provide daily global coverage, whereas the NP spectrometer captures the center of $∼16deg$ FOV to provide a vertical distribution of ozone in the stratosphere.

OMPS is equipped with two aluminum diffusers that are used for radiometric calibration purposes: one is observed once per 2 weeks and the other is observed once per 6 months in order to enable monitoring of optical degradation behavior in space. The different observational frequencies for the two reflectance diffusers allow for an accurate monitoring of exposure-dependent optical degradation.² Besides the use for radiometric calibration, the reference diffuser has proven to provide the best solar reference spectrum because of its low frequency of exposure. Suomi National Polar-Orbiting Partnership OMPS uses aluminum diffusers to direct solar radiation into its nadir viewing aperture. The diffuser-induced spectral and spatial features are relatively (at least one order of magnitude) larger when compared with a quasi-volume diffuser (QVD),³ the magnitude of the spectral feature shows very little impact for successful ozone data retrieval, and the aluminum diffusers are used to check the absolute calibration since these types of diffusers have a slower rate of degradation. In this paper, we describe two primary subjects that concern the solar calibration of the ozone mapping profiler nadir sensors: sensor channel spectral calibration and optical system degradation.

The OMPS orbital calibration provides an observed solar irradiance spectrum along with the measured Earth radiances. Neither the radiances nor the irradiances are fully calibrated products in the sensor data records (SDRs). Only the normalized radiances (NRs), in which sensor effects and solar flux changes cancel, fulfill the OMPS environment data record (EDR) requirement. The SDR algorithm does not compute NRs due to the fact that radiance and irradiance are measured with different spectral scales.

Figure 1 demonstrates optical layout of the OMPS nadir sensors. Light enters the telescope in one of the two modes, Earth observation or calibration view mode. In the Earth observation mode, the light enters directly into the telescope slit and the first bounce is off of the mirror 1 (M1) optic. In the calibration mode, the reflective diffuser is deployed and the first bounce is off of the diffuser and then the M1 optic. Beyond the M1 optic, the path is the same for Earth view and calibration view. In Earth view mode the light is collimated, and in calibration mode the light is diffused. A depolarizer in the telescope keeps the sensor’s linear polarization sensitivity $≤1%$, well below the 5% allocation. A dichroic beam splitter was optimized in the prelaunch ground test to reflect most of the 250 to 310 nm light to the NP spectrometer and transmitted most of the 300 to 380 nm light to the NM spectrometer. The use of a beam splitter allows the FOVs of the two spectrometers to be coincident. Each spectrometer is then preceded by a spectral flattening filter that compensates for the dynamic range inherent in the UV spectra.

The NM spectrometer design is referred to as an Offner or convex grating spectrometer. The optics consists of two concave spherical mirrors and a spherical convex holographic grating. The spectral smile at the detector is $<12μm$ over the full spectral and spatial range. The spectral sampling on the detector is $2.4pixels/nm$, about 0.45 nm spectral sampling across the wavelength range of 300 to 380 nm. The instantaneous FOV of each NM Earth view pixel measures 50-km cross track with an along-track reporting interval of 50 km. The NP spectrometer is a double-spectrometer based on a double Monk–Gillieson monochromator design. The optics for the first spectrometer consists of a toroidal mirror, a flat Sheridon-type blazed holographic grating, and an intermediate slit to limit the spectral band passed to the second spectrometer. The second spectrometer optics include a second flat grating identical to the first, a spherical concave mirror, and a cylindrical corrector lens. The spectral sampling on the detector is also $2.4pixels/nm$.

