Annual Moraines or nivo-glacial Ridges:
Depositional ridges surveyed in the glacial foreland of the Tarija Glacier (Fig. 1) are comparable to the well-known annual (recessional) moraines first observed and described in detail in Iceland, and termed “minor moraines2,3,4, being comparable to small “push moraines”. Icelandic minor moraines are initially formed by bulldozing pre-existing debris during seasonal (boreal winter) glacial advance; then, spring supraglacial and englacial streams add debris, until continued ablation over the summer causes the glacier to retreat, disconnecting the ridge from glacial terminus. These features were subsequently identified in various glacial forelands from the European Alps, North America and South America. Their formation during the annual ablation season resulted in naming them “annual moraines”5,6,7, involving a wide range of formation mechanisms. Annual moraines in Bolivia were mentioned regarding the last retreat of the Charquini Glacier7, but, unlike Tarija, they are not presently forming. Our study of the Tarija Glacier‘s seasonal and annual behavior (Fig. 2), indicates that the Tarija annual ridges form in a unique way, although having partial similarity to the Type B annual moraines of Lukas5.
Owing to its high altitude (terminus at 4800 m.a.s.l.), the Tarija Glacier is cold-based and attached to the underlying permafrost, in contrast to traditional polythermal alpine glaciers. Studies of active Bolivian rock glacier distribution8 also suggest the Tarija terminus lies above the 0ºC mean annual air temperature isotherm, where frozen ground begins, inhibiting sliding because of the frozen base and explaining the lack of seasonal or interannual advance. We only identify retreat rate variations.
Ridge formation process was observed using many high resolution satellite images over a ten-year sequence (2010-2019), confirming the formation of regular, annual debris ridges. The ridges of the Tarija Glacier foreland may not qualify as moraines sensu stricto since they are not carried down and deposited by the glacier itself. The observed glacier surface intra- and inter-annual changes indicate that these ridges formed during the wet season (summer), by the direct melting of seasonal snow, triggering a range of supraglacial transport processes such as gravity mass flows (micro-avalanches and debris flows) and fluvial processes. While maximum ablation occurs during summer, there is no visible glacier retreat, as ambient energy is consumed in melting the seasonal snow that enhances debris transport towards the glacial terminus. Supraglacial flows deposit their sediment load at the glacial terminus, forced by the sharp change in slope between the glacier front and the ground. Debris ridges form from December to March. The glacial surface remains quite stable from March to November, when retreat occurs--during the colder dry season (a feature of tropical glaciers), suggesting older ice ablation occurs mainly via sublimation and evaporation.
The internal structure of similar ridges was described for the Charquini Glacier foreland7 only 18 km away. They comprise well-bedded debris alternating between massive, matrix-rich and open-framework gravels, suggesting interplay between mass wasting (gravity) and fluvial processes, as we observed in Tarija. The asymmetry of these ridges suggests they form like pronival ramparts (“discrete debris accumulations found at the foot of firn fields” 9), but their relation to a glacier terminus relates them better to the type B annual moraines5, with the difference they are not sourced by glacial ice melting but by the immediate melting of seasonal snow fall. We suggest the term nivo-glacial ridges since this genesis has not been mentioned for any other annual glacio-related ridge.
The longest known annual glacier mass balance series
A glacier retreats when it has a negative mass balance. It loses more ice volume at lower elevation (ablation) than ice gained by snowfall or transport into the accumulation area. These areas are separated by a dynamic “equilibrium line”. In some cases, the mass loss does not cause a retreat but rather a widespread deflation of the ice body, as observed in arid and hyperarid reservoir glaciers that usually do not show an equilibrium line or exhibit flow10. The regularity of annual ridges surveyed in the Tarija Glacier, suggests it suffered a quasi-steady process of annual mass loss, making this unique ridge succession a window to observe how the mass of this glacier depleted for over 258 years.
The horizontal distance the glacier retreats yearly (D) relates to the vertical loss of ice thickness (H) by H=D.tan q, where q is ice slope, that today is 29°, and estimated as 22° during LIA maximum (Fig. 3). H progressively reduces to zero at the equilibrium line. A proper estimation of the annual mass balance would require determination of the position of the equilibrium line over time, which is closely connected to D: A year of less ablation or more accumulation will force the equilibrium line to move down within the glacier, and vice versa. Models exist to calculate the equilibrium line position, but we prefer to we use D as a proxy for the annual mass balance, as it is the only variable we can confidently measure.
Correlation between Tarija Glacier D and mass balance results from neighboring glaciers (glaciological or hydrological11,12) was not possible due to a lack of data. The Tarija Glacier experienced a steady mass loss year after year over the last 2.5 centuries, whereas other methods suggest some years with positive mass balances in nearby glaciers11,12. We consider D, however, to be a better indicator of the total glacial mass balance. Traditional glaciological methods are based on point measurements that are then averaged over the glacier. At present, the Tarija Glacier is unique in forming debris ridges every year, although many others in the region, such as the Charquini Glacier7,13 formed them in the past. Thus, the Tarija record of D, creates a unique opportunity to understand how these glaciers were retreating during last 3 centuries while providing some hints of the complex climate-surface interactions. It also may help understanding how this long-term retreat may end, by a further analysis of the conditions that caused the glacial advance until the LIA maximum, prior to our oldest ridge.
