I read in many places that "reliable records" go back to 1880, and that is the oldest date that is often used when comparing today's climate to a pre-industrial climate. My question is whether pre or post 1880 records are more useful. I could see data since 1880 helping to shore up the dataset that is currently being used, but I could also see value in pre-1880 data to help extend the historical record further into the past. Any thoughts from the scientific team?
TL:DR - we need them all, any time, any place.
The question 'which data is most useful?' is very hard to answer - we are building a dataset of fundamental observations which will be used by many people for many purposes. We guarantee that what we are rescuing is useful by choosing ships in times and places where we know there is demand from scientists now, but we try not to focus too much on what the scientists I know want today, we also have an eye on tomorrow's users and their requirements.
The statement 'reliable records go back to 1880' refers to records of global-mean, annual-mean, temperature. (See http://hadobs.metoffice.gov.uk/hadcrut3/diagnostics/comparison.html
- most of these start in 1880). This is the highest-profile scientific product made from our observations (and many others) - it's how we track global warming. When we make such products from all the observations we calculate not only our best guess at the answer, but also how uncertain that best guess is - we produce error-bars. But lots of users don't like uncertainty estimates, they say 'just give us the best estimate for times when that estimate can be treated as reliable, so we don't have to worry about the error-bars' and for those people we say 'just use the data after 1880 - there are still uncertainties, but we are pretty confident the best-estimate is not very wrong'.
That means that if we rescue observations after about 1880 they are unlikely to make a large difference in our estimates of global warming, and if we rescue observations from earlier than 1880, they might make a large difference. But in reality it's not a hard cut-off, we want to reduce uncertainties at all times.
Even more important, the global-mean, annual-mean temperature is not the only thing we are interested in. We also need to know the temperature at 11a.m on May 3rd in Tianjin - how has that changed over the last 100+ years? is spring getting earlier in that location? And for this specific question reliable data don't start in 1880 (more like 1960). And there's nothing special about Tianjin, or May 3rd; we need all times, all places, and the wind, pressure, rain, ..., as well as temperature.
So the early data are more useful for global-mean estimates - but the later data are more useful for local weather estimates, we really do need both. Whatever observations we rescue, someone will use them.