Terra MODIS On-Orbit Spectral Characterization ... - Semantic Scholar

4 downloads 0 Views 879KB Size Report
ment vendor, Santa Barbara Remote Sensing (SBRS), Goleta,. CA. The SpMA is ..... [4] C. L. Parkinson, “Aqua: An earth-observing satellite mission to examine water and other ... the University of Maryland, College Park. He is currently an ...
2198

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 44, NO. 8, AUGUST 2006

Terra MODIS On-Orbit Spectral Characterization and Performance X. Xiong, N. Che, and W. L. Barnes

Abstract—The Moderate Resolution Imaging Spectroradiometer (MODIS) protoflight model onboard the National Aeronautics and Space Administration’s Earth Observing System Terra spacecraft has been in operation for over five years since its launch in December 1999. It makes measurements using 36 spectral bands with wavelengths from 0.41 to 14.5 µm. Bands 1–19 and 26 with wavelengths below 2.2 µm, the reflective solar bands (RSBs), collect daytime reflected solar radiance at three nadir spatial resolutions: 0.25 km (bands 1–2), 0.5 km (bands 3–7), and 1 km (bands 8–19 and 26). Bands 20–25 and 27–36, the thermal emissive bands, collect both daytime and nighttime thermal emissions, at 1-km nadir spatial resolution. The MODIS spectral characterization was performed prelaunch at the system level. One of the MODIS onboard calibrators, the Spectroradiometric Calibration Assembly (SRCA), was designed to perform on-orbit spectral characterization of the MODIS RSB. This paper provides a brief overview of MODIS prelaunch spectral characterization, but focuses primarily on the algorithms and results of using the SRCA for on-orbit spectral characterization. Discussions are provided on the RSB center wavelength measurements and their relative spectral response retrievals, comparisons of on-orbit results with those from prelaunch measurements, and the dependence of center wavelength shifts on instrument temperature. For Terra MODIS, the center wavelength shifts over the past five years are less than 0.5 nm for most RSBs, indicating excellent stability of the instrument’s spectral characteristics. Similar spectral performance has also been obtained from the Aqua MODIS (launched in May 2002) SRCA measurements. Index Terms—Calibration, instrument, remote sensing.

I. I NTRODUCTION

T

HE MODERATE Resolution Imaging Spectroradiometer (MODIS) is one of the major instruments for the National Aeronautics and Space Administration (NASA) Earth Observing System (EOS) missions. The MODIS protoflight model (PFM) onboard the EOS Terra spacecraft has been in operation for over five years since its launch in December 1999. Flight model 1 (FM1) onboard the EOS Aqua spacecraft, launched in May 2002, has been in operation for more than three years. MODIS was designed with the capability to extend and enhance a number of heritage sensors’ observations and data records for studies of the Earth system and its land, oceans, and atmosphere [1]–[4]. There are approximately 40 science data products that

Manuscript received May 16, 2005; revised November 29, 2005. X. Xiong is with the Earth Sciences Directorate, NASA Goddard Space Flight Center, Greenbelt, MD 20771 USA (e-mail: xioaxiong.xiong.1@ gsfc.nasa.gov). N. Che is with Science Systems and Applications, Inc., Lanham, MD 20706 USA (e-mail: [email protected]). W. L. Barnes is with the University of Maryland Baltimore Country, Baltimore, MD 21250 USA (e-mail: [email protected]). Digital Object Identifier 10.1109/TGRS.2006.872083

Fig. 1.

MODIS scan cavity and OBCs.

are routinely produced from MODIS observations, including land/cloud boundaries and properties, ocean color, sea-surface temperature, and atmospheric water vapor. Each MODIS instrument makes observations using 36 spectral bands with wavelengths ranging from 0.41 to 14.5 µm at three spatial resolutions (nadir): 0.25 km (bands 1–2), 0.5 km (bands 3–7), and 1 km (bands 8–36). The number of detectors per spectral band is 40, 20, and 10 for the 0.25-, 0.5-, and 1-km spatial resolution bands, respectively. The detectors for the 36 spectral bands are located on four focal plane assemblies (FPAs) according to their wavelengths: visible (VIS), nearinfrared (NIR), short- and midwave infrared (SWIR/MWIR), and longwave infrared (LWIR). The VIS and NIR are uncooled FPAs while the SWIR/MWIR and LWIR FPAs are controlled nominally at 83 K. MODIS bands 1–19 and 26 with wavelengths from 0.41 to 2.2 µm are the reflective solar bands (RSBs), and the other 16 bands with wavelengths from 3.5 to 14.5 µm are the thermal emissive bands (TEBs). MODIS is a cross-track scanning radiometer that makes measurements using a two-sided scan mirror. It is equipped with a set of onboard calibrators (OBCs) for its on-orbit calibration and characterization. Fig. 1 shows the MODIS scan cavity and the locations of the OBCs. For radiometric calibration purposes, a solar diffuser (SD) and a SD stability monitor (SDSM) are used together for the RSB and a blackbody for the TEB [5]–[7]. The Spectroradiometric Calibration Assembly (SRCA) is primarily used for the instrument spatial and spectral characterization [8]–[11]. This paper presents MODIS spectral characterization approaches and results for the Terra MODIS RSBs. Typically, spectral characterization of the instrument is performed prelaunch. The characterization includes measurements of each spectral band’s center wavelength, bandwidth (BW), and relative spectral response (RSR). Often the RSR measurements include characterization of the band’s out-of-band

