Suomi NPP CrIS measurements, sensor data record algorithm, calibration and validation activities, and record data quality

Authors


Abstract

[1] The Cross-Track Infrared Sounder (CrIS) is a Fourier Transform Michelson interferometer instrument launched on board the Suomi National Polar-Orbiting Partnership (Suomi NPP) satellite on 28 October 2011. CrIS provides measurements of Earth view interferograms in three infrared spectral bands at 30 cross-track positions, each with a 3 × 3 array of field of views. The CrIS ground processing software transforms the measured interferograms into calibrated and geolocated spectra in the form of Sensor Data Records (SDRs) that cover spectral bands from 650 to 1095 cm−1, 1210 to 1750 cm−1, and 2155 to 2550 cm−1 with spectral resolutions of 0.625 cm−1, 1.25 cm−1, and 2.5 cm−1, respectively. During the time since launch a team of subject matter experts from government, academia, and industry has been engaged in postlaunch CrIS calibration and validation activities. The CrIS SDR product is defined by three validation stages: Beta, Provisional, and Validated. The product reached Beta and Provisional validation stages on 19 April 2012 and 31 January 2013, respectively. For Beta and Provisional SDR data, the estimated absolute spectral calibration uncertainty is less than 3 ppm in the long-wave and midwave bands, and the estimated 3 sigma radiometric uncertainty for all Earth scenes is less than 0.3 K in the long-wave band and less than 0.2 K in the midwave and short-wave bands. The geolocation uncertainty for near nadir pixels is less than 0.4 km in the cross-track and in-track directions.

1 Introduction

[2] The Cross-Track Infrared Sounder (CrIS) is a Fourier transform spectrometer (FTS) on board the Suomi National Polar-Orbiting Operational Environmental Satellite System Preparatory Project satellite (S-NPP), which was launched on 28 October 2011 into an orbit with an altitude of 824 km above the Earth surface, an inclination angle of 98.7°, and a 13:30 local time ascending node. S-NPP is the first in a series of next-generation U.S. weather satellites of the Joint Polar Satellite System (JPSS). CrIS is a major step forward in the U.S. operational infrared (IR) sounding capability previously provided by the High-resolution Infrared Spectrometer (HIRS) of the Television Infrared Observation Satellite (TIROS-N) Operational Vertical Sounder (TOVS) series. CrIS provides measurements of radiance spectra of 1305 channels in the three spectral bands: the long-wave IR (LWIR) band from 650 to 1095 cm−1, midwave IR (MWIR) band 1210 to 1750 cm−1, and short-wave IR (SWIR) band 2155 to 2550 cm−1, with spectral resolutions of 0.625 cm−1, 1.25 cm−1, and 2.5 cm−1, respectively. The CrIS high spectral resolution, large number of channels, and high signal-to-noise ratio combine to provide much improved vertical sounding resolution and accuracy in comparison to the HIRS [Smith et al., 2009]. CrIS will continue the measurements provided by the Atmospheric Infrared Sounder (AIRS) [Chahine et al., 2006] on the Aqua satellite, launched into orbit on 2 May 2002. It achieves a similar spectral resolution and spectral coverage on a similar early afternoon orbit as the AIRS grating spectrometer [Aumann et al., 2003]. The CrIS is the second FTS for operational weather applications next to the Infrared Atmospheric Sounding Interferometer (IASI) on the European MetOp-A satellite, launched into a morning orbit on 19 October 2006 [Chalon et al., 2001]. As demonstrated by the improvements in weather forecast accuracy through the assimilations of AIRS and IASI radiance data in the Numerical Weather Prediction (NWP) models [Le Marshall et al., 2006; Hilton et al., 2012], CrIS will provide critical global high vertical resolution IR sounding data for years to come.

[3] The concept of the high spectral resolution FTS IR sounding instrument for weather applications was first demonstrated in the mid-1980s by the High Resolution Interferometer Sounder (HIS) [Revercomb et al., 1988b]. That study showed that FTS technology and data processing techniques can provide unprecedented rich atmospheric sounding information for weather applications. The success of HIS motivated the 1997/2000 Phase-1 design of the CrIS system. It was followed by more detailed design and the ultimate system build by Exelis and ABB Corporation, with the primary objective of delivering the highest quality data product using the smallest and lightest sensor possible [Glumb and Predina, 2002; Kohrman and Luce, 2002; Stumpf and Overbeck, 2002]. The CrIS system includes both the instrument and ground processing software. The instrument provides interferogram measurements and calibration data in the form of Raw Data Records (RDRs). The processing software converts the interferogram measurements into calibrated and geolocated radiance spectra in the form of Sensor Data Records (SDR). The CrIS system is part of the S-NPP Cross-Track Infrared and Microwave Sounding Suite (CrIMSS). CrIS and the Advanced Technology Microwave Sounder (ATMS) together provide sounding data for the CrIMSS one-dimensional variational retrieval algorithm to derive atmospheric pressure, temperature, and water vapor profiles.

[4] The CrIS SDR calibration and validation (Cal/Val) process includes both prelaunch and postlaunch activities performed by a team comprising subject matter experts from government, academia, and industry. The prelaunch Cal/Val activities included characterizations of the instrument performance, validations of the SDR algorithm and software, and preparation for postlaunch Cal/Val activities. The prelaunch Cal/Val work successfully established the baseline instrument characterization and performance [Zavyalov et al., 2011] as well as the initial set of SDR calibration coefficients and parameters for postlaunch SDR processing. The objectives of the postlaunch Cal/Val activities are to safely and effectively configure CrIS for operational conditions, conduct RDR verification, checkout instrument functions, optimize instrument settings, refine CrIS algorithms, tune calibration parameters, and validate the CrIS SDR products. At the time of this writing (June 2013), the Cal/Val activities are still ongoing. So far, the postlaunch Cal/Val work has achieved great success. The radiometric and spectral performance specifications have been met with significant margin. In this special issue, there are five papers devoted to the CrIS SDR Cal/Val work. The instrument noise performance is summarized in Zavyalov et al. [2013], the spectral and radiometric performances in Strow et al. (Frequency calibration and validation of CrIS satellite sounder, submitted to Journal of Geophysical Research: Atmospheres, 2013) and Tobin et al. [2013a], respectively, and the geolocation accuracy assessment in Wang et al. [2013].

[5] In this paper we describe the CrIS measurement characteristics in section 2 and SDR processing algorithms in section 3 and summarize the postlaunch Cal/Val activities in section 4. In section 5 we provide an assessment of the CrIS SDR product quality. A list of acronyms in an alphabetic order is given in Appendix A.

