Early Radiometric Assessment of NOAA-21 Visible Infrared Imaging Radiometer Suite Reflective Solar Bands Using Vicarious Techniques

Abstract

The VIIRS instrument on the JPSS-2 (renamed NOAA-21 upon reaching orbit) spacecraft was launched in November 2022, making it the third sensor in the VIIRS series, following those onboard the SNPP and NOAA-20 spacecrafts, which are operating nominally. As a multi-disciplinary instrument, the VIIRS provides the worldwide user community with high-quality imagery and radiometric measurements of the land, atmosphere, cryosphere, and oceans. This study provides an early assessment of the calibration stability and radiometric consistency of the NOAA-21 VIIRS RSBs using the latest NASA SIPS C2 L1B products. Vicarious approaches are employed, relying on reflectance data obtained from the Libya-4 desert and Dome C sites, deep convective clouds, and simultaneous nadir overpasses, as well as intercomparison with the first two VIIRS instruments using MODIS as a transfer radiometer. The impact of existing band spectral differences on sensor-to-sensor comparison is corrected using scene-specific a priori hyperspectral observations from the SCIAMACHY sensor onboard the ENVISAT platform. The results indicate that the overall radiometric performance of the newly launched NOAA-21 VIIRS is quantitatively comparable to the NOAA-20 for the VIS and NIR bands. For some SWIR bands, the measured reflectances are lower by more than 2%. An upward adjustment of 6.1% in the gain of band M11 (2.25 µm), based on lunar intercomparison results, generates more consistent results with the NOAA-20 VIIRS.

1. Introduction

The Visible Infrared Imaging Radiometer Suite (VIIRS) is a multi-disciplinary instrument currently flown on the Suomi National Polar-orbiting Partnership (SNPP) and Joint Polar Satellite System (JPSS) series of spacecraft [1]. With the recent launch of the JPSS-2 (renamed NOAA-21 upon reaching orbit) spacecraft in November 2022, it became the third satellite in the JPSS series to house a VIIRS instrument, joining the SNPP and the NOAA-20. The VIIRS, a successor to the MODIS instruments, features 22 spectral bands ranging from 0.4 to 12.2 µm (

Table 1. Summary of spectral characteristics (in µm) and primary purposes of the VIIRS band.

Band	Wavelength (µm)	Bandwidth (µm)	Gain Mode	Primary Use
I1	0.64	0.08	Single	Imagery
I2	0.865	0.039	Single	Imagery
I3	1.61	0.06	Single	Imagery
M1	0.412	0.02	Low/High	Ocean color, aerosol
M2	0.445	0.018	Low/High	Ocean color, aerosol
M3	0.488	0.02	Low/High	Ocean color, aerosol
M4	0.555	0.02	Low/High	Ocean color, aerosol
M5	0.672	0.02	Low/High	Ocean color, aerosol
M6	0.746	0.015	Single	Atmospheric correction
M7	0.865	0.039	Low/High	Ocean color, aerosol
M8	1.24	0.02	Single	Cloud particle size
M9	1.378	0.015	Single	Cirrus/Cloud cover
M10	1.61	0.06	Single	Snow fraction
M11	2.25	0.05	Single	Clouds

), supporting a wide range of scientific applications [2,3]. All three satellites are in the afternoon orbit, with the NOAA-21 flying half an orbit, or 50 min, ahead of the NOAA-20 and the SNPP positioned between them. Starting April 2024, the NOAA-20 has operated in a 180-degree separated orbit relative to the NOAA-21, a configuration that was achieved by NOAA-20 spacecraft maneuvers.

After its launch, the NOAA-21 VIIRS underwent an intensive calibration and validation (ICV) phase that included three different types of maneuvers: a roll maneuver for lunar observations to monitor the calibration stability for the reflective solar bands (RSBs), yaw maneuvers to validate the solar diffuser’s bi-directional reflectance factor (BRF) as well as the solar attenuation screen transmission, and a pitch maneuver to validate the thermal emissive bands’ response versus scan angle characteristics. The data collected from these maneuvers, along with the data collected from regular calibrations, have been used to provide regular calibration updates in support of the Collection 2 (C2) Level-1B (L1B) products as produced by NASA Science Investigator-led Processing Systems (SIPS).