0196-2892/$20.00 © 2006 IEEE

XIONG et al.: TERRA MODIS ON-ORBIT SPECTRAL CHARACTERIZATION AND PERFORMANCE

(OOB) response. For a number of reasons, the sensors’ spectral characteristics could change from prelaunch to on orbit. The difference between the ground test environment and space and launch-related effects may introduce changes in instrument response. Environmental temperature variations often cause spectral response changes due to changes in the coating-film thicknesses and thus the associated reflectance or transmittance. Degradation of the optics over time during in-flight operation could also impact the sensor’s spectral characteristics. Tests have shown that spectral filters of the multilayer thin-film type can experience significant spectral response shifts [12]. Therefore, it is important to track the spectral characteristics throughout the sensor’s lifetime in order to assure accurate evaluation of its effect on instrument performance and to enable accurate radiometric calibration. On-orbit spectral calibration is often utilized with hyperspectral sensors. In the Hyperion hyperspectral imager onboard the NASA EO-1 satellite, spectral calibration is performed using the spectral features of a doped spectralon panel [13]. Measuring the reflected solar beam from the Hyperion cover and comparing it with solar and atmospheric absorption lines validates the spectral calibration. The spectral calibration for the Environmental Satellite (ENVISAT) uses an internal spectral line source and the calibration results are compared to solar Fraunhofer lines and atmospheric absorption features [14]. The atmospheric chemistry experiment (ACE) uses the absorption lines of the solar spectrum for its on-orbit spectral calibration. In addition, a laser diode is used as a spectral source. The wavelength shift of the diode is corrected for instrument temperature and current changes [15]. The response due to solar Fraunhofer lines is utilized for wavelength calibration of the limb-viewing broadband imaging spectrometer of the ozone mapping and profiler suite (OMPS) [16]. The spectral calibration of the atmospheric infrared sounder (AIRS) compares upwelling radiance spectra against precalculated radiance spectra. For verification purposes, the observed radiance spectra of the onboard spectral calibrator (a Parylene coating with spectral features) are also compared against precalculated radiance spectra [17]. The common feature of these hyperspectral instruments is that their BWs are very narrow and their spectral response depends on detector array location. For broadband multispectral instruments, accurately tracking spectral response is a challenge. For this purpose, the MODIS is equipped with a unique device that is capable of performing on-orbit spectral characterization of its RSBs (VIS, NIR, and SWIR bands). This appears to be the first such device for a remote-sensing multispectral instrument. We present in this paper a brief overview of MODIS prelaunch spectral characterization and describe the spectral calibration transfer from ground measurements to the OBC (SRCA). We discuss the SRCA spectral characterization algorithms and on-orbit results for the MODIS RSB, including center wavelengths, RSR, and dependence of center wavelength shifts on the instrument’s temperature. For Terra MODIS, the center wavelength shifts on orbit, over the past five years, are less than 0.5 nm for most RSBs, indicating excellent stability of the instrument’s spectral characteristics. Examples of Terra MODIS on-orbit spectral characterization results are provided and discussed.

2199

Fig. 2. Schematic of the SpMA.

