Glacier responses to recent volcanic activity in Southern Chile

Glaciers in Southern Chile (39–43°S) are characterized by frontal retreats and area losses in response to the ongoing climatic changes at a timescale of decades. Superimposed on these longer-term trends, volcanic activity is thought to impact glaciers in variable ways. Debris–ash covered Glaciar Pichillancahue-Turbio only retreated slightly in recent decades in spite of been located on Volcán Villarrica which has experienced increased volcanic activity since 1977. In contrast, the negative long-term Volcán Michinmahuida glacier area trend reversed shortly before the beginning of the explosive eruption of nearby Volcán Chaitén in May 2008, when Glaciar Amarillo advanced and a lahar type of mudflow was observed. This advancing process is analysed in connection to the nearby eruption, producing albedo changes at Michinmahuida glaciers, as well as a possible enhanced basal melting from higher geothermal flux. Deconvolution of glacier responses due to these processes is difficult and probably not possible with available data. Much more work and data are required to determine the causes of present glacier behaviour.


Introduction
Ice-capped active volcanoes of Southern Chile have experienced large retreat and shrinkage in historical times (Rivera et al 2005). This process has been well documented as a primary consequence of ongoing climatic changes (Carrasco et al 2008). Volcanic activity, however, is also affecting glaciers, by a combination of factors such as destruction of glacier areas during eruptions, increasing geothermal temperature at the glacier base inducing higher basal sliding, and ash deposition on top of the glacier modifying surface albedo. Salamantin et al (2000) indicated that in glaciers on top of volcanoes, 25% of the mass loss can be explained by the geothermal activity. The albedo effect has already been described for Volcán Villarrica (table 1) in Rivera et al (2008), where highest melting rates are observed downwind from the volcanic crater due to deposition of tephra fallout. However, depending on the ash thermal conductivity and thicknesses, it is possible that these layers could have enhanced snow and ice melting at the first stages of an eruption, but when a sufficient (though unknown) thickness threshold was reached, the layer would have reduced surface melting by insulating the snow and ice from solar radiation (Brook et al 2011). This kind of measurement was performed at Volcán Villarrica where it was shown that a very thin ash layer (less than 1 cm) could insulate the ice (Brock et al 2007). Another possible influence of the volcanic activity is the almost total destruction of the ice, for example within the caldera of Volcán Hudson (45 • 51 S, 72 • 55 W, 1905 m above sea level (m asl)) which in 1971 and 1991 experienced some of the most explosive eruptions recorded in Chile (González-Ferrán 1995).
This study includes (table 1, figure 1) the analysis of Volcán Villarrica, one of the two most active volcanoes in the Andes and largely covered by a glacier (Clavero and Moreno 2004); and Volcán Michinmahuida, an ice covered stratovolcano that last erupted 170 years ago but is located near Volcán Chaitén (42 • 50 S, 72 • 39 W, 1122 m asl), an ice-free caldera and dome which has experienced a sub-plinian eruption since May 2008 (Lara 2009). Also, two non-active volcanic glaciated mountains were included in the analysis for comparison purposes.

