Climatic data is gathered today with precision instruments, including satellites that can cover nearly the whole planet. Instrumental measurements are the gold standard, but there aren't enough of them for truly satisfactory science.
Thermometers were invented in the 1600s, and widespread temperature measurements date only from the mid-1800s. Even so, large areas of the world have poor data coverage for the pre-satellite era. The same is true for precipitation, air pressure, winds, humidity and sea ice, which are all essential data. One way to fill in the blanks is to assume that the world had weather patterns similar to today's in the past. Given the various climate oscillations discussed before as well as the possibility of long-term changes, this assumption is well-known to be a weak one. Estimates must be tested against other kinds of information that is a proxy or substitute for intrumental records.
Human Proxies
Documentary evidence is helpful, such as ships' logs, newspapers, government records and people's diaries and letters. These record softer data like dates of frost, planting and harvesting dates, significant weather events (including famine) and flowering dates. This information is clearly much less reliable than instrumental data. But it can help constrain the gross climatic conditions for large regions or for the hemisphere.
Human proxies are used elsewhere in geology—judging earthquake intensity is a prime example. Better is to find natural recorders of climate data, which are less dependent on human whims and occur all over the world including the ocean floor. There are many of these, but the ones most useful for our purposes can indicate climate year by year, or at least with a decade's resolution.
Natural Proxies
Tree ring data (dendrochronology) can yield information on seasonal temperature and wet or dry conditions. At best, tree rings are comparable to instrumental averages. However, suitable tree species are limited to parts of the temperate latitudes, and many trees need to be sampled and carefully assessed for good results. Even so, tree ring records are only useful for studying regional climates.
Corals can yield good data on water temperatures through isotopic techniques, although changes in salinity may affect it. Whereas tree rings can be used to construct records thousands of years back, coral chronologies are much shorter. But the ability to extract ocean climate information—again, on a regional basis—is promising.
Ice cores show promise for several different kinds of information: air temperature, precipitation, volcanic dust and sulfates, cosmic isotopes created by solar activity, and greenhouse gases in trapped air bubbles. Unfortunately glaciers cover a small portion of the Earth's surface.
Other lesser proxies include cavestone formations (speleothems) and the annual sediment layers, called varves, found in high-latitude lakes and rare places in the sea. Several other proxies are under study that are not ready for prime time.
Once the proxy data is collected, it must be calibrated against accurate instrumental records. But often the whole point of using proxies is that no instrumental records exist, so calibration can be a chancy affair with more assumptions than scientists would prefer.
The next step is to assemble the proxy data and use it to make estimates for the whole globe, or at least the northern or southern hemisphere. This step requires more assumptions and the liberal use of statistical methods. But even a fuzzy picture, which is what we get in the end, is better than no picture at all.
