JPSS Sea Surface Temperature Reanalysis

Visible Infrared Imager Radiometer Suite (VIIRS) is the new NOAA operational earth observing sensor, currently flown onboard SNPP since Oct'2011 and N20 since Nov'2017. Sea surface temperature (SST) is an important environmental variable produced from VIIRS clear-sky brightness temperatures (BTs), using the NOAA Advanced Clear-Sky Processor for Ocean (ACSPO) system.

The goal of SST reanalysis (RAN) is to uniformly reprocess all available VIIRS L1b data, after opening the cryoradiator doors (SNPP: Jan'2012; N20: Jan'2018), and produce a superior quality SST product in two GHRSST data formats: L2P (swath; 25 GB/day) and L3U (0.02°; 0.5 GB/day), validate against quality controlled in situ data (from NOAA iQuam system), monitor in the NOAA SQUAM, MICROS and ARMS web systems, and distribute to users via the NOAA CoastWatch (CW).

SNPP RAN1, performed in 2015 with ACSPO v2.40 in conjunction with UW CSPP group and covering a period from Mar'2012-Dec'2015, is publicly available to users on the CW website, with data from Dec'2015-on supplemented from near real-time ACSPO v2.41. Two RAN2's are underway, to reprocess the entire SNPP and N20 records with the latest ACSPO v2.60 and improved VIIRS calibration, and fix several issues identified in RAN1.

Here we share the status of VIIRS RAN2's, which commenced in June 2018. As of this writing, reprocessed are 13 months of SNPP and 1 month of N20 data. With the improved calibration, degradation of SST during the quarterly warm-up cool-down (WUCD) events (when the onboard blackbody temperature is varied over the course of three consecutive days) has been dramatically minimized. In particular, the quarterly 0.25K SNPP warm biases have been reduced to below several hundredths of a Kelvin, suggesting that the joint effort between the SST and NOAA VIIRS calibration teams resulted in a dramatic improvement in the thermal band calibration during WUCDs.

RAN2 SST data and matchups with in situ measurements (ground truth) serve as an excellent testbed for testing improved masking and SST algorithms. We will present development of experimental algorithms based on neural networks, with the aim of reducing seasonal and view-angle dependent biases in SST retrievals, and exploring SST information content in available infrared bands.