1. Introduction
Much evidence indicates that tropical glaciers were far more extensive in past centuries. This is especially true during the Little Ice Age (LIA), which is well documented for some mountain ranges in the Northern Hemisphere such as the European Alps and the Canadian Rockies (Reference Le Roy LadurieLe Roy Ladurie, 1967; Reference GroveGrove, 1988; Reference LuckmanLuckman, 2000), and also identified in the tropical Andes (Reference BroggiBroggi, 1945; Reference Hastenrath and BalkemaHastenrath, 1981; Reference ClappertonClapperton, 1983; Reference Ames and FrancouAmes and Francou, 1995; Reference Francou, Dollfus, Deler and MesclierFrancou, 2004). In the Cordillera Real, many moraines attributed to the LIA exist below the present glacier snouts, and the maximum extent has been considered the most prominent phase of glacier advance of the entire Holocene (Reference Gouze, Argollo, Saliege and ServantGouze and others, 1986). However, few studies have yet been able to accurately date glacier fluctuations over this period (Reference Solomina, Jomelli, Kaser, Ames and PouyaudSolomina and others, in press). The large uncertainties involved in dating these glacier fluctuations are partly due to the dating method using 14C (Reference Gouze, Argollo, Saliege and ServantGouze and others, 1986; Reference SeltzerSeltzer, 1992), an isotope with atmospheric production rates that have varied over recent centuries, precluding dating within the last 300 years (Reference ReimerReimer and others, 2004). Information obtained from ice cores retrieved in the Andes has made it possible to roughly estimate the LIA’s limits between the 15th and the late 19th centuries (Reference Thompson, Mosley-Thompson, Dansgaard and GrootesThompson and others, 1986). However more accurate dating of glacier evolution within this period requires other proxies.
Moraines are often used to date fluctuations and to reconstruct glacier areas and volumes with good accuracy. In the tropics, moraines are well preserved because reworking by frost penetration in the ground and snowmelt is extremely limited. Moreover, time lags between glacier mass-balance variations and the dynamic response of the snouts are generally short, i.e. one to a few years (Reference Francou, Dollfus, Deler and MesclierFrancou and others, 2004), particularly for the small glaciers (<0.5km2) that represent about 80% of the glacier area of the Cordillera Real, Bolivia (Reference JordanJordan, 1991). Thus the moraines deposited by these small glaciers are believed to represent the form of the glaciers in an equilibrium state.
The aim of this paper is to describe the glacier evolution since the LIA maximum for this tropical Cordillera using moraines dated by lichenometry and to infer climate changes from these fluctuations.
2. Study Area
Cerro Charquini (16°17’S, 68°09’W) is located 20 km northeast of La Paz, midway between Glaciar Zongo and Glaciar Chacaltaya where a permanent observation programme involving mass-balance, hydrological and energy-balance measurements has been run for more than a decade (Reference Francou, Ribstein, Saravia and TiriauFrancou and others, 1995, Reference Francou, Vuille, Wagnon, Mendoza and Sicart2003; Reference Wagnon, Ribstein, Francou and PouyaudWagnon and others, 1999).
Despite the relatively low elevation of the summit (5392 m a.s.l.), glaciers with similar morphology are present on the different slopes, allowing aspect effects to be analysed without interference from other factors. The Cerro Charquini glaciers are small (<0.5 km2 in 1997) and, together with their moraines, are all located within the same 4430–5300 m a.s.l. elevation range (Fig. 1). Moraines extend 1.5 km and 200–450m in elevation beyond the present snouts. As they have not yet been covered by vegetation, these moraines may be considered, a priori, as dating from the LIA. Moraines and bedrock are granodiorites and quartzites, rock types that favour growth of Rhizocarpon sp., a lichen commonly used to date moraines worldwide (Reference InnesInnes, 1985).
Cerro Charquini glaciers (Cordillera Real, Bolivia, 168 S). Although the ten moraines can be found on each glacier foreland, only the five biggest (M1, M3, M6, M8 and M9) are shown. Glacier limits in 1940, 1983 and 1997 are from stereophotogrammetric restitutions.
