The Level-3 weekly polar-gridded TB (at V and H polarizations) and SSS data are derived from the Aquarius Level-2 TB and SSS data distributed by NASA's Physical Oceanography Distributed Active Archive Center (PO.DAAC) and repackaged as weekly polar-gridded Level-3 products of TB and SSS. Derivation occurs only when Radio Frequency Interference (RFI) contaminated observations are excluded and a weekly average is calculated per EASE-Grid 2.0 cell.

The PO.DAAC Level-2 TB product is computed after empirical calibration of the measured antenna temperatures against a forward radiative transfer model over ocean surfaces (Le Vine et al. 2011a). More details about the Level-2 processing can be found in Wentz and Le Vine (2012), Le Vine et al. (2012), and Piepmeier et al. 2013.

Gridded data were produced using all the Aquarius observations flagged as RFI free, and recorded during nominal operation of the spacecraft.

TB observations and SSS retrievals at latitudes greater than 50 degrees in both hemispheres are averaged and gridded into a weekly product.

According to the orbit and sensor characteristics, the temporal resolution of the product was set to one week, corresponding to the time of revisit. Since the Aquarius sensors are in a push-broom alignment, a weekly-gridded product provides the largest spatial coverage.

Each radiometer was treated independently to produce three weekly-gridded TB products corresponding to observations from the ascending orbit, the descending orbit, and both orbit types combined. The distinction per radiometer is necessary due to the different incidence angles, which lead to different reflection intensities at the interface created by dielectric constant changes (e.g. snow/air interface). For each seven-day cycle, all valid observations from a given radiometer with footprint center within a given 36 km grid cell were averaged together, and the standard deviation was calculated. For each week and hemisphere, eighteen TB maps are produced and distributed, one for each of the three radiometers, each of the two polarizations, and each of the three orbit types (Brucker et al. 2014).

The sea ice fraction (ICEF) results from a combination of two elements: estimated sea ice concentration, and antenna characteristics. Sea ice concentration estimates are obtained from the analysis by NOAA's Marine Modeling and Analysis Branch (http://polar.ncep.noaa.gov/seaice/Analyses.shtml), available daily at a spatial resolution of 1/12 degree, and distributed by the U.S. National Centers for Environmental Prediction (NCEP) as the Global Forecast System (GFS) Global Data Assimilation System (GDAS) sea ice product.

Using the latest PO.DAAC Aquarius Level-2 data version 4.0:

• Select data at latitudes greater than 50 degrees.
• Reject data contaminated by RFI.
• Reject data collected during spacecraft maneuvers or anomaly periods as reported on the Aquarius status Web page.
• Grid and average data per seven-day cycle corresponding to the spacecraft time of revisit.
• Where less than six contiguous grid cells do not have data, apply a Delaunay triangulation with linear interpolation to the weekly-gridded values to spatially interpolate them in grid cells without observations/retrievals during the cycle.
• For a given one-week orbit cycle, all measurements within a same grid cell were averaged together, and the standard deviation calculated.

Version 5 of Aquarius L3 Weekly Polar-Gridded Brightness Temperature and Sea Surface Salinity utilizes Version 4 of the Level-2 Aquarius TB as input data.

The antenna temperatures are corrected for:

• the emission of extraterrestrial sources (Sun, Moon, Celestial Sky), that directly reaches the antenna through the side and back lobes
• the effect of the integration over the antenna gain patterns
• the Faraday rotation
• and atmospheric effects (upward emission, downward emission reflected at the surface, and attenuation of the signals from the surface)

These corrections provide the TBs at the Earth's surface. Over the oceans, the TB is corrected for the effects of surface roughness (including the reflected/scattered galaxy and Sun) using the scatterometer observations and a simulated wind speed. The remaining TB for a smooth surface is converted into SSS using ancillary data for the sea surface temperature, and a model for the sea water dielectric constant.

The sensitivity of TB is ~0.15 K and ~0.2 K per 1.4 s observation sample over the ocean and the dry ice sheet, respectively. Observations have shown stability within ±0.2 K over the ocean and Celestial Sky over several months (Dinnat et al., 2013; Lagerloef et al., 2013).

The requirement for the Aquarius SSS retrievals accuracy is 0.2 psu after temporal averaging over a month in global open oceans. While the algorithm used in the Aquarius Level-2 processing for retrieving SSS performs well in the tropics and mid latitudes (warmer) oceans (Lagerloef et al., 2013), L-band SSS retrieval in the polar (colder) oceans is challenging. SSS retrievals have not yet been specifically validated in cold water, and should be used with caution. L-band observations are less sensitive to salinity in cold waters. In addition, salinity retrieval is likely to be less accurate for very rough sea surfaces. For instance, in the Southern Ocean there are strong winds and the oceanic circulation is dominated by the Antarctic Circumpolar Current, which could reduce the quality of the SSS retrievals. Finally, the presence of sea ice and icebergs in the sensors field of view adds complexity to the retrieval of SSS in the high latitudes.

The weekly-gridded product contains SSS retrievals in ascending orbit and descending orbit. This distinction between the two orbit types may be required because differences have been identified (Lagerloef et al., 2013). The PO.DAAC Aquarius Level-2 product version 3.0 has reduced the differences, though not completely eliminated them. The reduced differences have been carried forward to the version 4.0 source data used for this current product. While the origin of these differences has not been established yet, it is likely to be due in part to the reflected/scattered galaxy, and RFI contaminations. It is possible that residual RFI and sky contaminations impact the empirical calibration performed over the oceans, and therefore create biases dependent on the type of orbit. The correction for the reflected galaxy has been found insufficient, and an empirical adjustment was introduced for PO.DAAC version 3.0, and carried forward to version 4.0. The galaxy contamination is very dependent on the type of orbit, because it is only significant when the contribution comes from a very limited region of the sky, for example the galactic plane.

The Aquarius calibration is valid over a narrow range of TB (oceans only) (Piepmeier et al. 2013). Biases were identified in PO.DAAC Aquarius 2.0 when observing the celestial Sky (during Cold Sky Calibration Maneuvers) and over land. The calibration coefficients will be evaluated after each Aquarius data release and there are plans to address the full range calibration in future releases.

Both passive and active L-band observations are impacted by RFI. In the weekly-gridded products, data associated with an RFI flag in the Level-2 products is eliminated (Brucker et al. 2014).