A correction, I have read the Kump & Pollard paper [Amplification of Cretaceous Warmth by Biological Cloud Feedbacks -- Kump and Pollard 320 (5873): 195 -- Science], which I discussed a little yesterday.
I looked at it again briefly, and the main things to add are (1) their experiment is what I thought, which is that they artificially change the effective droplet radius. This is fine, as it is artificially prescribed anyway. And (2) they provide a reference for the assertion that DMS is the primary non-anthropogenic CCN. That reference is actually a perspective in Science by Andreae(DOI: 10.1126/science.1136529). I've just read that piece, which is interesting, but I feel is deeply flawed. Rather than pick it apart, let me just say that my primary concern is that it does not seem to give a broad overview of the current measurements, but cherry picks a few studies that may or may not be designed well enough to get at the points being made. In the end, the firm conclusion is that pre-human aerosol distributions, specifically CCN-sized aerosol, is very poorly understood, and that no good way to guess the concentrations has been devised yet. Good points, and that is all I have to say of that article.
More on CCN later though, as this is a good area to explore!
2008-06-19
2008-06-18
biology in climate models
Just read a quick blurb in Nature [link] that kind of rankled me. The writer, David Beerling, worries about the fidelity of biological processes in climate models. Fair enough. He cites interesting work by Kump & Pollard that suggest a deficiency in climate modeling of the Cretaceous period might be (partly) due to lower numbers of cloud condensation nuclei because phytoplankton are more stressed and produce less dimethylsuphide, producing more cloud cover but thinner clouds.
In the last paragraph, Beerling writes that the results are unsatisfying because "the effects of heat on biological aerosol emissions need to be better described in their model for it to generate really solid conclusions."
I hardly know where to start. I think the comment is fair to an extent, but perhaps misguided.
Starting from the Kump & Pollard paper, which I admit I haven't read yet, I am not convinced that there's much evidence for this biological effect. If they've artificially changed the aerosol concentration or the CCN concentration, then it's almost a foregone conclusion that there'll be big effects in the simulation. That's one of those parameters which, in most large-scale models, is not well constrained and is set to help make a reasonable current climate. That probably means that it could be adjusted to make a "reasonable" past climate, too, but knowing what that means is a different story. The second issue is whether DMS is really such an important source of CCN. I know it is a source, but does stressing phytoplankton really have so much influence on the mean cloud field?
For that matter, this affect would mostly influence regions of low, stratiform cloud. Other regions are probably not that influenced by modest changes in CCN concentration, as the low clouds are mostly convective anyway (and I'm guessing salt would be their main source of CCN -- I could be wrong). So the result is dependent on the way these clouds are parameterized as well as the assumptions about the biological processes influencing CCN. Sounds shaky. I'll look at that paper and post an update. If I can find anything, I'll also post something about the source for CCN, and whether DMS is really that important.
As for adding biology to climate models, I'm very hesitant about the issue. There are potentially important climatic feedbacks. And if we could construct models for the biological interactions, that would help with long-term climate simulations, especially for future climate change and paleoclimate simulations. It'd be great. We're at the very nascent stages though. Current generation climate models incorporate some kind of land model that has simple biology, and usually isn't interactive (meaning the plants don't respond to changing climatic conditions). The ocean usually has no biological model, but there are some marine ecosystem models that exist and are being tested for the next generation of climate models. The complexity ranges dramatically from very simple to quite complex, but there is still much debate about the results from models incorporating these ecosystem components. This is a slippery slope though. Adding ecosystem models seems like a great idea, but if they are true models with prognostic equations, it usually means more expensive simulations (more computer time, more storage, and more human time to analyze the output). And where do we draw the line? Phytoplankton respond not just to temperature and salinity, which are the state variables in ocean models, but also to light availability (varies with depth) and micro- and macronutrients like nitrogen and phosphorous. Should we have include nutrient models, or prescribe the nutrient distributions. Well, if we try to prescribe them, then there will be biologists who (rightfully) will say that the models are missing the feedback between changes in nutrients and biology, which propagates upstream to the rest of the climate model. If we do somehow include nutrient models (perhaps by making the biological production simple diagnostics from available nutrients), then geologists and geographers and social scientists will argue that the processes that are the natural and anthropogenic sources and sinks of nutrients are not properly represented. For example, factories and power plants emit a lot of sulphur, but there's a lot of variability among factories and power plants, as well as seasonal and daily cycles in their emissions. Do we need to simulate these cycles to properly represent the emissions to the atmosphere? If so, that would mean that for climate change simulations, we'll have to model the changing energy needs of populations, since that will impact the amount of emissions from the powerplants. Meanwhile, current climate models to not properly represent processes like wave breaking at the ocean surface which is probably one of the main sources of CCN over the ocean; when will we get to add this? For that matter, different kinds of aerosol have totally different properties as CCN, and what happens chemically to these things in cloud droplets can have an influence on future cloud evolution, so should our microphysical models incorporate the chemistry of individual chemical species within cloud droplets? How do we do that?
