Jeffrey S. Whitaker

Ensemble Data Assimilation with the NCEP Global Forecast System

Real-data experiments with an ensemble data assimilation system using the NCEP Global Forecast System model were performed and compared with the NCEP Global Data Assimilation System (GDAS). All observations in the operational data stream were assimilated for the period 1 January–10 February 2004, except satellite radiances. Because of computational resource limitations, the comparison was done at lower resolution (triangular truncation at wavenumber 62 with 28 levels) than the GDAS real-time NCEP operational runs (triangular truncation at wavenumber 254 with 64 levels). The ensemble data assimilation system outperformed the reduced-resolution version of the NCEP three-dimensional variational data assimilation system (3DVAR), with the biggest improvement in data-sparse regions. Ensemble data assimilation analyses yielded a 24-h improvement in forecast skill in the Southern Hemisphere extratropics relative to the NCEP 3DVAR system (the 48-h forecast from the ensemble data assimilation system was as accurate as the 24-h forecast from the 3DVAR system). Improvements in the data-rich Northern Hemisphere, while still statistically significant, were more modest. It remains to be seen whether the improvements seen in the Southern Hemisphere will be retained when satellite radiances are assimilated. Three different parameterizations of background errors unaccounted for in the data assimilation system (including model error) were tested. Adding scaled random differences between adjacent 6-hourly analyses from the NCEP–NCAR reanalysis to each ensemble member (additive inflation) performed slightly better than the other two methods (multiplicative inflation and relaxation-to-prior).

Mechanisms Determining the Atmospheric Response to Midlatitude SST Anomalies

Abstract Previous GCM experiments demonstrated that a model atmosphere produces two different responses to a midlatitude warm SST anomaly over the Pacific under perpetual January and February conditions. To elucidate the mechanisms responsible for the different GCM responses and their dependence on the background flow, experiments with two idealized models are conducted. Experiments with a linear baroclinic model reveal that the GCM responses at equilibrium are primarily maintained by the anomalous eddy forcing. The anomalous flow induced directly by an idealized initial heat source exhibits little sensitivity to the background flow. Eddy feedbacks on the heating-induced anomalous flow are examined using a linear storm track model. The anomalous eddy forcing produced by the storm track model is sensitive to the basic state. The eddy forcing in January acts to shift the heating-induced upper-level ridge toward the northeast of the Gulf of Alaska, while in February it acts to reinforce the ridge. This sugge...

Ensemble Reforecasting: Improving Medium-Range Forecast Skill Using Retrospective Forecasts

Abstract The value of the model output statistics (MOS) approach to improving 6–10-day and week 2 probabilistic forecasts of surface temperature and precipitation is demonstrated. Retrospective 2-week ensemble “reforecasts” were computed using a version of the NCEP medium-range forecast model with physics operational during 1998. An NCEP–NCAR reanalysis initial condition and bred modes were used to initialize the 15-member ensemble. Probabilistic forecasts of precipitation and temperature were generated by a logistic regression technique with the ensemble mean (precipitation) or ensemble mean anomaly (temperature) as the only predictor. Forecasts were computed and evaluated during 23 winter seasons from 1979 to 2001. Evaluated over the 23 winters, these MOS-based probabilistic forecasts were skillful and highly reliable. When compared against operational NCEP forecasts for a subset of 100 days from the 2001–2002 winters, the MOS-based forecasts were comparatively much more skillful and reliable. For examp...

REFORECASTS An Important Dataset for Improving Weather Predictions

A “reforecast” (retrospective forecast) dataset has been developed. This dataset is comprised of a 15-member ensemble run out to a 2-week lead. Forecasts have been run every day from 0000 UTC initial conditions from 1979 to the present. The model is a 1998 version of the National Centers for Environmental Predictions (NCEPs) Global Forecast System (GFS) at T62 resolution. The 15 initial conditions consist of a reanalysis and seven pairs of bred modes. This dataset facilitates a number of applications that were heretofore impossible. Model errors can be diagnosed from the past forecasts and corrected, thereby dramatically increasing the forecast skill. For example, calibrated precipitation forecasts over the United States based on the 1998 reforecast model are more skillful than precipitation forecasts from the 2002 higher-resolution version of the NCEP GFS. Other applications are also demonstrated, such as the diagnosis of the bias for model development and an identification of the most predictable patt...

The Relationship between Ensemble Spread and Ensemble Mean Skill

Abstract Statistical considerations suggest that 1) even for a perfect ensemble (one in which all sources of forecast error are sampled correctly) there need not be a high correlation between spread and skill, 2) the correlation between spread and skill should be larger where the day-to-day variability of spread is large, and 3) the spread is likely to be most useful as a predictor of skill when it is “extreme,” that is, when it is either very large or very small compared to its climatological mean value. The authors investigate the relationship between spread and skill in an operational setting by analyzing ensemble predictions produced by the National Centers for Environmental Prediction. The geographical dependence of the spread–skill relationship is found to be related to the geographical dependence of day-to-day variability of spread. Dynamical mechanisms for spread variability are investigated using a linear quasigeostrophic model. Problems associated with the sample size needed to define what const...