Regional context
The volcanoes analysed in this work are located in the southern volcanic zone (SVZ) of the Chilean Andes, which comprises approximately sixty active volcanoes with well-known historical eruption records (Stern 2004). The SVZ is tectonically controlled by the convergence and subduction of the South American plate under the Nazca plate. The volcanoes are aligned for over 1000 km along the north-south oriented (figure 1) Liquiñe-Ofqui fault zone (LOFZ), which controls the location of most volcanoes in Southern Chile as well as those of numerous monogenetic cones (Cembrano et al 2000). During recent decades, several volcanoes within the region displayed explosive eruptive activity (Volcanic Eruption Indices (VEI) 1 to 3, Mason et al 2004), with permanent degassing, and material ejection and subsequent deposition over ice surfaces (Simkin and Siebert 1994). The study area is within the mid-latitude climatic belt located between the Mediterranean temperate regime of central Chile and the region under semi-permanent presence of the Westerlies (Rütllant and Fuenzalida 1991). This mid-latitude belt is characterized by dry and warm summers and wet winters with maximum precipitations of several hundreds of millimetres per year in the northern boundary of the region to a few thousands of millimetres per year in its southern part (Carrasco et al 2008). Unfortunately most of the regional stations are located at low altitude therefore very little is known about high-altitude precipitation and temperature changes, which have been analysed using radiosonde data (Falvey and Garreaud 2009).
Near-surface air temperatures in the study area showed a cooling trend between 1960 and 1970 at Temuco (38.5 • S) and Puerto Montt (Rosenblüth et al 1997); however, no noticeable temperature trends were obtained between 1979 and 2006 (Falvey and Garreaud 2009). The upper atmosphere (radiosonde data from Puerto Montt at 41 • S, measured between 1500 and 9000 m asl) indicates tropospheric warming of 0.12 • C/decade between 1958 and 2006 Rivera 2007, Carrasco et al 2008). Because the glacierized summits are well within this upper atmospheric zone, the latter trend is what is expected to control ablation of glaciers on these volcanoes below the equilibrium line altitude (ELA) ranging from 2800 m asl at 36 • S down to near 1600 m asl at 43 • S (Carrasco et al 2005). However, the surface temperature trends may influence the sources of atmospheric moisture and thus, precipitation. Hence, the way in which regional climate may affect the volcanoes' ice caps is a complex matter. High-altitude warming would suggest that the icecaps should be shrinking, but the controlling factor could be any changes, positive or negative, in precipitation.
The precipitation between 36 • and 46 • S Lat has shown a general decreasing trend in the period from 1900 to 2000, with particularly lower rainfall amounts occurring between 1970 and 2000 . North of 38 • S Lat, a relatively high frequency of wet years occurred in the 1940s-50s and 1980s-90s, whilst the opposite was evidenced in the 1960s-70s. Between 39 • and 47 • S a high frequency of dry years has occurred in the last three decades, which was preceded by a large number of wet years between 1960 and 1980 (Quintana 2004).

Remote sensing data sets
VNIR satellite imagery collected from Landsat MSS, Landsat TM, Landsat ETM+ and ASTER sensors, and SWIR data from ASTER, form the basis for our glacier inventories and for our analysis of recent variations on ice-capped volcanoes. The main selection criteria of satellite imagery included optical characteristics and acquisition dates . Optimal scenes contained minimal cloudiness and were acquired at the end of the ablation season, March-April in the austral hemisphere (Carrasco et al 2008). At this time of the year, the snow cover area is at a minimum extent, allowing a more precise delineation of glacier areas .

Georeferencing and image corrections
All maps and satellite images were converted to Universal Transversal Mercator (UTM) projection and World Geodetic System of 1984 (WGS84) datum allowing orthorectification using SRTM. The orthorectified Landsat MSS, TM and ETM+ imagery were co-registered to ASTER data by using the Cosi-Corr software (Leprince et al 2007) with subpixel error in co-registration between images.
Using histograms, adjusted by different threshold values, an accurate glacier classification was obtained. The best results were obtained with Landsat TM4/TM5  and ASTER VNIR 3/SWIR 4 band ratioed images (Kargel et al 2005). Furthermore, manual on-screen editing, essentially a manual smoothing, was performed to reduce the fractal structure along glacier boundaries (Rivera et al 2006). Once glacier areas are delineated, a comparison to previous datasets (aerial photographs) and historical records was completed in order to determine main glacier fluctuations. The glacier frontal changes are always calculated along the central flow line of each glacier.
Albedo changes of snow/ice facies due to volcanic ash deposition were analysed with VNIR ASTER data.
The distinction between areas experiencing albedo change due to volcanic activity and debris covered ice or rock outcrops/margins is difficult, but having a precise pre-eruption delineation of the glacier boundaries, a mask could be produced to restrict the albedo changes to the glacier surfaces allowing us to detect possible ablation enhancement of reduction, depending on the nature of the deposited ashes and, mainly, on the thickness of the falling material (Richardson and Brook 2010).

Results
Analysis of historical documents, aerial photographs and satellite images during recent decades indicated retreat for most of the regional glaciers (table 1).