II. P RELAUNCH S PECTRAL C HARACTERIZATION The MODIS prelaunch spectral characterization was made using a spectral measurement assembly (SpMA) in both ambient and thermal vacuum (TV) environments at the instrument vendor, Santa Barbara Remote Sensing (SBRS), Goleta, CA. The SpMA is essentially a double monochromator, as illustrated in Fig. 2. The source box of the SpMA holds gas emission-line sources for internal wavelength calibration. The combination of a lamp and glow bar serves as an illumination source that covers the spectral wavelength range of the MODIS bands. The beam, after passing through the selected diffraction order-sorting filter, is focused onto the entrance slit. The slit width is adjustable depending on the required spectral resolution and signal-to-noise ratio (SNR). Several gratings are utilized to cover the MODIS spectral range from visible to longwave infrared. Located in the optical path, just outside of the double monochromator, are reference detectors [a silicon photodiode (SiPD) and a pyroelectric detector] with known spectral responses that are used to monitor the monochromator’s output and to correct for the source spectrum impact on the RSR measurements. The beam from the SpMA’s double monochromator is then collimated and folded by a set of mirrors to fill the MODIS aperture. The SpMA output has nearly the same solid angle and area product as that from the Earth scene so that the spectral characterization can be properly measured. The SpMA reference detector signal in digital number dnref (λ) is directly proportional to its well characterized spectral response Rref (λ). Similarly, the MODIS instrument’s spectral response in digital number dn(b, d, λ) for a given band and detector, when characterized in the TV chamber, is directly proportional to its spectral response R(b, d, λ), the spectral reflectance of the fold mirrors ρmirror (λ), the transmittance of the TV chamber window τwindow (λ), and the transmittance of the atmosphere along the optical path from the SpMA reference detector to the TV chamber window τatm (λ). Therefore, the MODIS spectral response R(b, d, λ) can be determined by the following expression: dn(b, d, λ) R(b, d, λ)·ρmirror (λ)·τwindow (λ)·τatm (λ) = . (1) dnref (λ) Rref (λ) Normalizing the spectral response R(b, d, λ) to its peak response Rmax (b, d, λ) yields the relative spectral response RSR(b, d, λ) RSR(b, d, λ) =

R(b, d, λ) . Rmax (b, d, λ)

(2)

2200

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 44, NO. 8, AUGUST 2006

TABLE I RSB CWS AND BWS FOR TERRA AND AQUA MODIS

During MODIS spectral measurements, the environmental parameters were controlled and measured to enable correction for atmospheric absorption. In order to minimize water vapor absorption, the optical path region was purged with dry nitrogen. The atmospheric absorption was carefully evaluated and removed from the response profile. MODIS spectral bands have 10, 20, or 40 detectors in the along track direction corresponding to the 1-, 0.5-, or 0.25-km spatial resolutions. Additional corrections were applied in order to remove small wavelength differences along the SpMA exit slit (the smile effect) as the monochromator beam was projected (focused) onto different detectors within a band. From the relative spectral response RSR(b, d, λ) a given band and detector’s center wavelength λc (b, d) is then determined by  λ2 λc (b, d) =

λ1

λ · RSR(b, d, λ) · dλ

 λ2 λ1

RSR(b, d, λ) · dλ

(3)

where λ1 and λ2 are the wavelengths at which the RSR(b, d, λ) decreases to 1% of its peak. The BW is given by λR − λL where λR and λL are the wavelengths where the RSR(b, d, λ) is 50% of its peak, the full-width at half-maximum (FWHM). MODIS RSR data are available at the MODIS Characterization Support Team (MCST) web site.1 Table I lists the design specifications of the center wavelengths (CWs) and BWs for the MODIS RSBs together with the actual prelaunch measurement results (middle detector) from the SpMA for both Terra and Aqua MODIS. Examples of their in-band RSRs are plotted in Fig. 3 for bands 1, 3, 4, and 8–10. Band 8 is the only RSB with three peaks. The great similarity between the Terra and Aqua MODIS in-band RSRs is expected as both instruments were built using the same design parameters and nearly identical optical components and subsystems. The in-band RSR (numerical format) of 1 At ftp://ftp.mcst.ssai.biz/pub/permanent/MCST/PFM_L1B_LUT_4-30-99/ %20(Terra) and ftp://ftp.msct.ssai.biz/pub/permanent/MCST/FM1_RSR_LUT_ 07-10-01/%20(Aqua).

both Terra and Aqua MODIS (RSB and TEB) can be obtained from the MCST, NASA/Goddard Space Flight Center (GSFC). Prelaunch spectral characterization also included the OOB response measurements for each spectral band. Except for the SWIR bands (bands 5–7 and 26) that have noticeable thermal leaks and electronic crosstalk problems [18], the MODIS OOB responses are well within specification. The impact of the OOB response on the SWIR calibration and the corrections applied in the MODIS level 1B algorithm have been presented otherwise [19]. This paper only discusses the in-band spectral characterization. III. S PECTRAL C HARACTERIZATION T RANSFER F ROM G ROUND TO O N -O RBIT The MODIS onboard SRCA is an instrument by itself. It is capable of performing multiple functions as its name indicates. By slightly changing its configuration, the SRCA can be operated in three modes: radiometric, spatial, and spectral. When the SRCA is operated in spectral mode, it is configured as a monochromator with light sources and a collimator, as shown in Fig. 4. When it is in radiometric or spatial mode, a plain mirror replaces the grating and the monochromator becomes an optical relay [20]. This paper focuses on the SRCA’s function of performing on-orbit spectral characterization. The spectral characterization transfer from the ground measurements to the OBC was achieved prelaunch by operating the ground calibration device SpMA and the onboard SRCA in the spectral mode at nearly the same time and comparing their results. In the spectral mode, the SRCA light source consists of a spherical integration source (SIS) with six embedded lamps: four 10-W and two 1-W lamps with one of each used as the backup. The light from the SIS passes through a bandpass filter selected for a specific wavelength range and focuses onto the entrance slit of the monochromator. Different grating orders are used in order to cover the entire wavelength range. The exit slit of the monochromator is located at the focus of a Cassegrain system so that the light exiting the SRCA to the MODIS scan

