Glaciers in western North America advance or retreat in response to changes in winter accumulation and synoptic conditions during summer (e.g., Burbank, 1982; Bitz and Battisti, 1999; Moore and Demuth, 2001). The net mass balance of Sentinel Glacier in Garibaldi Park, for example, is positively correlated with winter precipitation (r²= 0.70) and negatively correlated (r²= -0.32) with summer temperature (Bitz and Battisti, 1999). Inter-annual to inter-decadal changes in winter accumulation in western North America are influenced by large-scale, ocean-atmospheric processes, including the El Niño-Southern Oscillation (ENSO) (Shabbar and Khandekar, 1996; Shabbar et al., 1997), and the Pacific Decadal Oscillation (PDO), an index of inter-annual to inter-decadal North Pacific sea surface temperature anomalies (Mantua and Hare, 2002). The PDO and ENSO exert their strongest control in the mountains of western North America during winter (Mantua and Hare, 2002).

Glaciers that have a persistent positive net mass balance will adjust their areas, lengths, and thickness to accommodate the increased mass. They typically leave moraines and trimlines, delimiting their former extents long after they have receded. Conspicuous moraines and trimlines in Garibaldi Park delineate the maximum recent cover of ice during the Little Ice Age (Fig. 2, 3).

The time of the Little Ice Age maximum in Garibaldi Park may differ from glacier to glacier by as much as decades due to differences in glacier response times, but Figures 2 and 3 give a sense of the amount of ice loss in the park since the most extensive advance of the Little Ice Age in the early eighteenth century. At least 630 km² (33 %) of the park was covered by ice in the early eighteenth century. Ice cover today is 390 km², a loss of 38 % in less than 300 years (Koch et al., 2003a). Moraines dating to the late nineteenth century are very close to Little Ice Age maximum ice limits, thus most of the ice loss has happened in the last 150 years. In fact, much of the thinning and retreat date to the last 80 years (Fig. 3) and occurred during two intervals, one between about 1930 and 1960 and another starting in the late 1970s and continuing to the present. The two retreat phases were separated by a stillstand or minor re-advance of many glaciers in the late 1960s and early 1970s (Fig. 3). Glaciers in other parts of the Canadian Cordillera have fluctuated in a similar manner to those in Garibaldi Park since the late nineteenth century (e.g., Osborn and Luckman, 1988; Luckman, 2000).

Ages of glacier advances can be estimated by dating plant remains in lateral and terminal moraines, or by determining the age of trees or lichens growing on the moraines (Fig. 4, 5;Luckman, 2000). Changes in tree growth may reveal the duration of an advance, as both react to the same forcing (Luckman, 2000). Cooler climate leads to glacier advance and, commonly, to narrow rings in temperature-sensitive trees (Luckman, 1996, 1998). Some tree species, for example mountain hemlock (Peterson and Peterson, 2001), may likewise form narrow rings during years when the snowpack is thick and lingers late into the summer.

The oldest tree on a moraine provides a minimum estimate for the time of moraine stabilization (Fig. 4; Lawrence, 1946). The tree ages need to be adjusted to account for the time between moraine stabilization and seedling germination ("ecesis"; McCarthy and Luckman, 1993; Winchester and Harrison, 2000). Photographic records and field studies indicate that the ecesis interval in Garibaldi Park is generally less than 20 years, but in some situations can be as long as 70 years (Koch, unpublished data).

The age of a glacial advance can also be estimated by dating living trees damaged during the advance, dead trees in growth position overridden by the glacier (in situ stumps), and detrital wood within moraines or lying on the surface in the glacier forefield (Fig. 5; Luckman, 1998, and references therein). The most valuable ages are those derived from in situ trees, especially if the rings can be cross-dated into living chronologies to obtain the year when the tree was damaged or killed. Detrital wood and damaged and dead trees in growth position are present in the forefields of glaciers in Garibaldi Park, allowing us to acquire a more complete Holocene glacier record than has been possible in other parts of the Coast Mountains to date. Radiocarbon dating of these materials has revealed advances around 6000 and 3000 years ago and during the last 500 years (Fig. 6; Koch et al., 2004). Evidence for these events is provided by in situ stumps and detrital wood from several glaciers in Garibaldi Park. Other, more poorly constrained advances occurred about 8000 (Menounos et al., 2004), 4500, and 1500 years ago. The 8000- and 4500-year advances are inferred from detrital wood in the forefields of two glaciers in the park. Evidence for the 1500-year-old event has been found at only one glacier in the park, but has been documented elsewhere in the southern Coast Mountains (Reyes et al., 2004).

Late Little Ice Age moraines in Garibaldi Park have been dated more precisely by lichenometry and dendrochronology. Most glaciers achieved their maximum Little Ice Age extent in the late seventeenth or early eighteenth century (Koch et al., 2004).