OMPS orbital solar measurements provide time-dependent solar irradiance, wavelength, and optics stability monitoring. The calibration observations are made via a reflective working diffuser for short-term monitoring and via a reflective reference diffuser at the telescope entrance aperture for a long-term monitor of sensor stability. As mentioned previously, the working diffuser is deployed once every 2 weeks to provide observed solar irradiances, as well as to monitor changes in sensor spectral wavelengths, spectral bandpasses, and radiometric sensitivities. The reference diffuser is nominally deployed every 6 months to monitor the stability of the working diffuser. The solar measurements take place near the Northern Earth Terminator—initiating Nadir solar calibration imaging with the “working diffuser”—as the S-NPP satellite observatory crosses over the sunlit side of the Earth from solar zenith angles of 80 to 100 deg. During the measurement, the diffuser moves through seven different positions to cover the entire FOV of 110 deg. Then, the diffuser response is mapped via seven different solar calibration sample tables over a solar azimuth range of 12 to 31.5 deg and an elevation range of $−10.4$ to 10.4 deg. The solar calibration sample table is defined by the extent of the illuminated patch for the on-board working and reference diffusers positions. Figure 2 shows solar images collected from seven different diffuser positions. The images have 151 pixels collected at positions 1, 2, 3, 5, and 6; 167 and 130 pixels collected at positions 4 and 7, respectively. The two adjacent images have at least 75 pixels overlapped, which allowed a repeatability check to ensure the measurement reliability. Position 4 illuminates the entire NP FOV of 16.7 deg.

A full-frame NM charge-coupled device (CCD) image array consists of 340 spectral columns by 740 spatial rows for the cross-track, plus $2×12$ serial overclock pixels at each edge in the vertical and $2×20$ parallel overclock smear pixels in the horizontal direction. These overclocked pixels can be used to estimate electronic offsets and define the dark current and image smear during the CCD readout and transfer time periods. The OMPS NM potential total readout size is thus 780 rows 364 columns, but the sampled solar image has the same spectral range as for Earth measurements. Nominally, for the NM, the spectral range spans sampled pixels from 101 to 296 as counted from the bottom, for a total of 196 contiguous pixels. The NP uses only one of the CCDs (the left one) to record its smaller FOV to form an image of 370 horizontal column spatial pixels and 147 contiguous pixels starting at spectral pixel 119 on a CCD. An NM solar image is constructed by calibrating each resulting solar image with a goniometry coefficient computed from the angle of the solar ray striking the diffuser and the Sun–Earth distance. After combining the images into one, a solar image also has spectral overclock and unused pixels are removed; the spatial shielded and smear pixels are removed as well. Finally, the image data is linearized first and then is corrected by dark current and electronic bias prior to computing measured solar flux. Figure 2 shows seven NM calibration images collected from the working diffuser during a solar measurement. Each image corresponds to a particular position of the diffuser, from position 1 to position 7, covering the entire NM Earth FOV of 110 deg. After correction of view angle variation (goniometry correction) and the Sun–Earth distance change, those images are stitched together into a full FOV solar image, which is then used to perform solar calibration.

OMPS solar measurements present challenges when performed using aluminum diffusers.² Due to the diffuser surface roughness and view angle variation, the collected OMPS solar view images exhibit fine features that are driven by the spatial nonuniformity. The magnitude of the structure is on the order of the CCD pixel size. Since the diffuser interference features are highly dependent on the viewing geometry of the diffuser, and since that geometry varies over a seasonal cycle, the percent difference exhibits inconsistencies of as much as 5% between two solar measurements that are taken at different times along that seasonal cycle. To mitigate the effect of these features, the orbital measurement sequence has been modified. The originated measurement deployed three individual exposures for each solar image and generated a total of 30 nadir solar calibration images (21 NM and 9 NP) within a single orbit. In the modified sequence, 23 individual exposures are collected and subsequently averaged for a given measurement. The measurement collects data within three consecutive orbits. The first orbit collects a total of 57 images from diffuser positions 1, 4, and 7; the second orbit measures positions 2 and 6 for NM and 4 for NP with 16 images for each position; and the last orbit collects a total of 52 images from positions 1, 3, 5, 7 with 9, 17, 17, and 9 images for each of the position, respectively. Since the interference features are highly dependent on the viewing geometry, and since the viewing geometry changes slightly for each exposure, the use of 23 rather than 3 exposures reduces the amplitude of the interference structure. A plot of the ratio between two solar images taken 1 month apart using the modified sequences (red) compared with the old sequences (black) is shown in Fig. 3(a) for NM spectral channels and Fig. 3(b) for NM spatial variation. Data was collected from working diffuser measurements. For sensor NM, diffuser position 3 data is displayed as an example. The overall inconsistencies between the two measurements in the new calibration sequences from position 1 to 7 are reduced by up to 5%.

