Researchers evaluate sensitivities of the NCEP Global Forecast System
Accurate weather and climate forecasts are important for environmental planning and safety, and atmospheric scientists are continually looking for ways to improve predictions. Broadly speaking, forecast systems have three basic elements: 1) the input observations, 2) the data assimilation (DA) method used to include those observations in the forecast initial conditions, and 3) the forecast model itself. In a new study published in Monthly Weather Review, CIRES and NOAA researchers at the Physical Sciences Laboratory evaluated the relative sensitivities of the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) to changes in these three basic elements.
NOAA and partner agencies deployed a ship, aircraft, and ground-based field campaign to collect in-depth observations of the atmospheric response over the equatorial Pacific during a strong El Niño event in early 2016. The broader goal of this El Nino Rapid Response (ENRR) field campaign was to help evaluate how well operational weather forecast systems represent key processes and features associated with tropical weather systems, in addition to whether the real-time assimilation of such field campaign data could improve operational weather forecasts. It was anticipated that detecting the impacts would require extensive post-campaign computer modeling experiments, such as those performed in this new study. Interestingly, the study found that the impact of the ENRR observations was much smaller than the impacts of improving the data assimilation system and the inclusion of “stochastic parameterizations” of chaotic physical processes in the forecast model.
In data-denial experiments, the researchers found that there was a slight improvement in 1- to 7-day forecast errors over the globe due to the addition of ENRR observations. Using two different data assimilation methods and comparing their forecast results, they also found that forecast errors in the GFS are much more sensitive to the data assimilation method, and using a more sophisticated “Hybrid 4D-EnVar” method is generally better. And finally, in turning on and off stochastic parameterizations, they found that stochastic parameterizations significantly reduce forecast errors in seven days in general, except for 200mb height in the tropics.
The study found that the impact of the inclusion of stochastic parameterizations in the forecast model and improvements to the data assimilation system were much larger than the impact of ENRR observations in reducing weather forecast errors in NCEP’s current Global Forecast System. Careful sensitivity tests of the type reported in the paper help to identify the factors that most impact weather forecast system performance in general.
Authors of the journal article Sensitivities of the NCEP Global Forecast System are: Aaron Wang, Prashant Sardeshmukh, Gilbert Compo, Jeff Whitaker, Laura Slivinski, Chesley McColl, and Philip Pegion of the ESRL Physical Sciences Laboratory.
Posted: April 30, 2019
Accurate weather and climate forecasts are important for environmental planning and safety, and atmospheric scientists are continually looking for ways to improve predictions. Broadly speaking, forecast systems have three basic elements: 1) the input observations, 2) the data assimilation (DA) method used to include those observations in the forecast initial conditions, and 3) the forecast model itself. In a new study published in Monthly Weather Review, CIRES and NOAA researchers at the Physical Sciences Laboratory evaluated the relative sensitivities of the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) to changes in these three basic elements.
NOAA and partner agencies deployed a ship, aircraft, and ground-based field campaign to collect in-depth observations of the atmospheric response over the equatorial Pacific during a strong El Niño event in early 2016. The broader goal of this El Nino Rapid Response (ENRR) field campaign was to help evaluate how well operational weather forecast systems represent key processes and features associated with tropical weather systems, in addition to whether the real-time assimilation of such field campaign data could improve operational weather forecasts. It was anticipated that detecting the impacts would require extensive post-campaign computer modeling experiments, such as those performed in this new study. Interestingly, the study found that the impact of the ENRR observations was much smaller than the impacts of improving the data assimilation system and the inclusion of “stochastic parameterizations” of chaotic physical processes in the forecast model.
In data-denial experiments, the researchers found that there was a slight improvement in 1- to 7-day forecast errors over the globe due to the addition of ENRR observations. Using two different data assimilation methods and comparing their forecast results, they also found that forecast errors in the GFS are much more sensitive to the data assimilation method, and using a more sophisticated “Hybrid 4D-EnVar” method is generally better. And finally, in turning on and off stochastic parameterizations, they found that stochastic parameterizations significantly reduce forecast errors in seven days in general, except for 200mb height in the tropics.
The study found that the impact of the inclusion of stochastic parameterizations in the forecast model and improvements to the data assimilation system were much larger than the impact of ENRR observations in reducing weather forecast errors in NCEP’s current Global Forecast System. Careful sensitivity tests of the type reported in the paper help to identify the factors that most impact weather forecast system performance in general.
Authors of the journal article Sensitivities of the NCEP Global Forecast System" are: Aaron Wang, Prashant Sardeshmukh, Gilbert Compo, Jeff Whitaker, Laura Slivinski, Chesley McColl, and Philip Pegion of the ESRL Physical Sciences Laboratory.
Posted: April 30, 2019