(Dyurgerov and Meier 2000, Dyurgerov 2003). Hence, it is reasonable to assume that precipitation changes induce variability (or noise) in the glacier length record but do not control its global mean pattern. Regional patterns, on the other hand, are in some cases dominated by precipitation or other variables. For example, in the late 20th century, increased snowfall caused glaciers in western Scandinavia to advance (Dyurgerov and Meier 1997a,b, 2000), whereas combinations of precipitation and temperature changes have induced retreat of the glaciers in equatorial Africa (Mölg et al. 2003, Kaser et al. 2004, Hastenrath 2005). Although warming in recent decades is an important factor driving glacier recession on Mt. Kenya and the Ruwenzori summits, the much higher altitude glaciers on Kilimanjaro may be shrinking primarily as a continuing response to precipitation changes earlier in the century. The magnitude and importance of recent warming are still being researched.
Temperature reconstructions based on glacier length and mass balance records are limited in their temporal and spatial resolution. They do not provide a year-to-year view of temperature change, but only averages over several years to decades (depending on the resolution of the length measurements and on the accuracy of assumptions in the physics). They do not provide any information about most of the globe prior to the 19th century. Only the North Atlantic and European Alpine regions have glacier records back to around A.D. 1600, and even in these regions there is little information prior to the 17th century. The time required for data collection, compilation, and reporting has so far prevented the most recent 15 years from being included in the analysis. In North America, many of the glacier records end between the mid-1970s and 1990 (Oerlemans 2005a), so for this region the late 20th century reconstruction is not yet reliable. Geographic limits arise from the obvious fact that glaciers do not exist everywhere, so the low and middle elevations of the low latitudes are entirely absent. There is also a paucity of data for the Southern Hemisphere. Finally, these reconstructions cannot be done for mountain glaciers in Antarctica because it is so cold there that melt is not the dominant mass loss process (it is iceberg production), so the connection to temperature is different.
Though not suitable for reconstructing temperature time series, other glacier indicators—such as melting on ice caps, organic material uncovered when glaciers melt, and disintegration of ice shelves—provide temperature information. An increase of summertime warmth over the last 150 years caused increased melt on Ellesmere Island’s major ice cap in the Canadian Arctic. This extent of melt had not occurred in the previous 1,500 years (Fisher et al. 1995).
The recent retreat of glaciers has exposed organic material that would have decomposed if not covered by ice, including a human body (the now famous “Ice Man” of the Alps) and plant material (Thompson et al. in press). Three of these finds have been dated (from the Alps, from Washington State, and from Peru) and all have ages greater than 5,000 years before present. This suggests rather strongly that the current deglaciation is unprecedented in the last few millennia at these widespread sites. Nonetheless, it is known from dating of organic material transported to the fronts of glaciers in the Alps that glacier recessions more extensive than the present one have occurred at some sites in Europe (Hormes et al. 2001), with dates ranging from A.D. 800 to