The OMPS SDR uses Sun spectra to normalize the Earth radiance spectra in order to obtain the absolute Earth reflectance spectra. These, in turn, are used to retrieve information about the Earth’s atmosphere. Depending on the type of the products, the retrieved information may be sensitive to the diffuser spectral and spatial features. The spectral features mainly introduced by a solar diffuser and the depolarizer affect products in the same way. The features introduced by the onboard diffuser employed to measure the reference spectrum interfere detrimentally with the Earth atmosphere retrieval techniques that make use of the high spectral resolution of the OMPS instrument. The polarization features cancel in first order in the Earth reflectance, because they are similar in both the radiance and irradiance optical paths of the instrument. Such is not the case for the diffuser features, which appear only in the irradiance optical path. The origin of the observed diffuser features can be found in the location of the diffusers in the optical system. The surface structure of the diffusers is imaged onto the CCD detectors, where it shows up as spectral and spatial nonuniformity in the observed solar images. This mechanism also applies to the reference diffuser, which is used semiannually for the reference spectrum observations.

Prelaunch calibration and characterization have established OMPS goniometry calibration coefficients, which are a set of $40×40$ grids at varying azimuth and elevation angles for each diffuser position and diffuser type. View angles at which the orbital solar irradiances strike and reflect off the diffusers vary from measurement to measurement; linearly interpolated values from the grids are applied to solar images to account for variations in irradiance sensitivity. Interpolation and/or extrapolation of a grid at the actual solar observation angles may have errors up to 1% for individual pixels.⁴ As measured in the laboratory, Fig. 4 shows the NM nadir view goniometry distribution for wavelength channel 310 nm at diffuser position 4. The azimuth angle ranges from 12.3 to 31.8 deg and the elevation ranges from 1.39 to 7.41 deg in the sensor goniometry coordinate.

OMPS spectral distortion in the across-track direction of the array, so called spectral smile, and its spatial equivalent, keystone, was found 0.3 nm for NM and 0.7 nm for NP from prelaunch lab test data. The smile and keystone have not noticeably changed since the beginning of OMPS operational life. The amount of the smile and keystone distortion on-orbit is low, on the order of $±0.25$ to $±0.8$ pixels and is ameliorated during calibration. The NP sensor has somewhat more spatial variation, but spectra for a given spatial index are almost as well behaved as those of the NM sensor.

An OMPS solar image, after goniometry calibration and spectral smile correction, still presents undesirable small-scale irregularities. The irregularities likely stem from (1) the surface roughness of the aluminum surface so that the incident light is not uniformly diffused after reflection and (2) the bidirectional reflectance distribution function (BRDF) error when interpolating the $40×40$ grid table to a specific local view angle. Figure 5 shows such nonuniformity observed from both NP and NM orbital solar images after the goniometry calibration and spectral smile correction. The images were collected on October 15, 2014. Analysis of spectral and spatial nonuniformity is also complicated by the entrance slit irregularities that show up on the CCDs as nearly horizontal stripes, as shown in Fig. 5. Both effects need to be taken into account when characterizing the spectral features. By computing a ratio of data collected from two independent measurements, all wanted features of spectrum were eliminated, and standard deviation is calculated per wavelength channel from a large number of independent spectra. The results can quantify the spectral features. Figure 6 shows such results of the ground BRDF magnitude and the orbital spectral features after BRDF correction. Data came from the working diffuser measurements for both sensors NP and NM. The BRDF values cover OMPS overall view angle variation, the azimuth angle ranges from 12.3 to 31.8 deg, and the elevation ranges from 1.39 to 7.41 deg in the sensor goniometry coordinate, for the sensor NP [Fig. 5(a)] and the sensor NM [Fig. 5(b)]. The magnitude of orbital spectral variation represents residuals of the BRDF correction. The amount of residual increases with wavelength and on the order of the CCD pixel size resulting from the diffuser surface roughness and optical path of the sensor, which has no significant impact on ozone retrievals but will affect $SO2$ retrieval with a higher resolution data.²

