Pre-Launch Day-Night Band Radiometric Performance of JPSS-3 and -4 VIIRS

Abstract

Following the success of Visible Infrared Imaging Radiometer Suite (VIIRS) instruments currently operating onboard the Suomi-NPP, NOAA-20, and NOAA-21 spacecraft, preparations are underway for the final two VIIRS instruments for the Joint Polar Satellite System 3 (JPSS-3) and 4 (JPSS-4) platforms. To that end, each instrument underwent a comprehensive sensor-level test campaign at the Raytheon Technologies, El Segundo facility, in both ambient and thermal-vacuum environments. Unique among the 22 VIIRS sensing bands is the day-night band (DNB)—a panchromatic imager that leverages multiple CCD detectors set at different gain levels to make continuous (day and night) radiometric observations of the Earth. The results from the JPSS-3 and JPSS-4 VIIRS DNB pre-launch testing are presented and compared against the design specifications in this paper. Characterization parameters include dark offset, gain, linearity, uniformity, SNR, and uncertainty. Performance relative to past builds is also included where appropriate.

1. Introduction

Three Visible Infrared Imaging Radiometer (VIIRS) instruments [1] are currently in operation, making continuous measurements of the Earth aboard each of their host spacecraft: Suomi-National Polar-Orbiting Partnership (S-NPP) [2,3,4,5], NOAA-20 [6], and NOAA-21 [7,8]. Each VIIRS instrument underwent extensive ground testing before launch to ensure proper operation and performance [9,10,11]. Similar test campaigns were performed and recently completed for the fourth and fifth VIIRS instruments, designated JPSS-3 (J3) [12] and JPSS-4 (J4) VIIRS [13], respectively.

VIIRS is a scanning radiometer, measuring earth scenes using twenty-two spectral bands spanning wavelengths from approximately 0.4 µm to out past 12 µm. Unique among those is the day-night band (DNB), which is in fact four separate imaging CCDs, with the output of each stage selectively chosen or combined to provide continuous Earth scene measurements over a dynamic range that spans seven orders of magnitude. Key characteristics and specifications for VIIRS DNB are presented in

Table 1. Key characteristics and specifications for VIIRS DNB.

Gain Stage	Lmin (W·cm−2·sr−1)	Lmax (W·cm−2·sr−1)	Subpixels Scan	Subpixels Track	Resolution (m)
LGS	N/A	0.02	1	672	742
MGS	N/A	N/A	3	672	742
HGS	3.00 × 10−9	N/A	250	672	742

. Each DNB CCD is an array with 672 pixels along-track, which are aggregated within the instrument processor to yield sixteen effective detectors. These can be treated much like the sixteen along-track detectors in each of the VIIRS mid-resolution bands. For the low-gain stage (LGS), this array is only a single pixel wide (along-scan), whereas for the mid-gain stage (MGS), it is three pixels wide. Both high-gain CCDs are 250 pixels wide. The additional width allows for time-delay integration of multiple measurements for a single scene to reduce the influence of noise in lower radiance scenes that require higher gains. Each high-gain CCD is an identical imager, referred to as high-gain stage A (HGA) and B (HGB) for ease of use. With such high sensitivity, a single high-gain CCD could easily be influenced by high-energy particle strikes. Dual high-gain stages allow for a direct comparison of each measurement. The VIIRS processor nominally takes an average of HGA and HGB outputs when reporting results. However, if the difference between those outputs exceeds a pre-defined limit, the lower value is reported as it is assumed that the higher value is anomalous. The processed result from each high-gain stage is referred to as the high-gain stage (HGS) measurement. While pre-launch testing can often separate the performance of HGA and HGB, once in nominal operation on-orbit, there is no indication which gain stage recorded each measurement. Because of this, depending on the metric, the performance of each gain stage may be addressed separately or as a combined HGS. A single focal plane assembly (FPA) houses the four DNB gain stages. To maintain the spatial resolution of the DNB detectors across a scan, multiple along-scan aggregation zones are defined along with the timing assigned to each zone. The timing applied to the aggregation is referred to as the aggregation modes. The aggregation modes and zones are symmetric about nadir, with the highest aggregation of pixel measurements occurring at the center of the scan (aggregation mode 1). Nominally, this yields an effective resolution of 742 m for each sample in both along-track and along-scan directions. After aggregation, a single DNB scan contains 4096 frames of data.

