Impacts of VIIRS SDR performance on ocean color products

Authors

  • Menghua Wang,

    Corresponding author
    1. NOAA/NESDIS Center for Satellite Applications and Research, E/RA3, College Park, Maryland, USA
    • Corresponding author: M. Wang, NOAA/NESDIS Center for Satellite Applications and Research, E/RA3, 5830 University Research Ct., College Park, MD 20740, USA. (menghua.wang@noaa.gov)

    Search for more papers by this author
  • Xiaoming Liu,

    1. NOAA/NESDIS Center for Satellite Applications and Research, E/RA3, College Park, Maryland, USA
    2. CIRA, Colorado State University, Fort Collins, Colorado, USA
    Search for more papers by this author
  • Liqin Tan,

    1. NOAA/NESDIS Center for Satellite Applications and Research, E/RA3, College Park, Maryland, USA
    2. CIRA, Colorado State University, Fort Collins, Colorado, USA
    Search for more papers by this author
  • Lide Jiang,

    1. NOAA/NESDIS Center for Satellite Applications and Research, E/RA3, College Park, Maryland, USA
    2. CIRA, Colorado State University, Fort Collins, Colorado, USA
    Search for more papers by this author
  • SeungHyun Son,

    1. NOAA/NESDIS Center for Satellite Applications and Research, E/RA3, College Park, Maryland, USA
    2. CIRA, Colorado State University, Fort Collins, Colorado, USA
    Search for more papers by this author
  • Wei Shi,

    1. NOAA/NESDIS Center for Satellite Applications and Research, E/RA3, College Park, Maryland, USA
    2. CIRA, Colorado State University, Fort Collins, Colorado, USA
    Search for more papers by this author
  • Kameron Rausch,

    1. Earth and Climate Science Directorate, The Aerospace Corporation, Los Angeles, California, USA
    Search for more papers by this author
  • Kenneth Voss

    1. Physics Department, University of Miami, Coral Gables, Florida, USA
    Search for more papers by this author

Abstract

[1] One of the primary goals for the Visible Infrared Imaging Radiometer Suite (VIIRS) on board the Suomi National Polar-orbiting Partnership is to provide the science and user communities with the data continuity of the Environmental Data Records (EDR) (or Level-2 products) over global oceanic waters for various research and applications, including assessment of climatic and environmental variations. The ocean color EDR is one of the most important products derived from VIIRS. Since ocean color EDR is processed from the upstream Sensor Data Records (SDR) (or Level-1B data), the objective of this study is to evaluate the impact of the SDR on the VIIRS ocean color EDR. The quality of the SDR relies on prelaunch sensor characterizations as well as on-orbit radiometric calibrations, which are used to develop the sensor F-factor lookup tables (F-LUTs). VIIRS F-LUTs derived from solar and lunar calibrations have been used in processing data from the VIIRS Raw Data Records (or Level-0 data) to SDR. In this study, three sets of F-LUTs with different generation schemes have been used to reprocess the SDR and then the ocean color EDR for product evaluations. VIIRS ocean color products are compared with in situ data from the Marine Optical Buoy and products from the Moderate Resolution Imaging Spectroradiometer (MODIS) on the satellite Aqua. It is found that the data quality of VIIRS operational ocean color products before 6 February 2012 is poor due to the inappropriate use of the at-launch F-LUTs for the SDR calibration, and that the recently updated VIIRS F-LUTs have significantly improved the SDR and ocean color EDR. Using reprocessed SDR with updated F-LUTs and including vicarious calibration, VIIRS ocean color EDR products are consistent with those from MODIS-Aqua in global deep waters. Although there are still some significant issues with VIIRS ocean color EDR, e.g., poor data quality over coastal regions, our results demonstrate that VIIRS has great potential to provide the science and user communities with consistently high-quality global ocean color data records that are established from heritage ocean color sensors such as MODIS-Aqua.

1 Introduction

[2] Satellite ocean color remote sensing products have long been used to study global ocean and atmospheric processes, such as ocean's global-scale biological and biogeochemical variability [Behrenfeld et al., 2006; Behrenfeld et al., 2001; Chavez et al., 1999; Shi and Wang, 2010], ocean response to a short-term weather event [Liu et al., 2009; Shi and Wang, 2007a, 2009b; Walker et al., 2005], effects of volcanic eruption on ocean environmental changes [Shi and Wang, 2011], mesoscale ocean processes [Chelton et al., 2011; Cipollini et al., 2001], phytoplankton blooms in the open ocean [Babin et al., 2004], floating green algae blooms [Hu, 2009; Shi and Wang, 2009a], sea surface temperature variability and ocean circulation [Nakamoto et al., 2000; Nakamoto et al., 2001; Subrahmanyam et al., 2008], ocean property variation in the Korean dump site of the Yellow Sea [Son et al., 2011], coastal environment changes and monitoring [Hu et al., 2004; Nezlin et al., 2008; Shi and Wang, 2009a; Son and Wang, 2012; Son et al., 2011; Warrick and Fong, 2004], harmful algae blooms [Carvalho et al., 2011; Stumpf et al., 2009; Tang et al., 2004], and inland fresh water environmental variations [Hu et al., 2010; Wang et al., 2011]. Much of the credit must go to the successes of various satellite ocean color missions including NASA's Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and Moderate Resolution Imaging Spectroradiometer (MODIS) on the Terra and Aqua, and the European Space Agency's (ESA) Medium-Resolution Imaging Spectrometer (MERIS) on the Envisat. However, both SeaWiFS and MERIS had stopped collecting data, and MODIS sensors are long past their expected lifetime.

[3] The Suomi National Polar-orbiting Partnership (SNPP) satellite was successfully launched into an 824 km sun-synchronous polar orbit on 28 October 2011. The satellite crosses the equator at around 13:30 local time. SNPP carries the Visible Infrared Imaging Radiometer Suite (VIIRS) [Schueler et al., 2002], a 22 band visible/infrared sensor that combines most features of the NASA ocean color sensors SeaWiFS and MODIS, the National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer, and the Defense Meteorological Satellite Program Operational Linescan System. One of the primary goals for the VIIRS mission is to provide the data continuity for the science community with Environmental Data Records (EDR) (or Level-2 products) over global oceanic waters to enable assessment of climatic and environmental variability [McClain, 2009; McClain et al., 2004]. The ocean color EDR is one of key product suites derived from VIIRS.

