2009-05-31

ClimateWizard.org: A web resource?



Having just ready this brief article [LINK] about a new web site called ClimateWizard.org, I was skeptical of its usefulness. Essentially, this is a web application that just makes reasonably nice map representations of some climate model runs (from the last IPCC report) guided by user input. The input is limited to what geographic region to map, which admittedly is the only way most people think to look at these kinds of simulations. The user can choose the USA and zoom in on specific states, or can choose "global" and choose individual countries and territories. Not everything seems to work (e.g., Guam doesn't show any data), but most of the buttons I clicked generated maps. The user also gets to choose a specific month, or annual average, and can plot the average or change in surface temperature or precipitation either for the past 50 years, the projections at 2050, or (for the USA) projections for 2100. As of writing this, the choices also including picking one of three models or the model average. It is unclear what exactly is being plotted for the past 50 years, the documentation indicates it is a combined observational data set, but there are two curves in a small timeseries plot that are not obviously explained(figure). All in all, it is a fun little application, though not the grandiose life-altering tool that the news article seems to suggest.



Now, a little criticism to go along with the overview. There are definitely really good things about this application, including using Google Maps as the underlying mapping framework. This is nice because it works well, looks good, and people know how to deal with it. That is a definite win. There are some limited download options, too, which is good so that people (especially students??) can just grab the image and put it into a document. They can also get the data in GIS format, which is good for some small set of people who know what to do with that. It would be nice to expand that to include some other common data formats, like HDF or netCDF. Of course, that would be mostly helpful to more experienced users, which gets to another point. This application, as a framework for quickly looking at data and downloading a small subset of it, seems like it could be really useful for all kinds of people, from policy makers to serious scientists, and back to students (and grad students) and teachers (and college profs). In its current state, it's a toy though. The limitation to these three models is curious, since the data from 20+ models should be readily available, and the limitation to surface temperature and precipitation are arbitrary. Yes, that is what most people will be interested in, but having an advanced tab and then having a lot more variables would be great. For example, maybe a high school or college student is doing a report on climate feedbacks, and wants to see how cloud fraction changes in the models. It's a trivial thing to add to this app, but its absence means the student will have to rely on secondhand information or be much more industrious. My other criticism is that the color choices for the contour maps are not that good. It's the standard blue-to-green-to-yellow-to-red rainbow. It works reasonably well for surface temperature and less well for precipitation. The real problem is that they apply it to the changes, which for most places are small making them light yellow or greenish. It's especially apparent in the precipitation, where it is often difficult to even discern the sign of the change.

Okay, one more important criticism. These global models do a poor job in representing regional climate, and there is little evidence that they accurately predict changes in regional climate to the level that this application implies. You can, for example, zoom in on Haiti and see that the model average temperature change at 2050 is about 1C (except for some whitish points that don't make sense on the color scale). But these are about 15 grid points, which are not the native grid points from the model, which have fraction amounts of land and almost no representation of the topography of Haiti. What I'm saying is that you can't trust a few points from one (or three) models to give an accurate portrayal of climate change unless you can verify the fidelity of the models at those points using observations and have an understanding of the limitations of the modeling framework. Looking at the past 50 years is helpful to ascertain some sense of the models' ability in the region, but as mentioned above, it isn't even clear what is being plotted for that in the application. If the blue line is the models and the black line is the observations, then the correlation is generally pretty poor for Haiti, right, but the trend is pretty decent, it's just that the average of the models doesn't have much variability (unless it is the other way around, or if the blue line is one of the models instead of the ensemble... but these are not explained).

So, overall, I like ClimateWizard.org, but I think there is a lot of room for improvement, both for the present purposes and for future expansions to more serious data exploration.

No comments: