Instrument records of temperature and rainfall reach back only a few centuries at best. Everything earlier is reconstructed from natural materials that recorded conditions as they formed.
Proxies record conditions indirectly
A proxy is any natural archive whose growth or chemistry responded to weather at the time it formed. It does not measure temperature; it responds to it in a measurable way.
The value of a proxy depends on whether that response can be calibrated. Researchers compare the proxy against overlapping instrument records to establish the relationship.
Once calibrated, the same relationship is applied backward, with the important caveat that conditions outside the calibration range are reconstructed less confidently.
Tree rings hold annual detail
Trees in seasonal climates add one growth ring per year, and ring width and density vary with the growing conditions of that particular year.
In arid parts of the American West, ring width tracks moisture closely, which has allowed drought reconstructions covering many centuries in some river basins.
Rings from living trees are matched against overlapping patterns in dead wood, timbers and archaeological beams, chaining the record further back than any single tree lived.
Ice cores trap the atmosphere itself
Snow that never melts compresses into ice while sealing bubbles of air. Those bubbles hold a physical sample of the atmosphere from the year the snow fell.
The chemistry of the ice itself carries a temperature signal, while layers of ash and dust mark volcanic eruptions and provide dating anchors.
Because layers thin under their own weight with depth, annual resolution gives way to coarser averaging in the oldest sections of a core.
Corals, sediments and caves fill the gaps
Reef corals add banded skeletal growth whose chemistry reflects sea surface conditions, making them useful for tropical ocean history where few other archives exist.
Lake and ocean sediments accumulate pollen, shells and mineral grains in layers, which record vegetation and ocean conditions over far longer spans at lower resolution.
Mineral deposits in caves grow slowly from dripping water and preserve rainfall signals in regions where trees and ice offer nothing usable.
Confidence comes from agreement, not any one archive
Every proxy has weaknesses. Trees respond to more than temperature, sediments can be disturbed, and corals cover limited spans of time.
The way around this is combination. When independent archives on different continents describe the same shift, the reconstruction rests on far more than one method's assumptions.
Uncertainty widens as reconstructions reach further back, which is why paleoclimate results are published with ranges rather than single lines.