Cross-calibration of MODIS and VIIRS long near infrared bands for ocean color science and applications
Introduction
Satellite ocean color instruments have provided otherwise unattainable data on spatial and temporal trends in the light field emanating from the world's oceans. However, the physical location of these sensors (on orbit) leads to difficulties in (1) discriminating oceanic from atmospheric signals in the measured total radiance signal; (2) calibrating and validating derived geophysical parameters (e.g., chlorophyll-a concentration; Ca); and (3) cross-calibrating multiple sensors towards longer-term datasets with minimal between-sensor uncertainties. Of these, the latter is particularly important in a climate context, as the usable life of any individual sensor (10 to 20 years at the absolute maximum) is likely too short to capture climate-scale variability (e.g., Lee et al., 2010).
Numerous previous works have investigated such climate-scale variability, using either single-sensor (Gregg et al., 2005; Henson et al., 2010; Siegel et al., 2013; Vantrepotte and Mélin, 2011) or merged-sensor (Gregg and Rousseaux, 2014; Lee et al., 2010; Signorini et al., 2015) datasets. These studies largely show no trend or a declining trend in global Ca, with large regional or basin-scale variability. Nevertheless, most of them note difficulties in statistical assessments of their respective datasets, owing to uncertainties in merging data from multiple sensors, or in drawing conclusions from datasets of too short duration. For example, Signorini et al. (Signorini et al., 2015) considered 16 years of MODIS/A [Moderate Resolution Imaging Spectroradiometer onboard Aqua] and SeaWiFS [Sea-viewing Wide Field-of-View Sensor onboard OrbView2] data. While they note general agreements between trends as derived using these two sensors (and a merged-sensor dataset), data from South Atlantic showed a positive trend using 11 years of SeaWiFS data, but a negative trend using either 11 years of MODIS data or the merged-sensor dataset. Gregg and Rousseaux (Gregg and Rousseaux, 2014) noted that Ca time series that switch from SeaWiFS to MODIS data always showed a negative trend due to a bias between the sensors, and thereby recommended other techniques to reduce cross-sensor differences (Gregg and Casey, 2010). In contrast, for waters offshore China, Zhang et al. (Zhang et al., 2006) noted strong continuity (with no significant bias) in Ca between the same two sensors.
Such cross-sensor comparisons, by definition, require aggregations of large quantities of data, with globally and/or regionally averaged statistics obscuring the complexities of the differences between sensors (Djavidnia et al., 2010; Mélin, 2010). Indeed, Djavidna et al. (Djavidnia et al., 2010) found overall continuity between MODIS- and SeaWiFS- derived Ca, but noted significant modulations of this relationship both regionally and seasonally. Similarly, Melin et al. (Mélin et al., 2016) found spatiotemporal patterns of discontinuities between SeaWiFS, MODIS, and MERIS [Medium Resolution Imaging Spectrometer onboard Envisat] remote sensing reflectance (Rrs, in sr−1) products. For demonstration of these discontinuities between MODIS and VIIRS/SNPP [Visible-Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (SNPP)], Fig. 1 shows the percent difference in Rrs and Ca as calculated using NASA default processing routines (specifically, [(MODIS-VIIRS)/MODIS], see Section 2.4. for processing details). Numerous potential sources of such discontinuities have been offered, including temporal differences in overpass times (Zhang et al., 2006), data quantity (Barnes and Hu, 2015), viewing geometry (Barnes and Hu, 2016), solar geometry (Djavidnia et al., 2010; Mélin et al., 2016; Zibordi et al., 2012), as well as Ca and aerosol optical thickness (Zibordi et al., 2012).
Amidst these potential sources of uncertainty, a fundamental assumption in system vicarious calibration (SVC) and atmospheric correction of satellite ocean color data concerns the long near infrared (NIRL) band. In effect, the pre-launch calibration of the NIRL band is considered sufficient for NASA's current operational SVC and AC procedures. This is supported by findings of Wang and Gordon (Wang and Gordon, 2002) that, based on radiative transfer simulations, moderate (±5%) errors in NIRL calibration have minimal impacts (2–3%) on subsequently derived remote sensing reflectance (Rrs) in the visible wavelengths. As such, the gain (also termed g-factor; g) for the NIRL band on any given sensor is not vicariously calibrated after launch, and is assigned a value of 1.0, with the radiance from this band then used “as-is” for calculation of atmospheric aerosol contributions to the measured total radiance prior to vicarious calibration of all other bands. While this overarching assumption may be true when performance is evaluated using discrete data points from either simulations or field measurements, recent research suggests that operational uncertainties resulting from this assumption are not equally distributed in space or time (Barnes et al., 2020).