When deciding which stage to record as the DNB output for a given sample, the onboard algorithm refers to the gain transition values stored in the software ID tables. These table values are nominally set to trigger a transition to a lower gain stage whenever the raw measured signal (without background subtraction) exceeds 95% of the output range. Transitioning back to higher gain stages occurs when the measurement falls below 90% of the higher stage’s dynamic range. The previously reported stage will continue to be used unless a transition is required. The overlap of measurement ranges between gain stages caused by different transition points for neighboring stages provides consistency for measurements near the transition point of one stage, not requiring switching back and forth. Simultaneous measurements among neighboring gain stages also provide the opportunity to perform in situ cross-stage calibration. This paper focuses on the radiometric calibration of these instruments. Separate works report the geometric [14,15] and spectral performance [16,17] of these instruments.

1.1. On-Orbit Calibration

As it is one of the sensors of the VIIRS instrument, the DNB is a radiometer and, like other VIIRS bands, requires gain coefficients to convert the measured detector response into radiance measurements. The DNB radiance calculation is shown as Equation (1):

$$L=c_1(DN-{DN}_0)$$

(1)

which shows that the radiance (L) is determined by applying a gain coefficient (c₁) to the measured value (DN) after background subtraction (DN₀). We use DN (capital letters) to denote direct instrument detector output, whereas later, we will refer to dn (lowercase letters) as detector output after background subtraction to present measured signal levels. Calibration parameters are required for each DNB detector, aggregation mode, and gain stage. Additionally, separate parameters are provided for each of the VIIRS rotating half-angle mirror (HAM) sides.

The LGS gain coefficient is determined on-orbit using measurements of the VIIRS solar diffuser, whose degradation is monitored by a solar diffuser stability monitor. MGS and HGS gains are found by leveraging measurements within the cross-stage overlap either using scheduled calibrations, defined within VIIRS operating procedures (VROPs), or other calibration targets.

The DNB background is a more complicated matter. To account for each aggregation mode, a flat field offset is applied to each Earth scene sample as part of the onboard processing. This occurs before checking for a gain transition and before the high-gain stages perform their comparison to reject any potential high-energy particle strike. The flat field offset brings all sample measurements for each gain stage to a common target value. For JPSS-3 and -4, the targets are 350 DN for HGS, 200 DN for MGS, and 60 DN for LGS (in 14-bit DN). To retain gain-stage information, a bit-flag is applied to each measurement before being stored. DNB measurements are recorded in 14-bit DN, and HGS measurements retain all 14 bits of data, since the HGS is the most sensitive of the gain stages. LGS and MGS measurements sacrifice one bit of sensitivity for differentiating between themselves. The result is that LGS and MGS Earth scene measurements are reported in 13-bit DN versus 14-bit DN for HGS measurements. This is important when applying parameters derived as part of pre-launch testing to operational results. The onboard flat fielding is included within the DN of Equation (1). The background subtraction is accomplished by determining DN₀, also referred to as the dark offset. Ideally, the DN₀ would be the target value for each stage. In practice, the DN₀ is affected by various component temperatures and increases as the detectors age. Several methods exist to characterize the DN₀ on-orbit, including the trending of onboard calibrator measurements, additional monthly VROPs, and an early mission maneuver where the sensor is pitched to allow for the instrument to fully view dark space.

2. Pre-Launch Calibration Methods

The first step in the pre-launch DNB calibration is determining how much of the measured detector response is due to the scene and what is the background. With the DNB, this is doubly important as there are two levels of background correction that are applied to DNB measurements, the flat-fielding onboard offset correction and the dark offset correction applied as part of the ground processing.

The onboard offset correction is determined for each system electronics side at each temperature plateau during instrument TVAC environmental testing. Component temperatures are given sufficient time to reach stability before taking any measurements. The telescope position is fixed and staring at a laboratory blackbody source. The spectral sensing range of the DNB does not extend into the infrared region so the blackbody can serve as a stable target surrogate for deep space. DNB processing is set to report the output from each gain stage separately and to not perform any onboard flat fielding. Results from multiple scans are collated to provide an average response (with outliers removed). HAM-side measurements are combined as there is no HAM-side dependence within the onboard tables. The target value for each gain stage is subtracted from these measurements and is used as the onboard offset correction. While the result of this test yields non-integer values, they are rounded to whole numbers as required by the instrument software. Key in the determination of these onboard offset values is to avoid offset values that reduce the saturation level of any sample below its transition point.