In the Cordillera Real, annual precipitation amounts to 800–1000mm at above 4000ma.s.l., which falls generally as snow on the glaciers. Precipitation comes with humid air masses from the northeast, according to a mechanism called the Amazonian monsoon, and presents a very contrasted seasonal pattern, with 80% falling from October to April (austral summer) and 65% from December to February. This marked seasonality of the climate plays an important role in the mass-balance fluctuations. During the austral summer, accounting for 90% of the annual mass-balance variability (Reference Francou, Vuille, Wagnon, Mendoza and SicartFrancou and others, 2003), the maximum net accumulation occurs in the upper part of the glaciers while ablation rates are highest in the lower part. Ablation is associated with a strong total radiation influx to the glacier surface (the sun is vertical and cloudiness is maximum from November to February), high humidity and low winds (which limit sublimation) and a relatively high air temperature. Precipitation controls the energy balance at the glacier surface through the albedo (Reference Wagnon, Ribstein, Francou and PouyaudWagnon and others, 1999). Every variation in the precipitation regime results in a change to the glacier surface albedo and consequently the ablation intensity (Reference SicartSicart, 2002).
On a regional scale, the El Niño–Southern Oscillation (ENSO) controls the majority of the interannual variability of mass balance (Reference Francou, Vuille, Wagnon, Mendoza and SicartFrancou and others, 2003). Warm phases are generally accompanied by a precipitation deficit of 10–30% (i.e. less accumulation leading to a lower albedo), a decrease in cloudiness (i.e. more shortwave radiation reaching the glacier surface) and an increase in the sensible heat flux (air temperature anomaly in the +1 ° to +3°C range) (Reference Wagnon, Ribstein, Francou and SicartWagnon and others, 2001). Consequently mass balance is very negative during El Niño years. An opposite situation occurs during the cold phases (La Niña), when mass balance shows near-equilibrium/positive values due to enhanced precipitation, colder air temperature and high cloudiness, which tend to decrease ablation strength and lower the ELA by 100–300 m compared with the warm phases. The climate during the LIA is believed to have been more extreme than at present, and a direct effect of the ENSO variability during the last centuries on glacier fluctuations and moraine deposits has to be considered.
3. Methods
3.1. Dating moraines by lichenometry
The dating of Charquini moraines by lichenometry with statistical analysis based on the extreme-value theory has been discussed elsewhere (Reference RabatelRabatel and others, 2005; Reference Cooley, Naveau and JomelliCooley and others, in press; Reference Naveau, Jomelli, Cooley and RabatelNaveau and others, in press) and only the basic principles will be presented here.
The basic premise of lichenometry is that the diameter of the lichens growing on a surface is proportional to the length of time that the surface has been exposed to colonization and growth in a specific environmental context. Thus the largest lichens can be used to date the deposit, the stabilization or the exposure of surfaces.
All measurements were carried out with the Rhizocarpon geographicum s.l. lichen, which presents a circular growth pattern and has a lifetime longer than the period to be dated (Reference InnesInnes, 1985). Our dataset of lichenometric measurements is composed of 10 previously dated and 48 further surfaces. The surfaces of known age are presented in detail by Reference RabatelRabatel (2005). All were found in the eastern Cordillera (up to 4500 m a.s.l.), and so are expected to be representative of lichen growth rate on the glacier forelands. Eight date from the 20th century (between 1910 and 1965), one from the mid-18th century and one from the mid-17th century. Only the oldest one is based on a 14C dating. All the others are dated using documentary sources. Undated surfaces correspond to the main moraines of each of the glaciers to be dated.
In the current study, a new method based on the extreme-value theory called generalized extreme values (GEV; Reference Katz, Parlange and NaveauKatz and others, 2002) was used to analyse lichenometric data. Maximum lichen diameters and average surface ages are assumed to have two different distributions, a GEV and a normal distribution, respectively. This statement is based on probability theory for maxima (extreme-value theory) and for averages (central limit theorem), respectively. These two distributions are linked through the relationship α = f(μ), where the function f represents the temporal variation of μ (for more details see the papers mentioned above).