Solar measurement determines the solar irradiance at the time of an Earth observation by correcting the baseline solar spectrum for subsequent changes in the sensor spectral registration. The OMPS sensor does not directly measure monochromatic irradiance, $I(λ)$, because of its finite bandwidth. Therefore, in generating calibration tables for the algorithm, slit-averaged values were computed for the central value of the 22 wavelengths using slit functions from the OMPS sensors as shown in Fig. 7. Mathematically, the slit-average irradiance is written as^5,6Display Formula

Consequently, the NR from measurements can be expressed as Display Formula

To account for the sensor spectral smile in the NM EDR total ozone algorithm, five different sets of the 22 wavelengths are used to generate the sensor table, spanning the range of wavelengths from the edge of the CCD to the center. Interpolation was made between the two sets of wavelengths that bracket the wavelength values provided by the SDR. Taking a high resolution spectrum, we convolve it with a set of instrument spectral functions that have been previously defined from OMPS ground laboratory test, together with their wavelength offsets from the central pixel. The functions are scaled by a half-width bandpass, interpolated to the spectral wavelengths, and then convolved with the solar spectrum. The computed irradiance is a linear interpolation across two high-resolution points bracketing the desired wavelength. Averaging spatially across the detectors, we compare the observed solar irradiance with the synthetic irradiance predicted using the current orbital wavelengths by computing a ratio of sensor measured irradiance to the synthetic irradiance over in Fig. 8. The data is selected to have a similar Earth viewing angle. Our results indicate very small drift in wavelength registration, while differences between the ratio over a half year point to features introduced by the diffuser itself. The synthetic solar irradiance is computed by applying the sensor bandpasses to a high spectral resolution reference solar spectrum and the instrument slit function based on ATLAS SUSIM and Kitt Peak National Solar Observatory.^7–11

Spectral response performance is dominated by two error sources, ground calibration and on-orbit bandpass shape broadening. The former introduces 2% error on irradiance and radiance calibration. Prelaunch calibration has observed that the NM bandpass full-width at half-maximum (FWHM) broadens as temperature decreases. The FWHM increase was spatially dependent, with a maximum increase about 10% at center of FOV. No significant NP FWHM change was observed. Root cause analysis attributed the broadening largely to slit “puckering.” The broadening causes 0.6 to 2.9% ozone measurement error. Wavelength shifting from ground to orbital was found about 0.12 nm for both sensors, NP and NM, due to a change in sensors’ thermal loading.¹² Correction for mean ground-to-orbit temperature change is about 2.5%. The orbital wavelength variation contributes about 0.2% error. However, a wavelength-dependent difference in sensor NM scan direction between the synthetic and observed solar irradiance, suggesting an unexpected wavelength variation across the scan direction, causes up to 2.4% local error in the far nadir positions. Although the root cause is still under study, the variations are correctable with an appropriate wavelength adjustment. OMPS NM has 35 ground pixels in the scan direction. Figure 9 shows such an example of the wavelength-dependent difference between a synthetic and observed solar irradiance for pixel #11 relative to pixel #17 before (left) and after (right) a wavelength nonlinear adjustment. Before the correction, the relative difference is independent to approximately linear wavelength variation. The left subfigure shows the relative difference when $±0.03$ and $±0.06nm$ linear shifts were applied and the linear trend is unchanged.

In general, a UV sensor’s wavelength scale is known to be sensitive to the thermal loading changes. The OMPS wavelength scale varies with the temperature of the optical bench. The temperature dependence of OMPS wavelengths was studied in-flight and found to be small, either intraorbital or seasonal, typically $±0.02nm$ shift over a change of $±1.25°C$ in telescope temperature. The intraorbital variation is about 0.025 nm,¹² whereas the seasonal change is about $±0.02nm$. Figure 10 shows a time-dependent NP orbital wavelength variation as well as a time-dependent sensor telescope temperature variation. Mathematically, the function of the time-dependent wavelength variation can be described as Display Formula