This work demonstrates the radiometric stability and calibration consistency of the newly launched NOAA-21 VIIRS RSBs in comparison with the VIIRS onboard the SNPP and NOAA-20 based on vicarious approaches using Earth-view targets. Earth targets with suitable characteristics can be used as references to verify the postlaunch radiometric performance of a given spaceborne sensor. The selection of these targets is generally based on the following considerations: the target surface should have high spatial uniformity, high surface reflectance, temporally invariant surface properties for reflectance, a low probability of cloudy weather and precipitation, or a minimum influence of aerosol and atmospheric water vapor. A number of test sites are endorsed by the Committee on Earth Observation Satellites (CEOS) Working Group on Calibration and Validation as being “reference standards” (https://calvalportal.ceos.org/web/guest/pics_sites, last visit on 2 July 2024). Among these sites, there are six pseudo-invariant desert sites in northern Africa, including Libya-4, and the Dome Concordia (Dome C) site located in the Antarctic. These sites have been used frequently as stable radiometric targets for monitoring the long-term performance of satellite instruments [4,5,6,7]. In addition to the use of Earth scene targets, simultaneous nadir overpasses (SNOs) have been used in recent years to determine the relative differences between a pair of polar-orbiting satellites [8,9]. Since there are no SNOs among the three VIIRS sensors, a third satellite, Aqua MODIS, is used as a transfer radiometer [10]. Our recent analysis based on the four independent vicarious approaches showed that the NOAA-20 VIIRS reflectances are systematically 2 to 4% lower than those of the SNPP VIIRS for most RSBs, with some of the shortest wavelength bands showing differences of up to 7% [10,11]. The assessment is performed using the NASA SIPS C2 L1B reflectance products from the three VIIRS instruments. Similar results were obtained using the VIIRS sensor data record (SDR) provided by the NOAA Comprehensive Large Array-data Stewardship System (CLASS) [12,13]. While the calibration of each VIIRS sensor is independent and traceable to standards maintained by the National Institute of Standards and Technology (NIST), these findings indicate a significant radiometric bias between the three VIIRS sensors. Therefore, it is critical to evaluate the radiometric calibration performance using vicarious approaches that do not rely on sensor-based components.

Section 2 briefly describes the vicarious approaches, including data extraction over each site, methods to mitigate the influence of viewing and illumination angular effects, and corrections for band spectral response differences. Section 3 presents VIIRS sensor trending and intercomparison results from each approach. In Section 4, the observed radiometric biases between the three VIIRS are discussed. Section 5 concludes this study, highlighting the expected on-orbit stability of the newly launched NOAA-21 VIIRS and its close radiometric agreement with the NOAA-20 VIIRS visible (VIS) and near-infrared (NIR) bands, while noting noticeable differences with the SNPP VIIRS. The short-wave infrared (SWIR) bands show a significant offset (2–4%) with both the NOAA-20 and the SNPP VIIRS. Additionally, this paper discusses the future role of the SI (International System of Units) Traceable Satellites (SITSats), such as the Climate Absolute Radiance and Refractivity Observatory (CLARREO) Pathfinder (CPF) [14], in enhancing our understanding of the absolute accuracy and the observed biases between the three VIIRS instruments [15].

2. Methodology

The VIIRS spectral radiance, L_EV, is calculated using the following cubic polynomial of the sensor response corrected for background:

$$L_{\mathrm{E}\mathrm{V}}=F\left(c_0+c_{1}d{n}_{\mathrm{E}\mathrm{V}}+c_{2}d{n}_{\mathrm{E}\mathrm{V}}^2+c_{3}d{n}_{\mathrm{E}\mathrm{V}}^3\right)/{RVS}_{\mathrm{E}\mathrm{V}}$$

(1)

where F is commonly known as the F-factor; dn_EV is the sensor response corrected for background and temperature; the quadratic coefficients c₀, c₁, c₂, and c₃ are characterized from prelaunch measurements; and RVS is the half-angle mirror (HAM) reflectivity with an incident angle at Earth view (EV) relative to its value at an angle corresponding to the center of the space view (SV) port. The VIIRS reflectance, r_EV, is given by

$$r_{\mathrm{E}\mathrm{V}}=\pi L_{\mathrm{E}\mathrm{V}}/E_{\mathrm{S}\mathrm{U}\mathrm{N}}cos{\theta }_{\mathrm{E}\mathrm{V}}$$

(2)

where E_SUN is the solar spectral irradiance and θ_EV is the solar zenith angle. In this study, the VIIRS reflectance data are obtained from data provided in NASA SIPS VIIRS C2 L1B products [16].

