ACSPO Global SST from VIIRS

These VIIRS SST data are produced using the NOAA Advanced Clear-Sky Processor for Ocean (ACSPO) SST system, from the afternoon NPP and N20 satellites, in two formats: L2P and L3U (uncollated). The data are reported in 10 min granule files in NetCDF4 format, compliant with the GHRSST Data Specifications v2 (GDS2). There are 144 granules per 24 hr interval, with a total data volume of ~25 GB/day for L2P, and ~0.4 GB/day for L3U.

The near real-time (NRT) data is operationally produced by the NOAA Data Exploitation (NDE) using the official enterprise ACSPO algorithms. The NRT L2P/L3U data are produced using ACSPO v2.61. A 2-week rotated buffer of the NRT data is provided here on Coast Watch, whereas the full data set is available from PO.DAAC and NOAA NCEI (see links under "Data Access"). Note that the NCEI archive has not been backfilled with ACSPO v2.61 and therefore containes SST data from earlier ACSPO versions before the operational implementation of v2.61 (23 Apr 2019), while PO.DAAC contains the full v2.61 archive.

The Reanalysis version 2 (RAN2) product was produced at STAR using ACSPO v2.61. L2P and L3U RAN2 data covers a period from 1 Feb 2012 - present 2019 for NPP and Jan 5 2018 - present for N20. A complete archive of L2P/L3U data is available on PO.DAAC, while the the coastwatch ftp only contains L3U data due to high data volume. The NOAA SST team worked with the NOAA calibration team and used an up-to-date converter from RDR (L0) to SDR (L1b), to address quarterly spikes in day-time SST bias which coincide with quarterly warm-up-cool-down calibration exercises onboard NPP/N20. The algorithms used in VIIRS RAN2, are similar to those described in (Ignatov et al., 2016) for AVHRR GAC, except fixed regression coefficients were used (in contrast with variable coefficients for AVHRR). Currently, the RAN2 data are available on this Coast Watch site (L3U) and PO.DAAC (L2P and L3U).

ACSPO retrievals are made in full VIIRS swath (~3,000 km). For data assimilation applications (such as production of L4 analyses, especially those that blend satellite and in situ data), correction for the Sensor-Specific Error Statistics (SSES; reported in ACSPO files; Petrenko et al., 2016) biases is strongly recommended.

In each valid water pixel (defined as ocean, sea, lake or river, up to 5km inland; note that in "invalid" pixels, defined as those with >5km inland, fill values are reported), the following layers are reported in both L2P and L3U, among others: SSTs derived using multi-channel SST (MCSST; night) and Non-Linear SST (NLSST; day) algorithms (Petrenko et al., 2014); ACSPO clear-sky mask (ACSM; provided in each pixel as part of l2p_flags; Petrenko et al., 2010); SSES bias and standard deviation (Petrenko et al., 2016); NCEP wind speed; and ACSPO SST minus reference (Canadian Met Centre L4 SST). For L2P, layers of brightness temperatures (BTs) in M12, 15, and 16 are also reported, for those users interested in direct "radiance assimilation" (e.g., NOAA NCEP, NASA GMAO).

Only ACSM "confidently clear" pixels (equivalent to GDS2 "quality level"=5; QLs are also reported for each pixel) should be used. The ACSM also provides day/night, land, ice, twilight, and glint flags. Note that users of ACSPO data have the flexibility to ignore the ACSM, derive their own clear-sky mask, and use BTs and SSTs in those pixels. They may also ignore ACSPO SST, and derive their own SSTs from the original BTs.

Both L2P and L3U SSTs are monitored and validated against in situ data (Xu and Ignatov, 2014) in SQUAM (Dash et al., 2010) and ARMS systems, and BTs are monitored in MICROS (Liang and Ignatov, 2011).