1. Introduction
The Visible and Infrared Imaging Radiometer Suite (VIIRS) onboard the Joint Polar Satellite System (JPSS)/Suomi National Polar-Orbiting Partnership (SNPP) satellite has 22 spectral bands, with 14 Reflective Solar Bands (RSB), 7 Thermal Emissive Bands (TEB) and 1 Day-Night-Band (DNB) [
1,
2]. VIIRS RSBs are calibrated using a full-aperture Solar Diffuser (SD) and the degradation of SD is monitored by a Solar Diffuser Stability Monitor (SDSM). Significant SD, SDSM detectors, and Rotating Telescope Assembly (RTA) mirror degradations were observed, especially early after launch. The rate of RTA mirror and SD degradations has decreased since mid-2013 and early 2014, respectively. Their impacts on the instrument performance are negligible (at 0.1% level) due to weekly updates of RSB calibration coefficients (F-factor) Look-Up Table (LUT) [
3,
4]. However, larger F-factor fluctuations have been observed since early 2014 in some shorter wavelength bands, with band M1 fluctuating the most (~3%). The underlying cause for the F-factor anomalies has not been identified so far. VIIRS RSB Sensor Data Records (SDR) provide input data for the retrieval of many Environment Data Records (EDR) products, such as ocean color, vegetation, cloud, and surface albedo. It is therefore critical to monitor the long-term stability of VIIRS RSBs to ensure quality EDR products.
VIIRS DNB is a panchromatic imagery band (0.5–0.9 µm) with dynamic range of approximately seven orders of magnitude [
5]. It has three gain stages: the Low Gain Stage (LGS, daytime), the Medium Gain Stage (MGS, twilight) and the High Gain Stage (HGS, nighttime). The onboard calibration of DNB LGS is similar to that for RSBs, with LGS gain derived using SD/SDSM data. DNB MGS and HGS are calibrated using LGS gain and MGS/LGS and HGS/LGS gain ratios estimated using Earth View (EV) data at the terminator orbit where EV observations at the three gain stages co-exist. VIIRS DNB is originally designed as an imagery band to continue the night observation heritage of the Operational Linescan System (OLS) onboard the Defense Meteorological Satellite Program (DMSP). DNB nighttime observations have been widely used in many areas such as detecting power outrage, monitoring city lights, urban expansion and fishing boats, as well as studying air glow, aurora, and lightning. Due to its superior spatial and radiometric performance, quantitative applications of the VIIRS DNB band have been developed or are under development, such as estimating population and economic output [
6,
7] and nighttime aerosol retrieval [
8].
The JPSS Interface Data Processing Segment (IDPS) ground processing team has produced nearly four years of NOAA operational VIIRS SDRs to date, with several major calibration changes applied since launch. The JPSS program is currently planning to reprocess the entire VIIRS SDR products. To better facilitate quantitative applications and the future reprocessing of the VIIRS RSB and DNB SDRs, it is important to use independent validation time series to evaluate their long-term post-launch calibration stability. Many efforts have been devoted to characterize the long-term stability of the NOAA operational VIIRS RSBs and DNB (HGS). Uprety
et al. [
9] and Uprety and Cao [
10] investigated the VIIRS onboard RSB radiometric performance using the extended Simultaneous Nadir Overpass (SNO-x) approach over desert sites, and the Antarctica Dome C site. Wang and Cao [
11] provided a preliminary assessment of the VIIRS bands M1–M5 and M7 calibration stability using monthly Deep Convective Clouds (DCC) time series. Liao
et al. [
5] evaluated DNB HGS radiometric calibration accuracy using vicarious calibration under lunar illumination. Ma
et al. [
12] investigated the DNB HGS calibration accuracy by comparing the observed VIIRS DCC radiance with nighttime DCC radiances simulated using a radiative transfer model. Cao and Bai [
13] investigated the feasibility of using point light sources for monitoring DNB HGS calibration stability.
The purpose of this study is to investigate the NOAA operational VIIRS RSB and DNB (LGS) long-term radiometric calibration stabilities using the DCC technique. It extends the previous study [
11], which provides preliminary results for the moderate resolution visible and near infrared (VIS/NIR) bands (M-bands) only, by using longer time series and analyzing more bands, including DNB, shortwave infrared bands (SWIR) and imagery resolution bands (I-bands). Moreover, semi-monthly DCC time series were also developed and analyzed for stability monitoring at higher temporal resolution. In this study, the DCC technique was applied to all RSB bands, including 10 M-bands and 3 I-bands. Stable DNB LGS radiometric calibration is a prerequisite for quality DNB nighttime calibration to support quantitative applications of DNB data. Previous studies have focused on characterizing HGS data directly, but the radiometric calibration stability of DNB LGS has not been well studied so far. This paper is organized as follows.
