2 ms·

In time series analysis picking the right time scale is everything. Let´s do a small experiment: One hot summer day I measure the temperatures in my home town

In time series analysis picking the right time scale is everything. Let´s do a small experiment: One hot summer day I measure the temperatures in my home town between 6:00 and 12:00 and I get some data like this: 6:00 - 10 C ... 12:00 - 32 C Now I make a small regression analysis and say that at 24:00 we are going to die a painfull death because the temperature will raise with a probability of 95% to 70-80 C. In this case I obviously forgot about the periodicity of daily temperatures. The global temperatures have a periodicity of ca. 400.000 years. Therefore any analysis based on a time period shorten than 400k years has the same problems as my summer experiment. >> the end of the century our planet might be experiencing environmental conditions never previously encountered by our species - 2*2 might be 5... - at the end of the century I will be dead and "our species" will be only one of million other species. For example the dinosaurs loved the 12 C more and had a happy life on the ice-free antarktis a couple of million years ago...

shoo 11y ago
So what are you arguing? That it is not possible to come to any conclusion regarding the earth climate without first waiting some multiple of 400k years? That it's not possible to separate a long-term 400k-year temperature variation trend from short term temperature variations due to unusual/novel forcing conditions? The thing your hot summer day analysis is missing is a clear physical explanation that explains why the trend you are observing is occurring, and why it is going to keep occurring until 24:00. We probably don't even need the explanation to be 100% accurate for this to be useful, merely accurate to a first or second order approximation. I believe "we" [1] have got enough understanding of the earth system to make some pretty reasonable inferences on limited data. Certainly not complete understanding, but it's not like a bunch of people are merely using e.g. some deep learning voodoo to perform regression on climate data and have literally no idea what the underlying physical processes involved might be. Here's an example of my own: consider the geoengineering "launch a bunch of disk-shaped satellites to block x% of the sun's light from hitting the earth" thing. Let's ignore the fact that this isn't necessarily a great idea, and assume someone has gone ahead and done it. We can predict with reasonable confidence what the primary consequences of this intervention will be: less light from the sun hits the planet, planet's mean temperature lowers a bit. This behaviour won't be at all explainable as part of a 400k year planetary periodicity. [1] not in the sense of you or i, necessarily, but the relevant scientists in the field.

shoo 11y ago

So what are you arguing? That it is not possible to come to any conclusion regarding the earth climate without first waiting some multiple of 400k years? That it's not possible to separate a long-term 400k-year temperature variation trend from short term temperature variations due to unusual/novel forcing conditions? The thing your hot summer day analysis is missing is a clear physical explanation that explains why the trend you are observing is occurring, and why it is going to keep occurring until 24:00. We probably don't even need the explanation to be 100% accurate for this to be useful, merely accurate to a first or second order approximation. I believe "we" [1] have got enough understanding of the earth system to make some pretty reasonable inferences on limited data. Certainly not complete understanding, but it's not like a bunch of people are merely using e.g. some deep learning voodoo to perform regression on climate data and have literally no idea what the underlying physical processes involved might be. Here's an example of my own: consider the geoengineering "launch a bunch of disk-shaped satellites to block x% of the sun's light from hitting the earth" thing. Let's ignore the fact that this isn't necessarily a great idea, and assume someone has gone ahead and done it. We can predict with reasonable confidence what the primary consequences of this intervention will be: less light from the sun hits the planet, planet's mean temperature lowers a bit. This behaviour won't be at all explainable as part of a 400k year planetary periodicity. [1] not in the sense of you or i, necessarily, but the relevant scientists in the field.