About this blog
Metamodel: a model that consists of statements about models.
This is a blog about the language, science, and philosophy of predictive modeling. The aim is to be more whimsical than polemical, and mostly non-technical. The discussions will generally be about climate models, which are arguably the most complex models ever built. But models from related fields will make a guest appearance now and then. The discussions will frequently venture beyond science because the prediction of long-term climate change is now so intertwined with the cultural milieu that it is impossible to discuss it within a purely scientific context.
As a climate scientist, I have worked with a variety of models for over three decades, ranging from very simple to highly complex. Every time you, your company, or your government plan for the future, you are relying on the output of models, whether or not you are aware of it. But models are imperfect abstractions of reality. You need to understand models to use them properly. You don’t need fancy mathematics or a massive supercomputer to understand models and their predictions. This blog will try to show that we can learn a lot from simple logical reasoning using basic physical and mathematical concepts.
Climate models are the essential tools used by IPCC for assessing our climate futures to help guide mitigation and adaptation. As statistician George Box observed, “all models are wrong; but some are useful.” It is only by analyzing how models are wrong can we figure out how best to use them. This blog will critically analyze climate and other models. The purpose of the critiques is not to diminish the serious threat of climate change, but to increase the efficacy of the urgent actions needed to mitigate it.
R. Saravanan
Twitter: @RSarava Website: r.saravanan.us Book: The Climate Demon
PS. This blog is mirrored on Substack, if you prefer to subscribe to it as a “newsletter”.
Long-form blogs seem passé in this age of short-form Twitter and Tiktok. But long-form articles are still important, because many complex issues cannot be discussed efficiently using a short format. At the other extreme, one can use the really long-format of a book to discuss the science and philosophy of modeling. But books are not free, they take time to read, and do not address current developments. Blog posts are free, relatively quick to read, and can address emerging issues.
There are a few climate blogs that are still around, but they are less active. Perhaps because climate denial has shifted from attacking the science to attacking the solutions. But models, climate and otherwise, continue to play an important role in climate solutions. There is perhaps still a role for a blog that discusses modeling and prediction.
The website metamodel.blog is the primary home of this blog. Keeping with current trends, posts will be announced on Twitter and you may also comment on posts by replying to the “official” announcement tweet. Due to algorithmic ranking, you may not see my tweets announcing new posts even if you follow me on Twitter. Also, not everyone is on Twitter. Therefore, all posts will be mirrored on metamodel.substack.com to provide a free subscription option for those who prefer to receive posts via email.
Human nature abhors a prediction vacuum. People always want to know, and often need to know, what may happen in the future. If a particular source (say, astronomy) can’t provide that information, people will tend to go to a different source that is willing to provide that information (say, astrology). Until the weather service started issuing seasonal forecasts using computer models, people relied on folksy predictions from the Old Farmer’s Almanac or groundhogs named Phil. Today, predictions from scientific models are used to make decisions that affect millions of people, often saving many billions of dollars.
Models, scientific or otherwise, will always be used to make decisions. But models are frequently misunderstood by the general public. All scientists can do is to help ensure that the most appropriate models are used and that their predictions are interpreted with the appropriate caveats.
The landscape of models is somewhat like the wild west – there’s the good, the bad, and sometimes even the ugly. It’s not easy for an outsider to figure out which is which because there are no clear rules. The word model itself can mean very different things in diverse fields such as economics, epidemiology, physics, and climate science. This often results in outsider misconceptions about how models in a particular field work.
In climate science, models are not used only for predicting the future, but also to improve our understanding of phenomena. For example, simple nonlinear models are used for qualitative understanding of amplifying climate feedbacks, such as the release of methane from melting permafrost or the increased reflection of sunlight due to melting icesheets. But such simple models aren’t necessarily good at making actionable quantitative predictions. Misunderstanding the limitations of models leads to people panicking about tipping points at specific temperature thresholds or believing in prophecies of imminent doom.
Complex models, which include numerous processes, are used by IPCC and others to make quantitative predictions of the future. But these complex models do not necessarily include all the climate feedbacks that can be studied using simpler models, because we often do not have sufficient data, or powerful enough computers, to accurately represent these feedbacks. To paraphrase a famous quote, we can only predict with the most comprehensive models we have, not with even more comprehensive models that we wish to have in the future.
As society has become increasingly reliant on models for climate risk assessment, there are many important questions need to be addressed, such as:
- What phenomena can be usefully predicted by models?
- How well can these phenomena be predicted?
- How to choose the best model(s) to use?
- Do we really need to use the most complex and expensive models for a particular problem of interest?
The purpose of this blog is to provide the background information to help answer these questions. What you can expect:
- new posts at roughly 2—3 week intervals
- some posts on fundamental, but unresolved, climate and modeling issues
- some posts on recent developments and publications
- “non-technical” discussions, featuring a mix of science and pop philosophy
- (guest posts on modeling-related issues are welcome)
As a programmer, I like to roll my own solutions and retain creative control (at the expense of inconvenience). This blog is implemented using open-source software on a small dedicated virtual linux server. It uses a static web site generator called Hugo, with the Blist theme and Nginx as the web server. The site is designed to be mobile and social-media friendly, in keeping with the times.
Comments on blog posts are handled using Remark42, a privacy-focused open-source commenting engine. Commenting on the site requires a “social login” to avoid spam. Alternatively, you can simply reply to the “official” tweet announcing the blog post to comment on it.
Modifications were made to the Blist theme and Remark42 integration to tweak the appearance and functionality of the blog. All the custom code modifications are available on Github, but not fully documented yet.
The simple markup language Markdown is used to format all the content on the local computer. After previewing locally, the content is pushed to the linux server. Markdown is also supported for comment entry.
Blog posts are mirrored on Substack as newsletters. Simply copying and pasting the Hugo-generated web output to the Substack editor appears to work fine for posting (except for additional formatting like footnotes). This extra bit of effort allows flexibility and avoids vendor lock-in, while still having access to the popular Substack platform.
Thanks to the miracle of the universal markup converter, Pandoc, all the posts on this blog are automatically converted to an eBook using the ePUB/PDF formats. You can download and view the eBook offline. A new book is created each time a new post is added. Individual articles are also downloadable as ePUB or PDF files.