2 CrIS Measurement Characteristics

[6] The CrIS instrument measures interferograms, an interference pattern introduced when the incoming radiation passes through the interferometer. The interferogram data are then converted to radiance spectra by the ground SDR processing system, a process that will be discussed in the next section. Figure 1 shows the signal radiance to detector flow of the CrIS optical system. The CrIS interferometer includes a beamsplitter, a porchswing moving mirror, a dynamic alignment (DA) mirror, and a metrology system (not shown in the diagram). The incoming radiation is divided by the beamsplitter into two beams along two paths. One beam travels toward the porchswing mirror, while the other to the stationary mirror. The two beams are reflected from the corresponding mirrors and recombined at the beamsplitter. The optical path difference (OPD) between the two beams is twice the physical path difference between the two mirrors in vacuum. As the OPD changes with the sweep of the moving mirror, a time-varying interference pattern is produced by the interferometer. The DA mechanism on the stationary mirror maintains an active alignment between the two beams over the full path and therefore a stable modulation of the signal. The CrIS interferogram measurements are double sided with the OPD varying from a negative maximum path difference (MPD), through the zero path difference (ZPD), to a positive MPD.

Figure 1.

Partially unfolded CrIS optical system showing flow of signal radiance to detectors.

[7] In order to accurately sample an interferogram at equal intervals in OPD as well as provide a reference to monitor and maintain the DA mirror alignment, a 1550 nm laser metrology system is used. The laser metrology injects a laser beam into the center of the interferometer optical path. Multiple metrology detectors convert the modulated laser fringes into electrical pulses used as interferogram sampling signals. As the laser wavelength may vary with time, a neon metrology system provides periodic precise measurements of the laser wavelength using spectrally ultrastable neon emission lines. The light from the neon lamp is injected in the same interferometer optical path as the metrology laser to generate the neon fringes. During a sweep interferogram measurement, counters are used to determine the number of fringes from both the metrology laser and neon lamp, and the recorded data are used to determine the laser wavelength through a process of counting and interpolating neon fringes relative to the laser. The neon data set is generated roughly once per orbit.

[8] After the interferometer modulation, the radiance is converged by the telescope onto the LWIR, MWIR, and SWIR detector focal planes through the Aft optics, which split the radiance beam coming from the telescope into the three beams arriving at the three detector focal planes [Stumpf and Overbeck, 2002]. The detector focal planes are cooled with a four-stage passive cooler maintaining temperatures of 205, 147, 98, and 81 K for stages 1 to 4, respectively. The LWIR focal plane operates at 81 K, while the MWIR and SWIR focal planes operate at 98 K. Nine detectors on each focal plane are arranged into a 3 × 3 grid. The size and position of the detector field stop define the field of view (FOV) for each detector. The combined 3 × 3 FOVs define the field of regard (FOR). The detectors convert the optical radiation into analog electrical signals, which are amplified and then sampled by 14 bit analog-to-digital (AD) converters triggered by the electrical pulses provided by the laser metrology. The OPD sampling interval is half of the laser wavelength.

[9] Figure 2 shows an example of a measured Earth scene (ES) LWIR band interferogram at the AD converter output. The digitized interferogram measurements are further processed by a complex finite impulse response (FIR) digital band-pass filter to reject out-band signals and noise. Unlike the predetection optical band-pass filter, the digital filter can attenuate features that may result from the following: out-of-band harmonic distortion caused by nonlinear photodiode response; second harmonic generation by modulated radiation that reflects off the focal plane and makes a second pass through the interferometer; and out-of-band electronic noise. In so doing, the digitally filtered interferogram can be resampled at a much lower rate (decimation) without undergoing any degradation from noise or aliasing. Since the complex FIR filter is designed to have high rejection of its image pass band (corresponding to the mirror image of the input spectrum that results from the Fourier transform process), the decimation factors (DF) in use are twice as large as those achievable with a real-number filter. The sample rate reduction from digital filtering/decimation is achieved without any loss of in-band information and without discernable signal-to-noise ratio degradation in band. Table 1 lists the numbers of interferogram data points before and after the decimation, the decimation factor, and the values of the radiometric MPD. After the FIR filtering and decimation, the complex interferogram is compressed with a bit trim encoding scheme. CrIS can also be operated in the full spectral resolution (FSR) mode, in which the MWIR and SWIR band interferograms are recorded with the same MPD as the LWIR band.

Figure 2.

Typical interferogram kernel of an Earth view acquired with the diagnostic mode (undecimated).

Table 1. CrIS Normal Mode Interferogram Samplesa
BandUndecimated Samples Spanning 2∙MPDSamples After DecimationDecimation FactorRadiometric MPD (cm)
  1. a

    Numbers of interferogram samples spanning 2 MPD before (column 2) and after (column 3) decimation, decimation factor (column 4), and MPD assuming an OPD sampling interval of 775 nm.

LWIR20,736864240.8035
MWIR10,560528200.4092
SWIR5,200200260.2015

[10] For radiometric calibration, CrIS also provides measurements of two known radiation sources. One is the Internal Calibration Target (ICT), a high-precision calibration blackbody, and the other is the Deep Space (DS), a source with negligible IR radiance. The CrIS ICT temperature is not thermally stabilized but instead is allowed to remain in thermal equilibrium with the rest of the instrument environment. This minimizes reflected radiance errors that could be associated with imperfect ICT emissivity. The ICT temperature is measured with two high-precision platinum resistance thermometers. The three radiation sources, ICT, DS, and ES, are selected by the scan mirror of the CrIS scene selection module (SSM), which is canted 45° relative to the cross-track rotational axis. The scan mirror can rotate in both cross- and in-track directions to make cross-track scan and in-track compensation for the spacecraft motion. A cross-track measurement scan sequence consists of 34 interferogram sweeps, including 30 ES, 2 DS, and 2 ICT measurements. One complete scan sequence takes 8 s. The step of the Earth view scan between two adjacent observation positions is 3.33°, and the full Earth view scan angle is ±48.33°. Half of the ES, DS, and ICT interferograms are measured when the porchswing mirror moves in a forward direction and the other half in the reverse direction. Radiometric calibrations are performed separately for different sweeping directions. Each ES, DS, or ICT interferogram measurement takes 0.2 s, including 0.167 s SSM dwell time for data acquisition and 0.033 s for repositioning. To compensate the in-track spacecraft motion, the scan mirror rotates slightly (within 0.08°) along the in-track axis so that a FOV footprint on the Earth surface is fixed during the measurement.

[11] The Earth scene motion due to Earth's rotation is not compensated and can cause small but detectable Doppler shift of the ES spectral frequency with a magnitude depending on the location of the observations. The shift can reach a value of 1.25 × 10−6σ, or 1.25 parts per million (ppm), where σ is the frequency of the spectrum for those FORs at the edges of the scan near the equator. A detailed analysis of the Doppler effect on CrIS measurements is given in Chen et al. [2013a].