Section 2 introduces VIIRS SDR products used in this study.
Section 3 presents the DCC technique for RSBs and DNB.
Section 4 presents and discusses monthly and semi-monthly DCC time series for calibration change and long-term stability monitoring. Finally
Section 5 summarizes the findings of the study.
2. VIIRS SDR Products Used
VIIRS has 11 RSB M-bands (M1–M11, 750 m) and 3 RSB I-bands (I1–I3, 375 m). M6 saturates over DCCs, therefore it was excluded from this study. TEB band M15 (10.763 µm) provides brightness temperature measurements that are required to identify DCC pixels for RSB M-bands and DNB. I1–I3 DCCs are identified using I5 (11.469 µm) brightness temperature. The center wavelength, spatial, and radiometric characteristics of all bands used in this study are summarized in
Table 1.
Figure 1 presents the Relative Spectral Response (RSR) functions for these bands.
Table 1.
Spectral, spatial, and radiometric characteristics of Visible and Infrared Imaging Radiometer Suite (VIIRS) spectral bands used in this study. M1–M5 and M7 SNRs at low gain stage and typical scene radiances are listed because Deep Convective Clouds (DCCs) are highly reflective; band M15 and I5 NEdTs is estimated at 205 K scene temperature.
Table 1.
Spectral, spatial, and radiometric characteristics of Visible and Infrared Imaging Radiometer Suite (VIIRS) spectral bands used in this study. M1–M5 and M7 SNRs at low gain stage and typical scene radiances are listed because Deep Convective Clouds (DCCs) are highly reflective; band M15 and I5 NEdTs is estimated at 205 K scene temperature.
| Band | Center Wavelength (µm) | Spatial Resolution at Nadir (m) | SNR/NEdT (Spec) | SNR/NEdT (On-Orbit) |
---|
VIS/NIR | M1 | 0.411 | 750 | 316 | 1045 |
M2 | 0.444 | 750 | 409 | 1010 |
M3 | 0.486 | 750 | 414 | 988 |
M4 | 0.551 | 750 | 315 | 856 |
M5 | 0.672 | 750 | 360 | 631 |
I1 | 0.639 | 375 | 119 | 214 |
M7 | 0.862 | 750 | 340 | 631 |
I2 | 0.862 | 375 | 150 | 264 |
SWIR | M8 | 1.238 | 750 | 74 | 221 |
M9 | 1.375 | 750 | 83 | 227 |
M10 | 1.602 | 750 | 342 | 586 |
I3 | 1.602 | 375 | 6 | 149 |
M11 | 2.257 | 750 | 10 | 22 |
DNB | DNB | 0.700 | 750 | ≥6@Lmin | >9 across scan after degradation |
TEB | M15 | 10.729 | 750 | 0.26 K | 0.10 K |
I5 | 11.469 | 375 | 1.7 K | 0.43 K |
Figure 1.
Relative spectral response functions (RSR) of all VIIRS bands used in this study.
Figure 1.
Relative spectral response functions (RSR) of all VIIRS bands used in this study.
The area of interests of this study is a region defined as 25°S to 25°N and 150°W to 60°W. This area covers a portion of the InterTropical Convergence Zone (ITCZ) over the western end of tropical Pacific Ocean and its adjacent South America Continent. DCCs over the same region were also used in previous studies [
11,
14]. VIIRS daytime Top of Atmosphere (TOA) reflectance/brightness temperature, and radiance were downloaded from the NOAA’s Comprehensive Large Array-data Stewardship System (CLASS), the NOAA/NESDIS/STAR Central Data Repository (a 4 months revolving mirror site of CLASS for SNPP data), and the NASA Atmosphere Science Investigator-led Processing Systems (SIPS). All VIIRS SDR products used in this study are generated by the NOAA operational ground processing unit at the JPSS IDPS. It is worth noting that no vicarious calibration is applied to the IDPS version of VIIRS RSB and DNB SDR products.
