I think 1750 would be ideal but starting in 1800 would include simulation of the major volcanic eruptions which is important for post-1850 climate. When we suggested these ideas for CMIP7 the modelling centres saw the value but were not keen on spending (limited) computational resources on this.

We performed parallel simulations using 1750 and 1850 frameworks.

When starting in 1750, the simulated climate for lengthy periods after 1850 is different from simulations started in 1850, even with the same radiative forcings.

This is mainly because of volcanoes & land use changes before 1850.

Might you want to be a volunteer rainfall observer and contribute to the ongoing monitoring of UK climate?
www.climate-lab-book.ac.uk/2020/wanted-...

Can’t possibly be as interesting as a discussion of Stephenson Screens, 5-inch copper rain gauges and 150-year old pieces of paper with hand written numbers on.

Wonder if @leohickman.carbonbrief.org knows?

Data rescue, often by digitizing old paper records, is one of the ways that climate science research improves our understanding of the recent past.

It's hard to know how much useful data is still hiding in unremarkable paper volumes in a government warehouse somewhere.

I estimate >1 billion daily or sub-daily observations in Met Office archives alone. This includes air temperature, pressure, rainfall, sunshine, cloudiness, winds etc on land and sea.

Congratulations Gavin - very well deserved!

We are experimenting with AI too, e.g. climatelabbook.substack.com/p/data-rescu... but for handwriting the volunteers are (at the moment) still more accurate, and that reduces the manual QC required. Lots of typeset sheets for the AI tools to rescue first while they improve further.