[4] The VIIRS ocean color EDR data processing uses VIIRS visible and near-infrared (NIR) moderate resolution (M) bands to derive normalized water-leaving radiance (nLw(λ)) [Gordon and Wang, 1994; IOCCG, 2010; Wang, 2007], chlorophyll-a (Chl-a) concentration [O'Reilly et al., 1998; O'Reilly et al., 2000], and inherent optical properties [Carder et al., 1999; Carder et al., 1991; Lee et al., 2002; Maritorena et al., 2002]. VIIRS has most of the capabilities of MODIS [Esaias et al., 1998; Salomonson et al., 1989], but offers a wider swath width and higher spatial resolution. Table 1 shows the spectral band nominal center wavelength and bandwidth for VIIRS and MODIS. The VIIRS operational ocean color data are officially processed by the NOAA Interface Data Processing System (IDPS), which is also responsible for processing all other atmospheric and land products, as well as products from the other sensors on board SNPP. There have been some preliminary evaluations of the VIIRS ocean color products [Arnone et al., 2012; Hlaing et al., 2013; Turpie et al., 2013]. In January 2013, VIIRS ocean color products from IDPS reached the beta status. Thus, the public can now access the VIIRS operational IDPS ocean color EDR through the NOAA Comprehensive Large Array-data Stewardship System (CLASS) (www.class.ngdc.noaa.gov).

Table 1. Comparison of the Ocean Color and Other Useful Spectral Bands for VIIRS and MODISa
VIIRSMODIS
  1. a

    Note that only MODIS bands corresponding to VIIRS are listed.

BandBand Center (nm)Bandwidth (nm)Band Center (nm)Bandwidth (nm)
M14102041215
M24431844310
M34862048810
M45512055110
M56712066710
M67451574810
M78624186915
M8123820124020
M9137815137530
M10161060164024
M11225050213050

[5] Since the ocean color EDR is processed from the upstream Sensor Data Records (SDR) (or Level-1B data), the performance of the SDR is critical to the performance of the VIIRS ocean color EDR products. VIIRS SDR contains the calibrated radiances and reflectances calculated from the raw data records (RDR) (or Level-0 data). The performance of the SDR relies on on-orbit instrument calibrations, including both solar and lunar calibrations [IOCCG, 2012]. Currently, solar calibration is the primary method of monitoring the on-orbit radiometric performance of the reflective solar bands (RSBs). VIIRS observes sunlight reflected by the Solar Diffuser (SD), and the change in radiometric sensitivity over time is computed from the SD data (the solar calibration scale factors, or F-factors). The solar calibration is also verified by off-line lunar calibrations. However, there are some noticeable discrepancies between VIIRS solar and lunar calibrations, particularly for VIIRS blue (M1–M3) bands [Xiong et al., 2013]. In addition, the Solar Diffuser Stability Monitor (SDSM) simultaneously observes the sun and the solar diffuser to monitor changes in the diffuser bidirectional reflectance distribution function (BRDF) over time (the H-factors). The SD time series must be corrected by the SDSM-derived BRDF change to yield the actual change in the instrument response. This solar calibration method yields a calibration of the instrument on a per band/per detector basis for the sensor two gain states and two mirror sides. The most up-to-date F-factors are generated periodically and stored in lookup tables (LUTs) that are used by the SDR data processing [JPSS-ATBD, 2012a].

[6] The Joint Polar Satellite System (JPSS) IDPS data processing system has been producing VIIRS RDR, SDR, and EDR since 21 November 2011, when the nadir door of the instrument was opened. However, VIIRS operational configurations, including the schemes to generate the F-factor LUTs (F-LUTs hereafter), have changed several times since the beginning of the mission due to various issues, such as significant sensor degradation in the NIR and shortwave infrared (SWIR) bands and incorrect at-launch onboard-calibration F-LUTs, which significantly affected the SDR data quality. In this study, the effect of the SDR performance on ocean color EDR based on the three different sets of F-LUTs (with different generation schemes) received from the VIIRS SDR team has been evaluated. With these F-LUTs, the SDR was reprocessed from IDPS RDR using the Algorithm Development Libraries (ADL) software package, and then the ocean color EDR products were reprocessed using both ADL and the NOAA Multi-Sensor Level-1 to Level-2 (NOAA-MSL12) ocean color science data processing system, for comparison and evaluation. Detailed descriptions of NOAA-MSL12 and ADL are given in section 2.2.

[7] To evaluate the impacts of SDR performance on EDR, nLw(λ) at VIIRS M1–M5 bands and Chl-a data produced by the NOAA-MSL12 and ADL are compared with in situ data from the Marine Optical Buoy (MOBY) in the waters off Hawaii [Clark et al., 1997]. The NOAA-MSL12 and ADL-generated EDR data are also compared with MODIS-Aqua data in four selected regions: Hawaii, South Pacific Gyre, the U.S. East Coast, and the Gulf of Mexico coastal site, which represent both open ocean and coastal waters. Since one of the important goals of the VIIRS ocean color EDR processing is to provide the ocean color community with continuous and consistent global data records established from the heritage sensors, e.g., SeaWiFS and MODIS, the VIIRS ocean color EDR is also compared with MODIS-Aqua data at global deep waters (>1000 m). The ultimate goal of this study is to evaluate and further improve the IDPS ocean color EDR.