3. The DCC Technique for VIIRS
DCCs are extremely cold clouds above which the absorptions due to water vapor and other gases are minimal in the VIS/NIR spectrum. DCCs are abundant over the ITCZ and can be simply identified using a single longwave infrared (LWIR) channel centered at ~11 µm brightness temperature (TB11 hereafter). Hu
et al. [
15] first demonstrated that DCCs have a constant mean albedo over the lifetime of the Clouds and the Earth’s Radiant Energy System (CERES) on board the Tropical Rainfall Measuring Mission (TRMM) satellite. The DCC technique outlined by Hu
et al. [
15] was further improved through various studies such as Doelling
et al. [
16,
17], Minnis
et al. [
18], and Fougnie and Bach [
19]. It has been widely used for post-launch calibration and stability monitoring in the solar reflective spectrum during the past decade [
11,
16,
17,
18,
19,
20,
21]. The DCC technique generally consists of the following steps: (1) collecting satellite data over an area of interest; (2) identifying DCC pixels; (3) correcting for the anisotropic effect in DCC reflectance; (4) calculating monthly DCC probability distribution functions (PDF); (5) generating and analyzing monthly DCC mean and mode time series. It is a statistical-based vicarious calibration method; therefore sufficient DCC samples need to be collected to ensure robust statistical analysis results.
In this study, VIIRS DCC pixels are identified using the Wang and Cao [
11,
14] method, which is similar to those described in Minnis
et al. [
18] and Doelling
et al. [
17]. The criteria for identifying VIIRS M-bands DCC pixels are summarized as follow: (1) M15 brightness temperature is less than 205 K; (2) standard deviation of TB11 of the subject pixel and eight adjacent pixels is less than 1 K; (3) standard deviation of TOA reflectance of the subject pixel and eight adjacent pixels is less than 3% relative to the mean reflectance of the nine pixels; (4) solar zenith angle is less than 40°; (5) sensor view zenith angle is less than 35°.
The identifications of DNB and I-bands DCCs are similar to that for M-bands. DNB radiances were mapped to M15 lat/lons and converted to TOA reflectance before DCC pixels were identified using band M15 brightness temperatures. Band I5 brightness temperatures were used for the I-band DCC identifications. Wang and Cao [
14] found that the mode and mean of DCC reflectance is a function of spatial resolution but not sensitive to brightness temperature difference on the order of 0.5 K. In this study, I1–I3 TOA reflectance and I5 brightness temperatures were down-sampled to M-band resolution before DCCs are identified to facilitate inter-comparison of DCC time series between M-bands and I-bands. Though different TB11 bands were used for M-bands/DNB and I-bands DCC identifications, the mean M15 brightness temperature is only ~0.2 K higher than the mean I5 brightness temperatures for the down-sampled I-bands DCCs, therefore its impact can be ignored.
RSB M-bands and DNB DCC TOA reflectance as well as TB11 brightness temperature datasets over the area of interest were generated using the DCC identification criteria described above from March 2012 to September 2015. The study period for RSB I-bands is from January 2014 to September 2015. Though DCCs have nearly Lambertian behavior, the anisotropic effect still exists in the DCC TOA reflectance and Angular Distribution Models (ADM) were developed to account for the effect [
15,
17,
22]. For each DCC pixel, the ADM-adjusted DCC reflectance datasets were also generated using an ADM developed by Hu
et al. [
15]. Monthly PDFs, as well as their means and modes, were calculated for the DCC TOA reflectance and the ADM-adjusted DCC reflectance (DCC reflectance hereafter) with a 0.003 increment (in reflectance) [
14].
Figure 2 shows examples of monthly PDFs for bands M1, M7, M9, and DNB ADM-adjusted DCC reflectance. Standard deviations (sd), minimum (Min), maximum (Max), and range (Max−Min) were calculated using DCC datasets for the selected months. Monthly DCC statistics for all bands during the entire study period are summarized in
Table 2 (before the ADM adjustment) and
Table 3 (after the ADM adjustment).
The monthly DCC statistics derived from this study are generally consistent with previous studies in the VIS/NIR spectrum. The averaged monthly DCC TOA reflectance is larger than 0.94 for all VIS/NIR bands and DNB, confirming that DCCs are highly reflective [
15,
17,
19]. The range of band M5 monthly DCC mean TOA reflectance is ~2.8% (see
Table 2), smaller then the ±2% range reported by Fougnie
et al. [
19]. The ADM-adjusted DCC reflectance is more invariant compared to the DCC TOA reflectance (see
Table 2 and
Table 3). The range of band M5 monthly DCC reflectance is further reduced after the ADM-adjustment, agreeing with the results from Doeling
et al. [
17]. Therefore, the ADM-adjusted monthly DCC reflectance (DCC reflectance hereafter) was used to analyze VIIRS VIS/NIR and DNB radiometric calibration stabilities. Detailed analysis of VIIRA VIS/NIR bands and DNB DCC time series are presented in
Section 4.1,
Section 4.2, and
Section 4.4.