2.1. Calibration of RSBs

The RSBs were radiometrically calibrated prelaunch and have been regularly maintained on orbit using the onboard solar diffuser (SD). Any on-orbit degradation in SD is determined by the onboard solar diffuser stability monitor (SDSM). Calibration changes and updates are achieved by regularly adjusting the F-factor [11,17]. The VIIRS also conducts near-monthly lunar observations, and the data can also be used to derive the lunar F-factors [18]. The long-term trends of the lunar-derived F-factors are compared to the SD-derived F-factors in order to derive a correction for any inadequacy in the SD degradation measurements. For the NOAA-21 VIIRS, the initial values of F-factor for the VIS and NIR bands range from 0.93 to 0.98, indicating substantially improved prelaunch measurements in comparison with the results from the SNPP and the NOAA-20 VIIRS [19,20]. The on-orbit performance of the F-factor showed excellent stability to well within 1% for the VIS and NIR bands. For the SWIR bands, however, their gains were adversely affected by the ice accumulated on the short/mid-wave infrared (S/MWIR) focal plane dewar walls, which caused a rapid increase in F-factors. To remove the ice, three outgassing activities were performed throughout the mission. Immediately after each outgassing, the F-factors for these SWIR bands returned to their respective at-launch values. Since the last outgassing, the SWIR band F-factors have maintained good stability.

2.2. Vicarious Approaches

This study relies on ground-based vicarious calibration sites for VIIRS calibration and intercomparison assessment. Our vicarious approaches include the Libya-4 desert site, deep convective clouds (DCCs), Dome C, and SNOs. Libya-4 (28.5°N, 23.4°E) is one of six CEOS reference pseudo-invariant calibration sites (PICSs) [6,21,22,23,24]. Near-nadir VIIRS overpasses over Libya-4 obtained from 16-day repeatable orbits are collected to ensure that each overpass is under the nearly same sensor view angles. A 25 × 25 km area centered at Libya-4 is used to extract pixels within the defined area. After the data extraction, these pixels are averaged for each band and half-angle mirror (HAM) side. A threshold of 2% in standard deviation is used to examine the uniformity of the pixel area for cloud contamination, with any overpasses larger than the 2% threshold being excluded. The impacts of the bi-directional reflectance function (BRDF) on the reflectance trending are corrected using a site-dependent semi-empirical BRDF model, with the coefficients derived based on the initial three years of observations from the SNPP VIIRS. The BRDF is a function of four angular variables that defines how light from a source is reflected off an opaque surface. The function takes an incoming light direction and an outgoing direction and returns the ratio of reflected radiance off the surface to the irradiance incident on the surface. Due to the local time differences of the overpasses from the SNPP and the NOAA-20 and -21 (less than 30 min), a BRDF model developed from one of the three sensors is applicable to the other two regarding the range of angular parameters. The BRDF normalized reflectances are used to perform radiometric stability and bias assessment for the three VIIRS sensors. The results from our previous studies [10,11] showed that temporal variations in reflectance in the VIS and NIR bands are around 1% after the BRDF correction. For a few of the SWIR bands, there are slightly large temporal variations (>1%), due to either a relatively low signal level or a high sensitivity to atmospheric water vapor [24].

DCCs are near-Lambertian for RSBs and are often found in tropical regions. They are excellent targets for calibration assessment because they are bright and located at high altitudes, where the impacts of atmospheric water vapor and aerosols are low. DCCs have been found to be useful for vicarious calibration across the entire RSB wavelength range [22,23]. The approach relies on large-ensemble statistics over monthly collected DCC pixels, which are identified over the tropics using the 11 µm infrared (IR) channel brightness temperature (BT) with a threshold of 205 K. A spatial uniformity criterion of 3% in reflectance at 0.65 µm and a standard BT deviation of 1.0K over a 3 × 3 pixel window are applied to eliminate impact due to optically thin clouds. Our previous studies applied a BRDF correction at the DCC pixel level [10,11]. An examination of the observed temporal variability of the monthly probability distribution functions (PDFs) indicates any impact on near-nadir results is not significant for the VIS and near-IR bands. For the SWIR bands, the application of the band-dependent empirical BRDF models can effectively reduce the temporal variability [24]. Since the focus of this study is on the assessment of the newly launched NOAA-21 VIIRS, no BRDF correction is applied to its DCC data. No BRDF correction is applied to the SNPP and the NOAA-20 VIIRS to allow a consistent intercomparison of results from all three sensors.