My point is not that we should not add more complexity to climate models, only that we don't know how to do it. Even if we stick to the ocean, ice, atmosphere, and land surface, we know we're missing huge chunks of well understood physics. To venture into the chemical and biological world, much less the anthropological, seem hasty at this point. I think we'll do it; I mean, it is being done, but we have to remember that climate models are tools for understanding the natural world. They won't be able to provide solid answers to questions without much thought; climate models can not be run as black boxes with the results taken as truth, no matter how much we add to them.
In the last paragraph, Beerling writes that the results are unsatisfying because "the effects of heat on biological aerosol emissions need to be better described in their model for it to generate really solid conclusions."
I hardly know where to start. I think the comment is fair to an extent, but perhaps misguided.
Starting from the Kump & Pollard paper, which I admit I haven't read yet, I am not convinced that there's much evidence for this biological effect. If they've artificially changed the aerosol concentration or the CCN concentration, then it's almost a foregone conclusion that there'll be big effects in the simulation. That's one of those parameters which, in most large-scale models, is not well constrained and is set to help make a reasonable current climate. That probably means that it could be adjusted to make a "reasonable" past climate, too, but knowing what that means is a different story. The second issue is whether DMS is really such an important source of CCN. I know it is a source, but does stressing phytoplankton really have so much influence on the mean cloud field?
For that matter, this affect would mostly influence regions of low, stratiform cloud. Other regions are probably not that influenced by modest changes in CCN concentration, as the low clouds are mostly convective anyway (and I'm guessing salt would be their main source of CCN -- I could be wrong). So the result is dependent on the way these clouds are parameterized as well as the assumptions about the biological processes influencing CCN. Sounds shaky. I'll look at that paper and post an update. If I can find anything, I'll also post something about the source for CCN, and whether DMS is really that important.
As for adding biology to climate models, I'm very hesitant about the issue. There are potentially important climatic feedbacks. And if we could construct models for the biological interactions, that would help with long-term climate simulations, especially for future climate change and paleoclimate simulations. It'd be great. We're at the very nascent stages though. Current generation climate models incorporate some kind of land model that has simple biology, and usually isn't interactive (meaning the plants don't respond to changing climatic conditions). The ocean usually has no biological model, but there are some marine ecosystem models that exist and are being tested for the next generation of climate models. The complexity ranges dramatically from very simple to quite complex, but there is still much debate about the results from models incorporating these ecosystem components. This is a slippery slope though. Adding ecosystem models seems like a great idea, but if they are true models with prognostic equations, it usually means more expensive simulations (more computer time, more storage, and more human time to analyze the output). And where do we draw the line? Phytoplankton respond not just to temperature and salinity, which are the state variables in ocean models, but also to light availability (varies with depth) and micro- and macronutrients like nitrogen and phosphorous. Should we have include nutrient models, or prescribe the nutrient distributions. Well, if we try to prescribe them, then there will be biologists who (rightfully) will say that the models are missing the feedback between changes in nutrients and biology, which propagates upstream to the rest of the climate model. If we do somehow include nutrient models (perhaps by making the biological production simple diagnostics from available nutrients), then geologists and geographers and social scientists will argue that the processes that are the natural and anthropogenic sources and sinks of nutrients are not properly represented. For example, factories and power plants emit a lot of sulphur, but there's a lot of variability among factories and power plants, as well as seasonal and daily cycles in their emissions. Do we need to simulate these cycles to properly represent the emissions to the atmosphere? If so, that would mean that for climate change simulations, we'll have to model the changing energy needs of populations, since that will impact the amount of emissions from the powerplants. Meanwhile, current climate models to not properly represent processes like wave breaking at the ocean surface which is probably one of the main sources of CCN over the ocean; when will we get to add this? For that matter, different kinds of aerosol have totally different properties as CCN, and what happens chemically to these things in cloud droplets can have an influence on future cloud evolution, so should our microphysical models incorporate the chemistry of individual chemical species within cloud droplets? How do we do that?
My point is not that we should not add more complexity to climate models, only that we don't know how to do it. Even if we stick to the ocean, ice, atmosphere, and land surface, we know we're missing huge chunks of well understood physics. To venture into the chemical and biological world, much less the anthropological, seem hasty at this point. I think we'll do it; I mean, it is being done, but we have to remember that climate models are tools for understanding the natural world. They won't be able to provide solid answers to questions without much thought; climate models can not be run as black boxes with the results taken as truth, no matter how much we add to them.
Subscribe to:
Posts (Atom)