2 Data and Method

2.1 Satellite Data

2.1.1 IDPS-Produced VIIRS RDR/SDR/EDR Data

[8] The SNPP IDPS-produced VIIRS RDR (Level-0), SDR (Level-1B), and ocean color EDR (Level-2) data are downloaded routinely from the SNPP central technical support infrastructure Government Resource for Algorithm Verification, Independent Testing and Evaluation (GRAVITE) and the NOAA CLASS. The IDPS RDR data contain raw digital numbers from Earth View observations and is the starting point of all data processing using the ADL and NOAA-MSL12 in this study. The F-LUTs are fundamental in the IDPS processing from RDR to SDR. The methods to calculate F-LUTs, as well as F-LUTs themselves, have been changed several times since the beginning of the SNPP VIIRS mission. The VIIRS IDPS SDR data include the sensor-measured top-of-atmosphere radiance/reflectance from M1–M11 (Table 1), geo-location data, cloud mask intermediate product [JPSS-ATBD, 2012b], bright pixel and onboard-calibration intermediate products. With the SDR, IDPS produces the ocean color EDR data files containing nLw(λ) spectra at VIIRS M1–M5 bands, Chl-a, and inherent optical properties (absorption and backscattering coefficients) at VIIRS M1–M5 bands, as well as various relevant quality flags.

2.1.2 ADL-Reproduced VIIRS SDR Data

[9] The JPSS ADL provides an algorithm framework to develop and execute existing IDPS algorithms or new IDPS-compatible algorithms. ADL contains the same codes as the IDPS to process VIIRS data from RDR to SDR, and SDR to EDR. Thus, the ADL data processing package helps scientists and algorithm developers to develop, test, and integrate new algorithm codes into IDPS more rapidly and efficiently. In this study, ADL version 4.1 is used to reprocess VIIRS SDR and ocean color EDR with improved F-LUTs and algorithms for data analysis and evaluation.

[10] The primary calibration constant for the VIIRS RSBs is the solar calibration scale factors (F-factors), which represent the system response variation or sensor gains changes. The F-factor is calculated in postprocessing for each detector in all RSBs, trended over time, and used to generate periodic updates for F-LUTs. VIIRS IDPS operational SDR data were produced with poor quality before 6 February 2012 due to unexpected issues and problems, e.g., a significant degradation anomaly in the VIIRS NIR and SWIR bands caused by a tungsten oxide contaminant in the mirror, some existing problems in calibration parameters, and particularly incorrect use of the at-launch F-LUTs for RSB calibration in the early postlaunch period. Consequently, the data quality of VIIRS ocean color EDR was poor before 6 February 2012. More frequent calibration strategies for updating F-LUTs (weekly/daily LUTs update, and the scan-by-scan automatic RSB calibration approach) were tested and adopted to mitigate the effect of the significant sensor NIR degradation. Some other major improvements to the VIIRS SDR include the implementation of the new scan-by-scan automatic online RSB radiometric calibration using a quadratic polynomial in the IDPS operational process. It also uses improved SD and SDSM attenuation screens transmission LUTs.

[11] The IDPS operational F-LUTs used in the official SDR data processing and the three sets of updated daily F-factor LUTs provided by the VIIRS SDR Team are briefly described below:

  1. [12] F-LUT-0 (IDPS F-LUTs): F-factor LUTs used in the IDPS operational VIIRS SDR data process (Collection Short Name (CSN): VIIRS SDR F-LUT and VIIRS SDR F-PREDICTED-LUT; source: NOAA Center for Satellite Applications and Research (STAR) VIIRS SDR Team archive). Before 8 August 2012, the VIIRS SDR F-LUTs were static F-factor tables generated off-line and updated weekly in the IDPS SDR processing. Beginning from 9 August 2012, the F-factors were calculated online scan-by-scan based on the VIIRS SDR F-PREDICTED-LUTs, which were generated off-line and updated weekly.

  2. [13] F-LUT-1: Aerospace updated (postprocessed) daily static F-factor LUTs (CSN: VIIRS SDR F-LUT; source: CasaNOSA (svn/viirs_early_release/trunk/aero/matlab/F_LUT/); creation time: 17 May 2012 to 24 September 2012). This set of VIIRS SDR F-LUTs was generated off-line with an introduction of a new SDSM screen transmission LUT derived from the yaw maneuver data, with noise artifacts being removed using Holt-Winters Averaging over 80–100 orbits.

  3. [14] F-LUT-2: Aerospace updated daily-predicted F-factor LUTs that include a corrected SD transmission table and Holt-Winters smoothing (CSN: VIIRS SDR F-PREDICTED-LUT; source: Aerospace, from CasaNOSA (svn/viirs_early_release/trunk/aero/matlab/PredFhw_LUT/); creation time: 30 January 2013). This set of VIIRS SDR F-PREDICTED-LUT contains the information needed to calculate the F-factor on scan-by-scan basis. It was generated off-line based on (a) using the full spectral H-factor curve generated by the SDSM rather than just using the H-factor curve evaluated at the VIIRS band center and (b) some newly updated SDR LUTs (such as the corrected SD transmission table).

  4. [15] F-LUT-3: Aerospace updated daily-predicted F-factor LUTs that include a new modulated sensor relative spectral response (RSR) (CSN: VIIRS SDR F-PREDICTED-LUT; source: Aerospace archive—VIIRS SDR F-PREDICTED-LUTS_orbits154_6524.zip; creation time: 2 April 2013). This set of VIIRS SDR F-PREDICTED-LUT was updated from F-LUT-2. It was generated off-line based on the new modulated RSR LUT that attempts to account for the sensor degradation caused by the tungsten contamination.

[16] Figure 1 provides the time series comparison of the IDPS F-LUTs and the three sets of updated daily F-LUTs for VIIRS bands M1–M5 (Figures 1a–1e) and M7 (Figure 1f). The F-LUTs for VIIRS M6 band have almost the same results as those shown in the M7 band (Figure 1f). These F-factor values are detector-averaged with the high-gain setting and from the instrument half-angle mirror (HAM) side A. Before 6 February 2012, IDPS operational F-LUTs (F-LUT-0) used the at-launch F-factors with the default value of 1.0 for all bands. From 6 February to 19 April 2012, the IDPS F-factors were adjusted significantly and updated more frequently (~weekly). After 20 April 2012, the trending of IDPS F-factor became smoother on a weekly basis and closer to the daily F-LUT-1. But, the F-LUT-1 (which ended on 19 September 2012) still differs remarkably from the updated F-LUT-2, which is the latest updated F-LUTs when we started VIIRS global ocean color data reprocessing in this study. Since the F-LUT-3 is actually very similar to the F-LUT-2 (Figure 1), and its major improvement in the new modulated RSR has little effect on ocean color EDR, it is not analyzed separately in this study. The sensor radiometric onboard calibration was carried out in the RDR to SDR data processing using the F-LUTs. Based on the F-LUT-1 and F-LUT-2, we have reprocessed VIIRS SDR using ADL with postlaunch updated SDR code and LUTs. The reprocessed SDRs are named ADL-SDR-1 and ADL-SDR-2. The F-LUTs and the corresponding SDR and EDR data generated based on these F-LUTs are all listed in Table 2.

