March 29-30, 2016
Santa Rosa, CA USA
In late March 2016, the two-day Aquarius Science Calibration/Validation (Cal/Val) workshop was held at Remote Sensing Systems (Santa Rosa, CA). The meeting focused on understanding the path towards completing the Aquarius Version 5 (V5) processing algorithm. Participants addressed key issues related to possible seasonal and regional biases in Aquarius data, the need for additional evaluation products to assess algorithm performance, and how salinity retrievals from the SMAP (Soil Moisture Active Passive) mission could help to improve Aquarius V5 processing.
AgendaDocuments: 18
Fore, A., Yueh, S., Tang, W. and Hayashi, A. [30-Mar-16]. This presentation began with comparing data from Aquarius and SMAP (Soil Moisture Active Passive) versus WindSAT wind speed in terms of surface emissivity and geophysical model functions (GMF). Below wind speeds of 20 meters per second, there is generally good agreement between Aquarius and SMAP. Sea surface salinity (SSS) data from Aquarius and SMAP were then compared with Argo data and the HYCOM model. Brightness temperatures were computed and gridded to estimate wind speed and salinity using constrained objective function minimization. In summary, SMAP can continue the record of global SSS maps from space using L-band. SMAP's SSS accuracy is not as good as Aquarius but is close.
Schanze, J. [30-Mar-16]. Aquarius-derived density (higher in cold and/or salty water) is available in the V4 product and spiciness (measure of hot and salty water) will be added to the V5 product. The definition of spiciness is now part of the Thermodynamic Equation of State 2010 (TEOS-10). Spiciness along isopycnals (i.e., line/surface of same density) is proportional to the water mass contrast and suited to study some mode water formation. In this presentation, density and spiciness were decomposed into thermal and haline components for Aquarius, along with Aquarius minus Argo. This study concludes that, in many areas, Aquarius provides significant temporal and spatial improvements over Argo. Continuity of L-band salinity measurements from SMAP (Soil Moisture Active Passive) and SMOS (Soil Moisture Ocean Salinity) allow the continuation of the density and spiciness record with the potential to gain new understanding of important questions related to mixing in the ocean.
Meissner, T. [29-Mar-16]. This presentation covered the evaluation of various sea surface temperature (SST) products including: NOAA Optimum Interpolation (Reynolds), WindSat, Canada Meteorological Center (CMC), Multi-scale Ultra-high Resolution (MUR), and UKMet Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA ). All three products that use microwave SST (WindSat, CMC, MUR) perform slightly better than Reynolds, particularly in cold water / at high latitudes. For future Aquarius algorithms, it is recommended to use the CMC product.
Brown, S. [29-Mar-16]. In this presentation, the focus and format of the workshop were defined along with objectives (e.g., state of current Aquarius V4.2 product, assess progress of yet-to-be-implemented V5 updates) and key questions to be addressed during the meeting.
Lee, T. [29-Mar-16]. The mission target accuracy of Aquarius SSS of 0.2 psu on a monthly time scale and a 150-km spatial scale includes time, mean and temporal variations. It does not isolate Aquarius' ability to characterize temporal changes on various space and time scales, which is more fundamental to science requirements. This presentation focuses on evaluating the consistencies of Aquarius V4 monthly gridded SSS using two Argo monthly gridded products on different space and time scales, and contrasts these consistencies with those between two Argo products.
Brown, S. [29-Mar-16]. The history of Aquarius algorithm development from launch until present. V1.3 (August 2012) took into account initial post-launch adjustments. V2 (February 2013) covered spacecraft pointing calibrations, updated antenna patterns, drift corrections, and radio frequency interference (RFI) algorithm parameters. In V2, "wiggles" were included as an offset and the roughness correction was based on the NCEP model. V3 (June 2014) included updated antenna pattern coefficient (APC) parameters, a roughness correction based on scatterometer data and horizontally-polarized brightness temperature, empirical ascending/descending symmeterization, application of empirical sea surface temperature (SST) bias adjustments, and simplified drift correction. V4 (July 2015) included a hybrid antenna pattern to adjust excess spillover, updated empirical SST bias correction and non-linear Stokes parameter (I/Q) coupling.