Equation (4) calculates a matrix of correlation coefficients $k$ from an input matrix constructed from wavelength shift $s(k)$ and telescope temperature $t(k)$ with a zero lag: Display Formula

The OMPS sensors are required to have 15% or less degradation in optical throughput due to contamination and particle obscuration in each channel of the instruments at the end of life. Orbital optical throughput degradation of OMPS nadir system, NM (300 to 380 nm), and NP (250 to 310 nm) has been observed and variations in degradation of $<1%$ are evident. The percent degradation of optical throughput was analyzed through a track of the change in sensors’ optical throughputs. Specifically, the sensors’ optical system is tracked with the reference solar diffuser, and degradation of the working solar diffuser is tracked by comparison of the working and the reference solar diffusers. For this study, the instrument that is the most sensitive to contaminants establishes the allowable contaminant levels of the entire suite after the first three-year operation. Figure 11 shows the optical throughput changes for six selected wavelengths, three on the NP and three on the NM, that well represent behaviors of all other wavelength channels. The data used in the analysis was collected from the measurements of the reference (green) and the working (red) diffusers and was fully calibrated. Results are presented for the lowest operating wavelength of each instrument, where contamination sensitivity is the highest. The NP is the most sensitive to contamination and has the most optics and the lowest operating wavelength; analysis indicates that the NP sensor has a higher degradation and, therefore, higher contamination levels. For the operating wavelengths of OMPS, the lowest wavelength corresponds to the highest absorption coefficients and, therefore, the highest throughput losses for the molecular contaminants. This, in turn, raises the degradation due to this type of contaminant. In addition, the NP has 13 optical surfaces, as opposed to 11 in the NM. Annually recurring features in both the NM and NP solar data have been observed from Fig. 11, which is associated with a variation of the solar beta angle. Comparable results are observed for other wavelengths. The feature is correlated with solar beta angle variation from diffuser measurements.

The reference data tracks the sensors’ optical system, and degradation of the working solar diffuser is tracked by comparison of the working and the reference solar diffusers. Given a fact that $<0.5%$ degradation was observed from the reference diffuser and $<0.5%$ to 2.5% degradation from the working diffuser, we have estimated that over the past 3.5-year operation, the working diffuser itself has degraded about $<2.0%$. Since the level of the solar diffuser degradation is highly dependent on the exposure time,² logically we expect that the degradation of the reference diffuser itself is about 0.2 to 0.3% because the reference diffuser has one-tenth the exposure of the working diffuser. Therefore, we have concluded that the OMPS optical system has $<0.3%$ degradation after 3.5-year operation, which is a high level of optical stability. Based on the existing measurement data, extrapolating the data trend was performed to predict and determine the sensor degradation contaminant levels in the next 15 years, which is below the 15% requirement.

To meet the mission performance requirements, prelaunch calibration and characterization have established OMPS sensors’ end-of-life signal-to-noise ratio (SNR) with allocations for a degraded signal due to CCD performance changes and contamination on the optical surfaces. The SNR derived from a sensitivity analysis of the NM and NP algorithms requires end-of-life performance for all 22 spectral channels used in both sensors’ measurements. Previous studies have shown that both Earth view and solar observations are in compliance with respect to the measurement SNR.^3,4 Our continued study shows that after 3 years in operation, OMPS sensors’ performance is as stable as expected and the current sensors’ SNRs still remain the same as what prelaunch predicted and what orbital observed for both sensors.

The solar calibration measurement, combined from all diffuser positions, is used to examine the observed irradiance uncertainties at each wavelength for both the NM and NP sensors. Each pixel band center wavelength used in the analysis was taken from the wavelength table derived from orbital data. For the NP, a single spectrum results, but for the broader NM image, there are spectral variations of $±0.2nm$ that increase from center toward the edges, resulting in extremes of about 0.7 nm, but still less than the spectral resolution. For our analysis, we used all of the individual pixels in the solar measurements and all of the images collected, correcting measured counts for variations in the Sun–Earth distance. Modeled synthetic solar spectrum is used as the absolute calibration standard by which irradiance accuracy can be accessed through the comparisons of the observed solar irradiance with the synthetic solar irradiance. Our analysis shows that the solar irradiance measurements with both the working and reference diffusers, at most wavelength channels, differ from the synthetic irradiances by less than the prelaunch 7% accuracy specification.