3.2. Reconstructing glaciers from moraines and aerial photographs
Glacier fluctuations during this period were reconstructed on the basis of moraines. All the moraines of the LIA and the beginning of the 20th century were dated by lichenometry. For the second part of the 20th century, five aerial photograph pairs dated 1956, 1963, 1974, 1983 and 1997 (Instituto Geográfico Militar, Bolivia) were processed using classical (non-digital) photogrammetric restitution techniques. This information allowed us to reconstruct glacier areas and to calculate volume changes. Aerial pictures from 1940 could not provide a three-dimensional restitution because the flight parameters were not available. Nevertheless, it was possible to calculate the glacier area of 1940 by plotting the limits on the 1/10 000 map based on other aerial photographs. Restitution was also used to obtain a complete 1/10 000 mapping of Charquini on which we represent the ten main LIA morainic stages for each of the five glaciers (see Fig. 2 which shows an example, Glaciar Charquini Sur). In order to validate the results of the restitution, a complete ground topographic survey of each moraine was performed on Glaciar Charquini Sur. Results of both methods are consistent and allow us to be confident that the mapping of glaciers and moraines for the whole Charquini area is accurate.
Glaciar Charquini Sur. Surface area reconstructed on the basis of the moraines until M10 and by aerial photography from 1940.
3.3. ELA reconstruction
The equilibrium-line altitude (ELA) was reconstructed for each morainic stage and each pair of photographs using the accumulation-area ratio (AAR) method and a ratio of 0.65 determined for Glaciar Zongo when mass balance is in equilibrium (1991–2004 series, see Reference SorucoSoruco and others, 2005). This 2 :3 ratio was also observed on other southfacing glaciers of the sector (Reference Rabatel and MendozaRabatel and Mendoza, 2004) and thus it can be considered as a realistic approximation for AAR0 on a regional scale. Nevertheless this ratio is valid only on glaciers that are in equilibrium, and may change significantly according to slope aspect on the scale of the Charquini area.
3.4. Ice-volume and mass-balance reconstruction
For the ten LIA stages, glacier volumes were reconstructed by contour line interpolation based on the moraine height and the bedrock morphology. Since depth to the bedrock below the current glaciers was unknown, we could only calculate volume fluctuations between the different stages. For the 20th century, the volume variations computed from the photogrammetric restitution of each pair of photographs were considered to be accurate. After dating the moraines and aerial photographs, it was possible to calculate the mean annual mass balance for each period, given by the following equation:
where 
is the mean annual mass balance for the period T to T
i+1, T
i is the glacier stage date (moraine or aerial photograph), V
i
, Si
are the volume and surface area for glacier stage i, and ρ is the ratio of ice density to water density (fixed at 0.9).
4. Results
4.1. LIA chronology on Charquini glaciers
Only the principal results are presented here. Details about the interpretation of moraine dating can be found in Reference RabatelRabatel and others (2005). On the five glaciers the ten most important moraines were considered to determine the chronology of glacier evolution since the LIA maximum. The ten moraines can be found on all five slopes and show the same relative position and morphology (Fig. 1). Five moraines are always prominent (M1, M3, M6, M8 and M9) whereas the others (M2, M4, M5, M7 and M10) are less marked. The morphologies of the prominent moraines are believed to reflect phases of stabilization or small advances, whereas the others represent only short stops in a continuous recession process. By dating the moraines (Table 1), we found that the prominent outer moraine M1, which corresponds to the LIA maximum, was formed during the second half of the 17th century, i.e. between 1648 and 1700, taking into account the five slopes and the margins of error. The moraines M2, M3, M4 and M5, formed during the 18th century, reflect a gradual recession of glaciers, only interrupted by an advance which deposited moraine M3, which dated from 1722–52 depending on the glacier. In the late 18th to early 19th centuries (1781–1818), M6 attests to another advance. Then, during the first half of the 19th century, the glaciers retreated slightly and did not display an advance which could remove the older moraines. The long distance, more than 250 m on average, between the last moraine of the 19th century, M9 (about 1870), and M10 (about 1910), suggests an acceleration of the retreat during the late 19th century.
