Supervolcanos seem to pose the highest risk, as we sit on the ocean of molten iron, oversaturated with dissolved gases, just 3000 km below surface and its energy slowly moving up via hot spots. Many past extinctions are also connected with large eruptions from supervolcanos.
Impacts also pose a significant risk. But, if we project the past rate of large extinctions due to impacts into the future, we will see that they occur only once in several million years. Thus, the likelihood of an asteroid impact in the next century is an order of magnitude of 1 in 100 000. That is negligibly small compared with the risks of AI, nanotech, biotech, etc.
The main natural risk is a meta-risk. Are we able to correctly estimate natural risks rates and project them into the future? And also, could we accidentally unleash natural catastrophe which is long overdue?
There are several reasons for possible underestimation, which are listed in the right column of the map.
1. Anthropic shadow that is survival bias. This is a well-established idea by Bostrom, but the following four ideas are mostly my conclusions from it.
2. It is also the fact that we should find ourselves at the end of period of stability for any important aspect of our environment (atmosphere, sun stability, crust stability, vacuum stability). It is true if the Rare Earth hypothesis is true and our conditions are very unique in the universe.
3. From (2) is following that our environment may be very fragile for human interventions (think about global warming). Its fragility is like fragility of an overblown balloon poked by small needle.
4. Also, human intelligence was best adaptation instrument during the period of intense climate changes, which quickly evolved in an always changing environment. So, it should not be surprising that we find ourselves in a period of instability (think of Toba eruption, Clovis comet, Young drias, Ice ages) and in an unstable environment, as it help general intelligence to evolve.
5. Period of changes are themselves marks of the end of stability periods for many process and are precursors for larger catastrophes. (For example, intermittent ice ages may precede Snow ball Earth, or smaller impacts with comets debris may precede an impact with larger remnants of the main body).
Each of these five points may raise the probability of natural risks by order of magnitude in my opinion, which combined will result in several orders of magnitude, which seems to be too high and probably is "catastrophism bias".
(More about it is in my article “Why anthropic principle stopped to defend us” which needs substantial revision)
In conclusion, I think that when studying natural risks, a key aspect we should be checking is the hypothesis that we live in non-typical period in a very fragile environment.
For example, some scientists think that 30 000 years ago, a large Centaris comet broke into the inner Solar system, split into pieces (including Encke comet and Taurid meteor showers as well as Tunguska body) and we live in the period of bombardment which has 100 times more intensity than average. Others believe that methane hydrates are very fragile and small human warming could result in dangerous positive feed back.
I tried to list all known natural risks (I am interested in new suggestions). I divided them into two classes: proven and speculative. Most speculative risks are probably false.
Most probable risks in the map are marked red. My crazy ideas are marked green. Some ideas come from obscure Russian literature. For example, an idea, that hydro carbonates could be created naturally inside Earth (like abiogenic oil) and large pockets of them could accumulate in the mantle. Some of them could be natural explosives, like toluene, and they could be cause of kimberlitic explosions. http://www.geokniga.org/books/6908 While the fact of kimberlitic explosion is well known and their energy is like impact of kilometer sized asteroids, I never read about contemporary risks of such explosions.
The pdf of the map is here: http://immortality-roadmap.com/naturalrisks11.pdf