Cross-posted to the EA-Forum 

Epistemic Status: I have spent less than 15 minutes thinking about / querying the Internet for answers on what I am going to ask here. As such, it might be the case that this idea has already been discarded due to implementation or theoretical difficulties, or due to it being a mostly useless exercise. The following idea constitutes an attempt to get more concrete results on how humans might react and adapt to a catastrophic event where anywhere between 1-99.9% of the population perishes. I believe the expected value of this line of work is high, even if the probability of success is low. With more searching, I might be able to produce a preliminary experimental design, but whether I do this depends somewhat on the reaction of those in this forum. 

The roughly 8 billion humans on Earth make up around 0.01% of Earth's life. Despite making up such a small proportion of life globally, humans have a strong presence across all continents (barring Antarctica), and roughly ~5% of the Earth's surface is habitable by humans. There has been increasing variability in Earth's climate over the last several million years, yet humans have historically adopted well to these changes. I imagine the past, current, and future climate-regions of Earth could be modeled (perhaps using some form of unsupervised learning). There are natural and anthropogenic global catastrophic risks that humanity faces. Many of these risks might leave regions of Earth uninhabitable to humans for some time. For example, nuclear war or super volcanic eruptions could leave many regions agriculturally unsustainable, engineered pandemics might result in large areas being quarantined indefinitely, and a large solar flare might result in some of Earth's climates becoming uninhabitable. These risks might also substantially shift the current climate-regions of Earth.

If a line of bacteria or virus could be bred to reflect human migration and habitation patterns (e.g., "Two-thirds of the world’s population is located within four major clusters: East Asia (China), South Asia (India and Indonesia), Southeast Asia, and Europe, with the majority in East and South Asia." would mean that roughly 2/3rds of the bacteria are located in 4 regions), then it might be possible to determine a rough estimate of humanity's robustness to the direct effects and subsequent environmental effects of a global catastrophe in terms of population resilience and migration patterns. The environment that these organisms exist in would have to reflect the distribution of the climatic regions (and their variability) that humans currently inhabit. An example could be modeling what a major (>10% of global population killed) asteroid strike in North America would do by shifting all the "climate-regions" of the culture appropriately and eliminating the culture directly near the strike, and observing how it reacts. 

Are there any prospects with this idea? 

I imagine the immediate problems would be (1) modeling the Earth's climate regions and climate variability in a culture is difficult (2) it is also probably difficult to create genetically some bacteria or virus that follows the parameters of human migratory and population change (3) bacteria don't have civilizations or technological progress, so using bacteria as a model of "humans across the globe", even if everything was implemented correctly, might only reflect the tendencies of human hunter-gatherers