Moraine dating by lichenometry with the associated error (years), surface area (km2) and ELA (ma.s.l.) for the LIA stages and six 20th-century aerial photographs of the Charquini glaciers. On Glaciar Charquini Sureste blocks were too scarce on the M2 and M5 moraines to allow lichenometric measurements
4.2. Ice surface area variations
Throughout the period extending from the mid-17th to the late 20th century, all the glaciers retreated over a distance of 950–1400 m. However, Glaciares Norte and Oeste lost much more (>78% of the maximum LIA area) than Glaciares Sur, Sureste and Noreste (<65%;see Table 1). Expressed in terms of annual means for the five glaciers, the area reduction was 0.15%a–1 between M1 and M9, increasing to 0.30%a–1 between M9 and M10. Afterwards, the retreat slowed during the first part of the 20th century (0.18% a–1) then accelerated after the 1940s, reaching 0.35% a–1. Since the beginning of the 1980s, the retreat rates (0.52% a–1) have been the highest recorded over the last four centuries (Fig. 3). Between the LIA maximum and the late 19th century, the greatest retreats occurred on Glaciares Sur and Oeste. These glaciers lost 30–50% of the area of the LIA maximum while the others lost only 16–25%. Nevertheless, during the 20th century, Glaciares Norte, Noreste and Sureste retreated more than the others (40–57% vs 29–36%).
Surface area lost for each period for the five Charquini glaciers. The area lost is given as a percentage of the initial area.
4.3. ELA fluctuations
Between the maximum LIA extent and the late 20th century, the ELA rise determined geometrically using the AAR method can be estimated at 158 ± 30 m (mean for the five glaciers) ranging from 115 m (for Glaciar Oeste) to 190 m (for Glaciar Noreste). The ELA rose from 4815–4990 m a.s.l. during the morainic stage M1 to 4960–5165 m a.s.l. during the last decades of the 20th century (Table 1).
Figure 4 shows that between the morainic stages M1 (~1665) and M3 (~1735) the ELA rise is very limited, less than 10 m. Its rise between M1 and M9 (~1870) is of the order of 50 m (ranging from 15 m on Glaciar Sureste to 70 m on Glaciar Noreste).
Fluctuation of the ELA since the LIA maximum. The zero represents the average altitude of the ELA over the whole period.
4.4. Volume variations and mass-balance changes
Figure 5 shows the mean annual mass balance, reconstructed for the following four periods: LIA maximum to late 19th century, late 19th to early 20th century, early 20th century to 1983 and 1983–97. This information leads to the same conclusions as the reconstructed areas. Between the LIA maximum and the late 19th century, ice loss was low, with a mean mass balance estimated at about –0.10mw.e.a–1. Between M9 and M10 (~1870 and ~1910), the mass balance of the five glaciers was, on average, four times more negative (–0.40mw.e.a–1). For the ~1910–1983 period, deficits were less pronounced, with an average of about 0.25 m w.e. a–1. The last two decades of the 20th century marked a shift in the deficits, which increased to 1.36 m w.e. a–1 (average for the five Charquini glaciers). This clearly represents the largest mass loss recorded over the past four centuries, and it is also for the shortest time-scale. In addition, we observe that Glaciares Oeste and Norte present the strongest recession for recent years (1983–97) with losses as high as 3 and 2.3mw.e.a–1, respectively. However, Glaciar Oeste is believed to be less representative than the others because of its smaller size (<0.15 km2) and because its low elevation puts it completely in the ablation zone. The three other glaciers facing east and south experienced clearly less negative mass balances that did not exceed –0.66 m w.e. a–1.