The Dome C site (75.1°S, 123.4°E), which is an invariant calibration site with surface uniformity covered by snow and ice sheets, is located on the Antarctic Plateau [25]. The site has been used for various studies due to its unique characteristics of high elevation, high reflectance, high percentage of cloud-free conditions, low water vapor content, aerosol, and dust content. Due to its location, Dome C is observed more frequently by polar-orbiting satellites and can be used to track and quantify sensor calibration performance. Similar to the Libya-4 site, pixels within a 25 km × 25 km area centered at the Dome C site are collected, and the pixel-level data are averaged for each band and HAM side. The same 2% threshold in the standard deviation is used to remove overpasses due to possible cloud impacts. A simple BRDF model, expressed as a linear relationship with the determined solar zenith angle, worked reasonably well for most RSB bands [9,10]. The BRDF model is determined based on observations from the SNPP VIIRS and applied to the reflectances of all three VIIRS sensors. The BRDF correction is able to reduce the angular-dependent fluctuations to within 2% for the VIS and NIR bands. For the SWIR bands, the results from the Dome C site are noisy due to a lack of sufficient signal-to-noise ratio and are not included in this study.

The SNO approach allows for a direct comparison of spectrally matching bands between two crossover sensors. The impacts of scene variability, solar illumination, and sensor view angles are largely reduced since the two sensors view the Earth and the atmosphere at nearly the same angle, time, and location but from different altitudes [8]. Because the three VIIRS sensors are in the same orbit and height but with different local overpass times, there are no simultaneous crossovers among them. However, there are abundant simultaneous crossovers between the MODIS and each of the VIIRS sensors [10,11]. We have selected Aqua MODIS as a transfer radiometer for intercalibrating the three VIIRS sensors. The SNO locations of the Aqua MODIS and VIIRS SNOs are not fixed but occur frequently in the high-latitude region. The geo-locationally matched and unsaturated VIIRS and MODIS pixel pairs within ±15.0^o off the nadir are collected. To allow more VIIRS bands to be used in the SNO approach, data are collected at the pixel level and therefore include more samples from MODIS high-gain bands 8, 9, and 15. An average of the VIIRS-to-MODIS reflectance ratio and the standard deviation are calculated once the number of matching pixel pairs is statistically sufficient (more than 10,000) for each SNO event. Having a large enough SNO sample size primarily helps reduce the impact of the pixel footprint difference between the VIIRS (750 m) and the MODIS (1 km). Pre-defined standard deviations of around 2.5% are applied to exclude those SNO events with uncertainties beyond the criteria.

2.3. Correction for Spectral Differences

Differences in the relative spectral response (RSR) between the same bands of the three VIIRS sensors can cause additional uncertainties for the radiometric assessment. The impact of these differences is corrected based on hyperspectral observations from SCIAMACHY, one of ten instruments aboard the European Space Agency (ESA) Environmental Satellite, ENVISAT [26]. SCIAMACHY provides hyperspectral measurements with a spectral resolution of 0.2 to 0.5 nm and a wavelength range of 240 to 1700 nm. Although the ENVISAT mission ended in 2012, it provided a high-quality dataset over a period from August 2002 to April 2012. The SCIAMACHY data used in this study are based on L1B version 7.03, SCI_NL_1P, over a one-year period in 2004. The correction is expressed as a theoretical ratio of reflectance, R, between any two sensors.

$$R_{\mathrm{1,2}}=\frac{{\int }_0^{\infty }{\rho }_1\left(\lambda \right){RSR}_1\left(\lambda \right)d\lambda /{\int }_0^{\infty }{RSR}_1\left(\lambda \right)d\lambda }{{\int }_0^{\infty }{\rho }_2\left(\lambda \right){RSR}_2\left(\lambda \right)d\lambda /{\int }_0^{\infty }{RSR}_2\left(\lambda \right)d\lambda }$$

