1. METR 5970.002 Advanced Atmospheric Radiation Dave Turner Lecture 7

2. From Thorsten Mauritsen et al. 2012

4. From a slide taken from Thorsten Mauritsen Understanding Clouds and Precipitation Workshop Berlin, Germany, February 2016

5. The LBLRTM It is a line-by-line radiative transfer model – Absorption line parameters from HITRAN – Continuum model is MT_CKD Developed by Atmospheric Environmental Research, Inc. (AER) Supported primarily by DOE/ARM, NASA, some European agencies, Air Force (in early days) Heavily validated with airborne, space-based, and ground- based spectrometer observations Instructions on how to use the model are provided on web Greg has a script to write input files (called TAPE5s)

6. The “Clough Plot” (Provided in the JGR 1995) Cooling Warming Spectral line absorption by different species Vertical concentration and gradient of absorbers T-dependence of the Planck function Clough et al., JGR 1995 Heavily smoothed spectrally!!

7. ~10% change ~18% change ~16% change ~80% change

8. Impact on Downwelling Zenith Radiance by Changing CH4 from 1.2 ppm to 1.8 ppm (Using US Standard Atmosphere)

9. Spectral Resolution of the Model LBLRTM provides “monochromatic” optical depths Determines the effective line width from the ambient pressure Model uses fixed number of points to resolve shape of the line Need to perform RT on 2641 spectral elements!

10. So Interpolation is Needed Need to interpolate (or average) to a common grid Details matter! Using this average: RT on 50 spectral elements

11. So Interpolation is Needed Need to interpolate (or average) to a common grid Details matter! Using this average: RT on 25 spectral elements

12. So Interpolation is Needed Need to interpolate (or average) to a common grid Details matter! Using this average: RT on 12 spectral elements

13. Instruments have finite spectral bandpass Atmospheric radiation is monochromatic Instruments require a non-zero spectral bandpass, otherwise the energy detected would be zero – The spectral characteristics of this bandpass is called the “instrument response function (IRF)”, “spectral response function (SRF)”, or “instrument line shape (ILS)” in the literature Effectively, the observed radiance signal R IRF is a convolution of the IRF with the monochromatic radiance R mono To properly model the instrument, monochromatic radiance should be computed and then the convolution performed The problem: Monochromatic has very high spectral resolution, and hence is computationally expensive – Can convolution be applied earlier to reduce the number of elements in the spectral domain used for the RT? – Ultimately, for GCMs will need to have pretty coarse spectral resolution

14. Applying the convolution earlier 1)Apply the IRF convolution to the monochromatic layer optical depths, and then perform the radiative transfer (RT) – Easily understood – Works well in spectral regions where optical depth is changing slowly 2)Applying the IRF convolution to the monochromatic layer transmittance (which was computed from the optical depth), then perform the RT – Improves the accuracy somewhat because there is a smaller dynamic range in transmittance space (0-1) than in optical depth space (several orders of magnitude) 1)Compute monochromatic layer-to-instrument transmittances, apply the IRF convolution, and then take ratios to rederive “effective layer optical depths” that are at the reduced spectral resolution, then perform the RT – Seems to provide best solution – Need to watch for “divide by zero” situations

15. Example of a monochromatic downwelling radiance spectrum This was computed from U.S. Standard Atmosphere, and will serve as my baseline For the following examples, I will assume that the IRF is a boxcar filter with different spectral widths

19. 19 Regularly spaced absorption lines Ellasser model Highly irregularly spaced absorption lines Random model In the CO 2 Absorption Band In the H 2 O Absorption Band

20. Correlated-K Method Region with Primarily CO 2 Absorption 20

21. Correlated-K Method Region with Primarily H 2 O Absorption 21

22. Why is is called correlated-k?

23. Typical Clear Sky Spectral Heating Rate Profiles in the Infrared RRTM (Rapid Radiative Transfer Model) uses correlated-k method – RRTM – 16 bands in LW (12 shown here) – RRTM – 13 bands in the SW – 16 g-values per spectral band (256 calcs for LW, 208 calcs for SW) – Extensively validated using ARM observations Cooling Warming Clough et al., JGR 1995

