The CEOS ARD for Aquatic Reflectance – Evolving From Inland and Near-Coastal Waters to Include Oceans

Abstract

CEOS Analysis Ready Data (ARD) are satellite data that have been processed to a minimum set of requirements to allow for immediate analysis with less additional user effort and interoperability both through time and with other datasets. Product Family Specifications (PFS) are the core component of the CEOS ARD concept and describe what is needed for each product and is ready for download and use from here: https://ceos.org/ard/ . In 2021 CEOS and GEO AquaWatch developed Version 1 of Aquatic Reflectance ARD PFS focused on inland and near-coastal waters, as the use of satellite imagery and associated products for aquatic applications is developing rapidly. On 23 February 2022, CEOS’s Land Surface Imaging Virtual Constellation (LSI-VC) endorsed a new Version 1 of the ARD PFS) for Aquatic Reflectance (AR) products. Version 1 is available here: https://ceos.org/news/ard-pfs-aquatic-reflectance/ Currently, four products from Sentinel 2, Landsat, EnMAP, Landsat and Oceansat-3 are under assessment for Aquatic Reflectance ARD compliance. Rapid developments in the inland and near-coastal EO capabilities and the desire to integrate all water bodies into one CEO ARD Aquatic Reflectance PFS has led to a new collaboration to develop Version 2 of the Aquatic Reflectance ARD PFS. As part of the collaborative Version 2 framework, an initial step established that an integrated CEOS ARD AR with a focus on inland and near-coastal waters (focused of Version 1) and ocean waters was needed. Thus, in 2024 a new committee was established with representatives from inland, coastal and ocean Earth observation as well as members of the CEOS ARD Oversight Committee. This committee defined the tasks needed to reevaluate some of the PSF established in Version 1 of the CEOS ARD AR. Version 1 followed the template for Land Surface ARD and focused on finer spatial resolution satellite imagery. The addition of ocean colour radiometry (OCR) coarser spatial resolution EO data from sensors such as MODIS, VIIRS, the OCM-2 &3, Sentinel-3 and PACE and others, including their multi-sensor products (e.g. Copernicus marine service and OC-CCI) added a dimensionality of spatial pixels to this AR PFS. Concurrently there is a broadening use of higher resolution sensors for ocean applications (e.g. Sentinel-2) and expected increases in resolution of ocean related missions (e.g. Sentinel-3 next generation). Thus, the distinctions between inland and near-coastal, and ocean applications; and what constitutes medium or fine resolution sensors. become increasingly blurred. Additional metadata fields (as compared to the Land Surface PFS) had to be inserted or modified such as: Land and water atmospheric adjacency effect correction; directional atmospheric scattering (accommodating global inland waters occur with an altitude from -430 m to + 6500 m); Ice mask; floating vegetation/ surface mask; optically deep or optically shallow waters assessment; sun glint and sky glint correction. For Version 1 the team needed to assess the threshold and goal levels (previously called targets). This took considerable time to do as it was essential to ensure that the proposed metadata field meets the state of the art so that any interested ARD provider has access to methods and algorithms to meet this requirement. In a fast-developing application domain such as inland and near-coastal waters EO, significant advances over the past two years, required revisiting and assessing many of these metadata and processing requirements. OCR, additionally has a much longer history of providing fully operational ocean information data using coarser spatial resolution but higher radiometric resolution EO data. As an example, for sun glint flagging and correction, the coarse spatial resolution EO OCR data can apply statistical methods for each pixel as the wave trains and facets are averaged over the pixel. However, finer spatial resolution resolves individual wave facets requiring a per pixel basis. A similar line of reasoning exists for identifying foam, algal scums, floating macrophytes etc. Other differences include the handling of the atmospheric adjacency effects (AE), which become more significant in satellite images of inland water bodies, since these are surrounded by land. Notably AE can range from positive to negative values depending on the spectral contrast between land and water. Land reflectance is generally higher than that of water (e.g., in the near-infrared for most land cover types, throughout the spectrum for highly reflecting land covers such as snow, white sand, dry vegetation and concrete), but in some cases the land can be lower reflecting than water, as in the blue for highly turbid water and green vegetation. OCR applications usually filter out (using bathymetry models) pixel areas where bottom reflectance might contribute to the surface water leaving signal. With fine spatial resolution coastal and inland EO scenarios need to cover both optical deep (lakes, reservoirs, estuaries, rivers etc.) and optically shallow waters (e.g. seagrass or macro-algae beds or coral reefs etc.). A guiding principle for the work to include and specify requirements for this kind of details, the PFS sets requirements on what shall be included in the processing, but not referring to any specific methodology how this should be achieved. This is left to the discretion of the data provider, to define the best methodology for their data and implementing the state-of-the-art methodology as this continuously evolves. At LPS 2025 we intend to present the final Version 2 of the CEOS ARD AR that integrates inland, near-coastal, coastal and ocean water reflectance applications.