XIONG et al.: TERRA MODIS ON-ORBIT SPECTRAL CHARACTERIZATION AND PERFORMANCE

Fig. 3.

2201

Terra and Aqua MODIS band responses (center detector).

zation, the SiPDs and MODIS detectors take data together at different grating angles (or step numbers). The calibration uses the following grating equation that relates the wavelength (λ) and corresponding grating angle (θ): λ=

Fig. 4.

SRCA layout in spectral mode.

mirror is collimated. As a spectral characterization device, the SRCA is capable of performing wavelength self-calibration in order to correct any wavelength bias or shift due to degradation of the mechanical/optical components. This self-calibration capability is achieved through the use of a didymium filter that has stable, well-calibrated spectral peaks and a pair of SiPD: one is the calibration SiPD, and the other the reference SiPD. The didymium filter is located next to the monochromator exit slit. Mounted just behind the didymium filter is the calibration SiPD (Fig. 4). The calibration SiPD’s response is proportional to the SRCA source spectrum, the didymiumfilter transmittance, and its own spectral responsivity. The reference SiPD, located at the secondary mirror of the Cassegrain collimator, is used to eliminate the source spectrum and the SiPD detector’s responsivity impact on the didymium peak measurements. When the SRCA performs spectral characteri-

2A · sin(θ + θOFF ) · cos β m

(4)

where A is the grating spacing, m is the grating diffraction order, θOFF is grating motor offset angle, and β is the halfangle between the incident and diffractive beams. The ratio of the calibration SiPD to the reference SiPD produces the didymium transmission profile as a function of SRCA grating angle and thus the relationship between the didymium peak wavelength and the grating angle. Three didymium peaks: one at 0.496 µm with grating order 3 and two at 0.551 µm with grating orders 2 and 3 are used in the SRCA wavelength calibration. Least square fitting [20] of the measured didymium peak profile to the nominal didymium peak profile determines the two parameters (θOFF and β) in the grating equation. The stability of two fitting parameters obtained during each calibration is directly related to the stability of the SRCA. The MODIS band/detector’s response to the SRCA illumination source during its spectral characterization, dn(b, d, λ), is proportional to its relative spectral response, RSR(b, d, λ), and the SRCA source spectrum. In order to remove the source spectrum, the reference SiPD’s response to the SRCA illumination, dnSiPD_ref (λ), is used. Thus, for a given MODIS band/detector, its relative spectral response, RSR(b, d, λ), is determined by RSR(b, d, λ) =

dn(b, d, λ) · RSRSiPD_ref (λ) dnSiPD_ref (λ)

(5)

2202

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 44, NO. 8, AUGUST 2006

TABLE II LAMP CONFIGURATIONS UTILIZED FOR THE RSB SPECTRAL CHARACTERIZATION

TABLE III TERRA ON-ORBIT SPECTRAL CALIBRATIONS

where RSRSiPD_ref (λ) is the known RSR of the reference SiPD. Therefore, the MODIS band/detector’s centroid wavelength λc (b, d) measured by the SRCA, is determined by  RSR(b, d, λ) · λ · dλ . (6) λc (b, d) = λ RSR(b, d, λ) · dλ λ

