Near-Real-Time Satellite-Based Bias Corrections for Ocean Model Forecasts

Atmospheric heat fluxes at the ocean surface are used to force operational ocean models. The required heat fluxes include solar radiation, longwave radiation, sensible heat flux, and latent heat flux. These fields come from atmospheric numerical weather prediction (NWP) models, which often have biases. Monthly or yearly bias corrections can be identified and applied to these fields prior to starting the model forecast. When changes occur in the atmospheric model, i.e. model version updates, new biases must be calculated which is often a time intensive process.

As an alternative to NWP model fields, the Naval Research Laboratory (NRL) ocean surface flux system (NFLUX) provides near-real time satellite-based three-hourly global gridded analysis fields of air temperature, specific humidity, wind speed, solar radiation, and longwave radiation. NFLUX is a complete end-to-end data processing, automated quality control, and 2D assimilation system. The state parameter fields are produced with satellite data primarily obtained from passive microwave sensors and scatterometers. The radiative flux fields are produced primarily with satellite-derived atmospheric profile products from the Microwave Integrated Retrieval System (MIRS) along with the rapid radiative transfer model for global circulation models (RRTMG).

The satellite-based NFLUX analysis fields can be used to determine NWP bias corrections in near-real time. This allows for the bias corrections to be much more responsive to changes in the atmospheric model, rather than having to reprocess a year of model output. In this study, the NWP model of interest is the Navy Global Environmental Model (NAVGEM). NAVGEM provides the surface forcing fields to several operational ocean model systems, including the Global Ocean Forecast System (GOFS). The NFLUX analysis fields are used to estimate NAVGEM bias corrections during the hindcast period. The recent history (days to weeks) of NFLUX and NAVGEM model differences are then used to determine long term (persistent) and short term (weather-dependent) bias corrections, which are applied to the forcing over the forecast period to extend the hindcast corrections into the forecast.

The averaging windows for the long term and short term bias corrections are determined from the NFLUX minus NAVGEM daily average time series for one year. For each ocean point, the variance of the n-day average difference time series is calculated, where n ranges from 1 to 30 days. The long term averaging window is identified when the calculated n-day variance became less than half of the original variance calculated from the 1-day average difference time series. All ocean points are averaged together to determine a single global long term bias correction averaging window for each parameter. The global long term bias averaging window is identified as 10 days for specific humidity, 11 days for air temperature, and 5 days for wind speed, longwave radiation, and shortwave radiation. The short term bias correction is then determined by subtracting the long term bias (calculated using the global long term bias averaging window) from the NFLUX minus NAVGEM daily average time series.

The long term and short term bias corrections are applied to the forecast period with different decorrelation time scales. For each ocean point in the same NFLUX minus NAVGEM daily average time series, the calculated long term and short term biases from the previous 30 days are used to determine the individual bias correction decorrelation times. The decorrelation times are averaged together to produce a single globally averaged long term and short term temporal decorrelation for each parameter. The long term bias correction, due to persistent model biases, is applied throughout the entire forecast. The short term bias correction, representative of weather features that change quickly, is applied with a decreasing weighting function from the analysis time out to 2 days for specific humidity and air temperature and 1 day for wind speed, longwave radiation, and shortwave radiation.