(3)

where ρ₁ and ρ₂ are the hyperspectral measured spectral reflectances at wavelength $\lambda$ for sensors 1 and 2, respectively, and RSR₁ and RSR₂ are their spectral responses. For the SNPP VIIRS, there are several RSR LUT updates due to the initial on-orbit spectral degradation in the Rotating Telescope Assembly (RTA) mirror reflectance. Since the degradation of the RSR has leveled off after 5 to 6 years of on-orbit operation and all comparisons in this study are for data after 2022, the latest update of the RSR is used. The hyperspectral data for Libya-4 and Dome C were collected based on the geo-location of the sites. For the SNO approach, since their locations are not fixed, an averaged result over different scene types is used. For the DCC approach, the SCIAMACHY-based spectral band adjustment factor (SBAF) is used, which is based on the tool provided by NASA-LaRC (see link at https://cloudsgate2.larc.nasa.gov/cgibin/site/showdoc?mnemonic=SBAF) (accessed on 15 March 2024) [27].

3. Results

3.1. Libya-4

Trends of near-nadir reflectance over the Libya-4 site for SNPP and NOAA-20/21 VIIRS M1, M3, M7, and M10 for a period starting in 2022 are shown in Figure 1. The four bands are selected to represent the VIS, NIR, and SWIR wavelength ranges. A site-dependent BRDF model derived using SNPP VIIRS reflectance data is applied to all three sensors. Trending results covering SNPP and NOAA-20 mission periods up to the end of 2023 indicate they are stable to within 1% for all bands [10,11]. The early trends for the newly launched NOAA-21 VIIRS show that they are stable, although the operational period is just over one year. A comparison of the BRDF normalized reflectances for their overlapping period provides an assessment of the relative consistency of radiometric calibration. Additional factors affecting the systematic biases include differences in spectral response, atmospheric effects, etc. These results show that NOAA-20 and -21 reflectances are consistent within 2% for the VIS and NIR bands, and both are systematically lower than SNPP’s by about 8%. For the SWIR band M10, although NOAA-20 reflectances are much closer to SNPP’s, NOAA-21’s results are systematically lower by about 4% before correcting for their spectral differences. Further adjustments to the SWIR band calibration in the NOAA-21 VIIRS are planned in a future L1B reprocess that includes other changes. In the current C2 L1B, the M11 band calibration has been adjusted by about 6% based on lunar calibration data, improving consistency with the NOAA-20 VIIRS.

3.2. DCC

For the DCC approach, the tropical DCC pixels are identified using similar angular, BT, and spatial homogeneity thresholds for all three VIIRS sensors. Because all of them are in a 13:30 sun-synchronous orbit, they have consistent local time sampling of the DCC cycle, which avoids any diurnal sampling differences in DCC properties between the three instruments. The impact of the BRDF on reflectances at near-nadir view angles (<40 degrees) is found to be not significant for the VIS and NIR bands due to the DCC’s highly Lambertian characteristics [28]. DCC pixel-level reflectances are compiled into monthly probability distribution functions (PDFs), and the mode of the PDFs is tracked over time to monitor the VIIRS sensor’s stability. The observed temporal natural variability of the monthly PDF mode timeseries in the VIS to NIR wavelength range is below 1%. However, for the SWIR bands, the DCC reflectivity is affected by the ice particle size, and as such, the DCC technique is more sensitive to the BRDF and the IR BT threshold, thereby resulting in large seasonal cycles in the DCC mode response timeseries [24]. The trends of DCC mode reflectance for M1, M3, M7, and M8 are shown in Figure 2. The mode reflectance timeseries for each VIIRS is normalized by the mean value obtained from the SNPP VIIRS over the same period. A normalized mode reflectance value of less than 1 indicates that the measured reflectance is darker compared to that of the SNPP VIIRS. Because the radiometric intercomparison is conducted by comparing the means of DCC mode timeseries for different VIIRS, BRDF correction is not necessary for this study. Like Libya-4 results, both NOAA-20 and -21 VIIRS DCC reflectances are consistent within 2% for the VIS and NIR bands, and both are systematically lower compared to the DCC reflectance values from the SNPP VIIRS. For the SWIR band M8, reflectances from the NOAA-21 VIIRS are noticeably lower than those from the SNPP and the NOAA-20 VIIRS.