Figure 1.

The IDPS F-LUTs and the three sets of updated daily F-LUTs as a function of Julian day in 2012 for VIIRS bands of (a)–(e) M1–M5 and (f) M7, respectively. These F-factor values are detector-averaged with the high-gain setting and from the instrument HAM side A.

Table 2. List of VIIRS SDR and EDR/Level-2 Data Sets in This Study
F-LUTsSDREDR/Level-2Description
F-LUT-0 (IDPS F-LUT)IDPS SDRIDPS EDREDR from IDPS
MSL12-0SDR from IDPS; EDR processed with NOAA-MSL12
F-LUT-1ADL-SDR-1ADL-EDR-1SDR processed by ADL using F-LUT-1; EDR processed with ADL
MSL12-1SDR processed by ADL using F-LUT-1; EDR processed with NOAA-MSL12
F-LUT-2ADL-SDR-2ADL-EDR-2SDR processed by ADL using F-LUT-2; EDR processed with ADL
MSL12-2SDR processed by ADL using F-LUT-2; EDR processed with NOAA-MSL12

2.1.3 NOAA-MSL12- and ADL-Produced VIIRS Ocean Color EDR Data

[17] The NOAA-MSL12 ocean color data processing system is based on the SeaWiFS Data Analysis System (SeaDAS) version 4.6, a multi-sensor comprehensive image analysis package for the processing, display, and analysis of ocean color data (seadas.gsfc.nasa.gov). However, it should be noted that the NOAA-MSL12 has been modified and improved to include (1) the SWIR-based ocean color data processing [Wang, 2007; Wang and Shi, 2007; Wang et al., 2009b], (2) improved aerosol lookup tables and more accurate Rayleigh radiance computations [Wang, 2002, 2005, 2006, 2007], (3) algorithms for detecting absorbing aerosols and turbid waters [Shi and Wang, 2007b], (4) implementation of an ice-detection algorithm for global and regional ocean color data processing [Shi and Wang, 2012a, 2012b; Wang and Shi, 2009], and other improvements, e.g., an approach to improve the performance of MODIS SWIR bands [Wang and Shi, 2012]. The NOAA-MSL12 has also been enhanced to include a function to process ocean color data for SNPP VIIRS. It has been used to routinely produce VIIRS global daily, 8 day, and monthly ocean color products since the instrument door open on 21 November 2011.

[18] In the NOAA-MSL12 software package for the current VIIRS ocean color data processing, the atmospheric correction algorithm uses the Gordon and Wang [1994], which uses two VIIRS NIR bands at 745 and 862 nm for aerosol reflectance estimation and correction [IOCCG, 2010; Wang et al., 2005]. The Chl-a concentration algorithm for VIIRS is OC3V, which is similar to OC3M for MODIS [O'Reilly et al., 1998; O'Reilly et al., 2000], but with coefficients tuned for the VIIRS spectral bands (M2, M3, and M4). In addition, the SWIR-based atmospheric correction algorithm [Wang, 2007] has been implemented in the NOAA-MSL12 package for VIIRS ocean color data processing and is currently in the evaluation process. The input data of NOAA-MSL12 include VIIRS SDR data from M1 to M7, geo-location data, and the onboard-calibration intermediate product. In addition, ancillary input data include ozone concentration, surface atmosphere pressure, sea surface wind speed, and water vapor, which are obtained from the National Center for Environmental Prediction [Ramachandran and Wang, 2011]. The NOAA-MSL12 output products include nLw(λ) at VIIRS spectral bands M1 to M5, Chl-a, and diffuse attenuation coefficient at 490 nm Kd(490) [Lee et al., 2005; Wang et al., 2009a], etc., as well as various Level-2 data quality flags. In this study, NOAA-MSL12 was used to generate ocean color Level-2 data from IDPS SDR and ADL-produced SDR, and the corresponding output data sets are MSL12-0, MSL12-1, and MSL12-2, which are listed in Table 2. In Level-1 to Level-2 data processing using NOAA-MSL12, the vicarious calibration (VC) gains derived from MOBY in situ data were applied [Franz et al., 2007; Gordon, 1998; Wang and Gordon, 2002]. It should be noted that NOAA-MSL12 is used to evaluate, understand, and improve the VIIRS IDPS-produced ocean color products.

[19] The ADL ocean color EDR processing code contains both the VIIRS atmospheric correction algorithm (ACO) and the ocean color and chlorophyll-a (OCC) EDR algorithm [JPSS-ATBD, 2012c], and can be used to process OCC EDR from VIIRS SDR data. Essentially, the ACO and OCC algorithms in ADL are the same as the algorithms in NOAA-MSL12, but are running in the ADL framework. We have modified the ACO and OCC codes in ADL for improved sunglint [Wang and Bailey, 2001] and cloud masking [Robinson et al., 2003] to retrieve more pixels in the ocean color data processing. In addition, the vicarious calibration gains using MOBY in situ data were applied to all OCC EDR processing using the ADL data processing system. To process VIIRS SDR to ocean color EDR using ADL, the geo-location data, various intermediate products, and onboard-calibration data are also required. We have processed ocean color EDR data from the two sets of ADL-produced SDR, and the corresponding outputs are ADL-EDR-1 and ADL-EDR-2 as listed in Table 2. They are evaluated and analyzed in detail in section 4.