With the onboard offset correction tables uploaded and applied, similar measurements are again performed while staring at the blackbody source. If generated and applied correctly, the average value for each aggregation mode and detector will be within five DN (for HGS) and two DN (for LGS and MGS) of the target value for each gain stage.

Nominally, gain transitions to lower gain stages will occur when the measured signal is greater than 95% of the higher gain stage dynamic range. Similarly, transitioning to a higher gain stage occurs when the measurement falls below 90% of its dynamic range. Assuming there is no pre-saturation, these values are simply 95% and 90% of the full 14-bit dynamic range (max DN = 16,383). With the gain transition values uploaded to the processing electronics and reporting in auto-gain, the SIS-100 (a 100 cm diameter spherical integration source with multiple halogen lamps) is placed at the center (nadir) of the VIIRS view. With the VIIRS telescope rotating, DNB measurements of the scene will begin by recording the dark laboratory environment in high gain. The auto-gain processing should switch to report MGS measurements as the telescope rotates towards the SIS-100 signal, eventually transitioning to LGS at the center of the scan. As the telescope rotates away from the SIS-100, the DNB measurements should transition to MGS and finally back to HGS. Analysis of these measurements should align with expectations for the set gain transition values.

The DNB gain coefficients are determined with the VIIRS telescope again in a fixed position staring at the SIS-100. A shutter is put in place between the SIS-100 output and VIIRS to provide regular background measurements throughout data collection. To capture the entire DNB dynamic range, twenty-four SIS-100 signal levels are measured. A combination of lamps is used, the brightest of which includes eighteen 200 W lamps, nine 45 W lamps, and ten 10.4 W lamps. Lower signal levels are produced using one 8 W lamp and/or one 0.5 W lamp driven at various de-rated current levels. These levels span the entire DNB specified measurement range (3 × 10⁻⁹ to 2 × 10⁻² W·cm⁻²·sr⁻¹), which encompasses all three gain settings. Values for expected measured radiance of each signal level are determined as part of the SIS-100 calibration, which occurs before TVAC testing. Additionally, in situ measurements of the SIS radiance are possible using a laboratory radiance monitor and/or Fiber Spectrometer at all but the lowest signal levels. These monitors measure radiance at multiple wavelengths, which, when combined with the DNB RSR, provide a more accurate measurement of what the DNB detectors are measuring and are used in place of the SIS-100 provided calibration values that were measured before the start of environmental testing, when available.

For each signal level, three data collections are performed: one primary measurement, another with an attenuator in place between VIIRS and the SIS-100, and, finally, one background collect. The gain determination uses the primary measurement data after subtracting background for each signal level, aggregation mode, gain stage, and detector. A fit of the measured response against the SIS-100 radiance provides the gain coefficients for each detector, aggregation mode, gain stage, and HAM side. These measurements are repeated throughout the test campaign for different electronics configurations at multiple environmental temperatures, with the measurement of primary electronics at the nominal temperature plateau being the primary source for gains used in the at-launch Lookup Tables (LUTs). Once operating on-orbit, the at-launch calibration parameters will be replaced with those derived according to the on-orbit calibration algorithm (as summarized in Section 1.1, above).

The data collected as part of gain determination testing are used for quantitative assessments of the gain linearity, SNR, and uniformity. Measurements of the attenuated SIS-100 signal are not included when determining the gain but can be used to support assessments of non-linearity in DNB detectors.

After three VIIRS instruments were successfully built, characterized, and launched, the program had used all the available DNB CCDs and required a new batch to be produced. The DNB CCDs used for JPSS-3 and -4 VIIRS are sourced from this new batch [12]. During JPSS-4 ambient testing in the fall of 2021, the DNB MGS showed non-responsive behavior and was subsequently replaced with a spare unit, and ambient characterization tests were performed in early 2022 for the new DNB unit. No issues were found with the new FPA, and no further changes to the DNB hardware were required. All JPSS-4 DNB results included in this work are using the current (functional) DNB FPA.

