Guest Post by Willis Eschenbach
Despite doing lots of research and investigations over the last few weeks, I’ve written little. Well, actually, I’ve published little, although I’ve written a lot. But I didn’t publish what I’d done, there was no wonder in it, no awe. So I tossed it all out and started “simply messing about”, as the Toad had it. For no apparent reason, I got to looking at the various methods for estimating the downwelling longwave radiation (DLR) based on surface conditions. DLR is the radiation emitted by the atmosphere which is directed downwards towards the earth. There’s a good summary of the various DLR estimation methods here.
In any case, I wanted to compare the estimated DLR to the DLR from the CERES satellite observations. I used the “Brunt” method, which calculates an “effective emissivity” from the vapor pressure. The vapor pressure in turn is calculated from the surface temperature. I subtracted the satellite observations from the Brunt estimate. Figure 1 shows the result.
Figure 1. Difference between the downwelling longwave radiation (DLR) as calculated by Brunt, and the downwelling longwave radiation dataset from the CERES satellite data.
I busted out laughing when I saw that graphic come up on the silver screen. I do my science visually, by painting the transformations and relationships in color. And in general, I have only the vaguest idea of what any given graphic will look like before it is displayed. So watching the graphics appear onscreen is like opening a line of scientific presents. Each one is unexpected, each one reveals new things.
This one was funny to me because it was such an excellent and detailed map of the desert and arid areas of the planet. From the Sahara to the Atacama Desert, the Gobi, the American Southwest, the Arabian Peninsula, the Namib Desert, it’s all laid out in precise detail. Heck, you can even see the green areas of Australia as a thin strip along the east and north coasts.
This is a curious result because the CERES satellite doesn’t measure water vapor … but what we have in Figure 1 is a map of water vapor. Over the desert areas we get less downwelling radiation than the estimate suggests, because water vapor is the main greenhouse gas. In the desert the air is so dry that more radiation escapes to space, and less is absorbed and radiated downwards (and upwards) by the atmosphere. It also shows the moistest areas of the planet (dark green and blue). These are in the equatorial tropical forests, where transpiration combines with evaporation. This leads to lots of water vapor, and a concomitant increase in DLR above what the estimate suggests.
This is the first time I’ve looked at the difference between a variable in the CERES dataset and an estimate of that variable. They say that all models are wrong, but some are useful. This model of downwelling longwave radiation is obviously wrong … but it’s useful because of exactly where and how much it is wrong.
Which leads to the final surprise for me, which was the size of the deviations from the expected DLR. From very dry regions to very wet regions is a range on the order of 100 W/m2 of downwelling LW radiation … I didn’t think it would be that big.
Anyhow, that’s the kind of thing I like to write about—the unexpected. For me, the adventure of science is never knowing which bush might be the one that hides the rabbit …
Regards to everyone,
w.
As Always: If you disagree with someone, please quote their exact words so we can all understand your objection.
A note on the Brunt Method: The Brunt method estimates the “effective emission” as a function of the form
a1 + a2 * sqrt( vapor_pressure )
Per the above citation, the canonical values for a1 and a2 are 0.51 and 0.066.
When I fitted the values, I got a1 and a2 as 0.65 and 0.029. I thought this might be a result of including the ocean. So I looked at just land, which gave a1 and a2 as 0.65 and 0.024. And looking at just the ocean I got 0.66 and 0.030. None of these are near the values given in the reference. However, they work quite well, and the canonical figures give much larger errors. Go figure.
