widespread climatic warming, which increased ice loss by melt (Oerlemans 2001, 2005a; Dyurgerov and Meier 2000; Oerlemans et al. 1998; Zuo and Oerlemans 1997; Oerlemans and Fortuin 1992). In the tropics, glacier retreat was apparently caused by both temperature changes and moisture changes, depending on the time and region (Kaser 1999, Francou et al. 2000, Mölg et al. 2003, Vuille et al. 2003b, Kaser et al. 2004, Mark and Seltzer 2005). Warming is the dominant factor in the tropical Andes (Thompson et al. 2000b, Vuille et al. 2003b, Mark and Seltzer 2005), whereas moisture change has an influence in equatorial Africa (Mölg et al. 2003, Kaser et al. 2004). On average, snowfall rates increased modestly on these glaciers during the latter half of the 20th century (Dyurgerov and Meier 2000), providing strong evidence against drying as a cause of retreat for this time period.
Records of glacier length changes during the 20th century and earlier have been analyzed to reconstruct past temperature changes (Oerlemans 1994, 2005a; Oerlemans et al. 1998). These analyses are based on glacier physics (Paterson 1994, Van der Veen 1999, Oerlemans 2001) and provide temperature information that is independent of other temperature reconstruction methods. In particular, the temperature reconstruction from glacier retreat is not calibrated against the instrumental record. Instead, it is based on glacier dynamics models that are highly generalized but calibrated against a small number of extensively studied glacier systems (Oerlemans et al. 1998, Oerlemans 2001) for which realistic glacier dynamics and energy balance modeling has been completed. Glacier-length-based temperature reconstructions are further significant because the influences of urban heat islands and land use changes are likely to be small.
Oerlemans (2005a) showed that temperatures increased by about 0.6°C from the late 19th to the mid-20th century and were persistently cool in the three centuries before this (Figure 7-1). Results also show a small cooling from 1950 to 1980, followed by renewed warming until the record ends in 1990. Data on recent and ongoing retreats have not yet been extensively compiled. It is known, however, that mass loss from mountain glaciers increased strongly in the 1990s due primarily to accelerated warming (Dyurgerov 2003, Meier et al. 2003, Dyurgerov and Meier 2005).
During their period of overlap (1850–1990) the glacier length reconstruction and the instrumental record are very similar in magnitude and pattern (Oerlemans 2005a). The glacier length record thus validates the quantitative accuracy of the instrumental composite for this period as a large-scale average. Reconstructed temperatures from the glacier length record are similar for low- and high-altitude glaciers (Oerlemans 2005a). The dominant pattern of cool climate for a few centuries followed by warming beginning in the late 19th century is seen in all geographic regions examined, though significant differences in details exist.
The most recent and comprehensive temperature reconstruction (Oerlemans 2005a) uses glacier length records for a large number of glaciers (Oerlemans et al. 1998, Oerlemans 2001). The information needed for detailed individual modeling for most of these glaciers does not exist, so the analysis instead uses an approximate and