The goal of the NOAA AVHRR GAC Reanalysis (RAN) project is to create long-term time series of uniform sea surface temperature (SST) retrievals (Level 2 and 3 products) from AVHRR data using the Advanced Clear-Sky Processor for Oceans (ACSPO) system. During Phase 1 (‘RAN1’), data of several AVHRR/3s from 2002-2015 were reprocessed. Ongoing Phase 2 (‘RAN2’) aims to cover the full period of AVHRR GAC data from 1981-on. At the time of this writing, we reprocessed five AVHRR/2s onboard NOAA-07, -09, -11, -12 and -14 and two AVHRR/3s onboard NOAA- 15 and -16, and created an initial “beta” RAN2 data set (‘RAN2 B01’) spanning ~22 years from 1981-2003. The ACSPO algorithms for cloud masking and training SST regression coefficients, initially developed for operational SST processing, required modifications to mitigate the issues, specific to the RAN2 period: multiple sensor issues, and insufficient number of in situ SST data and their degraded quality. Another derived complexity, also related to insufficient and poor quality of satellite and in situ data, is the limited availability and suboptimal quality of first guess SSTs, which is used in ACSPO for cloud masking and quality control, and employed in the right part of the Non-Linear SST equations. The paper describes modifications to the ACSPO algorithms made for the RAN2 B01, and demonstrates the resulting improvements in the retrieved SST.
Under the NOAA AVHRR GAC Reanalysis project (RAN), a global dataset of consistent sea surface temperature (SST) retrievals from 1981-on will be created from multiple NOAA AVHRRs using the ACSPO system. Following release of RAN1 dataset in 2016, the initial RAN2 Beta 01 (“RAN2 B01”) dataset was produced from NOAA-07, 09, 11, 12, 14, 15 and 16 from 1981-2003. This paper evaluates the initial RAN2 B01 dataset and compares it with two other SST datasets, the NOAA-NASA Pathfinder v5.3 (“PF”) and ESA CCI v2.1 (“CCI”). The time series of monthly global biases and standard deviations with respect to uniformly quality controlled in situ SSTs, and clearsky fractions (percent of SST pixels to the total ice-free ocean) are compared. ‘Skin’ and ‘depth’ SSTs, only available in RAN and CCI data sets, and sensitivity of ’skin’ SST to true SST, are also compared. The RAN B01 outperforms PF. Compared to CCI, it generally delivers more clear-sky observations, often with a better accuracy and precision for both ‘skin’ and ‘depth’ SSTs. The sensitivity to true SST is lower and more variable in RAN2 B01, than in CCI. The RAN2 B01 performance following large volcanic eruptions needs improvements.