3.3. Dome C

To reduce the impact of natural variability at the Dome C site, near-nadir reflectances obtained from the post-solstice period from 22 December 2023 to 15 March 2024 [29] are used to examine the calibration consistency among SNPP, NOAA-20, and -21 VIIRS. For each sensor, there is about one near-nadir overpass every two days, which provides enough data samples for the analysis. The reflectances are plotted as a function of solar zenith angle in Figure 3, showing the results of M1, M3, I1, and I2. A trend that decreases in reflectance while increasing in solar zenith angle indicates the impact of the BRDF. Given an uncertainty of ~1% [10,11], the relative differences among the three sensors for the VIS and NIR bands are generally consistent with those from the Libya-4 and DCC approaches. As discussed earlier, the results from the SWIR bands are excluded due to their exceptionally high uncertainties owing to the low SNR over Dome C.

3.4. SNO

Several SNO events in 2023 that cover a wide range of atmospheric and surface conditions are used to examine the VIIRS intercomparison results over a broad range of reflectance. The VIIRS-to-Aqua reflectance ratios from SNOs at the pixel level are binned at a reflectance interval of 0.02. Figure 4 shows the bin-averaged reflectance ratios as a function of reflectance for M2, M4, M7, and M10 of the three VIIRS sensors. For the VIS and NIR bands, the frequency of the geo-locationally matched SNO pixels with reflectances lower than 0.2 is significantly lower. The relative differences among the three sensors for these bands are consistent with those from the other three approaches. The impact of the band spectral difference between the VIIRS and Aqua MODIS is mainly seen within the SWIR bands; this is the main reason the ratios of VIIRS M10 to Aqua band 6 show a radiance-level-dependent trend.

The existing differences in the relative spectral response among SNPP and NOAA-20 and -21 VIIRS RSBs introduce systematic differences that would affect their radiometric assessment. The correction for the spectral differences is based on the theoretical ratio in reflectance derived using Equation (3) for the Libya-4, Dome C, and SNO approaches. For the DCC approach, it is based on the SBAF factor (SBAF provided by NASA-LaRC).

Table 2. SNPP and NOAA-20 VIIRS RSB comparison. Results are provided as percentage differences (SNPP reflectances relative to NOAA-20). Values after ± are the standard error as a percentage. Those marked “NA” are either not applicable or are excluded due to data variation.

Band	M1	M2	M3	M4	M5	M7	M8	M9	M10	M11	I1	I2	I3
Libya-4	7.85 ±1.25	6.16 ±1.32	4.22 ±1.40	3.20 ±1.21	4.47 ±0.87	2.67 ±1.10	2.44 ±1.25	NA	1.75 ±0.75	1.50 ±2.18	2.74 ±1.22	2.59 ±1.08	3.00 ±0.76
DCC	6.73 ±1.36	5.27 ±1.24	5.32 ±1.17	5.38 ±1.32	4.95 ±1.22	3.93 ±0.97	1.80 ±0.75	0.74 ±2.78	2.61 ±2.68	1.85 ±1.97	4.56 ±1.21	4.4 ±0.92	4.13 ±2.62
Dome C	7.99 ±1.30	6.42 ±1.15	4.73 ±1.48	4.51 ±3.38	4.88 ±2.33	2.17 ±2.49	NA	NA	NA	NA	3.01 ±3.25	2.42 ±2.50	NA
SNO	6.87 ±2.06	5.72 ±1.20	3.74 ±1.39	4.59 ±0.76	4.35 ±2.03	2.77 ±1.43	2.98 ±1.50	NA	2.93 ±3.27	NA	2.27 ±1.01	2.49 ±1.42	3.81 ±3.27

summarizes the absolute reflectance differences as percentages for all the VIIRS bands after accounting for the band spectral difference, determined by averaging the reflectances over their overlapping period since 2018. Comparisons before and after the correction indicate the differences are more consistent among the results from the four approaches [10]. For the shortest wavelength bands M1 and M2, the differences in reflectance between the SNPP and the NOAA-20 are the largest, ranging from 5 to 7%, with the SNPP’s reflectances being brighter. These differences are significantly larger than those reported based on VIIRS SDR products generated by NOAA CLASS [11,12]. One reason for the higher reflectances from the SNPP in the NASA SIPS C2 L1B product could be due to an initial gain adjustment (to account for the rapid degradation due to UV radiation after launch [30]). For the longer wavelength bands in the near-IR and SWIR regions, the differences reduce to around 2–4%, with results from the SNPP still higher than those from the NOAA-20.