2.1.4 VIIRS and MODIS-Aqua Global Level-3 Data

[20] For effective evaluations of VIIRS ocean color data quality, global VIIRS ocean color Level-3 data products are necessary. The Level-3 data processing algorithm is essentially the same as the one used for producing SeaWiFS and MODIS global Level-3 ocean color products [Campbell et al., 1995]. Specifically, in the Level-3 data processing, pixels containing valid Level-2 data are mapped to fixed spatial grids with resolution of 1, 4, or 9 km. The grid elements or bins are arranged in rows beginning at the South Pole. Each row begins at 180° longitude and circumscribes the Earth at a given latitude. Within each bin, statistics of mean for daily and median for 8 day and monthly are accumulated. In this study, NOAA-MSL12 Level-2, ADL-generated EDR, and IDPS-produced EDR are processed into Level-3 data for evaluation. Before the binning process, all standard flags in the ADL/IDPS data processing (e.g., sunglint, high sensor-zenith angle, high solar-zenith angle, etc.) are applied to remove these flagged data in the EDR product data, while three flags (high sunglint, high sensor-zenith angle, and high solar-zenith angle) are applied to the NOAA-MSL12 Level-2 data. The MODIS-Aqua Level-3 data were directly downloaded from the NASA Ocean Biology Processing Group (OBPG) website (oceancolor.gsfc.nasa.gov).

2.2 In Situ Data

[21] In situ radiometric data were measured at the MOBY site [Clark et al., 1997] moored off the island of Lanai in Hawaii (http://moby.mlml.calstate.edu/MOBY-data). The location of the MOBY site is usually in stable, clear-ocean waters with predominantly marine aerosols. The MOBY program has been providing consistently high-quality clear-ocean optics data since 1997, supporting various satellite ocean color missions, e.g., SeaWiFS, MODIS, VIIRS, etc. To evaluate and assess VIIRS SDR and ocean color EDR products, the in situ nLw(λ) measurements at the VIIRS-spectrally-weighted wavelengths beginning in November 2011 to May 2013 were obtained from the NOAA CoastWatch website (http://coastwatch.noaa.gov/moby/). Although some selected MOBY data have been used for the purpose of the on-orbit vicarious calibration [Eplee et al., 2001; Franz et al., 2007; Gordon, 1998; Wang and Gordon, 2002] for both NOAA-MSL12 and ADL VIIRS ocean color data processing, the high-quality MOBY time series data can be used for VIIRS SDR and EDR data quality monitoring. It is particularly useful to evaluate the performance of F-LUTs and the stability of SDR by comparing VIIRS-derived nLw(λ) with those from MOBY in situ measurements. With this purpose, MOBY in situ data are also useful to evaluate IDPS SDR performance as well as its data stability and quality.

3 Evaluation of NOAA-MSL12-Produced Ocean Color Level-2 Products

[22] As discussed earlier, three sets of F-LUTs have been used to process SDR, and then NOAA-MSL12 was used to process ocean color Level-2 products. The improvement in each F-LUTs set and their impacts on the ocean color products have been evaluated, and the feedbacks have been provided to the SDR team for further F-LUTs improvements. Due to the unexpected large degradation anomaly for the NIR bands in the early period after VIIRS launch, the NOAA-MSL12-produced ocean color products have significant errors when using the operational IDPS VIIRS SDR (MSL12-0). The VIIRS ocean color team reported this issue to the SDR team and requested more frequent weekly/daily calibration LUT updates to mitigate the effects of the significant sensor NIR degradation. Consequently, the VIIRS SDR team has implemented the daily F-LUTs since F-LUT-1. Our tests with the NOAA-MSL12 show that the F-LUT-1-based ocean color products (MSL12-1) have significantly reduced the data errors. More aggressive scan-by-scan updates have been implemented in F-LUT-2 and F-LUT-3, as well as in the current IDPS operational F-LUTs. In this section, we mainly focus our discussions on the comparison of two sets of NOAA-MSL12-generated ocean color data: MSL12-0 (based on the operational IDPS SDR) and MSL12-2 (based on the F-LUT-2 and ADL-SDR-2).

[23] The satellite-measured nLw(λ) and Chl-a data were extracted from 1 km resolution Level-3 file using an 11 × 11 bin box for comparison. The MOBY in situ data include nLw(λ) at VIIRS M1–M5 bands, but no in situ Chl-a data. For Chl-a comparison, MOBY in situ nLw(λ) data were used to derive Chl-a data based on the same OC3V algorithm as in the NOAA-MSL12 and ADL (IDPS). In effect, this compares nLw(λ) ratio values between satellite-derived and MOBY in situ-measured data. As expected, the MSL12-0 data have significant bias errors before 6 February 2012 in both nLw(λ) and Chl-a data (Figure 2). nLw(λ) at M2–M5 bands and Chl-a values are biased high compared with the MOBY in situ measurements, and values of nLw(410) from MSL12-0 are biased significantly low (not shown). These data are not useable. Figure 2 provides the time series of VIIRS-derived nLw(λ) at wavelengths of 443 (M2), 486 (M3), and 551 nm (M4), as well as Chl-a data compared with those from MOBY in situ measured (or derived) for the VIIRS period of 1 January 2012 to 20 April 2013. Results in Figure 2 clearly show very poor data quality for VIIRS ocean color products before 6 February 2012, due to the incorrect use of the at-launch F-LUTs for the SDR calibration. From 6 February 2012, corrected/improved LUTs were used in IDPS RDR to SDR data processing, and the noise and bias in VIIRS nLw(λ) and Chl-a data are significantly reduced and the values are reasonable (Figure 2). However, there were still some slightly high anomalies in MSL12-0 data after 6 February 2012, which was resolved by on-orbit vicarious calibration. The NOAA VIIRS ocean color team started working on the vicarious calibration using MOBY in situ data with the NOAA-MSL12 package in early 2012. In the MSL12-0 results, the vicarious calibration gains were applied after 20 April 2012 (indicated as a vertical line in the plots) and show significant ocean color data quality improvements (Figure 2). In fact, vicarious gains were derived using selected MOBY in situ data obtained from 6 February to 1 May 2012. Thus, the effect of the SDR data quality is shown in Figure 2 by comparing results before and after 6 February 2012, while importance of vicarious calibration is demonstrated by comparing results before and after 20 April 2012. Quantitative comparison of the NOAA-MSL12-derived ocean color data (MSL12-0) with MOBY in situ measurements is listed in Table 3. With excluding the data before 6 February 2012 and MOBY data used for deriving VC gains, it shows that average values of satellite/in situ ratio for nLw(λ) at VIIRS M1–M5 bands and Chl-a are 1.061, 1.049, 1.045, 1.055, 1.469, and 1.002, respectively.