The focus of this work is the intersection of noted cross-sensor differences in space and time (e.g., Fig. 1; Djavidnia et al., 2010; Mélin et al., 2016) and the sensitivity of Rrs(VIS) to g(NIRL) within individual sensor datasets (Barnes et al., 2020). As such, the overall objective of this work is to cross-calibrate the NIRL bands of MODIS/A and VIIRS/SNPP using their on-orbit measurements over global oceans, specifically scaling VIIRS/SNPP to match MODIS/A. Throughout this process, we seek to objectively identify characteristics of MODIS / VIIRS pixel pairs that are appropriate for such cross-calibration. Following this, we assess the impacts of cross-calibrating these NIRL bands on the continuity of downstream ocean color products including Rrs and Ca (Hu et al., 2012; O'Reilly et al., 2000), with the hypothesis that improving NIRL cross-calibration will result in greater continuity of downstream products. A fundamental question behind this work is, if MODIS were to replace VIIRS on SNPP (or vice versa, if VIIRS were to replace MODIS on the Aqua satellite), would they measure identical top-of-atmosphere (TOA) radiance over the same ocean pixels?
Section snippets
Cross-calibration approach
Data from five ocean gyres, representing the “clearest ocean waters” (Morel et al., 2010) were used in this study (Fig. 2). Compared to coastal waters, gyre targets are desirable for cross-calibration work as they more likely (1) include negligible water leaving radiance in the NIR (i.e., black-pixel assumption; Gordon and Wang, 1994), (2) are spatially homogeneous, (3) are overlain by primarily marine aerosols, and (4) are less optically complex than coastal waters, reducing uncertainties in
Results
While SSV were identified in each of the 14 overpass matches (Fig. 3), the quantity of SSV data differed greatly by gyre. Only two dates contained SSV in the NPG, with total data quantity of 4316 matchups. As such, NPG data are often excluded in subsequent analyses and discussion. Despite the SAG having only one overpass match, 40 dates of this overpass match included SSV, with total data quantity of 3.6e5 matchups. This is similar to the data quantity in SPG (4.6e5), where SSV were spread out
Discussion
Our analysis of all matchups which met the initial criteria for SSV showed general agreement between the gyres as to the median Lt(M'862)/Lt(V862) (0.95–0.98), with more substantial variability in MAD (Fig. 6). Of all the parameters tested, tightening of the spatial homogeneity metrics (CV and MMR) showed the most prominent impacts on MAD, with all gyres showing steady decreases with more stringent heterogeneity thresholds (Fig. 7). From this, we determined that an MMR threshold of 1.02 or a CV
Conclusions
In this work, we assessed the cross-sensor consistency of the NIRL bands of two mainstream sensors (MODIS/Aqua and VIIRS/SNPP) while weighing techniques commonly used to identify pixel matchups appropriate for satellite cross-calibration. In doing so, we note metrics for spatial homogeneity and relative azimuth are pivotal towards isolating high-quality simultaneous same view (SSV) matchups. Additionally, we found a ~ 3% difference in the pre-launch calibration of the NIRL bands of the two
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by funding from the United States National Aeronautics and Space Administration (NASA): NNX16AQ71G (BBB and CH), ROSES 17-TASNPP17-0065 (SWB and BAF), and contract 80GSFC20C0044 (NP). The authors wish to thank NASA for providing the data and processing software used in this work, as well as three anonymous reviewers, whose comments led to substantial improvements in this manuscript.
References (54)
- et al.
Dependence of satellite ocean color data products on viewing angles: a comparison between SeaWiFS, MODIS, and VIIRS
Remote Sens. Environ.
(2016) - et al.
Validation of VIIRS and MODIS reflectance data in coastal and oceanic waters: an assessment of methods
Remote Sens. Environ.
(2019) - et al.
Cloud adjacency effects on top-of-atmosphere radiance and ocean color data products: a statistical assessment
Remote Sens. Environ.
(2016) - et al.
On-orbit radiometric characterization of OLI (Landsat-8) for applications in aquatic remote sensing
Remote Sens. Environ.
(2014) - et al.
Uncertainties in coastal ocean color products: Impacts of spatial sampling
Remote Sensing of Environment
(2016) - et al.
Calibration of ocean color scanners: how much error is acceptable in the near infrared? Remote Sens
Environ.
(2002) - et al.
An improved in-situ bio-optical data set for ocean color algorithm development and satellite data product validation
Remote Sens. Environ.
(2005) - et al.
New aerosol models for the retrieval of aerosol optical thickness and normalized water-leaving radiances from the SeaWiFS and MODIS sensors over coastal regions and open oceans
Appl. Opt.
(2010) - et al.
Estimation of near-infrared water-leaving reflectance for satellite ocean color data processing
Opt. Express
(2010) - et al.