Diagnosing Sources of U.S. Seasonal Forecast Skill

In this study the authors diagnose the sources for the contiguous U.S. seasonal forecast skill that are related to sea surface temperature (SST) variations using a combination of dynamical and empirical methods. The dynamical methods include ensemble simulations with four atmospheric general circulation models (AGCMs) forced by observed monthly global SSTs from 1950 to 1999, and ensemble AGCM experiments forced by idealized SST anomalies. The empirical methods involve a suite of reductions of the AGCM simulations. These include uni- and multivariate regression models that encapsulate the simultaneous and one-season lag linear connections between seasonal mean tropical SST anomalies and U.S. precipitation and surface air temperature. Nearly all of the AGCM skill in U.S. precipitation and surface air temperature, arising from global SST influences, can be explained by a single degree of freedom in the tropical SST field—that associated with the linear atmospheric signal of El Nino–Southern Oscillation (ENSO). The results support previous findings regarding the preeminence of ENSO as a U.S. skill source. The diagnostic methods used here exposed another skill source that appeared to be of non-ENSO origins. In late autumn, when the AGCM simulation skill of U.S. temperatures peaked in absolute value and in spatial coverage, the majority of that originated from SST variability in the subtropical west Pacific Ocean and the South China Sea. Hindcast experiments were performed for 1950–99 that revealed most of the simulation skill of the U.S. seasonal climate to be recoverable at one-season lag. The skill attributable to the AGCMs was shown to achieve parity with that attributable to empirical models derived purely from observational data. The diagnostics promote the interpretation that only limited advances in U.S. seasonal prediction skill should be expected from methods seeking to capitalize on sea surface predictors alone, and that advances that may occur in future decades could be readily masked by inherent multidecadal fluctuations in skill of coupled ocean–atmosphere systems.

The midlatitude warming during 1998-2000

Large scale tropospheric warming of the Northern and Southern Hemisphere extratropics occurred during the period January 1998-January 2000. The anomalies were remarkable for their amplitude, their persistence, their zonal symmetry with warming occurring at all longitudes between 30-50 degrees latitude, and their axial symmetry with respect to the equator. The origin of the anomalies are investigated using atmospheric general circulation models forced by the global sea surface temperatures (SSTs) of this period. Two ensembles, based on different models, reproduce the key observed features indicating that the unusual climate state was forced by the global SSTs. However, the atmospheric states were unrelated to the strong and protracted La Nina of the east equatorial Pacific Ocean during this time. We speculate on the role of the Indo-Pacific warm pool, which itself experienced unprecedented warming in 1998 and 1999.

Subseasonal Variations of Tropical Convection and Week-2 Prediction of Wintertime Western North American Rainfall

Abstract A statistical prediction model for weekly rainfall during winter over western North America is developed that uses tropical outgoing longwave radiation (OLR) anomalies as a predictor. The effects of El Nino–Southern Oscillation (ENSO) are linearly removed from the OLR to isolate the predictive utility of subseasonal variations in tropical convection. A single canonical correlation (CCA) mode accounts for most of the predictable signal. The rank correlation between this mode and observed rainfall anomalies over southern California is 0.2 for a 2-week lag, which is comparable to correlation between a weekly ENSO index and weekly rainfall in this region. This corresponds to a doubling of the risk of extreme rainfall in southern California when the projection of tropical OLR on the leading CCA mode two weeks prior is extremely large, as compared with times when it is extremely small. “Extreme” is defined as being in the upper or lower quintile of the probability distribution. The leading CCA mode rep...

The Representation of Tropospheric Water Vapor Over Low-Latitude Oceans in (Re-)analysis: Errors, Impacts, and the Ability to Exploit Current and Prospective Observations

This paper addresses the representation of lower tropospheric water vapor in the meteorological analyses—fully detailed estimates of atmospheric state—providing the wide temporal and spatial coverage used in many process studies. Analyses are produced in a cycle combining short forecasts from initial conditions with data assimilation that optimally estimates the state of the atmosphere from the previous forecasts and new observations, providing initial conditions for the next set of forecasts. Estimates of water vapor are among the less certain aspects of the state because the quantity poses special challenges for data assimilation while being particularly sensitive to the details of model parameterizations. Over remote tropical oceans observations of water vapor come from two sources: passive observations at microwave or infrared wavelengths that provide relatively strong constraints over large areas on column-integrated moisture but relatively coarse vertical resolution, and occultations of Global Positioning System provide much higher accuracy and vertical resolution but are relatively spatially coarse. Over low-latitude oceans, experiences with two systems suggest that current analyses reproduce much of the large-scale variability in integrated water vapor but have systematic errors in the representation of the boundary layer with compensating errors in the free troposphere; these errors introduce errors of order 10% in radiative heating rates through the free troposphere. New observations, such as might be obtained by future observing systems, improve the estimates of water vapor but this improvement is lost relatively quickly, suggesting that exploiting better observations will require targeted improvements to global forecast models.