[12] The scan measurement sequence, FORs, and FOVs are depicted in Figure 3. The figure shows measurements for the ascending portion of an orbit with the scan direction from the left to the right. The nine small circles or ellipses represent the footprints of the FOVs on the ground for each of the three bands. The dashed circles or ellipses enclosing the nine FOVs represent the FORs. The FOV with index 5, FOV 5, is called the center FOV. The FOVs 2, 4, 6, and 8 are called side FOVs, whereas the remaining FOVs are called corner FOVs. The separation between a FOV and FOV 5 is measured with the angle between the centers of the two FOVs, which may be envisioned by the angle between the two lines from the centers of the two FOV footprints on ground to the satellite. The angle can be expressed with the two angles, one in the cross-track direction and the other in the in-track direction, referred to as the cross-track and in-track offset angles, respectively. As shown in Figure 3, the nominal cross-track and in-track offset angles are 1.1°. The actual values of the offset angles may slightly vary among different FOVs. The sizes of the FOVs are also measured with angle (= 0.963°), and at nadir the size of the footprints is about 14 km. The FOV size and offset angles are instrument line shape (ILS) parameters that together with the MPD parameters define the spectral response functions (SRFs) of the CrIS spectra prior to spectral calibration. Note that the FOV and FOR indexes defined in Figure 3 are consistent with the data array indexes used in the SDR product.

Figure 3.

A ground cross-track of 30 FOR viewed from the satellite on an ascending orbit. The ground footprints of the FOVs are represented by the solid small circles or ellipses and the FORs by the dashed large circles or ellipses. The indexes of the FOVs and FORs are consistent with indexes of the data arrays in the SDR product.

[13] The CrIS band-to-band coregistration defines how well individual FOV centroids are aligned between corresponding FOVs in the three IR bands. The requirement is that corresponding FOV centroids in the three different bands need to be aligned to within 1.4% of the FOV diameter. The radial offset positions with respect to the optical axis of the nine detectors on each of the three focal planes are determined postlaunch using the spectral calibration procedure discussed by Strow et al. (submitted manuscript, 2013). The largest difference between the center of a LWIR FOV with respect to the optical axis and the corresponding MWIR or SWIR FOV position was found to be 0.8% of the FOV diameter, well below the 1.4% coregistration specification.

[14] The data generated by the instrument are stored in various application packets in the form of RDRs, and the packets are broadcast from the spacecraft to the ground receiving station. On the ground they are packed into RDR granules. The RDRs processed by the CrIS SDR algorithm are science RDRs of three types: interferogram packet, 8 s science/calibration packet (SciCalP) and 4 min engineering packet (EP). An interferogram packet contains an ES, DS, or ICT interferogram measurement. An 8 s SciCalP is created every 8 s for each scan. It contains calibration data such as ICT and scan baffle temperature measurements. A 4 min EP is created every 4 min, or after every 30 measurement scans. It includes calibration parameters and tables (e.g., ILS parameters and ICT emissivity table) and the neon metrology measurements. The calibration parameters and tables are static and require an EP upload to change their content. The neon data are measurements taken roughly once per orbit and are automatically injected into EPs in space. A special feature of the CrIS calibration data set is that all the CrIS calibration parameters and tables are included in EPs, which are embedded in the data stream. A granule usually includes four scans of interferogram RDR packets, four corresponding SciCalPs, and one or no EP.

3 SDR Processing Algorithms

[15] The purpose of the SDR processing algorithms is to convert interferogram measurements to well-calibrated and geolocated Earth scene radiance spectra (SDR products) as specified in the requirements, some of which are shown in Table 2. In the following we will briefly describe the SDR algorithms as we follow the processing flow shown in Figure 4. Detailed descriptions of the algorithms are given in the CrIS SDR Algorithm Theoretical Basis Document [JPSS Configuration Management Office, 2011].

Table 2. CrIS SDR Requirements and Uncertainty Specificationsa
BandSpectral Range (cm−1)MPD (cm)Resolution (cm−1)Number of ChannelsNEdN mW/m2/sr/cm−1Radiometric Uncertainty (%)Frequency Uncertainty (ppm)Geolocation Uncertainty (km)
  1. a

    The radiometric uncertainty specification is defined as a percent of the287 K blackbody radiance for all the band channels, and if expressed as brightness temperature differences, the specified uncertainty is a function of channel frequency [see Tobin et al., 2013a, Figure 1].

LWIR650–10950.80.6257130.140.45101.5
MWIR1210–17500.41.254330.050.58101.5
SWIR2155–25500.22.51590.0070.77101.5
Figure 4.

CrIS SDR processing flow (see text for descriptions).

[16] The data input to the SDR processing software are science RDRs as well as spacecraft level ephemeris and attitude information for geolocation calculation. The preprocessing module unpacks the RDR granules and application packets, performs quality control tests, and establishes a 4 min moving window for averaging DS and ICT data. Usually, the moving window contains an EP and 30 consecutive scans of data, 14 scans before and 15 scans after the scan of data that will be processed for SDR output. A scan of data includes a SciCalP and a total of 918 interferogram packets from the 34 FORs (30 ES, 2 DS, and 2 ICT views) of the three spectral bands. The preprocessing module also computes laser metrology wavelength from neon calibration data stored in EP and determines the need to rebuild the Correction Matrix Operator (CMO) (which will be discussed shortly.) The Fast Fourier Transform (FFT) module converts the ES, DS, and ICT interferograms (IGMs) to raw spectra. The raw spectra have more spectral bins and a larger spectral range than the final SDR spectra. The set of extra bins at the beginning or end of the spectra is called guard band, in which the radiance signals have been damped by the optical and FIR digital filters. The guard bands are discarded at the end of the SDR process. The function of the Fringe Count Error (FCE) handling module is to detect and correct phase errors of the raw spectra due to interferogram sampling shift. Unfortunately, the FCE detection and correction module did not work robustly and therefore is disabled as of June 2013. Fortunately, the instrument has been working so well that not a single FCE event has occurred since instrument commissioning. An improved FCE algorithm has been formulated and will be implemented in the near future.

[17] The nonlinearity (NL) levels of all the LWIR and some MWIR band detectors of the S-NPP CrIS are high enough to require application of the NL correction algorithm on the raw spectra [Knuteson et al., 2013] to meet requirements. The NL correction module removes the second-order NL by scaling the raw spectrum with a factor applied for all wave numbers in the spectral band. The factor is a function of a2V, where a2 is the NL coefficient and V represents the direct current (DC) voltage at the detector preamplifier output due to the photon flux and detector dark current [JPSS Configuration Management Office, 2011]. The a2 coefficient is predetermined for each detector, while the DC component V is calculated dynamically for each DS, ICT, or ES spectrum. For DS spectrum, V is the output DC voltage Vinst, produced by the instrument background photon flux and detector dark current and is measured when the instrument views DS. For ICT or ES spectrum, a second term is added to V, representing the additional change of preamplifier output DC voltage as the instrument's view changes from DS to ICT or ES, and is estimated from the raw spectra.

