GEOS-FP Observation Impact Information

Millions of observations are assimilated into the GEOS-Forward Processing (GEOS-FP) atmospheric data assimilation system each day to provide accurate initial conditions for weather forecasts and other applications. The diagnostics displayed on this page provide a measure of the impact of these observations on GEOS-FP forecasts.

Methodology
Observation impacts are computed using the adjoint of the GEOS atmospheric data assimilation system, including the GEOS atmospheric model and Gridpoint Statistical Interpolation (GSI) analysis scheme. The calculation is based on the technique proposed by Langland and Baker (2004) and extended for nonlinear analysis schemes such as the GSI as described by Trémolet (2008) and Gelaro et al. (2010). Commonly referred to as the forecast sensitivity-based observation impact (FSOI), the technique efficiently estimates the impacts of all observations simultaneously on a selected measure of forecast error. The results can be easily aggregated by data type, location, channel, or other observation attribute. For global forecast systems such as GEOS-FP, the technique provides accurate estimates of observation impact for forecast ranges up to two days.

Impact Measure
The results shown are for a global measure of 24-h forecast error that combines errors in wind, temperature, specific humidity and surface pressure with respect to the verifying analysis from the surface to 1 hPa in terms of moist total energy (J/kg). Observation impact is taken to be the difference in this error measure between 24-h forecasts initialized from the analysis and corresponding background state, where this difference is due entirely to the assimilation of the observations. Only observations that have been actively assimilated are included in the calculation. The GEOS adjoint model currently includes parameterizations of surface drag, vertical mixing, moist convection and microphysics of clouds and precipitation (Holdaway et al. 2013; Holdaway 2015).

Plot Details
Currently, observation impacts in GEOS-FP are computed once each day for the 24-h forecast initialized at 00z. They reflect the impacts of all observations assimilated in the 6-h data assimilation window centered around this time. The results are plotted in two forms: summary diagrams showing average values for each observing system or channel over a selected time interval, and time series. For the summary diagrams, the values are averaged over the number of cases (00z analysis cycles) in the interval, and the color shading denotes the average number of observations for a given observing system or channel per case in this interval. For the time series diagrams, the plotted values are sums for each case, and the color shading denotes the magnitude of the plotted values. A universal color scale is used in the time series diagrams so that results for different observing systems can be easily compared. Results shown for bounded geographical regions, such as the Northern Hemisphere, show the contributions from observations in those regions (only) to the reduction of the global forecast error measure. Five diagnostic quantities are shown:

Total Impact -- calculated by summing the raw impact values (J/kg) for a given observation type in each analysis cycle. Negative values of total impact indicate that assimilation of the observation type reduced the 24-h forecast error norm compared to the background forecast and were therefore beneficial.

Impact per Observation -- calculated by dividing the total impact for a given observation type by the number of observations of that type in each analysis cycle.

Fractional Impact -- calculated by dividing the total impact for a given observation type by the combined impact of all observation types in each analysis cycle.

Fraction of Beneficial Observations -- calculated by dividing the number of observations of a given type that provide beneficial impact by the total number of observations of that type in each analysis cycle.

Observation Count -- calculated by summing the number of assimilated observations of a given observation type in each analysis cycle.



References:

Gelaro R., Langland R. H., Pellerin S., Todling R., 2010: The THORPEX observation impact intercomparison experiment. Mon. Weather Rev., 138, 4009-4025. [https://doi.org/10.1175/2010MWR3393.1]

Holdaway, D., R. Errico, R. Gelaro and J. G. Kim, 2013: Inclusion of linearized moist physics in NASA's Goddard Earth Observing System data assimilation tools. Mon. Weather Rev., 142, 414-433. [https://doi.org/10.1175/MWR-D-13-00193.1]

Holdaway D., R. M. Errico, R. Gelaro, J. G. Kim, and R. B. Mahajan. 2015. "A Linearized Prognostic Cloud Scheme in NASA's Goddard Earth Observing System Data Assimilation Tools." Monthly Weather Review, 143 (10). [https://doi.org/10.1175/MWR-D-15-0037.1]

Langland, R. H. and Baker, N., 2004: Estimation of observation impact using the NRL atmospheric variational data assimilation adjoint system. Tellus, 56A, 189-201. [https://doi.org/10.1111/j.1600-0870.2004.00056.x]

Trémolet, Y., 2008: Computation of observation sensitivity and observation impact in incremental variational data assimilation. Tellus, 60A, 964-978. [https://doi.org/10.1111/j.1600-0870.2008.00349.x]

[NASA logo] GMAO Experimental Forecast Suite


Responsible NASA Official: Steven Pawson, GSFC 610.1
Privacy Policy and Security Notice
Site Updated: 2024-05-03