To see the calibration accuracy, as shown in Fig. 12, we computed irradiance differences between the working diffuser measurements (red) and the synthetic data as well as irradiance differences between the reference diffuser measurements (black) and the synthetic data. The data was collected on August 15, 2014 as an example, when both the reference and working diffuser were employed. The two dot-dashed lines in the figure represent synthetic solar flux with $±7%$ tolerance uniform in wavelength. The dots represent working and reference diffusers’ solar irradiance differences from the synthetic solar flux.

The NP working and reference diffusers’ accuracy in the left subfigure exhibits $<7%$ tolerance for most of the spectrum but spreads slightly greater than $±7%$ tolerance at the shortest and longest wavelengths. At the shortest wavelengths, the calibration suffers low SNR, causing slightly higher uncertainties; on the longest wavelengths, the wavelength shifting plays a major role, as identified previously,⁴ not surprising given that the NP sensitivity is known to fall rapidly at the longest wavelengths. The wavelength shift was caused by (1) a temperature benchmark change in the OMPS sensors housing the dichroic beam splitter has a wavelength shift of about 0.12 nm¹² and (2) the prelaunch characterization has a large radiometric uncertainty on wavelengths from 300 to 310 nm. Such error is also found in the NM wavelength channels of 300 to 310 nm. For the NM, as shown in the right subfigure Fig. 12(b), all the irradiance accuracies fall within the $±7%$ boundary on wavelengths $>306nm$, indicating that the calibration is accurate to within specification at all wavelengths. Because the NM at each wavelength band has small variations in the wavelength band centers that are intrinsic to the spectrometer image on the focal plane, the points in the right subfigure are confined to a narrow range of wavelengths at each band center, resulting in the banded appearance of the NM plot.

The OMPS nadir sensors have demonstrated high-level optical system stability after more than 3 years of operation. The degradation evidenced is $<0.3%$. The absolute irradiance error compiles with the system requirement for most channels. The newly calibrated wavelengths effectively remove the NM scan direction spectral variations. The annual wavelength shift pattern, which causes about 0.1% radiance error, is well characterized by a fifth degree sine function and can be corrected for. The wavelength uncertainty will be improved from the current 2.5% to about $<1%$ with a proposed correction. Although the current solar calibration sequence does not completely eliminate the diffuser interference features due to the use of alumina solar diffusers, the features are considered to be small enough and do not cause significant error in ozone retrieval. For products that are sensitive to small calibration error, such as for $SO2$, the spectral features still are a factor that affects how well those retrievals perform. This is an area of active research to further mitigate the effects. The next Joint Polar Satellite System-1 (JPSS-1) OMPS will use a QVD, which will not exhibit diffuser interference features, for the OMPS sensors that will fly on JPSS-1.

This work was supported by National Oceanic and Atmospheric Administration (NOAA) Grant No. NA09NES4400006 (Cooperative Institute for Climate and Satellites-CICS) at the University of Maryland. The sensor data used in their analyses was provided by NOAA Comprehensive Large Array-Data Stewardship System. The manuscript contents are solely the opinions of the authors and do not constitute a statement of policy, decision, or position on behalf of NOAA or the U.S. government.

Chunhui Pan received her MS and PhD degrees in mechanical engineering from the University of Maryland at College Park in 2000 and 2002. Her research interests include remote sensing instrument calibration and characterization, algorithm development, mission planning, sensor performance evaluation and sensor data records validation. Her previous research experience at NASA/SSAI includes MODIS and VIIRS calibration and characterization from 2000 to 2009.

Larry Flynn received a BA in mathematics in 1978 from the University of Maryland, College Park, Maryland, an MA in mathematics in 1981, and a PhD in applied mathematics in 1987 both from the University of California, Davis, California. For the last 16 years, he has been a research scientist with NOAA. His duties include research and analysis for validation, algorithm development, and calibration of existing and next generation satellite ozone sensors.