Cross-sensor continuity of satellite-derived water clarity in the Gulf of Mexico: insights into temporal aliasing and implications for long-term water clarity assessment
IEEE Trans. Geosci. Remote Sens.
(2015)
Stray light in the SeaWiFS radiometer. NASA tech memo 104566
Sensitivity of satellite ocean color data to system vicarious calibration of the long near infrared band
IEEE Trans. Geosci. Remote Sens.
On Rayleigh optical depth calculations
J. Atmos. Ocean. Technol.
Measurements of molecular absorption spectra with the SCIAMACHY pre-flight model: instrument characterization and reference data for atmospheric remote-sensing in the 230-2380 nm region
J. Photochem. Photobiol. A Chem.
The marine optical buoy (MOBY) radiometric calibration and uncertainty budget for ocean color satellite sensor vicarious calibration
Temperature dependence of the ozone absorption spectrum over the wavelength range 410 to 760 nm
Geophys. Res. Lett.
An approach to cross-calibrating multi-mission satellite data for the open ocean
Remote Sens. Environ.
Comparison of global ocean colour data records
Ocean Sci.
gas_trans.c
Sensor-independent approach to the vicarious calibration of satellite ocean color radiometry
Appl. Opt.
CLARREO Pathfinder / VIIRS Intercalibration: Quantifying the Polarization Effects on Reflectance and the Intercalibration Uncertainty
Remote Sensing
Remote sensing of ocean color: a methodology for dealing with broad spectral bands and significant out-of-band response
Appl. Opt.
Clear water radiances for atmospheric correction of coastal zone color scanner imagery
Appl. Opt.
Retrieval of water-leaving radiance and aerosol optical thickness over the oceans with SeaWiFS: a preliminary algorithm
Appl. Opt.
Improving the consistency of ocean color data: a step toward climate data records
Geophys. Res. Lett.
Decadal trends in global pelagic ocean chlorophyll: a new assessment integrating multiple satellites, in situ data, and models
J. Geophys. Res. C Ocean
Recent trends in global ocean chlorophyll
Geophys. Res. Lett.
Cited by (17)
Leveraging Landsat-8/-9 underfly observations to evaluate consistency in reflectance products over aquatic environments
2023, Remote Sensing of EnvironmentMysterious increases of whiting events in the Bahama Banks
2023, Remote Sensing of EnvironmentDetermining pseudo-invariant calibration sites for comparing inter-mission ocean color data
2022, ISPRS Journal of Photogrammetry and Remote SensingCitation Excerpt :However, in addition to illumination-observation, meteorological, and synchronization effects, a multiplier effect could occur when synchronous image pairs over pseudo invariant calibration sites (PICS) are used for inter-mission comparisons (Mei et al. 2016; Devries et al. 2007). Nevertheless, many practical applications indicate that atmospheric and oceanic conditions in open ocean gyres are relatively stable and homogeneous compared to coastal waters (IOCCG 2006, 2019; Morel et al. 2010), and these gyres could be used as PICS (Barnes et al. 2021; Chen et al. 2020a; Fougnie et al. 2010). However, more evidence is needed to quantitatively support these ideas.
Spectral characteristics of sea snot reflectance observed from satellites: Implications for remote sensing of marine debris
2022, Remote Sensing of EnvironmentCitation Excerpt :The data were processed to generate Rayleigh-corrected reflectance (Rrc(λ), dimensionless) using the SeaDAS software package for OLCI, MODIS, and VIIRS, or the Acolite software package for MSI. These Rrc(λ) data were projected to an equidistant cylindrical projection (Barnes et al., 2021), and then used to generate Red-Green-Blue (RGB) and/or false-colour RGB (FRGB) composite images for visual inspection (Qi et al., 2020). Because sea snots show elevated NIR reflectance, traditional atmospheric correction approaches to obtain surface reflectance (R, dimensionless) will fail over sea snots.
Estimating the water-leaving albedo from ocean color
2022, Remote Sensing of EnvironmentCitation Excerpt :Thus, the use of αw_VIS could provide sufficient understanding of αw_broad contribution to αbroad regarding its spatial-temporal variations. However, further incorporation of αw(λ) in both the UV and NIR domains is certainly preferred for more accurate quantification of αw_broad, especially when VIIRS Rrs(λ) products at both the UV and NIR domains can be acquired (Barnes et al., 2021; Wang et al., 2020). Note that HydroLight simplifies the impact of wind speed on surface reflectance, such as reflectance of wind-induced foams and whitecaps are not considered (Mobley and Sundman, 2016), suggesting that the calculated αVIS from HydroLight simulations would be smaller than that measured in natural waters for the same observational condition.