NCEP Data Assimilation

Above: 30 years of Data Assimilation Progress: Timeline of significant milestones in NCEP’s Operational Global and Regional Atmospheric Data Assimilation from 1991-2021.


The backbone of NCEP’s data assimilation effort is the Gridpoint Statistical Interpolation (GSI), originally developed as a three-dimensional variational data assimilation (3DVAR) analysis system, which was implemented in the Global Data Assimilation System/Global Forecast System (GDAS/GFS) in May 2007 (Kleist et al., 2009). The GSI is an extension of the Spectral Statistical Interpolation (SSI) analysis/assimilation system which was first implemented into the NCEP GDAS/GFS in June 1991 (Derber et al., 1991). The use of the SSI in the GDAS/GFS led to improved model performance over the previously used optimum interpolation scheme due to many innovations, including the ability to use observations which are different from the model and analysis variables (satellite measure radiances (Derber and Wu 1998), total precipitable water, etc.), the analysis variables being different than the model variables and the use of all observations at once allowing a better dynamical constraint between wind and mass fields.

The GSI was developed at the Environmental Modeling Center at NCEP as part of an effort to create a more unified, robust, and efficient analysis scheme. The key aspect of the GSI is that it formulates the analysis in model grid space, which makes it straightforward for this single analysis system to be used across a broad range of applications, including both global and regional modeling systems and domains. In addition to its use in the GDAS/GFS, versions of the GSI are used in operational NCEP mesoscale and convective-resolving guidance systems such as the Rapid Refresh (RAP), High Resolution Rapid Refresh (HRRR), North American Mesoscale (NAM), NCEP’s real-time analysis systems (RTMA, URMA, RU-RTMA), and NCEP’s Hurricane modeling systems (HWRF). The GSI application in the GDAS/GFS includes the use of a Tangent Linear Normal Mode Constraint (TLNMC, Kleist(2009)) to reduce non-physical noise in the analysis system.

Many additional enhancements to the GSI analysis system have been made over the years. For example, in 2016 an upgrade was made in the GDAS with the replacement of the 3D Hybrid Ensemble-Variational to 4D Hybrid Ensemble-Variational Data Assimilation System. With this extension, 4D analysis increments at prescribed times are constructed through linear combinations of 4D ensemble perturbations. Also added to the GSI system were improved observational quality control (variational and platform specific) algorithms, the inclusion of additional observations (new satellite radiances and extensions to all-sky assimilation), more aircraft observations (with bias correction), GPS radio occultation observations and wind estimates from satellites, along with many other enhancements.

In 2016 an upgrade was made in the GDAS with the replacement of the 3D Hybrid Ensemble-Variational to a 4D Hybrid Ensemble-Variational Data Assimilation System. The 4-D increments are constructed by computing the best combination of 4-D ensemble perturbations. The weights for ensemble members are kept constant throughout the assimilation window, and the GDAS 4D-Hybrid uses uses 2 outer iterations with 50 and 150 inner iterations in those 2 outer iterations, with variational quality control turned on in every iteration. Tangent Linear Normal Mode Constraint (TLNMC) and Digital Filter Initialization (DFI) are used to constrain the analysis increment while additive error inflation is removed. Also, ozone cross covariances are in the 4D-Hybrid. Localization is recued to 50 percent in the troposphere and the weights for static and ensemble are kept at 12.5 percent and 87.5 percent, respectively.

More recent data assimilation changes included in NCEP GDAS/GFS upgrades are:

  1. In the 2017 GDAS/GFS upgrade the Near Sea-Surface Temperature (NSST) Analysis was implemented to replace Real-Time Global SST (RTG_SST), to provide more realistic ocean boundary conditions.
  2. In the 2019 GDAS/GFS upgrade:
    1. Upgrade specific humidity perturbation and statistics physics tendency perturbation with new parameter settings in the ensemble forecast for prescribing background errors.
    2. Remove digital filter and tropical cyclone vortex relocation
    3. Increase the horizontal resolution of the ensemble part of the hybrid data assimilation from 35 km to 25 km
    4. Update the Near Sea Surface Temperature scheme to a) apply Sea Surface Temperature climatology tendency to the foundation temperature and, b) reduce background error correlation length from 450~800 km down to 100 km.
  3. In the 2021 GDAS/GFS upgrade:
    1. Spinning up an offline land model with observed precipitation in the Global Land Data Assimilation System to provide improved land initial conditions.
    2. Replacing the Ensemble Square Root Filter with the Local Ensemble Kalman Filter (LETKF) that offers a model space localization and linearized observation operators.
    3. Employing the 4-Dimensional Incremental Analysis Update (4DIAU) technique.
    4. Adopting a stochastic kinetic energy backscattering (SKEB) perturbation technique in the ensemble forecast component used to prescribe background error covariances.
    5. Updating variational quality control.
    6. Applying the Hilbert curve to aircraft data


Community GSI Repository on GitHub

Radiance Monitoring

Ozone monitoring

Global Conventional Observation Monitoring

Minimization Monitoring

History of NCEP GDAS/GFS changes (2010-present)

NOAA Satellite Status Information

Joint Center for Satellite Data Assimilation

Observational Data Processing at NCEP

Near-surface Sea Surface Temperature (NSST)


Derber, J. C., Parrish, D. F., & Lord, S. J. (1991). The New Global Operational Analysis System at the National Meteorological Center, Weather and Forecasting, 6(4), 538-547.

Derber, J. C. and W.-S. Wu, 1998: The use of TOVS cloud-cleared radiances in the NCEP SSI analysis system. Monthly Weather Review, 126(8), 2287 - 2299.

Kleist, D. T., Parrish, D. F., Derber, J. C., Treadon, R., Wu, W., & Lord, S. (2009). Introduction of the GSI into the NCEP Global Data Assimilation System, Weather and Forecasting, 24(6), 1691-1705.

Whitaker, J. S., Hamill, T. M., Wei, X., Song, Y., & Toth, Z. (2008). Ensemble Data Assimilation with the NCEP Global Forecast System, Monthly Weather Review, 136(2), 463-482.