Mean annual mass balance reconstructed for each period for the five Charquini glaciers. The value for Glaciar Chacaltaya is shown for the 1983–1997 period for comparison.
5. Discussion
5.1. Comparison with other glaciers in the central Andes
The chronology obtained for the Charquini glaciers indicates that recession of glaciers in the Bolivian Andes since their LIA maximum is in good agreement with that of glaciers in the Peruvian Cordillera Blanca (Reference Solomina, Jomelli, Kaser, Ames and PouyaudSolomina and others, in press). After the LIA maximum, in the second half of the 17th century, the Cordillera Blanca glaciers receded nearly continuously, with only minor readvances, each less than that which deposited the outer moraine M1. This recession occurred even during the early 19th century, when many Northern Hemisphere glaciers grew and sometimes exceeded their 17th-century maximum (Reference GroveGrove, 1988). Surface area losses since the LIA maximum on Charquini glaciers are consistent with those obtained on the nearby south-facing Glaciar Chacaltaya (0.035 km2, 5395– 5150m a.s.l.), which in 1998 had lost 80% of its maximum LIA area (Reference RamírezRamírez and others, 2001). Similarly Reference Ames and FrancouHastenrath and Ames (1995a, Reference Hastenrath and Amesb) found that Glaciar Yanamarey (0.8 km2, 5200–4600ma.s.l., facing southwest) in the Cordillera Blanca lost about 53% of its surface area between the LIA maximum (undated) and the early 1980s, consistent with Glaciar Charquini Sur which lost almost the same area (51%) over the same period. The accelerated retreat and very negative mean annual mass balance during the M9–M10 period could be the consequence of a significant climate change. Both support the idea that the LIA ended between about 1870 and 1910. This result is consistent with evidence of rapid glacier decrease in other places of the tropical Andes in Peru and Ecuador, as reported by voyagers and scientists of the late 19th century (Reference BroggiBroggi, 1945; Reference Hastenrath and BalkemaHastenrath, 1981; Reference Ames and FrancouAmes and Francou, 1995; Reference Francou, Dollfus, Deler and MesclierFrancou, 2004).
Several studies of 20th-century recession have been performed in Peru (Cordillera Blanca and Cordillera de Vilcanota) by Reference BroggiBroggi (1945), Reference KinzlKinzl (1969), Reference Brecher and ThompsonBrecher and Thompson (1993), Reference Hastenrath and AmesHastenrath and Ames (1995a, Reference Hastenrath and Amesb) Reference Kaser and GeorgesKaser and Georges (1997) and Reference Thompson, Mosley-Thompson and HendersonThompson and others (2000). In the Cordillera Blanca, glacier retreat accelerated after the 1890s (Reference KinzlKinzl, 1969). Then it slowed down during the first half of the 20th century, with a small but marked readvance in the 1920s. This event was followed by another significant retreat in the 1930s–1940s (Reference BroggiBroggi, 1945; Reference Kaser and GeorgesKaser and Georges, 1997; Reference GeorgesGeorges, 2004). During 1950–1970, glaciers retreated very slowly (Reference Hastenrath and AmesHastenrath and Ames, 1995b), followed by a general acceleration (Reference Ames and FrancouAmes and Francou, 1995; Reference Kaser and GeorgesKaser and Georges, 1997). For the Cordillera de Vilcanota, Reference Brecher and ThompsonBrecher and Thompson (1993) and Reference Thompson, Mosley-Thompson and HendersonThompson and others (2000) noted that the Qori Kalis, a glacier issuing from the Quelccaya ice cap, receded between 1963 and 1998. The retreat increased after 1980. The area lost was 8.4 times greater between 1983 and 1991 than during the 1963–78 period, and the retreat accelerated dramatically during the 1990s. For the Bolivian eastern Cordillera, the recession of the Charquini glaciers can be summarized in four stages: (1) a major retreat starting in the late 19th century; (2) a relative slowdown during the 1910s/1930s; (3) an acceleration since the 1940s; and (4) a very strong recession during the 1980s/1990s. Thus, the consistency of glacier fluctuations observed throughout the tropical Andes between the late 19th and the 20th centuries highlights that glaciers appear to respond to the same climatic signal on a regional scale.