Brown, S. [30-Mar-16]. This study assessed the dependence of atmospheric boundary layer stability on the excess emission generated by a rough ocean surface, one of the largest corrections required in the sea surface salinity retrieval algorithm. The difference in temperature between the air and sea is a measure of boundary layer stability, shown in previous work to be a key factor in air-sea interaction including wave and foam formation. Microwave sea surface excess emissivity has been shown to have a significant dependence on the air-sea temperature difference at C-band and higher, which might be the cause of the observed seasonal biases at high latitudes in Aquarius data. The model function developed for this study relates backscatter/emission as a function of wind speed, wind direction, sea surface temperature, significant wave height, polarization and incidence angle. Results show a strong correlation between salinity bias and air-sea temperature difference. The bias is suitable for an empirical correction but more work is needed to completely understand the physics.
Misra, S. and Brown, S. [29-Mar-16]. An overview of the "wiggle correction" based on locking of the Aquarius voltage-to-frequency converter (VFC) locking was given. Each step of the correction implementation was described: (1) derive correction using difference between consecutive reference load count samples, (2) normalize mean correction to zero, (3) provide wiggle correction table with respect to reference load count, (4) use table to generate a time series for each channel, and (5) correct the reference load counts to calculate the antenna temperature. In conclusion, the wiggle correction has a zero bias with respect to the reference load and thus is not responsible for annual, seasonal, or other anomalous trends observed in the data.
Kao, H-Y. [29-Mar-16]. This presentation compared sea surface salinity (SSS) retrievals using NOAA Reynolds and the Canada Meteorological Center (CMC) sea surface temperatures (SSTs). It began with a comparison of Aquarius SSS data against Argo float matchups, which indicates that Aquarius SSS is biased high in the higher latitudes and low in the tropics. Time series plots for daily global average Aquarius (Beams 1, 2, 3) minus Argo SSS were shown along with histograms for Aquarius minus buoy differences. Overall, no obvious difference is found from the validation of SSS retrieval with different SST products. The same analysis was conducted for SSTs using Reynolds versus co-located buoy data and CMC versus co-located buoy data. SST from CMC shows smaller biases than Reynolds SST when comparing with in situ data, especially in low-mid latitudes. The last portion of the talk focused on the rain impact model (RIM) adjustment to Aquarius data.
Wentz, F.J. and Meissner, T.
[30-Mar-16]. The Aquarius galaxy model is fairly complex. It integrates the antenna over a rough ocean surface based on an existing astronomy map. SMAP (Soil Moisture Active Passive), which looks fore and aft (i.e., "dual looks"), provides the means to directly determine the galaxy map for SMAP. About one year of data from SMAP were investigated in two cases: (1) averaged over all winds, and (2) using only winds below five meters per second. Next step is to assess whether the differences observed are due to antenna patterns or scattering models. This will be done by fine tuning the modification, implementing in the Aquarius test bed, symmetrizing the galaxy correction, and assessing the improvement.
Soldo, Y., Le Vine, D., Dinnat, E., de Matthaeis, P., Hong, L., Gales, J., and El-Nimri, S. [30-Mar-16]. This presentation addressed proposed changes and issues for V5 Aquarius data processing. There was a focus on three items: (1) additional radio frequency interference (RFI) flagging, (2) emissivity model, and (3) beam 2. Existing flagging misses some "noise-like" RFI thus it is recommended to add a flag based on fixed thresholds. This would be applied to retrievals over land (not the ocean) and not remove any data. The proposed change to the Aquarius V5 land emissivity model would make it more consistent with the model used by SMAP (Soil Moisture Active Passive). Inter-beam differences for the three Aquarius beams have been examined with respect to sea surface salinity and Faraday rotation angle, showing possible issues with Beam 2.
Jacob, M., Santos-Garcia, A., and Jones, L. [30-Mar-16]. This presentation began by addressing sea surface salinity (SSS) differences between Aquarius data and the HYCOM, which have been observed during simultaneous rain events. A majority of these events have shown significant reductions in satellite-derived SSS. Instantaneous rainfall causes strong SSS reduction and salinity stratification at the surface. Following the rain event, wave mixing and diffusion reduce the salinity over a time period of several hours. Aquarius SSS measurements in the presence of rain can be significantly fresher than the bulk salinity at greater than one meter of depth. For this work, the Aquarius salinity model has been applied to SMAP (Soil Moisture Active Passive). The weighting function for Aquarius footprint has been modified to better match SMAP's instantaneous field of view (IFOV). RIM's performance with SMAP data was very good at predicting SSS, compared to AQ and SMOS (Soil Moisture Ocean Salinity) retrievals, providing a robust quality flag for identification of salinity stratification.