[18] After the NL corrections, a complex radiometric calibration is performed on the ES spectrum using the DS and ICT spectra as from two known radiation sources [Revercomb et al., 1988a]. The calibrations are done separately for each moving mirror sweep direction. In the radiometric calibration process, the DS and ICT spectra are averages of the spectra within the average moving window, and the ICT Planck radiance is adjusted to take into account of the beam divergence effect, known as self-apodization. To accurately determine the ICT radiance, the radiometric calibration module includes an ICT model to take into account the amount of radiance reflected by the ICT that originates from the surrounding environment such as the scan mirror baffle surface [JPSS Configuration Management Office, 2011]. The contribution of reflected radiance to the CrIS radiometric calibration uncertainty is described in Tobin et al. [2013a]. To handle the situations when the DS view includes the Moon, the SDR software includes a scheme to detect the affected DS spectra and remove them in the average moving window. The ES spectra calibrated during a lunar intrusion event are marked with the Lunar Intrusion flag for information purpose. These ES spectra are still valid because the loss of one to three DS views due to brief lunar intrusion out of 30 DS views normally used for calibration does not significantly impact data quality. The radiometric calibration module also computes the noise equivalent differential radiance (NEdN) for each ES spectrum using the 30 radiometrically calibrated ICT spectra within the 4 min moving window centered at the ES spectrum [Zavyalov et al., 2013].

[19] A Fourier transform operation results in a real and imaginary spectral output. For ideal spectra with no noise, the radiometric calibration would result in a spectrum with the imaginary part of the radiance equal to zero for all channels. This property is very useful to identify if the real part of the spectra is erroneous, since contaminated interferogram data will produce large unexpected magnitudes in the imaginary part of spectra. A new quality control (QC) algorithm was implemented on 15 October 2012 in the SDR software to invalidate a spectrum if the magnitude of its imaginary radiance values exceeds a predefined threshold (see Table 3).

Table 3. Imaginary Radiance Threshold Values Used for Identifying Invalid Spectra and Mean Imaginary Radiance Values Over the Specified Spectral Ranges for Typical Imaginary Radiance Spectra
BandSpectral Range (cm−1)Threshold Value (mW/m2/sr/cm−1)Mean Value (mW/m2/sr/cm−1)
  1. a

    When the magnitude of the imaginary radiance in the spectral range is larger than the threshold, the spectrum is flagged as invalid.

LWIR800–9801.50.08
MWIR1500–17000.50.05
SWIR2250–23500.050.003

[20] Following the radiometric calibrations, the ES spectra enter the process of spectral correction. The process is completed with the three operations enclosed in the dashed box in Figure 4. First, a digital band-pass filter (BPF) is applied to attenuate noise signal in the guard bands which were amplified during the radiometric calibrations. This prevents the guard band noise from leaking into the two subsequent spectral correction operations. Next, the spectral resampling operation maps the spectrum from the instrument's spectral grid (determined by laser metrology wavelength and the radiometric MPD shown in Table 1) onto a common user's spectral grid defined by the SDR MPD and spectral ranges as defined in Table 2. Finally, the ILS correction operation removes the interferometer self-apodization effect from spectra so that the CrIS channel SRF is an ideal Sinc function defined by the SDR MPD of the band, as shown in Figure 5. The CrIS SDR channel radiance is given by a convolution of the SRF with the monochromatic radiance at the entrance to the interferometer. To be computationally efficient, the above three spectral correction operations are combined into a single matrix multiplication of the spectrum. The matrix is referred to as Correction Matrix Operator (CMO) as mentioned earlier. The CMO is automatically rebuilt if the cumulative change of the metrology laser wavelength exceeds a predefined threshold (currently set to 2 ppm) or if the ILS parameters in an EP are updated. A CMO file can be operationally provided at the initialization of the SDR process, and if it is not presented at the initialization, the CMO will be computed using the data in EP from the input data stream.

Figure 5.

CrIS SDR channel spectral response functions of the LWIR (solid curve), MWIR (dotted curve), and SWIR (dashed curve) bands as a function of wave number. FWHM, full width at half maximum.

[21] The geolocation module computes the CrIS line-of-sight (LOS) pointing vector relative to the spacecraft body frame for each FOV and scan position, using a host of information such as the mapping and ILS parameters in EP, time stamp from the interferogram packet, and the scan mirror cross- and in-track servo errors in SciCalP. The LOS vector is then passed to the spacecraft level geolocation algorithms to compute the FOV center location, where the LOS intercepts the WGS84 earth reference ellipsoid. The geolocation product includes the geodetic longitude and latitude of each individual FOV with no terrain correction.

[22] The operational CrIS SDR algorithms are coded in C++ programming language. The software has two packages, which share the same processing code. One package is for operational use, running on the Interface Data Processing Segment (IDPS). The other package is the Algorithm Development Library (ADL), which can run on multiple computing platforms including the Linux operating system. The ADL provides a framework for algorithm and software development and refinement, bug fix, and anomaly investigation. ADL uses file based inputs and outputs, while IDPS uses a Data Management Subsystem (DMS) to manage the inputs and outputs.

[23] The CrIS SDR product generated by the SDR software includes complex ES radiance spectra, geolocation data, and various QC variables and flags, such as NEdN and the SDR overall Quality Flag (QF). The channel SRF is a Sinc function as mentioned earlier (Figure 5). In other words, the radiance spectrum is unapodized. The Sinc SRF has large side lobes which alternate between negative and positive values. Because of the large side lobes, it is not unusual to see negative radiance values in some SWIR band channels. To reduce the side lodes, apodization functions may be applied to an interferogram [Barnet et al., 2000]. The Hamming apodization function is a reasonable and efficient function to use for CrIS because the channel SRF after the apodization has side lobes less than 1% of the central lobe. The Hamming apodization may be performed in the spectral domain on a spectrum with a running mean of the three-point smooth filter {a, 1-2a, a}, where a = 0.23. The cost of the Hamming apodization is an increase of the channel SRF width by 50.4%. To facilitate an apodization with a Hamming or other apodization functions, two additional spectral channels are added to each side of a CrIS band in the SDR product.

[24] Figure 6 shows an example of IDPS-produced ES spectrum (red curve). In the same figure, an FSR spectrum is also shown to demonstrate that the CrIS system is able to provide FSR spectra. The FSR RDR data were collected during a short period when the CrIS was operated in the FSR mode and processed by a modified ADL code.

Figure 6.

An example of CrIS SDR LWIR, MWIR, and SWIR spectra (in red curve) produced by IDPS. The spectra in black curve were produced with a modified ADL code from full spectral resolution RDRs, collected when the CrIS was operated in the full spectral resolution mode. During mission operation, the IDPS only produces SDRs with spectral resolutions specified in Table 2.

4 Postlaunch Cal/Val Activities and Developments

[25] The CrIS postlaunch SDR Cal/Val process comprises three phases: Early Orbit Checkout (EOC), Intensive Calibration and Validation (ICV), and Long Term Monitoring (LTM). The EOC phase started on 18 January and ended on 23 February 2012. Activities included sensor activation, instrument checkout and optimization, and RDR data verification. To support these activities and those in the next two phases, both normal and special mode data were collected. The special mode data sets include diagnostic interferogram measurements (interferograms before FIR filtering and decimation) and a set of 22 orbits of FSR data. An EP update (version 32) was uploaded to CrIS on 31 January 2012 with updates of the Programmable Gain Amplifier settings and bit trim table. At the end of the EOC phase, the instrument was in an excellent condition, ready to provide mission (normal mode) data for ICV Cal/Val activities.