3. Pre-Launch Calibration Results

Instrument TVAC testing took place in El Segundo, California in the fall of 2020 and 2023 for JPSS-3 [18] and -4 [19] VIIRS, respectively. Testing for both instruments followed the same baseline plan, which was based on the test plan from previous VIIRS instruments.

3.1. Dark Current Noise

Testing for both instruments yielded applicable onboard offset and initial ground offset tables for most cases. However, JPSS-3 and JPSS-4 HGB detectors for aggregation modes 30–32 had flat-field measurements above 1169 DN, which would necessitate onboard offsets above 819 DN (after accounting for the HGB target value of 350 DN). This level of flat fielding would bring the saturation point below the transition level for these aggregation modes and prevent the DNB logic from transitioning to use a lower gain stage. Figure 1 presents the measured results of electronic side A from the initial flat field measurement before removing the target values for each gain stage of JPSS-3 VIIRS DNB. Comparable results for JPSS-4 are shown in Figure 2. In both figures, for HGB, the transition limit line (shown in dotted blue) is exceeded for detectors at the beginning and end of the scan. Electronics side B measurements also show similar behavior in both instruments. As a result, the HGS to MGS transition level was reduced from 95% of the dynamic range to 93% of the dynamic range for both builds. The reduction provided sufficient margin to use the derived onboard offsets. Measurements of the stable laboratory blackbody source with onboard offsets applied confirm the accuracy of the flat field for each sample of each gain stage within ±2 DN of the target value.

3.2. Gain Transition

While simultaneous measurements are performed for all DNB CCDs, the output from only one gain level is reported in nominal operations. To determine the proper gain stage to output, a stage selection algorithm is used by the onboard software. The current gain stage output continues to be reported unless the measurement crosses a gain transition point. Gain transition points are nominally set at 5% and 95% of the stage dynamic range (in digital number after onboard offset subtraction). As mentioned earlier, the HGS gain transition point is set at 93% for both JPSS-3 and -4 VIIRS DNB since their required onboard offsets are higher than typical. Key to note is that a gain transition depends on the measurement made by the active stage. To measure proper adjustments between reported gain stages, measurements are performed in ground testing with the SIS-100 near-nadir, the DNB set to auto-gain, and the telescope rotating. Figure 3 presents the measured results from this scanning test. Following the data trend from left to right, high-gain data are reported for samples up to 108 before nadir, at which point the scattered light from the SIS-100 is bright enough to cause a transition to the mid-gain stage and continuing in mid-gain until, again, the signal becomes too bright for that stage, and the DNB transitions to the low-gain stage when viewing the SIS output directly. This behavior then repeats in reverse as the measured signal decreases and the stages shift to mid-gain and finally back to high gain at the end of the scan. Because the active stage is different on either side of the scan, the transitions pre- and post-nadir occur at different signal levels (DN). The results shown match expectations with the set gain transition values and show that the DNB can transition appropriately.

3.3. Gain Determination

The DNB detector gain is determined using a test that looks to measure the relationship between detector response and input radiance for twenty-four signal levels. This is a staring test, where the telescope is locked in a single position and the electronics cycle over each aggregation mode. This test is performed at multiple temperature plateaus. Due to the long test duration and tight testing schedules, it was not performed for all permutations of temperature plateaus and electronics sides for either JPSS-3 or JPSS-4 VIIRS. Linear fits of JPSS-4 DNB HGS measurements initially showed signs of non-linearity at low radiance levels for all aggregation modes when using the vendor provided SIS-100 radiance values. A subsequent investigation found that there was an issue when calibrating the lowest SIS-100 radiance values, which led to incorrect radiances. Empirical radiance values are used in place of those radiances. The influence of these changes will be accounted for when assessing the gain calibration uncertainty (see Section 4.1).

Figure 4 shows an example of the measured dn (measured signal after background subtraction) vs. radiance for a single DNB detector using multiple aggregation modes over the three gain levels. Only select aggregation modes are shown for clarity. The figure illustrates how the same radiance level can be used to characterize aggregation modes across multiple gain stages. The gains themselves are calculated using a linear fit between dn and radiance, assuming there is no offset term (fit line passes through the origin). This matches how the DNB gain is applied in the VIIRS DNB on-orbit calibration (no offset term in gain). This test determines the gain for each detector for every aggregation mode separately for each gain stage and HAM side. For each gain stage, all non-saturated points are used that fall within the stage radiance limits. As the measured signal levels will not align exactly with these limits, an additional point past the lower limit is included in the fitting, when possible.