5.2. Climatic interpretation
From the mid-17th to the late 19th century
Whatever their aspect, the Charquini glaciers have evolved in a similar way since their LIA maximum and their retreat has been practically continuous since the second half of the 17th century. The presence of many morainic ridges shows that the retreat was interrupted by several periods of stabilization or small advance, but these advances were never large enough to bury the moraines deposited previously. Being similar in size, morphology and elevation, the main parameter which may determine fluctuations of different magnitudes for these glaciers is their aspect, which controls precipitation amounts, cloudiness and radiative fluxes. Thus, variations from one glacier to another can be interpreted in terms of climate influence.
According to the sensitivity analysis carried out by Reference Hastenrath and AmesHastenrath and Ames (1995a) on Glaciar Yanamarey in Peru and assuming melting to be the only ablation process (with a latent heat of melting Lm = 33 × 104Jkg–1), the Charquini glaciers’ mean annual balance value between the LIA maximum and the late 19th century of –0.10mw.e. translates to increasing energy absorbed by the glacier surface of about 1 Wm–2. This excess energy could have been the consequence of a rise in temperature of 0.15°C, an additional humidity of 0.05 gkg–1, a decrease in cloudiness of about 0.1/10, or a combination of these variables. The increase in incoming solar radiation caused by a decrease in cloudiness can explain the strong sensitivity of glaciers to a change in this parameter. However, the resulting decrease in incoming longwave radiation can offset this effect (Reference AmbachAmbach, 1974). It is thus difficult to determine the exact influence of changes in cloudiness on the glacier surface energy balance. A change in cloudiness can also affect precipitation and thus the glacier surface albedo (assuming that precipitation falls in mainly solid form on the glacier) but such a small change is hardly quantifiable. However, these two parameters (less cloudiness plus less precipitation) are linked and even with a slight shift they may sum their effects to increase melting at the glacier surface. If the glacier retreat between the LIA maximum and the late 19th century was the result of a 0.15°C rise in temperature, the resulting ELA shift according to Kaser’s model linking ELA to climate variables (Reference KaserKaser, 2001) would have been only 20 m, inconsistent with the 50m found on Charquini glaciers. In the model of Reference KaserKaser (2001) a change of the ELA can be interpreted in terms of changes in accumulation, energy balance and temperature or any combination of these variables. Considering change in accumulation only, this 50 m ELA shift between the LIA maximum and the late 19th century (of which 42 m was between M3 (~ 1 735) and M9 (~1870)) could have been the consequence of a decrease in annual accumulation of about 340 mm. Taking into account both temperature and precipitation changes, with a temperature rise of 0.10°C (i.e. two-thirds of the sensitivity analysis from Hastenrath and Ames) the decrease in accumulation between M1 (~1665) and M9 (~1870) becomes 210 mm. This change represents a decrease of 20–25% of the present accumulation amount. This is consistent with the results of Reference Thompson, Mosley-Thompson, Dansgaard and GrootesThompson and others (1986) based on the Quelccaya ice core that showed an accumulation deficit of about 20% of the present average amounts during the 1720–1860 period.
Furthermore, as noted in section 4, Glaciares Oeste and Sur exhibited the greatest retreat and the most negative mass balance between the LIA maximum and the late 19th century, a trend which was particularly marked during the 19th century. Precipitation and cloudiness coming from the northeast are crucial for mass balance (section 2). Lower accumulation on the leeward slopes than on the northeastfacing slopes (Reference Rabatel and MendozaRabatel and Mendoza, 2004) leads to a reduced albedo on Glaciares Sur and Oeste. Since cloudiness follows the same pattern, these slopes experience a higher shortwave radiation influx which, combined with the reduced albedo, causes greater energy absorption by the glacier surface. In the tropics, the energy available to melt snow and ice is particularly high on southern slopes during summer when the sun is at its highest (Reference Francou, Ribstein, Saravia and TiriauFrancou and others, 1995). The recent energy balance measured year-round on Glaciar Zongo by Reference Wagnon, Ribstein, Francou and PouyaudWagnon and others (1999, Reference Wagnon, Ribstein, Francou and Sicart2001) shows that the consequences of a precipitation/cloudiness deficit during the austral summer are considerable for the glacier. Therefore, considering the importance of precipitation deficits on glacier mass balance nowadays, the assumption that the mid-18th to late 19th centuries could have been dry is reasonable.