[26] Following the EOC is the ICV process, which is still ongoing at the time of this writing (June 2013). The main objectives of ICV activities are to improve and validate the CrIS SDR products. Activities include SDR algorithm and software improvement and validation, calibration parameter refinement, and characterizations of instrument performance and radiometric, spectral, and geolocation uncertainties. During the ICV phase, the SDR product is validated and released to the public users at three maturity levels, marked with Beta, Provisional, and Validated, respectively, The Beta product is an early release product. It is calibrated but minimally validated and may still contain significant errors. The Provisional product is an improvement over the Beta product, but it may not be optimal and incremental improvements are still occurring. At the Validated maturity level, the SDR product is well calibrated and validated and uncertainties are characterized over a range of representative conditions.

[27] Table 4 lists important developments so far during ICV. Below we will briefly describe their significances in improving SDR product quality. The web-based CrIS SDR LTM system was established during EOC and opened to public on 1 March 2012 [http://www.star.nesdis.noaa.gov/icvs/NPP/ipm_telemetry_npp_cris.php], as part of the large Integrated Calibration and Validation System which covers many operational weather satellite sensors. Over 80 CrIS RDR and SDR parameters and quality flags have been trended, monitored, and displayed on a continuous basis. It has since provided critical information on instrument health and RDR/SDR data quality status. On 2 April 2012, IDPS began to produce valid CrIS SDR product after two critical SDR software problems were fixed. The software problems prevented IDPS from producing valid SDRs, although the RDRs were in valid status. On 11 April 2012, the NL coefficients and ILS parameters for radiometric and spectral calibrations were updated with EP version 33. The update was the result of the efforts to refine the calibration coefficients and parameters with on-orbit normal and special mode data [Tobin et al., 2013a; Strow et al., submitted manuscript, 2013]. This update significantly improved the spectral and radiometric calibration absolute accuracies and relative agreement among the nine FOVs. Figure 7a shows the brightness temperature (BT) differences for three selected channels between the CrIS observations under cloud-free conditions on 15 May 2012 (processed by ADL with the use of EP version 33) and the simulations that were calculated using the Community Radiative Transfer Model (CRTM) [Han et al., 2006] with collocated European Center for Medium range Weather Forecasting (ECMWF) 3 h forecast/analysis fields as CRTM inputs. The data set in Figure 7a does not include pixels over ice and snow surfaces due to large errors in modeling ice and snow surface emission and reflection. The mean and root mean square BT differences between the observations and simulations for pixels over ocean are −0.2 K and 0.5 K for the 900 cm−1 window channel, 0.13 K and 1.5 K for the 1598.75 cm−1 water vapor channel, and −0.005 K and 1.3 K for the 2520 cm−1 window channel. To show the effects of the NL coefficient and ILS parameter update for the LWIR and MWIR bands, the mean differences of the observed and simulated spectra for each FOV are plotted in Figure 7b. The SWIR band result is not included in the figure since the SWIR detectors have little NL, and ILS parameters remain unchanged. The black curves in Figure 7b are the differences between the observed and simulated spectra when the original NL coefficients and ILS parameters in EP version 32 are used. The blue curves are the differences after the NL coefficients are updated, while the red curves are those after both NL coefficients and ILS parameters are updated. In the comparisons, a channel-dependent bias averaged over the nine FOVs between the CrIS observations and model simulations was removed from the radiance data, similar to the radiance bias corrections commonly applied in the satellite data assimilation systems [Derber and Wu, 1998]. As Figure 7b shows, the updates of the NL coefficient and ILS parameters significantly improved the uniformity of FOV-to-FOV radiometric and spectral performances. Our goal is to achieve an agreement among the nine FOVs in spectral and radiometric performance, which allows the NWP models to assimilate data from all of the FOVs without special treatment for different FOVs.

Table 4. Important ICV Developments
DateDescription
1 March 2012CrIS SDR LTM system open to public
2 April 2012IDPS first SDR product after major software error fixes
11 April 2012NL coefficients and ILS parameters updates
18 April 2012CrIS on-orbit FIR digital filter update
19 April 2012SDR product achieving Beta maturity status
27 June 2012First temperature drift limit updates
15 October 2012Geolocation error correction and imaginary QC algorithm in operation
25 October 2012Second temperature drift limit updates
31 January 2013SDR product achieving Provisional maturity status
Figure 7.

Cloud-free BT differences between CrIS observations and CRTM simulations on 15 May 2012: (a) BT difference images in the three channels of (top) 900 cm−1, (middle)1598.75 cm−1, and (bottom) 2520 cm−1, respectively, and (b) BT differences as a function of channel wave number in the LWIR and MWIR bands. In Figure 7b, a bias common to all the FOVs is removed. The black curves in Figure 7b are from the SDR data produced with the NL coefficients and ILS parameters in EP version 32. The blue curves are from the data after the NL coefficients are updated but with the ILS parameters unchanged. The red curves are from the data after both of the NL coefficients and ILS parameters are updated. The sizes of the data sets in the comparisons shown in Figure 7b for each FOV are over 30,000. Both of the observed and simulated spectra are apodized with the Hamming apodization function.

[28] On 18 April 2012, a new FIR filter was uploaded to CrIS to replace the corrupted one, which was discovered during the analysis of both normal and diagnostic mode data. The replacement eliminated the radiance biases (up to 0.2 K), which had a dependency on the interferometer mirror sweep direction. The SDR product achieved Beta status on 19 April 2012.

[29] On 27 June and 25 October of 2012, the values of the temperature drift limits used in the SDR QC algorithms were updated with EP versions 34 and 35, respectively. The SDR QC algorithms use these temperature drift limits to detect excess temperature drifts of the monitored objects such as ICT and scan baffle. The updates were to correct the prelaunch errors that set the temperature drift limits to values 2 or 4 times smaller than the required ones, resulting in over 60% of the valid spectra flagged with a “degraded” status. On 15 October 2012, a software bug in the geolocation calculation module was removed. The fix corrected geolocation errors that ranged from 4.5 km at nadir to 12 km at the edge of the scan positions. On the same day, the imaginary QC algorithm (described in the previous section) became operational. The purpose of this QC algorithm implementation is to identify corrupted spectra caused by software and SDR processing errors before the errors can be fixed. The SDR product achieved Provisional status on 31 January 2013. The product is expected to reach Validated status in early 2014.

5 CrIS SDR Record Data Quality

[30] In this section we describe the quality of the CrIS SDR record data produced by IDPS from three aspects. First, we provide an overall view of the CrIS noise performance and calibration uncertainties. Second, we examine cases of CrIS measurements taken from “hot” and “cold” scenes, including deserts, sun-glint areas, and deep convective clouds (DCCs). Third, we provide an assessment of the radiance spectra that are flagged with an invalid or degraded status. Results of the assessment will be useful not only for CrIS SDR users but also for those planning to reprocess CrIS data in the future. The IDPS processed CrIS SDRs as well as RDRs are available at NOAA's Comprehensive Large Array-data Stewardship System.