With such a large dynamic range, the calibration uncertainty specification for the DNB is defined differently at four different radiance levels. It is 5% at half-maximum radiance (0.5 L_max), 10% at the LGS–MGS transition, 30% at the MGS–HGS transition, and 100% at the minimum specified radiance (L_min). This specification provides two uncertainty test levels applicable to each gain stage. Figure 5 and Figure 6 show the derived gain for all four gain stages of both VIIRS builds along with the fitting residuals of the data presented in Figure 4. The vertical lines denote the overall limits of the DNB dynamic range. While the specification is defined only at the limits for each gain stage, we have included sloped dashed lines to provide a visual approximation of how the specification changes with radiance level across each gain stage. The magnitude of the fitting residuals increases as expected with lower radiance levels.

3.4. Non-Linearity

Non-linear behavior in the DNB has been observed in previous VIIRS builds [10]. This is most common for the high-gain stages at low radiance but has also been found in the other stages. For NOAA-20 VIIRS, the most significant impact was for the HGS in aggregation modes 21–32, where less aggregation takes place. An operational decision was made to adjust the aggregation allocation for NOAA-20 DNB samples that nominally use aggregation modes 22–32 to all use aggregation mode 21. As this was one of the potential options to address non-linearity, it is colloquially referred to as Option-21. Option-21 reduced the non-linearity for measurements at the leading and trailing edge of each scan, with the tradeoff of reducing the spatial resolution at those samples.

Figure 7 shows the gain as the slope of a linear fit of the data points (left plot) and the residuals to that fit (right plot) for JPSS-3 DNB LGS detectors using aggregation mode 29. By crossing into any one of the four areas bound by red dashed lines, the residual for a detector would be seen to exceed the DNB radiometric uncertainty specification. This can be seen for detectors 7 (shown as cyan plus signs) and 10 (blue triangles) at low LGS radiance levels near the MGS transition. Similar non-linearity is observed from JPSS-4 DNB LGS detector 10 using aggregation mode 29 as shown in Figure 8.

Using the ratio of attenuated measurements to un-attenuated measurements provides an additional point of comparison when identifying non-linearity in the DNB detectors [20]. This ratio removes any dependency on the provided or calculated radiance values for each measurement. Regardless of the radiance, the ratio between attenuated and un-attenuated measurements should be constant (assuming the attenuator exhibits a linear behavior within the measured range of signal levels, wavelengths, etc.). Non-linear behavior is noted when ratioed measurements deviate from the expected attenuation factor. Figure 9 displays the ratio of attenuated to un-attenuated measurements for detector 10 of JPSS-3 (left plot) and JPSS-4 (right plot) DNB LGS for each aggregation mode. The results from this analysis confirm the non-linearity in aggregation mode 29 (shown as black squares in Figure 9). The results also show a spread in the measured attenuation ratios amongst other aggregation modes at low LGS radiances. Aggregation mode 14 (shown as cyan asterisks) and mode 32 (green asterisks) are out-of-family for measurements within the specification range for both instruments. This points to non-linearity for detector 10 when using these modes but not to the same degree as mode 29. Figure 10 provides a similar analysis for detector 10 from JPSS-3 and JPSS-4 VIIRS HGA stages using all aggregation modes. While the results for most aggregation modes show little change in measured attenuation across radiance levels, aggregation mode 29 and 32 measurements indicate non-linearity at lower radiance levels for JPSS-4 VIIRS that is not found in JPSS-3 VIIRS. The characterization of these non-linearities can provide insight into the detector’s behavior when evaluating each instrument’s on-orbit performance.