Melnichenko, O. and Hacker, P.
[29-Mar-16]. This study compared versions of Aquarius data processing (e.g., V4.0, V4.1.0, V4.2.1) in various ways, including: box analysis (Indian and Pacific Oceans), latitude-time distributions of zonally averaged differences between Aquarius weekly maps and Argo buoy data; ascending minus descending Aquarius data (three-year mean) with respect to Argo; time-varying bias for ascending and descending data (April and September 2013); and empirical orthogonal function (EOF) decomposition of ascending and descending orbits. The conclusion is that some static and time-varying biases persist, which need to be characterized and quantified (e.g., spatial maps of amplitude and phase) for science and applications users.
Meissner, T. and Wentz, F. [29-Mar-16]. This presentation focused on Aquarius V4.1 ("instrument only calibration"), which shows spurious annual signal (plus drift after October 2014 in some of the channels). The approach taken was to go back to V4.0 (ocean calibration, global HYCOM) and analyze seasonal and regional biases. Various components were investigated: inter-beam differences, Aquarius versus SMAP (Soil Moisture Active Passive), galactic reflection, surface roughness correction, dependence on residual sea surface temperature and wind (speed, direction), ancillary sea surface temperature input, and air-sea temperature differences. The conclusion was that the spurious annual signal is likely instrument related.
Lagerloef, G., Kao, H-Y., and Carey, D.
[30-Mar-16]. A major science objective is to build a long term satellite salinity climate data record (CDR) to measure climatic trends in ocean salinity and marine water cycle. This requires decadal measurement records and long-term calibration stability, stretching the present data to its limits. This presentation focuses on which trends may be observable with our present satellite record. Empirical orthogonal function (EOF) decomposition was used to analyze variance of Aquarius monthly salinity data (V4.0, V4.1), Argo data, and HYCOM model salinity. The outcome of this analysis a recommendation to: discontinue using HYCOM as a calibration target; use Argo in areas where its EOFs 1 and 2 have minimum variance; and use individual Argo matchups for channel calibrations (Aquarius Level 2 swath data).
Brown, S. and Misra, S. [29-Mar-16]. This study looked into the possibility of using differences in vertical and horizontal polarization data from the Amazon rainforest, the ocean, and Antarctica as Aquarius calibration model sources. Possible relationships between instrument temperature seasonal variations (e.g., noise diodes, feedhorns, reflector, Dicke load) and regional biases were also investigated. Third Stokes errors were analyzed for regional patterns (e.g., impacts on Q for descending minus ascending orbits). We found an apparent small (about 0.1K) temporal variation present in third Stokes brightness temperatures over the Amazon and Congo rainforest regions. It is concluded that empirical corrections in earlier Aquarius algorithm versions may be impacting the physical understanding of the observed residuals.
Tsontos, V. and Vazquez, J.
[30-Mar-16]. This presentation was designed to promote discussion about the large number (143) of Level-3 (L3) datasets with further additions expected upon inclusion of sea surface salinity uncertainty. The PO.DAAC suggests possible re-packaging and/or consolidation of data variables currently in separate files within a given file; for example, place ancillary and other variable (e.g., uncertainty) fields embedded in L3 file as additional parameters (variable array). In addition, the PO.DAAC suggests implementation of two metadata standards frameworks for the V5 dataset, the "Attribute Conventions for Data Discovery" (ACDD) and the "Climate Forecast" (CF) convention, which are important in the remote sensing context.
Hong, L.
[29-Mar-16]. The overview covered updates to Aquarius V4.2 data including the addition of spiciness, the rain impact model (RIM), updated products for random / systematic uncertainty, and the Canada Meteorological Center (CMC) sea surface temperature ancillary data set. Two sub-versions of V4.2 (V4.2.1 and V4.2.1) were discussed: V4.2.0, which included an exponential drift correction and HYCOM-derived bias wiggle correction, and V4.2.1, which included an exponential drift correction and instrument-based wiggle correction. In summary, sea surface temperature from CMC reduces the standard deviation of residual sea surface salinity retrieval error while orbital biases are unchanged. Neither conventional wiggle flattening nor instrumental wiggle correction change latitudinal variation, which changes over season. There are some residual galactic effects in the southern hemisphere, especially in ascending passes.