5.1 CrIS Radiance Noise Level and SDR Uncertainties

[31] Characterizations of on-orbit radiance noise and radiometric, spectral, and geolocation uncertainties have been among the major activities during ICV. The results obtained so far are described in detail in Zavyalov et al. [2013] for CrIS noise performance and in Tobin et al. [2013a, 2013b], Strow et al. (submitted manuscript, 2013) and Wang et al. [2013] for radiometric, spectral, and geolocation uncertainties, respectively. The overall radiance noise levels (NEdNs) are 0.0982, 0.0359, and 0.0033 in milliwatts per square meter per steradian per inverse centimeter (mW/m2/sr/cm−1) for the LWIR, MWIR, and SWIR bands, respectively. These numbers were derived from the ES spectra using the principal component analysis technique [Zavyalov et al., 2013]. The radiance noise levels were also estimated from ICT or DS spectra with the method described in section 3 and Zavyalov et al. [2013]. The results are similar to those derived from the ES spectra. Note that the SDR product includes the NEdN data derived from ICT spectra. It can be seen by comparing the NEdNs with the specifications in Table 2 that the CrIS radiance noise levels are significantly lower than the requirements. The NEdNs have been very stable since monitoring began at the end of January 2012; see, for example, Zavyalov et al. [2013, Figure 7].

[32] The LWIR and MWIR band spectral (channel frequency) calibration uncertainties were estimated by measuring the correlation between the observed spectra and simulated spectra that are offset in frequency until the maximum correlation is found (Strow et al., submitted manuscript, 2013). The absolute spectral uncertainties of the two bands range from 2 to 3 ppm, significantly lower than the specifications of 10 ppm given in Table 2. The effect of a 2 ppm frequency shift on radiance may be evaluated through simulations with a radiative transfer model. The maximum radiance uncertainty due to 2 ppm spectral uncertainty is estimated less than 0.12 K for unapodized spectra and below 0.05 K for spectra smoothed with a Hamming apodization function. The variation of the spectral difference between the observed and simulated spectra over the past 12 months was less than 2 ppm, where the pattern and magnitude of this 2 ppm variation are detected by the neon calibration system but generally not used in the SDR processing, which only updates the SDR frequency calibration matrix, CMO, when the neon measurement of the metrology laser changes by more than 2 ppm. The metrology laser wavelength varies by about 2 ppm peak to peak with an apparent cycle of a year. The SDR frequencies are accurate to better than 1 ppm when averaged over 1 year. For the SWIR band, due to the coarse spectral resolution, it is difficult to evaluate the absolute spectral uncertainty using the method described above. The maximum relative spectral calibration difference among the nine FOVs for each band is 0.4 ppm, 0.6 ppm, and 1.2 ppm for the LWIR, MWIR, and SWIR bands, respectively (Strow et al., submitted manuscript, 2013).

[33] Requirements on the CrIS radiometric uncertainty (RU) are listed in Table 2. Converted to 3 sigma (99% confidence interval) brightness temperature uncertainties [see Tobin et al., 2013a, Figure 1], the RU requirements are frequency dependent with values greater than 0.7 K, 0.6 K, and 0.5 K for the LWIR, MWIR, and SWIR bands, respectively. Estimates of the on-orbit CrIS RU and its dependence on Earth view scene are described in Tobin et al. [2013a]. The RU is estimated for each spectral channel as the root sum square of the individual contributors using a perturbation method applied to the CrIS radiometric calibration algorithm. The values of the algorithm parameters and their uncertainties used in the perturbation process are estimated from preflight thermal/vacuum testing data and on-orbit measurements. The RU estimates are performed for various Earth scenes, including warm (clear sky) and cold (high thick cloud) Earth scenes. The overall 3 sigma RU estimates are less than 0.3 K in the LWIR band and less than 0.2 K in the MWIR and SWIR bands. Assessments of radiometric calibration accuracy are underway through comparisons of CrIS observed radiances with simulations and radiance measurements from other satellite sensors, including IASI, AIRS, and the S-NPP Visible Infrared Image Radiometer Suite (VIIRS). Results obtained so far are in general consistent with the RU estimates [Wang et al., 2012; Tobin et al., 2013a; Tobin et al., 2013b].

[34] The geolocation uncertainty was estimated by comparing CrIS observations with measurements from VIIRS IR image band I5 (11.4 µm) over highly inhomogeneous scenes such as DCCs over ocean [Wang et al., 2013], taking advantage of the high spatial resolution (375 m) and accurate geolocation of the VIIRS IR band. The method works well for scan positions with a zenith angle less than 30° but is difficult to apply for scan positions with larger zenith angles due to the bowtie deletion of VIIRS pixels. For the nadir FOVs, the 1 sigma geolocation uncertainty of the spectrum is determined to be better than 0.4 km. For off-nadir FOVs with zenith angles less than 30°, the estimated uncertainty is below 1.3 km. Efforts are underway to estimate the geolocation accuracy for FOVs with zenith angles larger than 30° and fine tune the geolocation mapping parameters for better geolocation accuracy.

5.2 Measurements Over Desert, Sun-Glint, and DCC Scenes

[35] In this subsection, we examine typical cases of measurements taken from deserts, sun-glint areas, and DCCs to see how CrIS perform over these “hot” and “cold” scenes. Figure 8 shows comparisons of CrIS radiances with VIIRS LWIR and SWIR window channels M16 (12.01 µm) and M13 (4.05 µm) over the Australian deserts. To make the comparisons, the CrIS spectral radiances are spectrally integrated over the VIIRS channel SRFs, and the VIIRS radiances are spatially averaged over the CrIS footprints using the method described in Wang et al. [2013]. No data points were excluded in the plots. It can be been seen that the two sensors at both the LWIR and SWIR channels are in good agreement over the full range of the brightness temperature, which reaches a maximum value of 332 K in this case. The spread of the data points around the mean differences in Figure 8 is mainly due to nonuniform scenes and imperfect collocations of the two sensors.

Figure 8.

Radiance comparisons between CrIS and VIIRS channels (a) M16 and (b) M13 over the Australian deserts on 29 January 2013. The CrIS data in the figure are spectrally averaged over the VIIRS spectral response functions, and VIIRS data are spatially averaged on the CrIS FOV footprints. Data are plotted as (left) CrIS versus VIIRS and (right) CrIS-VIIRS versus VIIRS.

