Tropical cyclones and climate change
When looking at tropical cyclones in relation to climate change, it is normal to look at both their intensity and their frequency. This IPCC SREX summary report states that there is low confidence in any increases in frequency or intensity of these events:
There is low confidence in any observed long-term (i.e., 40 years or more) increases in tropical cyclone activity (i.e., intensity, frequency, duration), after accounting for past changes in observing capabilities.
However, in a recent paper: Downscaling CMIP5 climate models shows increased
tropical cyclone activity over the 21st century (Emanuel, 2013), a case is made that both the intensity and the frequency of tropical cyclones could increase over the next few decades. This paper uses 6 models from the CMIP5 ensemble of General Circulation Models (GCMs), and downscales them (embeds a model of finer resolution into the GCM) so as to be able to better resolve the storms themselves. From the paper:
From this, they conclude that running the model at a higher resolution provides a better estimate for the overall storm intensity. The results that they found, namely that intensity and frequency will increase by 8-80% and 11-41% respectively, are interesting because they are at odds with a similar study carried out on an early generation and current generation models - CMIP3 and CMIP5 - in this paper: Dynamical Downscaling Projections of Twenty-First-Century Atlantic Hurricane Activity: CMIP3 and CMIP5 Model-Based Scenarios (Knutson et al., 2013). In this paper, the CMIP3 models showed that the intensity of storms would increase by 87%, but that the frequency of tropical cyclones would decrease by 27%. The more recent CMIP5 models found similar results: storms would increase by 39-45% and frequency would decrease by 20-23%.One distinct advantage of our downscaling technique is that it captures the full spectrum of storm intensity
In an earlier review article: tropical cyclones and climate change (Knutson et al, 2010), both current (in 2010) generation models run at normal resolution, and higher resolution models are considered. It finds that high-resolution models project that storms will become stronger by 2-11% by 2100, which is in qualitative agreement with the above papers, but the magnitude of the change is much smaller. It was written before both of the above papers, which could explain the discrepancy. The article goes on to say that for the current generation of models, storm frequency is projected to decrease by 6-34%. Whereas higher resolution models "project substantial increases in the frequency of the most intense cyclones". The article points out that the use of downscaling can have a noticeable effect on the results in terms of frequency of stronger storms:
There is a clear tendency among the models, particularly at higher resolution (60-km grid spacing or less), to project an increase in the frequency of the stronger tropical cyclones
Conclusions
There seems to be a fairly broad consensus that the intensity of the largest storms will increase over the next century. However, there seems to be considerable variation in the projections for frequency between different models on one hand, and different studies on the other. The role of downscaling appears to be important to the outcomes of these studies, with models run at a higher resolution tending to project higher frequencies of storms.
Interesting that changes to not even the model, but just the grid size can have such a dramatic difference in the predictions generated. I guess you, and the rest of the climate modelling community, have your work cut out to improve the robustness of these models!
ReplyDeleteIs it always considered the case that a smaller mesh size leads to a better model prediction (and current sizes are bounded by compute power) or are there cases when making the mesh size smaller and smaller could lead to inherent instabilities in the model? If the former then it would be interesting to see if running the same scenarios with smaller and smaller mesh sizes leads to convergence of the predictions for storm intensity and frequency.
It is interesting isn't it, and another interesting question is just how sensitive are these models to the size of grid cells that are used. For looking at cyclones in particular, it does make sense that the grid cell size has a big effect: if the grid cell size is larger than the extent than your cyclone, you're not going to see it! Another question is whether modelling cyclones affects the behaviour of the model at a larger scale.
DeleteI think it's fair to say that most climate modellers would prefer to run at higher and higher resolutions, but they are, as you say, constrained by compute power. The question that you asked about instabilities is a good one, and my hunch is that it could lead to instabilities if you make the resolution fine enough. Even if there are instabilities at a small cell size, there might still be no point in running models with smaller and smaller cell sizes due to diminishing returns, perhaps running at 20km cell size is almost as good as running at 5km, in which case why use the extra resources?! I haven't been able to find any literature on this yet, but I'll have a look and write a post about it if I find something good.
You begin by referring to the lack of empirical evidence of a trend in cyclone intensity and frequency over the last few decades, citing the IPCC SREX Summary Report. However, I have heard various claims in the media that there is such a trend. Indeed, no less an authority than Prince Charles has claimed this, indeed relating it to man-made climate change (http://www.dailymail.co.uk/news/article-2510660/Prince-Charles-Philippines-Typhoon-Haiyan-direct-result-man-climate-change.html).
ReplyDeleteIs it your view that the IPCC SREX report represents the most considered analysis of all the available evidence, and therefore trumps other studies which have come to a different conclusion? Or could it be that they are being ultra cautious, and that the balance of empirical evidence does indeed point to a trend?
Robert Muetzelfeldt
The IPCC is, to my knowledge, known for being conservative in its estimations. As an example, Stefan Rahmstorf argues that they were likely underestimating sea level rise back in 2007. Also, the IPCC only accepted submissions for papers for AR5 up until the 31st of Jan 2013, so it certainly couldn't have considered the two 2013 papers cited above, which might have meant that the IPCC could have used stronger language with regards to cyclone intensitiy/frequency.
DeleteI am of the opinion that the IPCC has to be ultra cautious about what it says, because if there is any inaccuracy (as there was about the projected melting of glaciers in the 2007 report) then rightly or wrongly this will undermine the rest of its findings. So I wouldn't be surprised if e.g. the actual levels for sea level end up being at the top end or higher than the IPCC projected values.
The issue that Prince Charles touches upon, of whether one particular event can be attributed to climate change, is an interesting one. He certainly went too far in saying Haiyan was a 'direct result' of climate change given what we know at the moment. But scientists are developing new ways of attributing specific events to the changes in climate, so it is possible that a future study will prove him right! I'll be looking into this in future posts.
An interesting post - I'd come across the Knutson paper before. However, it seems to me that it's not the global number, or intensity, of cyclones that matters but regional changes. What is the state of knowledge concerning e.g. Indian vs. Indonesian cyclones? Could the global average intensity rise, yet Caribbean storms decrease in intensity? If some areas see decreasing cyclone activity what are the implications for climate politics?
ReplyDeleteIn the analyses that I've come across so far, the two main regions of study have been the North Atlantic and the Pacific, so they are not nearly fine grained enough to say what changes are expected on a country to country scale. Having said that, it's possible, maybe even likely, that different countries would experience different effects from these changes, and it would obviously be of great importance to these individual countries. So I'd expect to see some more targeted analysis in the near future, so as the countries affected can start to make plans for how to adapt to these changes.
DeleteIf some areas see a reduction in activity and others see an increase, then it would seem fair that the lucky area should take measures to help the unluckier area, possibly through some sort of regional insurance scheme. The debate is mired in uncertainties though: how to attribute and quantify the reduction in one area to climate change, how to enforce any agreements. It reminds me of a similar issue, that of heat waves vs cold snaps. It is likely that heat waves will go up as the climate warms, and that cold snaps will similarly decrease. This overwhelmingly benefits countries that are colder to begin with, i.e. predominantly countries at higher latitudes. Should they then be liable to help out countries that are suffering the effects of worse heat waves? A thorny issue indeed!