Strengthening the findings of Reference Thompson, Mosley-Thompson, Dansgaard and GrootesThompson and others (1986) about a dry 19th century, Reference Valero-Garces, B.L., Delgado-, Navas, Edwards, Schwalb and RattoValero-Garces and others (2003) have pointed out a precipitation deficit during the late 19th to early 20th century from palaeohydrological reconstructions based on sedimentological, geochemical and isotopic records in the Bolivian Altiplano. This evidence is also consistent with the historical ENSO records (Reference Quinn, Neal, Bradley and JonesQuinn and Neal, 1992; Reference OrtliebOrtlieb, 2002) which revealed that the last decades of the 19th century were marked by strong and long-lasting ENSO warm events. As pointed out in section 2, melting accelerates dramatically during the ENSO warm phases. The El Niño occurrence during this period is thus consistent with the accelerated retreat observed on the Charquini glaciers (section 4.2 and 4.4) and throughout the central Andes.
The 20th century
For the 20th century, a 100 year record (from 1898) of precipitation in La Paz (San Calixto station, 3660ma.s.l., 20km south of the Charquini massif) recently analysed by Reference Gioda, L’Hote, Drake, Ronchail and PouyaudGioda and others (in press) did not display any significant trend over this period. In consequence, the estimated –0.4mw.e. a–1 mean annual mass balance for all the Charquini glaciers over the 20th century, corresponding to an increasing energy at the glacier surface of 4 Wm–2, could have resulted principally from a rise in temperature of 0.57°C according to the mass-balance sensitivity analysis. According to Kaser’s model, the 110m rise of the ELA (mean of the five glaciers) corresponds to a rise of 0.6°C, assuming only a change in temperature. Note that both results of air-temperature rise are consistent with the 0.55°C warming generally estimated in the literature over the same period (Reference Lean, Beer and BradleyLean and others, 1995). Thus, in contrast to the mid-17th to late 19th centuries, the recession of glaciers during the 20th century appears to have been the result of increasing air temperature.
Consistent with this hypothesis is the more severe retreat throughout the 20th century observed on Glaciares Norte and Oeste. Since the snouts there were 150 m lower at the end of the 19th century than on southern and western slopes, a rise in temperature, which controls the altitude of the snow/rain limit, might have led to higher ablation rates and consequently to a more rapid retreat of these tongues.
During the last two decades of the 20th century, the acceleration of glacier retreat is here, as elsewhere in the tropical Andes, coincident with the Pacific climate shift of 1976 after which the El Niño frequency and intensity increased significantly (Reference Francou, Vuille, Wagnon, Mendoza and SicartFrancou and others, 2003, Reference Francou, Dollfus, Deler and Mesclier2004). The mean annual mass balance over the 1983–97 period was about –0.51 mw.e. a–1 for Glaciares Sur, Noreste and Sureste and as high as –2.64mw.e. a–1 for Glaciares Norte and Oeste. This represents an increase in energy available at the glacier surface of about 5Wm–2 for Glaciares Sur, Sureste and Noreste and 28Wm–2 for Glaciares Norte and Oeste. Depending on the slope, the possible causes of the rapid glacier retreat in the late 20th century appear to be a rise of 0.8–3°C in air temperature (enhancing sensible heat transfer to the glacier), an increase in specific humidity of 0.3–1.5gkg–1 (reducing sublimation and increasing melting), a decrease in cloudiness of less than 0.5/10 to 1/10 (increasing solar radiation) or any combination of these processes.