[36] For the SWIR band, sun glint may be encountered when sunlight is specularly reflected by the ocean surface into a CrIS FOV. Figure 9a shows an image of the 2457.5 cm−1 CrIS window channel. The sun-glint-affected pixels are roughly those with BT greater than 300 K over ocean, and there are a total of 3023 such pixels in this example. Among these affected pixels, there are 51 invalid measurements due to saturation of the AD converter near the interferogram ZPD. The rest of the pixels are valid measurements. Figure 9b shows an example of a valid spectrum located in this sun-glint area, whose window channel BT reaches 375 K. The figure also shows a simulated spectrum (dotted curve), calculated with CRTM and colocated NWP profiles, sea surface temperature, and near-surface wind vector. In the CRTM model, a nonlocal thermodynamic equilibrium correction scheme for the 4.3 µm CO2 absorption band and a bidirectional reflectance distribution function for sunlight reflection were recently implemented [Chen et al., 2013b]. As the figure shows, the spectral shapes and the radiance values in the opaque channels agree well between the two spectra. The large radiance differences in the window regions are likely due to the CRTM's sea surface reflectivity model which underestimates the surface reflectivity.

Figure 9.

(a) CrIS radiance image of the window channel 2457.5 cm−1 over an ocean area South of Sri Lanka with some pixels affected by sun glint on 14 March 2013 and (b) a valid SWIR spectrum in solid curve. In Figure 9b, the dashed curve is a simulated spectrum using CRTM and collocated NWP predicted atmospheric profiles and surface variables.

[37] For measurements taken over “cold” scenes, we chose a case of DCCs and compared CrIS radiances with VIIRS channel M13 and M16 measurements. Figure 10a shows an image of the CrIS window radiance at 2457.5 cm−1. The DCCs are at the locations of the pixels with BT below 220 K. The radiance comparisons are shown in Figures 10b and 10c. It can be seen that for the LWIR channel (M16), the good agreement at the low radiance regions indicates that the CrIS LWIR window channels and VIIRS channel M16 have good sensitivity at low radiance. For the SWIR window channel, there is a good agreement between the two sensors above BT 200 K, although the data are noisier at the low radiance. Below 200 K, the VIIRS channel has no sensitivity.

Figure 10.

(a) CrIS radiance image at the window channel of 2457.5 cm−1 over an area with DCCs near Solomon Islands on 5 February 2013 and radiance comparisons between CrIS and VIIRS channels (b) M16 and (c) M13 with data collected over the same area. The CrIS data in the figure are spectrally averaged over the VIIRS spectral response functions, and VIIRS data are spatially averaged on the CrIS FOV footprints. Data in Figures 10b and 10c are plotted as (left) CrIS versus VIIRS and (right) CrIS-VIIRS versus VIIRS.

5.3 Invalid and Degraded SDRs

[38] The IDPS system produces about 8.7 million Earth view CrIS spectra each day on a global scale. Among them, only a very small number of the spectra are reported as invalid or degraded. In this subsection we will give an assessment of these invalid or degraded SDR data collected over the past 12 months. To provide data quality status for each of the spectra, CrIS SDR processing software includes QC algorithms to evaluate the quality of the spectrum and reports the data quality information in various Quality Flags (QFs). Among the QFs, the overall SDR QF may be the most useful one for CrIS SDR users. It is a three-level status flag, named as QF3, with a value of 0, 1, or 2 assigned to each spectrum, representing good, degraded, or invalid quality status. The various conditions employed in the SDR QC algorithms to set the QF to invalid or degraded status are documented in JPSS Configuration Management Office [2012] and the CrIS SDR User Guide currently in preparation. A spectrum with an invalid status is erroneous and should not be used for applications, while a spectrum with a degraded status may have quality issues and should be used with caution by checking other QFs. Examples of the degraded spectra include cases of invalid geolocation data or when the actual size of the DS or ICT moving window average used for calibration drops below half of the full window size.

[39] The sources of the invalid spectra may be categorized into the three groups: invalid instrument measurements, missing RDR packets, and SDR processing errors. Invalid SDRs caused by invalid measurements are usually not recoverable, while those due to missing packets and processing errors are usually recoverable in reprocessing using the SDR software with a version after July 2013, when we believe most of the software issues will have been resolved.

[40] Over the past 12 months, the occurrences of invalid instrument measurements were extremely low and were mainly due to the signal saturation of the SWIR band in the strong sun-glint areas over oceans as discussed in the previous subsection. The number of the invalid SW spectra due to measurement saturation is about 0.0001% of the total number of the SWIR spectra. In an area affected by the sun glint, about 2% of the SWIR spectra are invalid. So far, no FCE event has occurred, and no ice contamination on the detectors has been observed from the NEdN time series in the spectral region of 650 to 700 cm−1, which contains the spectral feature of ice absorption. There have been no invalid LWIR and MWIR spectra due to invalid measurements.

[41] Missing data packets and SDR processing errors have been major sources of the invalid spectra during the ICV thus far. For various reasons, including end of communication contacts between the ground receiving station and the spacecraft, sensor and spacecraft mode changes, and simple loss of data, an RDR granule may contain only a subset of the required science data packets when it is processed. For the CrIS SDR process, a missing packet with a potential impact on an ES spectrum can be any of the interferogram, SciCalP, and EP packets within the 30-scan moving window. For example, missing ES interferogram packets are a direct source of invalid spectra. For such cases, the preprocessing creates placeholder spectra with the Fillvalue (–999.8). However, missing packets might also introduce corrupted spectra through SDR processing software, which failed to appropriately handle certain cases of missing packets, not anticipated at the time when the software was developed. In addition, corrupted SDRs occurred in some cases when IDPS mishandled the repair of corrupted granules. For each incomplete granule, a repaired granule was usually issued at a later time and tasked for reprocessing if certain conditions were met. Most of the reprocessing of repaired granules was successful. However, some of the reprocessed granules resulted in corrupted spectra and occasional SDR anomalies in which a number of consecutive SDR granules were corrupted. The root causes were difficult to analyze because the processes in creating the corrupted IDPS data were not repeatable. Figure 11 shows a corrupted spectrum and its imaginary part of the spectrum, caused by a combination of a software error and mishandling of the repaired granules. The corrupted spectrum has a large imaginary radiance spectrum. The spectrum (shown by grey curves) was recovered by reprocessing the RDR data with ADL after the error sources were removed.

Figure 11.

An example of a corrupted SDR LWIR complex spectrum (dark curves) with (a) a distorted real part of the radiance spectrum and (b) a large imaginary part of the spectrum. The distortion of the spectrum is caused by a processing error. The spectrum in grey color is the correct spectrum produced by the ADL with the processing error removed.

[42] The number of invalid spectra such as those described above has been reduced over time with a decrease in the number of incomplete RDR granules, improvements of the CrIS SDR software, and improved IDPS handling of missing packets and repaired granules. Figure 12 shows the time series of the number of repaired RDR granules per day, which is roughly equal to number of incomplete RDR granules. Figure 13 shows the daily percentage of invalid LWIR band spectra relative to the daily total number of LWIR band spectra over the period from April 2012 to April 2013. The daily percentage of invalid spectra in general is less than 0.1% over the 12 month period. It can be seen that there is a correlation between the two time series shown in Figures 12 and 13. After July 2012, the number of repaired granules was significantly reduced, to about 27 per day, and so was the number of invalid spectra. For the MWIR and SWIR bands, the situations are similar. In the Provisional product, during the first 3 months of 2013, the daily percentage of invalid spectra on average is about 0.0011% and is about 0.15% in a few SDR anomaly cases.

