El Niño in the Climate Forecasting System: T62 vs. T126 ......El Niño in the Climate Forecasting...
1
El Niño in the Climate Forecasting System: T62 vs. T126 or Dance, Dance, Resolution!!! Cécile Penland (CDC-PSD1/ESRL/NOAA) and Suranjana Saha (EMC/NCEP/NOAA) Linear Inverse Modeling (LIM) uses a combination of lagged and contemporaneous covariance matrices to provide the best- fit linear operator to the multivariate SSTA field. If the system is linear, then this operator does not strongly depend on the lag at which it was estimated. We estimate the leading eigenvalue of the corresponding right singular vector as a function of lead time for operators estimated with lags of 3 months (dotted lines), 4 months (solid lines) and 5 months (dashed lines). Blue: T62. Green: T126. Red: COADS, estimated with a lag = τ₀ 4 months. This eigenvalue represents the field variance growth factor over a specified lead time, given the right singular vector (“optimal structure”) as an initial condition of the linear model, and this curve is called a Maximum Amplification Curve. In both models and data, the optimal structure is dominated by 3 normal mode pairs. Unfortunately, the data optimal is contaminated by a trend. Trend mode COADS: The optimal structures (left) and what they grow into (right): If we reconstruct the SSTA data using only the 3 leading normal mode pairs, the Maximum Amplification Curve Changes a bit and so do the optimal structures. Now, the models agree somewhat better with the observations, but there are still differences. Not shown: the optimals still grow into El Niños and the character of the Niño 3.4 spectra don't change a lot. Niño 3.4 spectra: Red: COADS. Blue: T62. Green: T126. Conclusion: The resolution of the atmospheric component of the model matters a lot! The T126 does get the El Niño spectrum about right. Speculation: Should we trust process studies using models whose atmos- pheres don't vary enough? Does increasing the resolution of the CFS's atmospheric component affect the model's El Niño? Yes! To see this, we prepare CFS output as we do COADS data: project SSTs onto a 4° 10° grid, subject to a 3-month ╳ running mean, and then project onto 20 leading EOFs. The Niño 3.4 SSTA time series: COADS(top; red), T62(left; blue), T126 (right, green).
Transcript of El Niño in the Climate Forecasting System: T62 vs. T126 ......El Niño in the Climate Forecasting...
El Niño in the Climate Forecasting System: T62 vs. T126 or Dance, Dance, Resolution!!!Cécile Penland (CDC-PSD1/ESRL/NOAA) and Suranjana Saha (EMC/NCEP/NOAA)
Linear Inverse Modeling (LIM) uses a combination of laggedand contemporaneous covariance matrices to provide the best-fit linear operator to the multivariate SSTAfield. If the systemis linear, then this operator does not strongly depend on the lagat which it was estimated.
We estimate the leading eigenvalue of the corresponding rightsingular vector as a function of lead time for operators estimatedwith lags of 3 months (dotted lines), 4 months (solid lines) and 5months (dashed lines). Blue: T62. Green: T126. Red: COADS,estimated with a lag =τ₀ 4 months. This eigenvalue represents thefield variance growth factor over a specified lead time, given theright singular vector (“optimal structure”) as an initial condition ofthe linear model, and this curve is called aMaximum AmplificationCurve.
In both models and data, the optimal structure is dominated by 3 normalmode pairs. Unfortunately, the data optimal is contaminated by a trend.
Trend mode
COADS:
The optimal structures (left) and what they grow into (right):
If we reconstruct the SSTAdata using only the 3leading normal mode pairs, the MaximumAmplification Curve Changes a bit
and so do the optimal structures. Now, the modelsagree somewhat better with the observations, butthere are still differences. Not shown: the optimalsstill grow into El Niños and the character of theNiño 3.4 spectra don't change a lot.
Niño 3.4 spectra: Red: COADS. Blue: T62. Green: T126.
Conclusion:
The resolution of the atmosphericcomponent of the model matters a lot!The T126 does get the El Niño spectrumabout right.
Speculation:
Should we trustprocess studiesusing modelswhose atmos-pheres don'tvary enough?
Does increasing the resolution of the CFS's atmosphericcomponent affect the model's El Niño? Yes!
To see this, we prepare CFS output as we do COADS data:project SSTs onto a 4° 10° grid, subject to a 3-month╳running mean, and then project onto 20 leading EOFs.
The Niño 3.4 SSTAtime series: COADS(top; red), T62(left;blue), T126 (right, green).
![Ba Rainstar T32-T62 ab 2014 engl - bauer-at.com0].pdf · Operation manual for BAUER RAINSTAR Series T32-T62 2 Introduction Thank you for buying BAUER RAINSTAR T ! The present manual](https://static.fdocuments.in/doc/165x107/5e27e304686c345e202790e2/ba-rainstar-t32-t62-ab-2014-engl-bauer-atcom-0pdf-operation-manual-for-bauer.jpg)