For the newly launched NOAA-21 VIIRS, the agreement with the NOAA-20 is within 2% for the VIS and NIR bands, except for the results for bands M1 and M2 from the Dome C site, as shown in

Table 3. NOAA 20 and 21 VIIRS RSB comparison. Results are provided as percentage differences (NOAA-20 reflectances relative to NOAA-21). Values after ± are the standard error as a percentage. Those marked “NA” are either not applicable or are excluded due to data variation.

Band	M1	M2	M3	M4	M5	M7	M8	M9	M10	M11	I1	I2	I3
Libya4	0.90 ±1.16	1.37 ±1.45	1.29 ±1.49	0.91 ±1.10	−0.36 ±0.95	0.91 ±1.10	3.06 ±1.15	NA	4.62 ±0.82	2.05 ±2.23	−0.46 ±1.28	0.51 ±1.09	2.34 ±0.88
DCC	0.79 ±1.05	1.17 ±1.23	0.62 ±1.08	−2.23 ±1.04	−1.18 ±1.05	0.31 ±0.75	2.47 ±0.79	1.86 ±2.31	2.97 ±2.50	−0.24 ±1.99	−0.93 ±1.06	−0.42 ±0.82	0.74 ±2.59
Dome C	3.83 ±1.02	4.28 ±1.24	1.46 ±1.48	−0.81 ±3.48	0.47 ±2.39	−0.99 ±2.38		NA	NA	NA	−0.81 ±3.23	−1.10 ±2.40	NA
SNO	1.80 ±2.28	2.11 ±1.24	1.64 ±1.52	1.11 ±0.85	−0.10 ±1.98	1.21 ±1.59	1.09 ±1.68	NA	3.33 ±2.60	NA	0.59 ±1.08	1.19 ±1.58	2.73 ±2.43

. The reason for the relatively larger bias at the Dome site is likely due to the conditions of the site, and an examination of the results between the SNPP and NOAA-20 indicates that the differences are slightly larger than those from the other three approaches. For the SWIR band comparison, reflectances from the NOAA-21 are lower than those from the NOAA-20 by 2 to 4%, depending on the band. The results from NOAA-21 VIIRS M11 (2.25 µm) showed much lower reflectance in the earlier version of L1B. As a result, an upward adjustment of 6.1% in the calibration coefficient, i.e., the F-factor, is applied to M11 over the entire mission in the current version of L1B, which is obtained based on lunar intercomparison results.

Table 2 and Table 3 indicate that the results among the different approaches show differences up to 2.0%, with the DCC approach having relatively low standard errors for the VIS/NIR bands in comparison with the other approaches, while the Dome C results show the largest uncertainties. This indicates potential improvement in the use of these methodologies. We expect residual errors due to the BRDF to be approximately 1% for these bands, plus an additional 0.5% due to the lack of an observed real-time spectral correction among sensor-to-sensor comparisons.

4. Discussion

This study extended our previous results [10,11] by including the newly launched NOAA-21 VIIRS. As discussed in the early study, the systematic biases between the SNPP and NOAA-20 VIIRS RSBs are likely due to errors in the prelaunch measurements of the solar diffuser BRF. The results from the NOAA-21 showed that the reflectances from a few SWIR bands are systematically lower than those from the NOAA-20. The reason for the biases is not clear; however, a possible cause could be errors in the results of the SD BRF determined from prelaunch tests. Because there were no direct measurements of the SD BRF in the SWIR wavelength region due to the limitations of the test conditions, a model-based extrapolation was applied to determine the BRF for the SWIR bands. The largest bias, which occurs in M11 (the longest wavelength band before the on-orbit adjustment in the F-factor was applied), indicates errors that are wavelength-dependent due to extrapolation.

