Physical Basis

There is a considerable amount of scientific literature, both observational and theoretical, suggesting that UV reflectance of the ocean surface is sensitive to wind speed and ocean color; [Li2010] , [DeKloe2008], [Koepke1984] and [Menzies1998], just to name a few. Figure (1) indicate the different components which contribute to the signal reflected off the ocean surface and measured by by the Aeolus instrument. The reflection by white caps, the surface itself and the multiple scattering of sea salt particles are function of the wind speed. The sensitivity of UV reflectance to ocean color increases at larger incidence angles away from nadir viewing geometry (e.g. [Li2010]). However, much of the published literature on ocean surface reflectance avoids estimation of the effects of ocean color. One reason for this is that space-based ocean color observations are typically confined to the wavelength range from ca. 400 to 900 nm, i.e. from the visible (VIS) to the near infrared (NIR), due to the increasing difficulties in separating the oceanic from the atmospheric contribution at shorter wavelengths in the UV. Therefore, only limited information on the upward light field emanating from the ocean surface water layer is available.

(1)\[\textbf{UV Ocean Reflectivity}\]

Detailed radiative transfer calculations offer a way to overcome this deficiency. To be able to do so, absorption and volume scattering of pure sea water and the water constituents (i.e. typically phytoplankton, detritus, colored dissolved organic matter, and, close to the coast, also inorganic suspended matter) have to be known. Unfortunately, very little observational data of the radiative properties in the spectral domain employed by Aeolus exists. For instance [Morel2007] reports on optical properties of oligotrophic waters found in the South Pacific Gyre (SPG). An important step in the early project phases will therefore consist in reviewing the scientific literature on optical properties of sea water in the UV spectral domain.

For a more quantitative evaluation of the dependence of surface reflectance on wind speed, there are a number of available model functions. For example, [Li2010] compare the 3-component ocean reflectance model of [Menzies1998] to aircraft LIDAR sea surface returns at 355 nm for a variety of incidence angles and wind speed conditions.

The results by [Li2010] indicated that the modeled subsurface contribution to total reflectance dominates for an off-nadir viewing angle of 35 degrees, increasing the reflectance by about two orders of magnitude for a 5 m s-1 sea surface wind speed. For low ocean surface winds, the ocean surface forward scatters most of the LIDAR signal away (specular reflection) and there are virtually no whitecaps to increase the surface reflectance. But for higher sea surface winds, the magnitude of the subsurface contribution becomes comparable to the sum of other contributions (i.e., whitecaps and specular reflection).

Figure (2) shows the contribution of whitecaps to total modeled reflectance using equation (4) from [Li2010]. This is in very good agreement with the contribution of whitecaps to a much more sophisticated surface reflectance model, the spectral bi-directional reflectance distribution function (BRDF) used by [Sayer2010], part of the Oxford-RAL Aerosols and Clouds (ORAC) retrieval scheme (cf. Fig. 2 of [Sayer2010]).

Modeled whitecap reflectance for 35 degrees off-nadir viewing for sea surface winds up to 20 m s-1.A close up of a mans face Description automatically generated

(2)\[\textbf{Modeled white cap reflectance}\]

The primary retrieval method will rely on an empirical geophysical model function (GMF), to be developed during this research program, that relates the Aeolus sea surface return signal strength to a corresponding wind speed. Because of the confounding contributions of reflections from the sea surface and subsurface, ancillary data will be needed to stratify Aeolus collocations in the DP, such as ocean color, presence of sea ice, and approximate wind speed (e.g., from global atmospheric analyses, or other satellites that produce estimated ocean surface winds).


Here is a basic outline of the approach to develop an empirical Aeolus ocean surface wind GMF:

  1. Identify a number of geo-physical oligotrophical locations where it can be expected that the sub-surface contribution to the Aeolus surface return is relatively small. For there regions prepare a collocated dataset e.g. Auxiliary NWP model wind speed, collocated ASCAT, CYGNSS and SSMI winds and ocean color data from imagers like MODIS or Sentinel as available.

  2. Divide the collocated DP of Aeolus data into three categories for processing by ocean color and approximate wind speed range: high wind speed/low ocean color, all wind speeds/low ocean color, and high wind speed/all ocean color.

  1. Compute the mean Mie spectrometer counts of the sea surface return for each BRC whose quality control flags are clear, and compare mean intensity to collocated, independent wind speed sources, The returns that contain the sea surface will be corrected to remove the atmospheric signal in that portion of the range gate above water, as in [Li2010], for example.

  2. Iterate validation steps (2A & 2B) until bias and std dev statistics are stable for the high, medium and low probability wind retrieval categories.

Illustration of the Approach

A simplified illustration is provided in figure (3)

(3)\[\textbf{Aeolus surface return, collocated with surface windspeed and ocean color}\]


Two key goals of this product prototype study are

– find the range of conditions under which prototype winds may be retrieved from Aeolus sea surface returns with error statistics comparable to other satellite-derived ocean surface winds (e.g., 2 m s-1 RMS or 10%), and

– determine the technical advances that will facilitate wind or ocean color or joint wind/ocean color retrievals in the future from Aeolus data under a wider range of ocean surface/subsurface conditions.