Alpine environments are ideal sites for dendroclimatic studies. Trees near treeline produce annual rings that vary in width and density in response to changing environmental conditions. Temperature and precipitation are the principal environmental factors affecting tree growth near treeline (Fritts, 1976; Schweingruber, 1996). Subalpine coniferous trees are particularly useful for reconstructing climate over the last millennium because they respond to changes in temperature and precipitation, yet can live for hundreds of years. Subalpine fir and mountain hemlock, for example, can attain ages of about 300 and 900 years, respectively.

Ring-width chronologies of these species correlate positively with summer temperature and the PDO, and correlate negatively with winter precipitation, spring snowpack depth, and glacier mass balance (Lewis, 2001;Peterson and Peterson, 2001; Gedalof et al., 2002; Peterson et al., 2002). Mountain hemlock ring-width chronologies can also be correlated over much of the Pacific Northwest, including Garibaldi Park (Gedalof and Smith, 2001), which allows us to apply them in our study.

Sediments in many lakes consist of distinctive annual couplets, termed varves. Varves yield valuable hydroclimatic information that cannot be obtained from other paleoenvironmental datasets. Flood events and seasonal hydrology, for example, can be inferred from detailed sedimentological analysis of varves (Menounos, 2002). The lower, coarser grained portion of a varve is deposited during late spring and early summer snowmelt. Other coarse layers may record severe rain-on-snow storms in the watershed, typically in summer or fall (Fig. 7). The basal, coarse layer of a varve is commonly overlain by microlaminated silt and clay, reflecting variable streamflow during summer. These sediments, in turn, are capped by clay deposited during winter when inflow to the lake is lowest and the lake is ice-covered.

Reconstructing hydroclimatic events from lake sediments is complicated by changes in sediment supply (Desloges and Gilbert, 1995), which result, for example, from fluctuations in glaciers (Leonard, 1997) or summer temperature (Leonard, 1986) on decadal and centennial timescales. Acquisition of varved sediment records from several lakes in a region can reveal the common hydrological signals contained in the records.

We have obtained and analyzed cores of varved sediments from Cheakamus, Green, and Glacier lakes (Fig. 2). Some of the conclusions that we have drawn from these records are 1) varve thickness is most highly correlated with the intensity of the annual flood during the period of streamflow monitoring; 2) varve thickness correlates with inter-annual changes in air temperature prior to the period of rapid glacier recession in the early twentieth century; and 3) temporal patterns of sedimentation are consistent with published records of glacial activity in the Canadian Cordillera and with our moraine chronologies based on tree rings and lichens. The lake sediment records suggest that ice cover was more extensive 3000 to 2500 years ago and between AD 1700 and 1920 than at present.

Changes in treeline provide important insights into past climate. Treeline is a dynamic boundary reflecting a balance between environmental conditions favouring and limiting tree growth (Arno and Hammerly, 1984). Temperature and the depth and duration of snowpack determine the location of treeline. Trees do not germinate successfully where the mean temperature of the warmest month is less than ca. 10°C (Arno and Hammerly, 1984). Thick snowpack and late-lying snow also determine vegetation distribution in the subalpine zone, as snow inhibits the successful germination of trees (Arno and Hammerly, 1984). Differences in aspect and local topography result in a variable vegetation cover near treeline, with islands of trees separated by meadows, an ecotone referred to as parkland. For this reason, the term "treeline" represents a transitional zone rather than an exact altitudinal line (Holtmeier, 1985).

Alpine treeline fluctuated throughout the Holocene, at all timescales and in all mountain ranges around the world (Rochefort et al., 1994; Kullman, 1995). Treeline was lowest during the final phase of the Little Ice Age in the nineteenth century. Since then, treeline has advanced upslope, reaching heights not attained for several centuries or even millennia (Kullman, 2001). Recent studies indicate that treeline is more sensitive to small changes in climate than previously thought and that paleotreeline identification may prove to be a useful tool for reconstructing past climate (Kullman, 1998). The Pacific Northwest shows great potential in this regard (Woodward et al., 1995; Rochefort and Peterson, 1996).

Treeline in Garibaldi Park has not been disturbed by humans, and apparently has not been affected, at least over the last several hundred years, by large forest fires (Cashman, 2004) or insect outbreaks. Therefore, recent shifts in treeline can be attributed to changes in climate. We have determined the position of treeline at a few locations in Garibaldi Park in the twentieth century from photographic records dating back to the 1910s (Fig. 8). We have also sampled trees in areas with long photographic records to establish germination dates and growth rates (Koch et al., 2003b). Similar studies in the Canadian Rocky Mountains have provided important information on climate change (Luckman and Kavanagh, 2000) and for use in land management (Rhemtulla etal., 2002). Preliminary findings from our study indicate that treeline in some parts of Garibaldi Park has risen about 40 m over the last 100 years, and that many alpine meadows have been invaded by trees over this period (Koch et al., 2003b).