24. Example of Longwave and Shortwave Heating Rate Profiles 24 Longwave Shortwave Net Radiative Heating

25. RRTM is Too Computationally Expensive! Radiation codes have long been recognized to be very slow, even when significant spectral simplifications are made – E.g., band models, correlated-k models, etc Solution: do not run RT model on each dynamical model time step Model Grid Number Model Time Steps RT model every n th step Frequency

26. RRTM With Monte Carlo Sampling RRTM too slow for routine GCM or NWP models Solution: Randomly select a subset of the g-values for each band – Model is called RRTM-G Greatly accelerates code, but adds random noise Noise has virtually no impact; key is that there is no bias in RT Model Grid Number Model Time Steps RT model every n th step Frequency

27. Different cases Current Status of LW RT Accuracy Intercomparison of 11 different LW RT models from GCMs against observations Some models still off by ~3%, which is large considering CO2 doubling is a 4 W/m 2 signal 27 Percent errors Different parameterizations RRTM and RRTM-G look very good!

28. Current Status of SW RT Accuracy 28 Percent errors

29. The Next Step: MC Sampling in Time and g As dynamical model spatial resolution increases, RT processes become more important (especially in cloudy scenes) Only running RT every nth time step no longer adequate Moving towards running RRTM-G every model time step, but with fewer g- values (being tested now) Model Grid Number Model Time Steps

30. 1)The concentration of CO2 in preindustrial times was approximately 280 ppm. (Today it is nearly 400 ppm.) Depending on the how you want to predict human/government behavior, the CO2 concentration will reach 560 ppm sometime in the next 100 years. For each of the 6 standard atmospheres encoded in the LBLRTM (i.e., tropical, mid-latitude summer, mid-latitude winter, sub-arctic summer, sub-arctic winter, and US Standard), compute the change in the infrared (5-100 µm) flux (W/m 2 ) caused by the CO2 doubling for both (a) downwelling at the surface and (b) upwelling at the TOA radiation. Assume that the surface of the earth is black (emissivity of 1) with a skin temperature of 300, 295, 273, 288, 258, and 289 K for each of the standard atmospheres, respectively, for the upwelling calculations. 1)The preindustrial concentration of methane (CH4) was about 1.2 ppm in 1750. (Today it is over 1.8 ppm.) If CH4 doubles from its preindustrial concentration, what would be the impact on the upwelling and downwelling fluxes at the TOA and the surface for the 6 standard atmospheres, respectively? 1)The PWV of the 6 standard atmospheres are 4.1, 2.9, 0.9, 2.1, 0.4, and 1.4 cm for tropical, mid-latitude summer, mid- latitude winter, sub-arctic summer, sub-arctic winter, and US Standard atmospheres, respectively. Does this help explain the results in the above two problems? Why or why not? 1)There have been many studies showing that if CO2 doubles that the average surface temperature on the planet would increase somewhere between 3 and 7 degC. (Even Arrhenius in the late 1800s predicted an increase of about 4 degC for a CO2 doubling). Based on what you saw with problems (1)-(3), are you surprised? 1)For one of the standard atmospheres, compute the monochromatic optical depths for a model atmosphere that extends from the surface to 70 km AGL. Compute the upwelling (nadir) radiation monochromatically, assuming a black surface with the appropriate temperature given in (1) above. Then compute the upwelling radiance for different boxcar widths (0.2, 1.0, 5.0 cm-1) where you perform (a) average in layer optical depth space, (b) average in layer transmission space, and (c) average in layer-to-instrument transmission space. Compare these three estimates with the monochromatic truth that was boxcar averaged after the RT was performed. How similar / different are these results from the downwelling calculations? Assignment #3 Due 5 April