Figure 2.

The time series of VIIRS-derived nLw(λ) at wavelengths of (a) 443 (M2), (b) 486 (M3), and (c) 551 nm (M4), as well as (d) Chl-a data compared with those from MOBY in situ measured (or derived) for the VIIRS period of 21 January 2012 to 20 April 2013. Two lines with 6 February 2012 (change the at-launch F-LUTs) and 20 April 2012 (vicarious gains applied) are shown in the plots.

Table 3. Average (Avg), Standard Deviation (SD), and Number of Data (No) of the Ratio of VIIRS/MOBY In Situ Data for nLw(λ) at VIIRS M1–M5 Bands and Chl-ac
ProductsMSL12-0aMSL12-2IDPS EDRbADL-EDR-2
AvgSDNoAvgSDNoAvgSDNoAvgSDNo
  1. a

    Excluding the data before 6 February 2012.

  2. b

    No vicarious calibration applied.

  3. c

    Note that data used in vicarious calibrations are excluded in matchup comparisons.

nLw(410)1.0610.170980.9920.1511011.0370.154870.9930.17998
nLw(443)1.0480.169980.9970.1371010.9880.164871.0140.16198
nLw(486)1.0440.146980.9940.1171011.0000.144871.0040.15398
nLw(551)1.0550.306980.9690.2231011.0560.261870.9800.26097
nLw(671)1.4830.881981.0260.601950.7230.585741.0280.68683
Chl-a1.0040.330980.9440.2741011.0800.322870.8630.29997

[24] The MSL12-2 nLw(λ) and Chl-a data show good consistency with the MOBY in situ measurements since 2 January 2012, and the noise and bias errors before 6 February 2012 are significantly reduced. Figure 3 shows the scatterplot of MSL12-2 nLw(λ) and Chl-a data with MOBY in situ measurements. However, it is noted that the MOBY data used for vicarious calibration are excluded in the comparison. The quantitative comparison of MSL12-2 ocean color data with the MOBY in situ data matchup is also listed in Table 3. Average satellite/in situ ratios for nLw(λ) at M1–M5 bands are 0.992, 0.997, 0.994, 0.969, and 1.026, respectively. Chl-a values are matched almost perfectly with in situ-derived data, and the satellite/in situ-derived Chl-a ratio is 0.944. It should be noted that the vicarious calibration coefficients in MSL12-2 were derived based on ADL-SDR-2 data. Specifically, the gain coefficients are 0.9746, 0.9746, 0.9697, 0.9577, 0.9700, 0.9802, and 1.0 for VIIRS M1–M7, respectively. The MSL12-2 vicarious gains were derived using selected MOBY in situ measurements from 2 January 2012 to 31 January 2013.

Figure 3.

Comparison of VIIRS MSL12-2 ocean color products with MOBY in situ measurements for (a) nLw(410), (b) nLw(443), (c) nLw(486), (d) nLw(551), (e) nLw(671), and (f) Chl-a. Note that the data used for vicarious calibration are not included in the comparison.

[25] The above analysis shows that the ocean color data based on F-LUT-2/ADL-SDR-2 have much better data quality than those from the operational IDPS SDR, and the SDR performance has significant impact on the ocean color EDR. Based on the daily-predicted F-LUTs generation scheme with improved smoothing functions and the H-factor correction, the F-LUT-2 corrected the onboard-calibration error prior to 6 February 2012. In addition, due to significant sensor degradation in the NIR bands in the early VIIRS operation, static F-LUTs in the operational IDPS SDR data processing produced significant noise errors in the VIIRS ocean color EDR. The new method of daily-predicted F-LUTs significantly reduced the data errors. It can be concluded that the VIIRS ocean color products are highly sensitive to the quality of upstream SDR data, and the use of the correct F-LUTs in the VIIRS RDR to SDR data processing is critical to ocean color EDR. Furthermore, for the VIIRS ocean color products to be reliable and consistent, it is required to reprocess the ocean color EDR using ADL from the beginning of the VIIRS mission.

4 Evaluation of IDPS and ADL-Produced Ocean Color EDR

[26] Similar to the analysis in the previous section, in this section, we focus on the analysis of operational IDPS EDR and the ADL-EDR-2 (based on F-LUT-2 and ADL-SDR-2), and compare the two sets of ocean color EDR data with MOBY in situ measurements. To show the consistency of the VIIRS ocean color EDR with long-term ocean color data records established from MODIS, the IDPS and ADL-produced EDR ocean color data are also compared with MODIS-Aqua data in four selected regions and in global deep waters (open ocean).

4.1 Comparison With the MOBY In Situ Data

[27] For the reasons discussed previously, there are no valid data in IDPS-EDR before 6 February 2012. The IDPS-EDR nLw(λ) at M1–M4 bands is in reasonably good agreement with MOBY in situ measurements since 6 February 2012, but nLw(671) (band M5) is significantly biased low. Quantitative comparisons of IDPS-EDR ocean color data with MOBY in situ data matchup are listed in Table 3, i.e., average satellite/in situ ratios for nLw(λ) at M1–M5 bands are 1.037, 0.988, 1.000, 1.056, and 0.723, respectively. Note that no VC gains were applied in IDPS-EDR results.