3.5. Uniformity

Another metric for tracking DNB performance is detector uniformity. For each SIS-100 signal level measured as part of gain determination testing, the calculated gain can be applied to derive the radiance measured by each detector. The standard deviation of these calculated radiances is used to monitor the uniformity among DNB detectors. DNB uniformity is characterized in two ways: detector uniformity within an aggregation mode (intra-aggregation uniformity) and detector uniformity between aggregation modes (inter-aggregation uniformity). The expectation for intra-aggregation uniformity is the derived radiance for all detectors using the same aggregation mode should agree to within the requirement, which is half the noise or 0.002 times the radiance, whichever is larger. Intra-aggregation non-uniformity would lead to horizontal striping in the calibrated data product when viewing a uniform scene. Figure 11 presents an example of the intra-aggregation uniformity performance of JPSS-3 and JPSS-4 VIIRS DNB HGA for aggregation mode 1. The vertical lines again denote the specified radiance range for the gain stage, and the red dotted line is the specification line, which dynamically changes with radiance level.

Vertical striping in a uniform scene may be a sign of inter-aggregation non-uniformity. Pre-launch characterization of inter-aggregation uniformity similarly uses the derived gains to produce measured radiance values for all detectors. In this evaluation, the radiance for each detector is compared with its measurement from the neighboring aggregation mode and measured against the specification. Figure 12 shows an example of inter-aggregation uniformity performance within specification for all 16 DNB detectors of JPSS-4 HGA for the boundary between aggregation modes 1–2 and modes 10–11. Figure 13 displays uniformity compliance matrices for both JPSS-3 and -4 VIIRS DNB determined as part of pre-launch thermal vacuum testing at the nominal performance plateau. Non-compliances to the specification are shown in red. Previous VIIRS builds also showed non-compliances in uniformity to a similar degree. Aggregation mode-dependent calibration coefficients have successfully mitigated non-uniformities among and between modes for the VIIRS instruments currently operating on-orbit. We expect the same to hold true for these instruments.

3.6. SNR

The signal-to-noise (SNR) is an important metric to characterize for the VIIRS DNB since each aggregation mode uses different sampling times. Design specifications require the SNR to be a minimum of six at L_min for samples within ±53° of nadir. For JPSS-3 and JPSS-4 VIIRS DNB, that range of samples is measured using aggregation modes 1 through 27. Samples outside of that range are expected to have an SNR of at least five at L_min (aggregation modes 28 to 32). The gain determination test collects fifty scans at each radiance level. A power fit of the SNR for the lowest radiance levels is evaluated at L_min and compared against specification for both HGA and HGB outputs separately for each detector and aggregation mode. An example from JPSS-3 and JPSS-4 VIIRS is shown in Figure 14. All detectors for both instruments were found to perform above the specified SNR requirement.

4. Discussion

4.1. Uncertainty

The analysis of the test results yields not only a characterization of the VIIRS DNB parameters comparable against the design specification but also provides insight into the uncertainty of those results. Compiling the uncertainties associated with each parameter can lead to a quantified uncertainty of the calibration. As each DNB gain stage is intended to measure over a different range of radiance levels, the designed radiometric calibration uncertainty of the DNB is defined differently for each gain stage as shown in

Table 2. Worst-case pre-launch calibration uncertainties (in %) for JPSS-3 and JPSS-4 VIIRS DNB using electronics side A. Each gain stage is evaluated at two radiance levels. Cells are shown in red highlight where the calculated uncertainty exceeds the specification.

	LGS Lmin		LGS Lmax		MGS Lmin		MGS Lmax		HGS Lmin		HGS Lmax
Instrument	J3	J4	J3	J4	J3	J4	J3	J4	J3	J4	J3	J4
Gain Cal.	12.4	6.7	12.4	6.7	15.6	9.7	13.0	9.7	12.5	8.9	13.4	9.8
Offset Cal.	0.8	2.8	0.5	0.5	0.8	1.8	0.5	1.8	9.5	12.6	0.5	0.5
Stability	5.1	4.7	0.7	0.9	0.3	0.2	0.5	0.2	0.4	52.9	0.6	3.9
Measurement	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.8	1.8	1.0	1.0
Total	13.5	8.7	12.5	6.9	15.7	10.0	13.1	9.8	15.8	55.1	13.5	10.6
Specification	10	10	5	5	30	30	10	10	100	100	30	30

. The uncertainties associated with the VIIRS DNB calibration are separated into four sub-categories based on their relation to the final calibration coefficients. All calculated uncertainties are given the same weight.