Volcán Villarrica
At Volcán Villarrica (figure 2), the debris and ash covered Glaciar Pichillancahue-Turbio experienced an ice area reduction of 2.1 km 2 between 1961 and 2011 representing 13% of the 1961 glacier area, with a mean frontal retreat rate of −24.5 m yr −1 during the same period of time. The glacier areal and frontal changes at Villarrica are among the smallest compared to the other studied glaciers in the region (table 1), in spite of being one of the most active volcanoes of the country with several eruptive events in recent years when a large volume of fumarolic and eruptive materials were ejected (Clavero and Moreno 2004). This smaller change rate compared to other glaciers in the region (Masiokas et al 2009), is more likely related to the thick ashes and debris covering the

Monte Tronador
This non-active volcano (figure 3) is characterized by several glacier tongues flowing west and east from the Chile-Argentina border, some of them being heavily debris covered, especially the northern flank Glaciar Casa Pangue experiencing a significant decadal thinning ) and a long-term frontal retreat between 1911 and 2007 (Masiokas et al 2009). In more recent years, the glacier continued thinning with the smallest area change rate compared to the other studied glaciers.

Monte Inexplorado
The glaciers of Monte Inexplorado have undergone strong retreats in recent decades (Masiokas et al 2009). Before the early 1990s, the main glacier (figure 4) was reaching low altitude at Río Blanco being fed by two ice accumulation basins. Since then, the glacier retreat resulted in two separated ice bodies (Inexplorado II and III), leaving behind a proglacial lake.
The combined area change of Glaciers Inexplorado II & III between 1961 and 2011 was −0.13 km 2 yr −1 , leaving a total ice area in 2011 of 26 km 2 (25% of the 1961 area).
The glacier with maximum retreats between 1961 and 2011 is Inexplorado II, with a rate of −125 m yr −1 .

Volcán Michinmahuida
At Volcán Michinmahuida (figure 5(a)) the four main glacier basins had a total ice area in 2011 of 48.1 km 2 , representing 89% of the total glacier area estimated for 1961. The maximum frontal retreat observed at Volcán Michinmahuida took place at Glaciar Amarillo (figure 5(b)), a glacier tongue aligned to the LOFZ (figure 1) that has been shrinking as all the other local glaciers (table 1) 6). These reductions were more significant on the western slopes of Michinmahuida, which were more directly receiving Volcán Chaitén ashes. The lower part of the glaciers, some of them already debris covered before the eruption, did not have a significant albedo change, and some upper areas were probably covered by snow during or after the eruption, therefore the albedo changes in these areas were smaller.