The reference SiPD’s response can only be used for the normalization of the bands with wavelengths below its spectral response cutoff at about 1 µm. This covers all MODIS RSBs except bands 5–7 and 26 (the SWIR bands). For this reason, there is no on-orbit wavelength characterization requirement for these four bands. The ground-based SpMA is typically operated at very narrow slit widths in order to reduce its impact on the RSR measurements. Due to illuminating source limitations, the SRCA is normally operated with much larger slit widths for the purpose of having better SNRs or source throughput. The SRCA measured band/detector response is therefore the convolution of the SpMA measured response with the SRCA slit function. The result is that the SRCA measured band CW differs from that of the SpMA. The partial filling of the MODIS aperture by the SRCA and their differences in solid angle and area product further bias the CW difference. In general, it can be reasonably assumed that this bias is unchanged from prelaunch to on orbit. With its internal wavelength self-calibration capability, the SRCA can therefore be used to track the MODIS spectral performance from prelaunch to on orbit and throughout its life. IV. MODIS O N -O RBIT S PECTRAL C HARACTERIZATION R ESULTS A. CW Trending The SRCA spectral calibration is performed on each RSB using either a 30- or 10-W lamp configuration (Table II). The reason for using different lamp configurations is to avoid saturation while maximizing the SNR of each band. The SRCA was characterized in a TV chamber at both sensor and spacecraft levels before launch. The SRCA was operated on-orbit before the MODIS nadir door was opened. Two months after the Terra MODIS nadir door was opened, the SRCA commenced normal spectral mode operations. A similar approach was used for Aqua MODIS. The SRCA is routinely operated in the spectral mode every three months. Table III shows the year and day of the 25 Terra MODIS spectral calibrations during the last five years.

Fig. 5. Trending of the Terra MODIS SRCA monochromator parameters, beta angle (top) and offset angle (bottom).

The challenge of running the SRCA on-orbit is that the spectral mode must be performed when both the Earth and the atmosphere are dark (called space dark). The reason stems from the high-gain setting of the reference SiPD. Notice that the reference SiPD is located after the monochromator’s exit slit so that the spectral irradiance on the SiPD is significantly reduced. Unlike the environmentally controlled prelaunch calibration in TV, the MODIS instrument is exposed on orbit to both earthshine and scattered atmospheric illumination. When the scan mirror is rotating, part of the irradiance passes through the SRCA aperture and is retro-reflected back onto the SiPD. This unwanted light adds significant noise to the signal during daytime calibrations. Earth scene illumination passed to the MODIS detectors during SRCA calibrations can also adversely affect the calibration. Duration of the space-dark portion of the orbit ranges from 32.9 to 34.7 min annually. Hence, the spectral calibrations

XIONG et al.: TERRA MODIS ON-ORBIT SPECTRAL CHARACTERIZATION AND PERFORMANCE

2203

TABLE IV TERRA MODIS ANNUAL CW SHIFTS (AVERAGE OVER DETECTORS AND MIRROR SIDES)

(> 120 min in duration) require four orbits to be completed. The first two orbits use the 30-W lamp configuration and the others the 10-W lamp. For the same lamp configuration but different orbits, the SRCA is in standby during the day and resumes when space dark occurs. The SRCA spectral measurement range includes three didymium peaks and the bands assigned to each lamp configuration. The SRCA-measured band RSR ranges vary depending on the RSR shape, BW, and SNR level. For a majority of the RSB, the SRCA measured RSR does not cover the entire wavelength range, but only the middle section of no less than 0.3 RSR. This is due to limitations in measurement time and inaccuracy near the RSR tails. The SRCA on-orbit performance is very stable. The monochromator parameters β and θOFF in (4) are determined during each spectral calibration. Their trends are shown in Fig. 5. β is in the range of 15 ± 0.1◦ and θOFF is 0.003◦ ± 0.004◦ over the five years. In the SRCA constant radiance mode, an embedded SiPD inside the SIS controls the broadband output to be constant to ±0.1% (unless otherwise noted, uncertainties correspond to one standard deviation) and ±0.5% for the 30- and 10-W configurations, respectively. Table IV provides the yearly averaged spectral band CW shift for all RSB (λ < 1 µm). Band 9 is used as an example to illustrate the shift from each detector and mirror side (Table V). The variation in the shift between detectors is ±0.06 nm and the mirror side difference is ±0.03 nm. Except for band 8, slightly over 0.6 nm, the CW shifts over the past five years are in general less than 0.5 nm. Presented in Fig. 6 are the Terra MODIS CW shifts from storage/launch through the last five years. It provides a comprehensive view of the spectral characterization performance on orbit. Two measurements made immediately after launch are not included in this trending because the system was not fully stabilized at the time. The CW shifts of all channels are plotted versus day excluding SWIR bands (5–7 and 26). Fig. 6 has used the same shift range (1.3 nm) for all plots for comparison purposes. The data point at day zero corresponds to prelaunch (zero shift). The X axis is the days from year 2000 and the Y axis is the CW shift. The error bar marks the range of shifts for all detectors. Along the X axis, two vertical lines, on days 305 and 549, mark electronic changes from A-side to B-side and then from B-side back to A-side. Although the changes of electronics impacted the detector gains