Figure 2.
Probability distribution functions of VIIRS bands M1, M7, M9, and Day-Night-Band (DNB) Angular Distribution Models (ADM)-adjusted monthly DCC reflectance. Standard deviation (sd), minimum (Min), maximum (Max), and range (Max−Min) were calculated using DCC data from selected months.
Figure 2.
Probability distribution functions of VIIRS bands M1, M7, M9, and Day-Night-Band (DNB) Angular Distribution Models (ADM)-adjusted monthly DCC reflectance. Standard deviation (sd), minimum (Min), maximum (Max), and range (Max−Min) were calculated using DCC data from selected months.
Table 2.
Statistics of monthly DCC Top of Atmosphere (TOA) reflectance (before the ADM adjustment) mean and mode time series for Reflective Solar Bands (RSB) M-bands (M1–M5, M7–M11, March 2012–September 2015), DNB (May 2013–September 2015), and RSB I-bands (I1–I3, January 2014–September 2015).
Table 2.
Statistics of monthly DCC Top of Atmosphere (TOA) reflectance (before the ADM adjustment) mean and mode time series for Reflective Solar Bands (RSB) M-bands (M1–M5, M7–M11, March 2012–September 2015), DNB (May 2013–September 2015), and RSB I-bands (I1–I3, January 2014–September 2015).
Band | DCC Mode | DCC Mean |
---|
Avg | sd (%) | Max−Min (%) | Avg | sd (%) | Max−Min (%) |
---|
VIS/NIR | M1 | 1.018 | 1.0 | 4.124 | 0.977 | 1.0 | 3.825 |
M2 | 1.009 | 1.1 | 4.461 | 0.967 | 1.1 | 5.107 |
M3 | 1.005 | 1.0 | 4.178 | 0.963 | 1.0 | 4.187 |
M4 | 0.973 | 0.9 | 4.009 | 0.931 | 0.9 | 3.534 |
M5 | 1.002 | 0.8 | 3.294 | 0.959 | 0.7 | 2.804 |
I1 | 0.963 | 0.8 | 3.740 | 0.919 | 0.9 | 3.544 |
M7 | 0.991 | 0.6 | 2.725 | 0.955 | 0.7 | 2.637 |
I2 | 0.991 | 0.6 | 1.816 | 0.954 | 0.8 | 3.355 |
SWIR | M8 | 0.704 | 1.0 | 3.834 | 0.698 | 1.0 | 3.250 |
M9 | 0.655 | 1.7 | 6.387 | 0.625 | 1.7 | 6.777 |
M10 | 0.230 | 3.1 | 14.332 | 0.232 | 3.1 | 10.342 |
I3 | 0.230 | 3.2 | 13.930 | 0.232 | 3.1 | 10.050 |
M11 | 0.371 | 2.2 | 8.398 | 0.371 | 2.2 | 7.199 |
DNB | DNB | 0.982 | 0.8 | 3.361 | 0.944 | 0.9 | 3.110 |
Table 3.
Statistics of monthly DCC reflectance (after the ADM adjustment) mean and mode time series for all RSB M-bands (M1–M5, M7–M11, March 2012–September 2015), DNB (May 2013–September 2015), and RSB I-bands (I1–I3, January 2014–September 2015).
Table 3.
Statistics of monthly DCC reflectance (after the ADM adjustment) mean and mode time series for all RSB M-bands (M1–M5, M7–M11, March 2012–September 2015), DNB (May 2013–September 2015), and RSB I-bands (I1–I3, January 2014–September 2015).