Discussion
The ongoing climatic changes favour the long-term glacier variations in this region, characterized by ice thinning , area shrinkage and frontal retreats (Masiokas et al 2009). Glaciers located on active volcanoes have similar negative trends (figure 7) but different rates when compared to the variations of other glaciers located on top of non-volcanic mountains affected by the same climatic conditions.
For example, between 1961 and 2007, the main glacier of Monte Inexplorado retreated at a rate very similar to those of Volcán Michinmahuida. This mountain is only located 150 km north from Volcán Michinmahuida (figure 1), and its summit is only 170 m lower than that of Volcán Michinmahuida (table 1). However, the Glaciar Amarillo advance at Michinmahuida between 2007 and 2009 was not mirrored by the glaciers at Monte Inexplorado, which kept retreating until 2011, with an even higher rate of −339 m yr −1 (figure 7).
During recent years/decades, neither the regional climate record nor the glacier record of unceasing glacier retreat on non-volcanic Monte Inexplorado and Monte Tronador would appear to support a possible glacier advance in the region. Therefore, the observed glacier advance between 2007 and 2009 could be explained as a response to other factors including a possible surge event, or more likely, the active condition of Volcán Michinmahuida and the recent eruption of Volcán Chaitén.
The ashes produced by the Volcán Chaitén eruption have been spread by wind mainly towards the east creating a thick layer of deposits over the region, where the glaciers at Volcán Michinmahuida have also been affected (Watt et al 2009, Alfano et al 2011. The ash cover has produced a drastic reduction in surface albedo (figure 6) as expected from any glacier surface (characterized by higher spectral reflectance) when covered by minerals, ash and dust (Brock et al 2007). This albedo reduction at Volcán Michinmahuida was confirmed by comparing an ASTER image collected from before the eruption, on 27 January 2006, with an ASTER image from 19 January 2009, where widespread tephra layers covering the ice surface were detected, yielding a significant decrease in surface albedo, especially on altitudes between 1500 and 2000 m asl and western slopes of the volcanic edifice. Considering the above, a glacier advance could have been explained by an increase of snow melting at the surface of the glacier (due to albedo reduction) that led to increased basal sliding when melt water penetrated down to the glacier bedrock, thereby reducing basal shear stress. However, glacier advances at Michinmahuida began several months before the eruption inception, therefore the albedo changes are not responsible for the observed advances. Even more, the ablation changes due to the albedo decreases are not expected to be major, because the reported ash thickness are, more likely, insulating the underlying snow/ice. An alternative hypothesis is related to enhance geothermal heat flux at some bedrock spots of the Michinmahuida icecap. If this is the case, increasing meltwater will provoke higher basal water pressure leading to its rapid release (Gudmundsson et al 1997). Higher amounts of meltwater at the glacier bed can lead to enhance basal sliding and therefore produce higher velocities or glacier advances. However, in temperate glaciers like Michinmahuida, this process does not necessarily result in glacier surges, because the bed is already wet and sliding is a predominant factor. In spite of this temperate condition, ice velocities at the lower part of Glaciar Amarillo were obtained by applying a feature tracking procedure (Scambos et al 1992) based upon ASTER images (figure 8). The resulting vectors show higher velocities during the advancing period of 2007-09 compared to 2009-11, when the glacier was more stable. The lower coherence between images obtained during the advancing period is responsible for the small number of velocity vectors (left panel) compared to the more recent period. The decrease in ice velocities  The existence of hot spots at Michinmahuida is possible, because it is an active volcano located at the LOFZ, and because it is linked to nearby Volcán Chaitén by geologic faults and parasitic cones. If the detected glacier advance is connected to enhance subsurface geothermal fluxes prior the eruption, this process will be only a 'temporal' response. However, a unique feature among the downstream valleys of Michinmahuida is the lack of vegetation at the Glaciar Amarillo valley, indicating frequent mudflows coming from the glacier. During the Chaitén eruption, the satellite image from 19 January 2009 clearly shows a lahar type of deposit at this valley, which must have been generated by a sudden melt of the ice. This flood event was described by Alfano et al (2011) who visited the area in January 2009. This process could be related to the existence of enhanced subglacial volcanic activity at Michinmahuida producing high quantities of meltwater, or to rainfall removing ashes deposited on top of the glacier, which combined with higher ablation resulted in a flood.

Conclusions
Glaciers primarily respond to prevailing atmospheric conditions, with a time delay dependent upon glacier length, mass balance and ice thickness. On a decadal timescale, the glaciers of Southern Chile are responding to the changing climate of the region that includes reduced precipitation and tropospheric warming.
There are no significant recent climatic anomalies or trends which could explain the ice advances observed at Volcán Michinmahuida. This suggests that the behaviour of these glaciers is also driven by non-climatic factors, namely volcanic activity associated with the eruption of Volcán Chaitén located a few kilometres to the west of Volcán Michinmahuida.
As the Michinmahuida glacier advances began a few months before the eruption of Volcán Chaitén, the increased ice flow could be a response to higher subsurface geothermal heat fluxes prior to the inception of the eruption. If heating at the bedrock of this glacier increased, leading to the generation of basal meltwater, it is possible that enhanced basal sliding would occur, resulting in a sudden glacier advance. This hypothesis is also sustained by the lack of vegetation at the downstream valley of Glaciar Amarillo, indicating lahar kind of mudflows due to sudden melting events. No other downstream valleys of Michinmahuida have similar features, indicating a feature possibly connected to a geothermal heat flux spot. In spite of the almost synchronous events (volcanic eruption and glacier advance separated by a few months), much more work is required to determine the causes of present glacier behaviour.
The described results indicate that the primary factor in the glacier behaviour in Southern Chile is clearly the regional climate change observed during most of the twentieth century. The secondary factor, but maybe dominating at certain volcanoes, is the volcanic activity, which may have a significant role in altering long-term trends or disrupting linear responses as suggested from anomalous ice advances in recent years.
Although the satellite imagery analysed in this research is a key contribution to better understanding glacier variations in volcanic environments, more in situ field studies are required to determine the precise relationships between glacier behaviour and current regional volcanism, as well as testing the presented hypothesis.
improved the letter. Claudio Bravo and Sebastián Cisternas collaborated with this research. Two anonymous referees also improved the letter.