TABLE V TERRA MODIS ANNUAL CW SHIFTS (NANOMETERS) FOR BAND 9 DETECTORS AND SCAN MIRROR SIDES

and the SRCA lamp levels, no impact has been identified in the spectral characterization. The changes of the detector gains and the lamp levels only impact the signal levels. They do not affect the measured (relative) response profiles. The CW can shift during instrument storage and/or following launch-related environmental changes. Fig. 6 also shows that the CWs for the majority of the bands (except for bands 8–10, 16, and 18) had a positive shift immediately after launch. The CWs subsequently shifted toward negative. This mainly happened during the first 500 days and the shifts became fairly stable afterward. The overall trending is consistent among detectors within a band. Fig. 7 shows an example of the CW shift trending for all the detectors in band 13. This is the case for all the RSBs except for band 8 due to its small signal and low-SNR values. The standard deviation from the trending is typically within ±0.02 nm. B. Temperature Impact Prelaunch tests demonstrated that use of ion-assisted deposition (IAD) filters greatly reduces band response shifts from the ambient to TV environments for VIS/NIR bands. However, the environmental temperature could still induce CW shifts

2204

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 44, NO. 8, AUGUST 2006

Fig. 8. CW temperature coefficients measured prelaunch and on orbit with standard deviations.

Fig. 6. Terra MODIS CW trending over five years using the SRCA.

the RSBs with λ < 1 µm. The bars are the standard deviation of the coefficient measurements. We notice that the averaged temperature coefficient is 0.011 nm/K prelaunch and 0.008 nm/K on orbit. Although the temperature coefficients measured prelaunch and on orbit are both very small, with the on-orbit results showing more variation. The variation of temperature coefficient for prelaunch measurement appears to be related to the BW. Bands 1, 2, 17, and 19 have broad BWs (35–50 nm) in comparison with other narrower BWs (10 nm) bands. Low-SNR in the SRCA spectral characterization causes large variation for bands 3 and 8. In comparison with the temporal changes, the impact of instrument temperature change is minor. C. Recovered Band Response

Fig. 7. Terra MODIS CW shift trending for band 13 (all detectors).

to some degree. The MODIS temperature trending on orbit contains two components: slow increases over time (2 K over five years) and an annual variation with an amplitude of ±2 K. Prelaunch SRCA spectral calibrations at the Terra MODIS sensor level were completed when the instrument was at the nominal temperature plateau (about 268 K). During spacecraft level testing, the CWs were measured with the instrument at both the cold and hot plateaus. Therefore, the prelaunch temperature coefficients for CW shifts were calculated using the spacecraft level data. On orbit the CW shift has a temporal and an environmental temperature component. The CW trending line in Fig. 6 includes both effects. Excluding the slow change over time, the instrument temperature coefficient can be derived from the variation in CW relative to the trending line as a function of the instrument temperature variation. Fig. 8 illustrates the bandaveraged prelaunch and on-orbit temperature coefficients for

It was observed prelaunch that the RSR varied with polarization for the shortest wavelength band (band 8). The potential of on-orbit variation of the RSR profiles with changes in polarization is also a concern. It is understood that the SRCA measured detector profile is the convolution of the spectral response measured by the SpMA (narrow slit) and the SRCA slit function [20]. Because the SRCA is generally very stable, it can be assumed that the prelaunch measured SRCA slit function is fixed. Thus, the on-orbit RSR can be recovered from the SRCA measured profile using the SRCA slit function measured prelaunch. Fig. 9 shows examples of the recovered RSRs for bands 1, 3, 4, and 8–10 after applying a Fourier deconvolution approach [20]. The curves are band-averaged RSRs. There are four curves for each band, including the results from prelaunch (by the SpMA) and that from on-orbit spectral characterization (by the SRCA). It is observed that the RSR profiles are fairly stable. Small changes have been detected for some bands, particularly in band 8 (0.412 µm) at its shorter wavelength subpeak. Band 9 displayed some BW narrowing. As mentioned earlier, the on-orbit measurements by the SRCA do not cover the entire RSR range. Considering all the difficulties and challenges involved in the spectral characterization, such as wavelength calibration and reference transfer, and changes in the on-orbit operational environment, such as the temperature and electronics configuration, the Terra MODIS SRCA has been functioning well, providing useful information on the sensor’s spectral performance. The

XIONG et al.: TERRA MODIS ON-ORBIT SPECTRAL CHARACTERIZATION AND PERFORMANCE

Fig. 9.