[28] The ADL-EDR-2 nLw(λ) data are significantly improved (with VC gains applied), especially for nLw(671). Figure 4 shows the scatterplots of ADL-EDR-2 nLw(λ) at bands M1–M5 and Chl-a data compared with the MOBY in situ measurements. Quantitative comparisons of ADL-EDR-2 data with MOBY in situ measurements are listed in Table 3, i.e., average satellite/in situ ratios for nLw(λ) at M1–M5 bands are 0.993, 1.014, 1.004, 0.980, and 1.028, respectively. With the improved SDR (reprocessed), the incorrect onboard-calibration issue before 6 February 2012 has been resolved. Thus, it can be concluded that, same as in the tests with NOAA-MSL12, the F-LUT-2 has significantly improved the data quality of the SDR and ADL-produced ocean color EDR. It should be noted that the vicarious calibration coefficients were applied in deriving ADL-EDR-2 results. These vicarious gains were derived based on the ADL-SDR-2 data using MOBY in situ measurements. Specifically, these gains are 0.9775, 0.9852, 0.9787, 0.9651, 0.9730, 0.9750, and 1.0 for M1–M7, respectively. In the previous analysis, however, the MOBY in situ data used for vicarious calibration are excluded in the comparison.

Figure 4.

Comparison of VIIRS ADL-EDR-2 ocean color products with MOBY in situ measurements for (a) nLw(410), (b) nLw(443), (c) nLw(486), (d) nLw(551), (e) nLw(671), and (f) Chl-a. Note that the data used for vicarious calibration are excluded in the comparison.

4.2 Comparison With MODIS in the Four Selected Regions

[29] nLw(λ) spectra and Chl-a concentration of IDPS EDR and ADL-EDR-2 were also compared with MODIS-Aqua data in four selected regions: Hawaii, South Pacific Gyre (SPG), the U.S. East Coast (USEC), and the Gulf of Mexico coastal site (GOM), which are bounded by 1° × 1° box centered at (21.0°N, 157.0°W), (25.0°S, 119.0°W), (34.0°N, 74.0°W), and (28.0°N, 90.5°W), respectively. Figure 5 provides examples of the scatterplot comparison for VIIRS nLw(λ) at M2–M4 bands and Chl-a data with MODIS-Aqua measurements in the U.S. East Coast region. The regional averages of VIIRS/MODIS ratio and VIIRS-MODIS correlation coefficients of nLw(λ) spectra and Chl-a are listed in Table 4. In general, the IDPS EDR Chl-a data are higher than those of MODIS-Aqua. ADL-EDR-2 Chl-a data are significantly improved and are much closer to MODIS-Aqua data. It also shows a very good correlation with MODIS-Aqua data for seasonal variations. The IDPS-EDR nLw(443) and nLw(486) are generally lower than those of MODIS-Aqua, but nLw(551) is higher. Since VIIRS has slightly different RSR (and band centers) and bandwidths from MODIS-Aqua, nLw(λ) will show some minor differences between the two sensors. With that considered, the nLw(λ) comparison with MODIS is consistent with Chl-a comparison because underestimations of nLw(443) or nLw(486) and overestimations of nLw(551) will result in overestimation of the IDPS-EDR Chl-a data. Thus, the corrections of nLw(443), nLw(486), and nLw(551) in the ADL-EDR-2 make Chl-a converge with MODIS-Aqua data. This is mainly due to the inclusion of the vicarious gains in the ADL data processing as well as improved SDR. However, for coastal regions (e.g., Gulf of Mexico coastal site), results in Table 4 show that there are still some significant issues with IDPS/ADL ocean color data processing. In fact, IDPS/ADL does not have the capability now to deal with productive or turbid ocean waters [JPSS-ATBD, 2012c]. The NIR water-leaving radiance correction algorithm for coastal waters such as those reported by Bailey et al. [2010], Ruddick et al. [2000], Siegel et al. [2000], Stumpf et al. [2003], and Wang et al. [2012] has not yet been implemented in IDPS/ADL.

Figure 5.

Scatterplot comparisons of VIIRS-derived ocean color products with MODIS-Aqua measurements in the U.S. East Coast region for (a) nLw(λ) at M2–M4 bands and (b) Chl-a. In Figure 5a, VIIRS data were from ADL with vicarious gains applied, while Figure 5b shows both VIIRS data from IDPS (no vicarious gains) and ADL (with vicarious gains).

Table 4. Regional Averages of Ratio for VIIRS/MODIS and VIIRS-MODIS Correlation of nLw(λ) at VIIRS M1–M5 Bands and Chl-a in Hawaii, South Pacific Gyre (SPG), U.S. East Coast (USEC), and Gulf of Mexico Coastal Site (GOM) Regions
RegionHawaiiSPGUSECGOM
ProductIDPS EDRADL-EDR-2IDPS EDRADL-EDR-2IDPS EDRADL-EDR-2IDPS EDRADL-EDR-2
nLw(410)Mean Ratio0.9410.9170.9450.9220.9300.8670.8540.870
Correlation0.5580.5900.8750.9130.9370.9130.6380.802
nLw(443)Mean Ratio0.8710.9000.8990.8880.8420.8730.8290.899
Correlation0.6390.6960.8220.8930.9420.9490.6330.852
nLw(486)Mean Ratio0.9510.9720.9590.9360.9330.9420.9460.986
Correlation0.6950.6640.6960.8090.9050.9090.6410.768
nLw(551)Mean Ratio1.0991.1201.1270.9731.1811.0761.1691.137
Correlation0.3910.0830.1170.0700.1240.0420.2790.193
nLw(671)Mean Ratio0.4372.0700.3780.5520.7541.0140.7251.729
Correlation0.0660.3690.1950.1500.0980.0290.0220.064
Chl-aMean Ratio1.1321.0931.3950.8661.3611.0681.3701.112
Correlation0.1860.0830.5770.3360.8970.8740.8500.836

4.3 Global Image Comparisons

[30] To understand how VIIRS ocean color products compared with MODIS-Aqua on a global scale, VIIRS global Level-3 composite images were generated for visual inspection and comparison. Figure 6 provides the ADL-EDR-2 global nLw(443) monthly composite images for January 2012 (Figure 6a) and July 2012 (Figure 6c) compared with the corresponding monthly results from MODIS-Aqua (Figures 6b and 6d). From a qualitative visual inspection, VIIRS global nLw(443) images are consistent with those from MODIS-Aqua, with similar features in global spatial nLw(443) distributions for January and July 2012. Figure 7 shows comparisons of global Chl-a distributions between VIIRS from ADL-EDR-2 data and MODIS-Aqua. Figures 7a, 7c, 7e, and 7g are VIIRS-derived global Chl-a images for months of January, April, July, and October of 2012, respectively, while Figures 7b, 7d, 7f, and 7h are the corresponding global Chl-a monthly images derived from MODIS-Aqua. Results in Figure 7 show that both VIIRS and MODIS-Aqua produced very similar global Chl-a maps for the months of January (winter), April (spring), July (summer), and October (fall) in 2012, e.g., showing low Chl-a data in mid-Atlantic and South Pacific Gyre, highs in high latitude of the Northern hemisphere and equatorial regions. However, it is noticed that data over all inland lakes are masked out in VIIRS-derived ocean color products, while there are retrievals in these regions from MODIS-Aqua, e.g., the Great Lakes. This issue has already been identified and resolved in the current IDPS data processing.