The uncertainty of the gain calculation is made up of several uncertainty factors but, in this case, is primarily driven by the sample-to-sample variation in non-linearity in cross-calibration between gain stages. DNB gain is a linear calibration (with no offset term), so any non-linear behavior induces uncertainty into the derived radiance product. The level of LGS cross-calibration uncertainty leads to the higher-than-specified levels of overall uncertainty for the LGS of both instruments shown in Table 2. The cause of JPSS-3 LGS exceeding specification at L_min is due to the cross-calibration uncertainty measured with the MGS at its L_max; we see that the JPSS-3 MGS at L_max also exceeds the specification. Once operating on-orbit, the gain calibration will be performed using the on-board solar diffuser instead of the coefficients derived as part of the pre-launch testing. We expect the on-orbit gain calculation to meet specifications for uncertainty.

The largest contributor of the offset calibration uncertainty is the variation in derived offset values due to temperature changes in the test environment. While the DNB electronics are temperature controlled, there are differences of up to 5 DN for some samples/detectors between the cold (−5 °C) and nominal thermal vacuum temperature environments. The largest differences were measured using the high-gain stages due to their higher degree of sensitivity. Table 2 shows the most significant uncertainty from the offset calibration category for the HGS of both instruments at L_min.

The stability category of uncertainty quantifies the variation in samples across the gain stage. This can be seen as a combination of the inter- and intra-aggregation uniformities. The adjustments to the radiance levels used for JPSS-4 HGS gain determination to account for incorrectly derived SIS-100 radiances, described earlier, are incorporated here and are the major source of the large uncertainty associated with JPSS-4 HGS at L_min. The final category includes all remaining sources of error, mostly related to measurement uncertainty, including uncertainties associated with the DNB spectral response and those due to optical non-uniformity (response versus scan angle—RVS). It also includes a factor for stray light, which is only considered at the lowest radiance levels for each HGS.

4.2. Stray Light

The VIIRS instruments onboard S-NPP [21,22,23,24,25], NOAA-20 [26], and NOAA-21 [27] have seen evidence of sunlight contaminating Earth observations, causing an unwanted increase in the measured radiance and a ‘washing out’ of the scene. Stray light is most significant when light enters the instrument via the solar diffuser screen on the leading face of the spacecraft as the spacecraft moves from night to day and again through the nadir port after the spacecraft moves from day to night. On-orbit observations have shown the impact of stray light on DNB images for NOAA-20 VIIRS DNB to be less significant than is seen for S-NPP VIIRS. Additional baffling was included in the design for NOAA-21 VIIRS (called JPSS-2 during testing) along with planned efforts to identify, measure, and patch remaining stray light paths. The result was an even further reduction in stray light seen in NOAA-21 VIIRS DNB images. The improvements developed for NOAA-21 VIIRS were implemented in the design and test of both JPSS-3 and JPSS-4 VIIRS. Since there is no design specification for DNB stray light performance, pre-launch testing for JPSS-3 and JPSS-4 VIIRS DNB stray light involved qualitative assessments, which verified the improvements were implemented successfully. Based on the improved performance observed in NOAA-21, we expect the stray light performance of JPSS-3 and JPSS-4 VIIRS DNB to be improved when compared against NOAA-20 VIIRS.

5. Conclusions

The instrument-level ground test campaign has been completed in the thermal vacuum environment for both JPSS-3 and JPSS-4 VIIRS. These tests provide comprehensive characterization for each DNB and yield at-launch calibration coefficient LUTs. Onboard dark offsets require an adjustment in the default gain transition tables, as was seen for JPSS-2 VIIRS DNB. The derived ground offset and gain coefficients meet expectations, with non-linearity only seen in one end-of-scan aggregation mode of the LGS at low radiance. While we do not propose any adjustment to the calibration coefficients due to the observed instances of non-linearity, the pre-launch characterization can provide insight on how to best address and mitigate the impact of these non-linearities in the case they impact on-orbit measurements. The assessment of each band’s uniformity showed aggregation modes with higher-than-specified responses but fewer than has been seen in past VIIRS builds. The SNR for both DNB was measured at levels better than specified for all detectors, gain stages, and aggregation modes. In general, the JPSS-3 and JPSS-4 VIIRS DNB are expected to perform at the same level, if not better, than previously characterized VIIRS DNB currently operating on-orbit.