OSA's Digital Library

Spotlight on Optics

Spotlight on Optics


  • November 2012

Optics InfoBase > Spotlight on Optics > Dynamic range and sensitivity requirements of satellite ocean color sensors: learning from the past

Dynamic range and sensitivity requirements of satellite ocean color sensors: learning from the past

Published in Applied Optics, Vol. 51 Issue 25, pp.6045-6062 (2012)
by Chuanmin Hu, Lian Feng, Zhongping Lee, Curtiss O. Davis, Antonio Mannino, Charles R. McClain, and Bryan A. Franz

Source article Abstract | Full Text: XHTML | Full Text: PDF

Spotlight summary: Routine measurements of sunlight reflected from the ocean have been made continuously from space since 1997 by a variety of US, European, Indian, Chinese and South Korean instruments. These imaging spectroradiometers in low Earth and geosynchronous orbit have provided frequent, wide field of view observations of the global ocean, offering unprecedented opportunities to study ocean biogeochemistry, map ocean circulation, detect and monitor Harmful Algal Blooms (HABs) and pollution such as oil spills, track coastal water quality changes and enable better informed management decisions on coral reef ecosystems and fisheries.

Ocean color measurements can be challenging, because the relatively low, few percent diffuse reflectance signal from the ocean is overwhelmed by light scattered by the atmosphere. Simultaneous, high signal-to-noise ratio (SNR) measurements are required at different wavelengths to separate the water leaving radiance from sunlight scattered by the atmosphere. Furthermore, to correct ocean color measurements for contamination by light from nearby clouds, the maximum signal level that can be measured by the instrument must be very high so that the high reflectance cloud component can be measured without saturation and its effects subtracted from the spectral radiance measured at the top of the atmosphere. The resulting combination of high dynamic range and high sensitivity requirements has been difficult to achieve in previous instruments without specialized electronic and focal plane assembly designs.

To enable design studies for future instruments and compare performance of past and current instruments, reference signal levels for typical spectral radiance (Ltyp) expected from clear ocean scenes and maximum measurable spectral radiance (Lmax) for highly reflective clouds must be specified along with SNRs needed to enable clean separation of water leaving radiance from the atmospheric radiance and deliver the high sensitivity ocean spectral radiance measurements needed for Earth science research. Design studies for current and previous systems defined SNR, Ltyp, and Lmax in different ways, making it difficult to compare performance of these instruments directly and to have a clear basis for defining typical measurement conditions over the ocean for future missions such as NASA’s Pre-Aerosol-Clouds-Ecosystems (PACE) and Geostationary Coastal and Air Pollution Events (GEO-CAPE) missions.

This important and timely paper resolves these inconsistencies by describing a coherent set of SNR, Ltyp and Lmax for 16 spectral bands spanning 412 nm to 2130 nm in wavelength based on measurements at various solar zenith angles (SZAs) by NASA’s Moderate Resolution Imaging Spectroradiometer onboard the Aqua satellite (MODISA). Ltyp values were based on selecting radiometrically calibrated pixels from clear sky and relatively clear water ocean scenes with the same SZA using an objective, statistical method that assures scene homogeneity. Clear water pixels were defined as having a low chlorophyll concentration of <0.7 mg m–3, which is characteristic of water with almost no nutrients. Spectral radiance from each uniform scene was determined using well established radiometric calibration methods for MODISA. Lmax was established in a similar way using sets of cloudy pixels in unsaturated MODISA spectral bands. SNR was calculated from the statistical properties of uniform scenes – a tricky process to say the least because apparent variations in a scene could be due to system noise or intrinsic scene variability or a combination of both. This uncertainty was resolved by looking at maximum/minimum radiance ratios within small windows in real data and in simulations and establishing thresholds for that ratio below which the SNR appeared stable and the noise appeared to be Gaussian, based on a fit to the histogram distribution of standard deviations in 3x3 windows. In some cases, there was no threshold level that yielded noise with a Gaussian distribution.

Results of this paper show that among the most widely used past or present ocean color sensors with global coverage, MODISA ocean bands (processed to 1-km pixels) showed 2-4 times higher SNRs than comparable bands in another NASA system called SeaWiFS (1-km), which operated during 1999-2010. The MODISA SNRs are about the same as SNRs from the European MERIS-RR (reduced resolution, 1.2-km) system, which operated from 2002 to 2012. SNRs of all MODISA ocean bands and SeaWiFS bands (except the SeaWiFS near-infrared or NIR bands) were greater than pre-launch sensor specifications after adjusting the input radiance to Ltyp values in this paper.

An especially interesting conclusion of this paper is that it may be possible to relax SNRs in the visible wavelength range for future ocean color instruments like GEO-CAPE, based on analysis reported here and radiative transfer work described elsewhere in Applied Optics (Volume 46, pp. 1535-1547 [2007]) by Menghua Wang at NOAA. Hu et al. argue that the following SNRs should provide sufficient sensitivity for the GEO-CAPE mission: >1000 for spectral bands between between 350 and 720 nm, >600 for 720 – 900 nm, and >100-200 for the SWIR bands. These recommended SNRs are much lower in the visible and slightly lower in the NIR but significantly higher in the SWIR than MODISA SNRs determined for Ltyp at a SZA of 45 degrees. Given most data product noise results from atmospheric correction bands in the NIR and SWIR and not from the ocean signal blue-green bands, performance of GEO-CAPE in resolving small changes in ocean properties should be comparable to or better than MODISA, if the same processing approaches are used with the NIR bands. If SWIR bands are used for atmospheric correction of both MODISA and GEO-CAPE, significantly improved performance should be achieved from GEO-CAPE over MODISA because of higher SNR in the SWIR bands.

--Jeffery J. Puschell

Technical Division: Information Acquisition, Processing, and Display
ToC Category: Image Processing
OCIS Codes: (010.4450) Atmospheric and oceanic optics : Oceanic optics
(280.0280) Remote sensing and sensors : Remote sensing and sensors
(280.4788) Remote sensing and sensors : Optical sensing and sensors

Posted on November 09, 2012

Add Comment
You must log in to add comments.

Browse Journals / Lookup Meetings

Browse by Journal and Year


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Previous Spotlights