Figure 6.

(a and c) VIIRS-derived (ADL-EDR-2) global nLw(443) monthly composite images compared with those from (b and d) MODIS-Aqua for cases of (Figure 6a) January 2012 for VIIRS, (Figure 6b) January 2012 for MODIS-Aqua, (Figure 6c) July 2012 for VIIRS, and (Figure 6d) July 2012 for MODIS-Aqua.

Figure 7.

(a, c, e, g) VIIRS-derived (ADL-EDR-2) global Chl-a monthly composite images compared with those from (b, d, f, h) MODIS-Aqua for cases of (Figures 7a and 7b) January 2012, (Figures 7c and 7d) April 2012, (Figures 7e and 7f) July 2012, and (Figures 7g and 7h) October 2012 for VIIRS and MODIS-Aqua, respectively.

4.4 Chl-a Data From Global Deep Waters

[31] VIIRS ocean color EDR is also compared with MODIS-Aqua data at global deep (>1000 m) waters to show the consistency of ocean color products from the two sensors. Five months of global ADL-EDR-2 data were reprocessed for January, April, July, and October 2012, as well as January 2013, and are intended to capture the seasonal variations. Only 5 months of global ocean color data were reprocessed with improved ADL-EDR-2 data due to the limitation in processing time. Since the IDPS cloud mask intermediate product has no valid data until 20 January 2012, we have modified the ADL code to use the heritage cloud mask [Robinson et al., 2003; Wang and Shi, 2006] for the EDR data reprocessing. Figure 8 shows the comparison of IDPS-EDR (red) and ADL-EDR-2 (black) with MODIS-Aqua (blue) for daily mean Chl-a in the global deep waters since January 2012. MODIS-Aqua has data available during the entire period of time, but the IDPS has no valid data before 6 February 2012. Results show that the IDPS-EDR significantly overestimates Chl-a in global deep waters due to errors in IDPS nLw(λ). With improved SDR data quality and particularly with vicarious calibration gains applied, Chl-a data from ADL-EDR-2 global deep water show a very good agreement with MODIS-Aqua in terms of mean, variation, and data correlation. Thus, all evaluation results show that VIIRS has great potential to provide the ocean community with consistent global ocean color data records established from heritage ocean color sensors. Furthermore, it is also demonstrated that, to fully utilize the ocean color data from the beginning of the VIIRS mission, VIIRS data need to be reprocessed from the RDR to SDR using updated F-LUTs and then from SDR to EDR including vicarious calibration coefficients.

Figure 8.

Time series comparison of daily mean chlorophyll-a concentration from global deep waters (>1000 m) with data from VIIRS IDPS-EDR (red), VIIRS ADL-EDR-2 (black), and MODIS-Aqua (blue).

5 Conclusions

[32] In this study, impacts of the VIIRS SDR performance on the ocean color EDR products have been assessed and evaluated. The performance of the SDR relies on prelaunch sensor characterization and on-orbit radiometric calibrations. Currently, solar calibration is the primary method of radiometric calibrations for VIIRS, and it is maintained by F-LUTs in the VIIRS RDR to SDR data processing. Three sets of F-LUTs (one from the IDPS and other two from the VIIRS SDR team) were used to reprocess SDR, and the reprocessed SDR was then used to produce VIIRS ocean color EDR with both NOAA-MSL12 and ADL data processing systems. To address various issues, the IDPS F-LUTs have been changed several times since the beginning of the VIIRS mission, and the two sets of F-LUTs received from the VIIRS SDR team have greatly improved data generation schemes and increase the update frequency. The reprocessed ocean color EDR is compared with MOBY in situ measurements and MODIS-Aqua ocean color products. The comparison with the MOBY in situ data demonstrated that the ocean color EDR products are highly sensitive to the SDR data quality (as expected), and that the most recent F-LUTs (F-LUT-2 received on 30 January 2013) significantly improved nLw(λ) and Chl-a in both NOAA-MSL12 and ADL-produced Level-2/EDR ocean color products.

[33] The IDPS-EDR and ADL-produced EDR products were also compared with MODIS-Aqua data in four selected regions and the global deep waters. Comparison results show that, with vicarious calibration, the most recently received F-LUTs from the VIIRS SDR team significantly improved the ocean color EDR products, and that ocean color EDR quality in the open ocean is consistent with MODIS-Aqua. Thus, it can be concluded that VIIRS has great potential to provide science and user communities with consistent global ocean color data records established from SeaWiFS and MODIS. However, it should be noted that there are still some important issues and problems with VIIRS SDR and ocean color EDR, e.g., some large discrepancies between solar and lunar calibrations, poor data quality over coastal regions, incorrect and inappropriate IDPS ocean color EDR flags, etc. Significant efforts are required to address these issues in order to have high-quality VIIRS ocean color products consistent with those from SeaWiFS and MODIS.

Acknowledgments

[34] The work was supported by the Joint Polar Satellite System (JPSS) funding. We thank two anonymous reviewers for their useful comments. We thank the MOBY team for providing the in situ data. The MODIS-Aqua data were from NASA OBPG ocean color website. The views, opinions, and findings contained in this paper are those of the authors and should not be construed as an official NOAA or U.S. Government position, policy, or decision.

Ancillary