5.3. The imbalance of Charquini glaciers
The assumption that moraine deposits represent glaciers in an equilibrium state is valid for small glaciers. It is therefore appropriate to determine the ELAs ‘geometrically’ from glacier areas between the LIA maximum and the first part of the 20th century. However, the method is spurious when glaciers are in a constant and rapid recession and hence markedly unbalanced, as occurred during the last decades of the 20th century. In this situation, the snout fluctuations reflect ablation conditions in the lower zone rather than oscillations of the ELA, which are generally located at the top of the glacier or above. In that case a comparison of the ELAs calculated using the AAR method with those determined from field observations (ablation stakes) shows how unbalanced these glaciers are in relation to the present climate and how rapid their shrinkage will be in the near future (Reference RamírezRamírez and others, 2001). From ablation stakes during the 1991–2004 period, the ELA0 is observed on Glaciar Zongo (facing south to east) and Glaciar Chacaltaya (facing south) at 5215ma.s.l. (Reference SorucoSoruco and others, 2005). Calculated for glaciers of similar aspect on the Cerro Charquini using the AAR method for 1997, the ELA is found at 4960–5095 m a.s.l. That is a difference between direct measurements and the AAR calculation of about 120m for the southern slopes and about 250m for the southeastern slopes. This average difference of 175 m represents the imbalance of the Charquini glaciers under current climatic conditions. Due to the small gap existing between their mean ELA and the maximum elevation of their upper reaches, these glaciers could recover equilibrium in the future only by a drastic reduction in their area. This implies that the Charquini glaciers, and in general all the glaciers of the Cordillera Real extending below ~5300ma.s.l., are unbalanced under the present climate and could disappear in the near future.
6. Conclusions
A new chronology and a quantification of the glacier recession since the LIA maximum have been presented in this study from the Charquini area in the Bolivian Andes. Surface area, ELA, volume and mass balance were reconstructed using well-preserved moraines dated by lichenometry and using photogrammetry for the last six decades of the 20th century.
The LIA culminated during the second half of the 17th century and glaciers subsequently retreated nearly continuously. Retreat slowed down during the early 19th century but no important readvances occurred.
The major acceleration of glacier recession observed between M9 and M10 suggests the LIA ended between about 1870 and 1910, consistent with evidence found throughout the Andes and in many places in the Northern Hemisphere.
The retreat between the LIA maximum and the late 19th century was probably associated with a significant decrease in precipitation and cloudiness, particularly during the 19th century.
In contrast, a sensitivity analysis shows that glacier retreat during the 20th century has been principally due to a rise in air temperature of about 0.57°C. In order to reverse the present deficit and to balance these glaciers, a decrease in energy available for melting ice of about 28 Wm–2 would be required. This exceeds the current climatic variability.
The AAR method suggests the ELA might have been on average 158 m lower during the second half of the 17th century (LIA maximum) than during the last decades of the 20th century, assuming that current glacier geometry corresponds to a steady state of the glaciers.
The ELA measured on nearby Glaciar Zongo for the 1991–2004 period (i.e. 5215 m a.s.l.) differs from the ELA geometric computation based on the AAR method by 175 m; this difference shows that the current glacier extent does not correspond to a balanced situation. It can be deduced that the Charquini glaciers, as well as all the small glaciers of the Cordillera Real below 5300 m a.s.l., are not in equilibrium with the present climate and that they could disappear in the next few decades. This confirms the assertion from Glaciar Chacaltaya (Reference RamírezRamírez and others, 2001).
Acknowledgements
We thank all those who took part in the field measurements (lichenometry and ground topography), the French Eclipse program (INSU–CNRS) managed by V. Masson-Delmotte (LSCE, CEA–CNRS), and P. Naveau (LSCE–CNRS) and D. Grancher (LGP–CNRS) for analysis using the Lichen Power® software. We also thank R. Naruse, editor of this paper, and two referees, H. Kerschner and B. Mark, for their helpful comments and suggestions.