Band | DCC Mode | DCC Mean |
---|
Avg | sd (%) | Max−Min (%) | Avg | sd (%) | Max−Min (%) |
---|
VIS/NIR | M1 | 0.950 | 0.8 | 3.157 | 0.911 | 0.8 | 2.961 |
M2 | 0.942 | 0.8 | 4.461 | 0.902 | 1.0 | 4.782 |
M3 | 0.938 | 0.8 | 4.158 | 0.897 | 0.9 | 4.232 |
M4 | 0.908 | 0.6 | 2.972 | 0.868 | 0.7 | 3.606 |
M5 | 0.936 | 0.4 | 1.924 | 0.894 | 0.6 | 2.756 |
I1 | 0.898 | 0.5 | 2.004 | 0.856 | 0.8 | 3.407 |
M7 | 0.924 | 0.4 | 1.299 | 0.891 | 0.7 | 2.803 |
I2 | 0.924 | 0.4 | 1.624 | 0.889 | 0.8 | 3.227 |
SWIR | M8 | 0.656 | 1.1 | 3.938 | 0.650 | 1.1 | 3.378 |
M9 | 0.611 | 1.9 | 7.860 | 0.582 | 1.8 | 6.628 |
M10 | 0.214 | 3.5 | 14.478 | 0.216 | 3.3 | 10.862 |
I3 | 0.214 | 3.3 | 14.037 | 0.216 | 3.3 | 11.024 |
M11 | 0.345 | 2.6 | 10.427 | 0.346 | 2.5 | 8.065 |
DNB | DNB | 0.917 | 0.5 | 2.291 | 0.881 | 0.8 | 3.701 |
The monthly DCC statistics derived from this study also generally agree with previous studies in the SWIR spectrum [
17,
20]. DCCs are less reflective in the SWIR spectrum, with the averaged monthly DCC mean reflectance ranging from ~0.2 to ~0.7 for VIIRS band M8–M11 and I3. DCCs are also less stable in the SWIR spectrum. The standard deviations of ranges of SWIR bands monthly mean and mode of DCC TOA reflectance are much larger than those for the VIS/NIR bands in majority of cases (see
Table 2). Our results indicate that the Hu
et al. [
15] ADM used in this study is ineffective for the SWIR bands. The ADM-adjusted DCC reflectance has larger variance than the DCC TOA reflectance before the ADM adjustment (see
Table 2 and
Table 3). Therefore, the DCC TOA reflectance without ADM adjustment was used to characterize the VIIRS SWIR bands calibration stability. More effort is needed to develop ADM for the SWIR bands in the future. In-depth analysis of VIIRS SWIR bands DCC time series are presented in
Section 4.3 and
Section 4.4.
5. Summary and Conclusions
The NOAA operational VIIRS RSBs and DNB calibration stability using the DCC technique is demonstrated in this study. Monthly DCC time series for bands M1–M5, M7–M11, DNB, I1-I3 were developed and extensively analyzed. The mode of monthly DCC reflectance (after ADM-adjustment) was used for calibration stability monitoring for the VIS/NIR bands and DNB. Our results show that the NOAA operational radiometric calibration for bands M5 and M7 are generally stable, with stabilities of 0.4% and ranges less than 1.9% during the entire time period. The stabilities of bands M1–M4 are 0.6%–0.8%, with ranges of 3.0%–4.5%. Large fluctuations in bands M1-M4 monthly DCC reflectance were observed since early 2014, correlated with F-factor trend changes during this period. DNB is stable since May 2013 (after the RSR update), similar to bands M5 and M7. The calibration stability of DNB is 0.5% from May 2014 to September 2015. The mean of monthly DCC TOA reflectance (without ADM adjustment) was used for the SWIR bands calibration stabiwality monitoring. The calibration stabilities for M8–M11 are from 1.0% to 3.1%, similar to the natural DCC temporal variability at the SWIR spectrum. The calibration stabilities of I1-I3 are very close to those of their spectral equivalent M-bands (M5, M7, and M10).
VIIRS RSBs inter-channel relative calibration stability was analyzed using DCC mean band ratio time series. I1/M5, I2/M7, and I3/M10 bands ratio time series show consistent calibrations for the 3 band pairs from January 2014 to September 2015. The May 2014 C0 = 0 calibration change is clearly observable in the I2/M7 and I3/M10 band ratio time series. Band ratio time series reveal relative calibration changes for M1/M4 and M5/M7, which may have negative impacts on long-term trend analyses for ocean color and vegetation EDRs if the calibration changes are not accounted for.
The DCC time series were also compared with the IDPS F-factor time series, VIIRS validation site time series, and the VIIRS-MODIS SNO-x time series. Comparison results further support that the DCC time series are capable of detecting sub-percent calibration changes in VIIRS VIS/NIR RSB and DNB bands. Semi-monthly DCC time series were also developed and our results show the monthly and semi-monthly DCC time series generally agree with each other. The semi-monthly time series are slightly noisier than the monthly time series, especially for bands M1–M4, in which large F-factor and DCC reflectance fluctuations exist but the DCC reflectance are known to be more invariant. Our results indicate that semi-monthly DCC time series are useful for stability monitoring at higher temporal resolution. The VIIRS DCC time series developed in this study may contribute to the EDR long-term trend studies and the future SNPP VIIRS SDR reprocessing. The monthly DCC time series used in this study are updated each month and available online [
29].