The results from the Dome C site show relatively high uncertainties in comparison with the other three approaches. An examination of the trending results indicates that the Dome C reflectances show noticeable seasonal and interannual variability, particularly for bands with wavelengths longer than 0.488 µm. A recent study by Doelling et al. [29] showed that the Dome C reflectance in some visible channels seems to be well correlated with the monthly ozone concentrations. Regressing the ozone concentration with the reflectance and applying the ozone regression coefficients could reduce the Dome C trend standard error.

Understanding and addressing differences and changes in radiometric accuracy regarding gains and biases is important for multi-sensor data to be used coherently. This is particularly useful for observations made from the series of VIIRS sensors since their instrument design is nearly identical. The observed systematic biases in reflectance among the three VIIRS sensors are currently up to 8%, which is largely attributed to errors in prelaunch measurements. Although the calibration of each VIIRS sensor is independent and traceable to standards maintained by the NIST, there is a large concern within the community about which VIIRS sensor represents the absolute radiometric scale most accurately. NASA’s CPF mission includes an Earth-viewing reflected solar (RS) spectrometer that will be onboard the International Space Station (ISS) to provide Earth-reflected solar radiation measurements with an unprecedented SI-traceable radiometric uncertainty of 0.3–0.6% (k = 1) [31]. The CPF will be the first SITSat to provide high-accuracy benchmark measurements to serve as an intercalibration reference for other RS instruments in orbit. The CPF intercalibration team has developed a state-of-the-art approach to intercalibrate the RSBs of the VIIRS instrument onboard the NOAA-20 satellite against the CPF benchmark measurements with a targeted methodology uncertainty no greater than the radiometric uncertainty of the CPF [14]. To achieve this high level of intercalibration accuracy, the CPF team has devised strategies to mitigate the effects of spatial, spectral, and angular disparities between the intercalibration footprints of the CPF and the target instruments. Additionally, the CPF team has developed Polarization Distribution Models (PDMs) to characterize the polarization state of Earth-reflected radiance as a function of the intercalibration spatial resolution, Earth scene type, solar and viewing geometry, and wavelength. The PDMs will assist in identifying low-polarized scene radiances to meticulously intercalibrate the polarization-sensitive VIIRS short-wave channels against the significantly less polarization-sensitive CPF instrument.

The precise intercalibration between the CPF and the NOAA-20 VIIRS will unveil the absolute radiometric accuracy of the VIIRS. Once the NOAA-20 VIIRS’ absolute calibration is established through the CPF, the VIIRS can function as an improved transfer radiometer, enabling the extension of the CPF’s SI-traceable reference calibration to other reflected solar sensors, including other VIIRS sensors. This extension can be accomplished using the same vicarious methods outlined in this paper, as depicted in Figure 5.

5. Conclusions

The recently launched VIIRS instrument on the NOAA-21 is the third sensor in the VIIRS series, with two earlier similar instruments operating on the SNPP and the NOAA-20 spacecrafts. The three VIIRS sensors have been in combined operation for nearly 12 years, providing the global science community with unique and high-quality Earth observations to facilitate a broad range of scientific studies related to land, atmosphere, cryosphere, and oceans. This study provides an early radiometric assessment of the NOAA-21 VIIRS RSBs using multiple vicarious approaches and its calibration consistency with its predecessors. The data used in the analyses were derived from NASA SIPS VIIRS C2 L1B reflectance products. The results indicate that the radiometric calibration of the NOAA-21 VIIRS is comparable with that of the VIIRS onboard the NOAA-20 for the VIS and NIR bands. For the SWIR bands, the NOAA-21 VIIRS reflectances are lower reflectances (2–4%) compared to those on the NOAA-20. The observed systematic biases in reflectance among the three VIIRS sensors are currently up to 8%. Despite the noticeable inconsistencies in calibration, all three VIIRS sensors exhibit excellent temporal radiometric stability over their missions. The direct intercalibration of the NOAA-20 VIIRS with the high-accuracy SI-traceable calibration reference of the CPF will enhance the absolute radiometric accuracy of all three VIIRS. The results of this study provide useful information about potential impacts on the use of downstream science products. They also provide valuable information that can be helpful for diagnosing and improving future calibration algorithms.