2205

Recovered RSRs from the SRCA measurement.

results show that the Terra MODIS spectral characterization of the RSBs has been very stable. V. L ESSONS L EARNED F ROM SRCA S PECTRAL O PERATIONS The Terra MODIS SRCA has been operated on orbit in its spectral mode for over five years. This unique device provides valuable information about sensor spectral response from prelaunch to on orbit for the VIS/NIR bands, including CW shifts and response profile variations. Currently, the CW shifts have not been applied in the MODIS level 1 B (calibrated radiometric product) for two reasons: 1) SRCA is a new device and there is not sufficient information on the CW shifts (and their magnitudes) from precursor sensors and 2) spectral changes observed in Terra MODIS are within the SRCA’s calibration uncertainty. Therefore, it was decided to make the SRCA results available to the MODIS science communities for their independent review, evaluation, and applications. The MODIS spectral performance has been valuable for use in the design of other remote sensing instruments especially for its succession instrument, the Visible and Infrared Imaging Radiometer Suite for the National Polar-orbiting Operational Environmental Satellite System. The SRCA stability is normally tracked by evaluating the SRCA engineering parameters, such as lamp current and voltage, and output signals. On-orbit Terra MODIS SRCA operation in its spectral mode over five years shows that the SRCA has been operating normally and is stable. The on-orbit operation demonstrates that the goals of the SRCA design have been achieved including the calibration approach, algorithms,

and expected precision [20]. The operation of the SRCA in its spectral mode also shows as follows. 1) SRCA can provide precise on-orbit trending of the sensor’s spectral performance. 2) Sensor’s RSR, CW and BW can change during storage prior to launch and during on-orbit operations. 3) CW shifts change with time and instrument temperature variations. For Terra MODIS time is the dominant factor because the instrument’s temperature has been very stable on orbit with an annual increase of less than 0.5 K. VI. C ONCLUSION This paper provides a brief description of the MODIS spectral characterization performed prelaunch using the groundbased SpMA and the onboard SRCA and comparisons of the CW and the RSR for both the Terra and Aqua MODIS spectral bands with wavelengths less than 2.2 µm. Observations from over five years of on-orbit operation show that the SRCA has been functioning with good stability and measurement repeatability. The Terra MODIS CWs have shifted since prelaunch characterization, slightly toward shorter wavelengths during the first 1.5 years of on-orbit operation, for most of the bands, and have remained relatively stable since. Except for band 8, the CW shifts for all the Terra MODIS VIS/NIR bands are less than 0.5 nm. If the initial changes from prelaunch to shortly after launch are ignored, the on-orbit shifts over the last five years have been within 0.2 nm on average. It seems that the use of IAD technology stabilizes the band CWs. The lessons learned from the MODIS on-orbit spectral characterization

2206

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 44, NO. 8, AUGUST 2006

have not only benefited the MODIS science communities, but continue to provide valuable information for the design of future remote sensing instruments. ACKNOWLEDGMENT The authors wish to thank their Raytheon SBRS colleagues for sharing expertise and knowledge, for their technical support, and for invaluable discussion. R EFERENCES [1] W. L. Barnes and V. V. Salomonson, “MODIS: A global image spectroradiometer for the Earth Observing System,” Crit. Rev. Opt. Sci. Technol., vol. CR47, pp. 285–307, 1993. [2] V. V. Salomonson, W. L. Barnes, X. Xiong, S. Kempler, and E. Masuoka, “An overview of the Earth Observing System MODIS instrument and associated data systems performance,” in Proc. IGARSS, 2002, pp. 1174–1176. [3] W. L. Barnes, X. Xiong, and V. V. Salomonson, “Status of Terra MODIS and Aqua MODIS,” in Proc. IGARSS, 2002, pp. 970–972. [4] C. L. Parkinson, “Aqua: An earth-observing satellite mission to examine water and other climate variables,” IEEE Trans. Geosci. Remote Sens., vol. 41, no. 2, pp. 173–183, Feb. 2003. [5] X. Xiong, K. Chiang, J. Esposito, B. Guenther, and W. L. Barnes, “MODIS on-orbit calibration and characterization,” Metrologia, vol. 40, no. 1, pp. 89–92, Feb. 2003. [6] X. Xiong, J. Sun, J. Esposito, B. Guenther, and W. L. Barnes, “MODIS reflective solar bands calibration algorithm and on-orbit performance,” Proc. SPIE, vol. 4891, pp. 95–104, 2002. [7] X. Xiong, K. Chiang, B. Guenther, and W. L. Barnes, “MODIS thermal emissive bands calibration algorithm and on-orbit performance,” Proc. SPIE, vol. 4891, pp. 392–401, 2002. [8] H. Montgomery, N. Che, and J. Bowser, “Determination of the spatial characteristic by using the Spectro-Radiometric Calibration Assembly (SRCA) of MODIS (Part I. Along-scan),” Proc. SPIE, vol. 3439, pp. 226– 237, 1998. [9] ——, “Determination of the spatial characteristic by using the SpectroRadiometric Calibration Assembly (SRCA) of MODIS (Part II. Alongtrack),” Proc. SPIE, vol. 3439, pp. 238–246, 1998. [10] X. Xiong, N. Che, and W. L. Barnes, “Terra MODIS on-orbit spatial characterization and performance,” IEEE Trans. Geosci. Remote Sens., vol. 43, no. 2, pp. 355–365, Feb. 2005. [11] N. Che, X. Xiong, and W. L. Barnes, “On-orbit spectral characterization of the terra MODIS reflective solar bands,” Proc. SPIE, vol. 5151, pp. 367–374, 2003. [12] J. Heaney, K. Stewart, R. Boucarut, P. Alley, and A. Korb, “The cyrotesting of infrared filters and beamsplitters for the Cosmic Background Explorer’s instruments,” Proc. SPIE, vol. 619, pp. 142–147, 1986. [13] L. Liao, P. Jarecke, D. Gleichauf, and T. Hedman, “Performance characterization of the Hyperion Imaging Spectrometer instrument,” Proc. SPIE, vol. 4135, pp. 264–275, 2000. [14] M. Dobber and E. Zoutman, “Calibration of the SCIMACHY instrument on the ESA-ENVISAT satellite,” Proc. SPIE, vol. 3750, pp. 407–418, 1999. [15] Y. Dutil, S. Lantagne, S. Dube, and R. Poulin, “ACE-FTS level 0 to 1 data processing,” Proc. SPIE, vol. 4814, pp. 102–110, 2002. [16] M. Dittman, J. Leitch, and M. Chrisp et al., “Limb broad-band imaging spectrometer for the NPOESS Ozone Mapping and Profiler Suite (OMPS),” Proc. SPIE, vol. 4814, pp. 120–130, 2002.