Figure 12.

Number of daily repaired CrIS science RDR granules from April 2012 to April 2013.

Figure 13.

Daily percentage of SDR LWIR band spectra with an invalid quality status from April 2012 to April 2013.

[43] The spectra flagged with a degraded status encountered different issues. Most of them are actually valid spectra and were mistakenly given the degraded status by the QC algorithms that used incorrect temperature drift limits, an issue discussed in section 4. These QC algorithm errors were corrected by the two EP updates on 27 June and 25 October of 2012. The effects of the corrections can be seen in Figure 14, which shows a time series of the daily percentage of the spectra flagged as degraded. The daily percentage of “degraded” spectra was over 60% before the first correction and reduced to about 0.014% on average after the second correction. The cause of the remaining “degraded” spectra is under investigation.

Figure 14.

Daily percentage of SDR LWIR band spectra with a degraded quality status from April 2012 to April 2013. The high percentage of spectra labeled as degraded before 25 October 2012 is due to errors in the QC algorithm. These data are actually valid spectra.

[44] The geolocation data before 15 October 2012 were actually degraded by the geolocation errors described in section 4. Unfortunately, the degradation status of these data was not included in the SDR data.

[45] Our experiments showed that with the exception of the invalid SWIR spectra due to the measurement errors in sun-glint areas, about 99% of the invalid and degraded spectra can be recovered by reprocessing the RDRs after all the repaired granules are collected. Alternatively, most of the cases in which SDR data are corrupted by data and processing errors can be avoided when a latency period of 24 h is applied to the incoming RDRs [DeSlover et al., 2013].

5.4 Two Remaining Issues

[46] Two radiometric and spectral calibration issues have been identified and are currently under investigation. One issue is that SDR spectra (unapodized) contain spectral ringing artifacts, which takes the form of every other spectral point jumps up and down throughout the spectra but are significant in only some narrow spectral regions. For the LWIR band, the largest ringing magnitude occurs at the long-wave end of the spectral band and can reach maximum amplitude of 0.6 K in worst cases. The ringing magnitude is reduced to a level below 0.1 K when a Hamming apodization is applied to the spectra. Our investigation effort is to understand the source of the ringing and seek a solution to reduce the ringing artifacts for the original spectra. The second issue is that there are significant radiance differences between CrIS LWIR band channels and VIIRS channels M15 [Tobin et al., 2013b]. The differences become large in cold regions and reach a value about 0.25 K at a scene BT of 210 K, relative to the differences seen in the warm regions.

6 Summary

[47] Since the CrIS instrument was powered up, a well-planned series of Cal/Val activities has been carried out with the goal of providing a well-calibrated and characterized SDR product. The SDR product reached Beta maturity status on 19 April 2012 and Provisional status on 31 January 2013. We are confident that after completing the remaining ICV work, the SDR quality and calibration accuracy will be further improved for the Validated SDR product.

[48] The CrIS has been providing science RDR data since February 2012. Due to a data corruption error in the onboard FIR filter, the RDR data before 18 April 2012 are slightly degraded, resulting in a radiance error of up to 0.2 K. This error was eliminated with a digital FIR filter update to the CrIS on-orbit signal processor. The instrument performance has been stable with radiance noise (NEdN) levels basically unchanged from instrument commissioning up to the preparation of this manuscript (June 2013). So far, no FCE events have occurred and no ice contamination on the detectors has been observed from our monitoring systems. The SWIR band measurements may saturate over ocean scenes with strong direct sun glint into the CrIS LOS. Both the RDRs and the corresponding SDRs are appropriately labeled as invalid when sun-glint saturation occurs.

[49] The ground CrIS SDR processing software performs radiometric and spectral calibrations and geolocation calculations with tunable calibration and mapping parameters. The update of the NL coefficients and ILS parameters on 11 April 2012 significantly improved the radiometric and spectral accuracies and the uniformity of the radiometric performance among the nine FOVs. For both of the Beta and Provisional products, the LWIR and MWIR band spectral uncertainties are better than 3 ppm, and the 3 sigma radiometric uncertainties are better than 0.3 K for the LWIR band channels and 0.2 K for the MWIR and SWIR band channels. The geolocation uncertainty is better than 1.3 km (1 sigma) for FOVs with zenith angles less than 30°.

[50] The IDPS has been producing operational CrIS SDRs since 2 April 2012. The success rate to produce valid SDRs in general is better than 99.9%. The degraded status of the SDR data before 25 October 2012 is likely a false alarm, caused by errors in the SDR QC algorithms. The number of degraded SDRs after the correction of the algorithm errors is around 0.014%. Following the FIR filter update on 18 April 2012, the majority of the invalid and degraded SDRs have been due to missing RDR packets, software errors, and mishandling incomplete or repaired RDR granules by the IDPS processing system. These SDR errors can be removed through reprocessing with a post-July 2013 version of the SDR software. We expect the number of invalid and degraded SDRs to be reduced further for the Validated product with the additional improvement of the IDPS processing system and software.

Appendix A

Acronyms

[51] 

AD

analog-to-digital

ADL

Algorithm Development Library

AIRS

Atmospheric Infrared Sounder

ATMS

Advanced Technology Microwave Sounder

BPF

band-pass filter

BT

brightness temperature

Cal/Val

calibration and validation

CMO

Correction Matrix Operator

CrIMSS

Cross-Track Infrared/Microwave Sounding Suite

CRTM

Community Radiative Transfer Model

DA

dynamic alignment

DC

direct current

DCC

deep convective clouds

DF

decimation factor

DMS

Data Management Subsystem

DS

Deep Space

EOC

Early Orbit Checkout

EP

engineering packet

ES

Earth scene

FCE

Fringe Count Error

FFT

Fast Fourier Transform

FIR

finite impulse response

FOR

field of regard

FOV

field of view

FSR

full spectral resolution

FTS

Fourier transform spectrometer

ICT

Internal Calibration Target

ICV

Intensive Calibration and Validation

IASI

Infrared Atmospheric Sounding Interferometer

IDPS

Interface Data Processing Segment

IGM

interferogram

ILS

instrument line shape

JPSS

Joint Polar Satellite System

LOS

line-of-sight

LWIR

long-wave infrared

MWIR

midwave infrared

MPD

maximum path difference

NEdN

noise equivalent differential radiance

NL

nonlinearity

OPD

optical path difference

ppm

parts per million

QC

quality control

QF

Quality Flag

RDR

Raw Data Record

SciCalP

science/calibration packet

SDR

Sensor Data Record

SRF

spectral response function

SSM

scene selection module

SWIR

short-wave infrared

VIIRS

Visible Infrared Image Radiometer Suite

ZPD

zero path difference

Acknowledgments

[52] The authors would like to thank the JPSS Program Office for the support of the CrIS calibration and validation (Cal/Val) activities. The authors are also grateful to Raytheon for the contribution to the CrIS SDR software development. 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.

Ancillary