Skip to content. | Skip to navigation

Personal tools
You are here: GES DISC Home Hydrology Additional Features Science Focus Seasonal Forcast

Seasonal Forcast

GLDAS improves sub-seasonal weather forecasts


Typical numerical weather forecasts attempt to predict atmospheric conditions several days into the future.  This capability is highly valuable for aviation, severe weather alerts, and everyday weather-related decisions.  However, predictions on such a short time horizon cannot inform decisions relevant to water resource planning, agriculture management, drought preparation, or other activities that require knowledge of weather conditions weeks to months in advance.  For application to these problems, sub-seasonal and seasonal forecasts are required.  The effort to predict weather several weeks to several months in advance is complicated by the fact that the “memory” of atmospheric conditions lasts only about ten days.  For forecasts longer than this, the initial, known atmospheric conditions provide little skill towards prediction.  Instead, the initialization of surface conditions takes on a larger significance.  Land surface states such as soil moisture, snow cover, and vegetation vary slowly relative to the atmosphere, such that the memory of an initial anomaly in one of these states can influence atmospheric processes for weeks or months. 

Estimates of land surface states and fluxes produced by GLDAS can be used to initialize numerical weather forecasts.  This is of particular value to the sub-seasonal and seasonal forecasts that are most sensitive to initialization of land surface states. 

Sample Application

Scientists at NASA GSFC performed a retrospective forecast study to assess the value of GLDAS for sub-seasonal weather prediction systems.  In this study, GLDAS was used to initialize 1-month forecasts with the seasonal prediction system of the Global Modeling and Assimilation Office (GMAO) at GSFC.  The skill of these forecasts was then evaluated against ground observations of precipitation and air temperature.  Over 75 simulations, it was found that GLDAS led to a substantial improvement in forecast skill relative to forecasts that did not use GLDAS to initialize land surface conditions.  GLDAS was particularly important for the second half of the monthly forecasts (i.e., three and four weeks into the future), after the memory of initial atmospheric states had already faded from the system. 


GLDAS ppotential to improve forecasting accuracy


Figure 1: Potential for GLDAS to contribute to skill in 1-month forecasts of air temperature.  Top panel shows the potential skill (assessed in the context of other sources of prediction uncertainty) for monthly forecasts that include GLDAS initialization of land surface conditions.  The middle panel shows the same metric for simulations that lack GLDAS initialization.  Bottom panel shows the difference, and thus the potential contribution of GLDAS to model skill. [Figure from Koster et al. (2004)]


GLDAS improvement in forcasting accuracy


Figure 2: Actual improvement in prediction skill for air temperature due to GLDAS, evaluated against field observations.  The geographic extent of the evaluation is limited by the lack of quality surface observations over much of the globe. [From Koster et al. (2004)]

Data Used

GLDAS has the flexibility to simulate land surface conditions using a number of advanced Land Surface Models (LSMs).  For this application, GLDAS was used to produce 15 years of surface states using the Mosaic LSM, as this is model is used in the GMAO seasonal prediction system.  The simulations utilized atmospheric forcing data produced by Berg et al. (2003).  These forcing data are of high quality and are available globally.  Both GLDAS-Mosaic outputs and Berg atmospheric forcing data are available for download through the GES DISC GLDAS interface.


Berg, A. A., J. S. Famiglietti, J. P. Walker, and P. R. Houser, 2003: Impact of bias correction to reanalysis products on simulations of North American soil moisture and hydrological fluxes. J. Geophys. Res., 108, 4490, doi:10.1029/2002JD003334.

Koster, R.D. and Coauthors, 2004: Realistic initialization of land surface states: impacts on subseasonal forecast skill. J. Hydrometeor., 5, 1049-1063.

Koster, R.D. and M. J. Suarez, 2003: Impact of land surface initialization on seasonal precipitation and temperature prediction. J. Hydrometeor., 4, 408–423.

Rodell, M., and Coauthors, 2004: The global land data assimilation system. Bull. Amer. Meteor. Soc., 85, 381–394.

Relevant Links

General information: The Global Land Data Assimilation System

GLDAS data at the GES DISC:

Document Actions
NASA Logo -
NASA Privacy Policy and Important Notices
Last updated: Mar 14, 2011 03:01 PM ET