[17] T. Pagano, T. Hearty, and S. Gaiser et al., Atmospheric Infrared Sounder (AIRS) Level 1B Visible, Infrared and Telemetry Algorithms and Quality Assessment (QA) Processing Requirements, 2003. version 2.2, JPL-D20046. [18] X. Xiong, K. Chiang, and F. Adimi et al., “MODIS correction algorithm for out-of-band response in the shirt-wavelength IR bands,” Proc. SPIE, vol. 5234, pp. 407–418, 1999. [19] X. Xiong, K. Chiang, J. Sun, N. Che, and W. L. Barnes, “MODIS on-orbit calibration: Key issues and approaches,” Proc. SPIE, vol. 5542, pp. 24–34, 2004. [20] H. Montgomery, N. Che, K. Parker, and J. Bowser, “The algorithm for MODIS wavelength on-orbit calibration bands using the SRCA,” IEEE Trans. Geosci. Remote Sens., vol. 38, no. 2, pp. 877–884, Mar. 2000.

X. Xiong received the B.S. degree in optical engineering from the Beijing Institute of Technology, Beijing, China, and the Ph.D degree in physics from the University of Maryland, College Park. He is currently an Optical Physicist with NASA’s Goddard Space Flight Center, Greenbelt, MD, working on the EOS Terra/Aqua MODIS project and NPOESS/VIIRS instrument calibration and characterization. Before he joined NASA’s instrument calibration program, he was in the fields of nonlinear optics, laser/atomic spectroscopy, and mass spectrometry at private industry and at the National Institute of Standards and Technology.

N. Che received the engineering degree from the Beijing Institute of Technology, Beijing, China, in 1963. He is currently a member of the MODIS Calibration Support Team and the NPOESS/VIIRS Instrument Calibration Support Team. He was an Associate Professor with the Beijing Institute of Technology in 1985. He was with Swales Aerospace, Inc., USDA Beltsville Agriculture Research Center Remote Sensing Laboratory, Beltsville, MD, and is currently with Science System and Applications, Inc., Lanham, MD. He has been working in remote sensing and laboratoryrelated measurement, instrument calibration, and data applications since 1982 and has worked on TM, AVHRR, SPOT, MODIS, and NPOESS/VIIRS.

W. L. Barnes received the B.S. and M.S. degrees from the University of North Texas, Denton, in 1960 and 1961, respectively, and the Ph.D. degree from Florida State University, Tallahassee, in 1970, all in physics. He is currently a Senior Research Scientist with the Joint Center for Earth Systems Technology, University of Maryland Baltimore County, Baltimore, and an Emeritus Research Scientist with the Earth Sciences Directorate, NASA’s Goddard Space Flight Center, Greenbelt, MD. He was a MODIS Sensor Scientist and was a member of the MODIS Science Team for more than 12 years. He led the MODIS Characterization Support Team for two years and has over 30 years experience in the development and radiometric calibration of Earth-observing imaging radiometers.