Calibration of the record
The record was built from 254 successive annual ridges counted back from 2016 (Fig. 1), when the best satellite image was taken for mapping (pixel resolution, contrast, sharpness, cloudiness). Four additional ridges were added to bring the record to 2020, completing the 258-year record. We interpret the sequence to be continuous where ridge sizes are similar. Larger ridges are interpreted as potential time gaps. Thus, 258 years is the minimum timespan for this record, during which the Tarija glacier retreated 1.79 km at an average rate of c. 5.6 m/yr, while the rate during the 10 years analyzed in detail (2009-2019, Fig. 2) was 11.5 m/yr. The fastest retreat occurred between 1974 and 1994, peaking in 1987 at 20 m/yr. Some intervals are complex, such as between 1896 -1916 due to the separation of the Tarija Glacier from the north tributary and between 1972-1980 due to changes of the terminal tongue shape. Between 1896-1916 the glacier acquired a narrower frontal tongue, while between 1926 -1956 a wider, spatulate front developed.
Simple ridge counting suggests the oldest ridge formed in 1762 (minimum age). Readvances or periods of stagnation may introduce time gaps in our proposed 258-year record. Ridges larger than those observed to be annual would involve more time for their formation because larger sediment volume would require more time to reach glacier terminus. This might be the case for the initial glacier retreat after the LIA maximum. For instance, ridge #240 (Fig. 1) is larger than the typical annual ridge. Other exceptional ridges (red lines, Fig. 1), numbered #144, #100 and #36, may also indicate brief hiatuses, a possibility that needs to be taken into account to compare this record with others. A correlation is proposed between two prominent ridges that converge to form ridge #240 (1776 C.E., southernmost glacier foreland) and moraines M3 (1734 ± 21) and M4 (1765 ± 17) of the Charquini Glacier7,13, suggesting no more than 2 decades of time loss after #240 ridge.
Changes in the glacial terminus geometry between ridges #36 and #37 (Fig. 1), suggests that D can be laterally highly variable. We measured D along a central axis where it should be maximum. Distances diminish progressively to almost zero at the margins of the snout. Significant surface erosion, however, limited this task so the reconstructed record is biased by the lack of a continuous succession along a central line. Measurements were done along the segments indicated in Fig. 1 that allowed the only continuous succession. The most problematic segment is near the LIA maximum, as older ridges are more degraded and they are absent along the centre-line of the retreating glacier. At that time, the Tarija was joined by a tributary glacier descending from the north that left a continuous succession of annual ridges.
We applied 5-, 11-, and 50-year moving averages to depict the main tendencies (Fig. 4). From the initial ridge at 1762 or earlier, until 1976, the average D remained between 4 and 6 m/yr, with some intervals increasing to 8 m/yr, centered at 1888, 1911, 1925, and 1933. Since 1976, D increased rapidly to exceed the 12 m/yr, between 1981-1988. After 1988, retreat slows to 6 and 8 m/yr, and after 2003, it increased to c. 10 m/yr. Annual retreat today is slower than the maximum rate observed between 1981-1988, a tendency shown by all curves.
Understanding this record
It is important to emphasize that glacial retreat is not only an effect of temperature, but is also significantly affected by precipitation. Older glacial ice does not melt if more seasonal snow falls and consumes ambient energy to melt; years with higher precipitation diminish glacial melting and retreat. There is no simple relation between retreat, temperature, and precipitation, but a detailed analysis of these factors using annual glacial ridges suggests that the main controlling factor on D is summer temperature6.
To differentiate the effects of temperature and precipitation we compare the Tarija Glacier retreat to Lake Titicaca, which receives glacial meltwater. The Titicaca lake-level record begins in 1914. Its maximum level of 3812.55 m.a.s.l. (December 1986)14 corresponds exactly with the maximum annual retreat (20 m/yr) of the Tarija Glacier, evinced by the ridge formed from January-March 1987 mirrored by the glacial terminus melting between April-December 1986. This suggests a direct contribution of meltwater to the rising Titicaca lake-level. The lowest recorded lake level (April 1943) correlates with the period of slowest retreat rate at 5 m/yr (11-yr average). Other peaks of Titicaca lake-level recorded in 1924, 1934, 1980, 1989, and 2005 roughly align with faster glacial retreat rates. Elevated temperatures causing faster glacial retreat apparently did not produce a sufficient increase in evaporation from the lake to counterbalance the meltwater surplus.
Some lake-level anomalies occur almost independently from changes in the glacial retreat rate,. Such variation suggests another independent and first order control on lake level, which we interpret to be precipitation.. The e1 triple peak (Fig 4) is well shown by lake level, but is very subtle in the glacial retreat history. Increased precipitation would slow the retreat rate under the same thermal regime. Thus, the e1 triple peak (1949 to 1967) suggests only a precipitation increase without temperature change.
A more distal, independent large-scale surface system to compare with the Tarija record is Lago Mar Chiquita, the largest interior (Pampean) lake in central Argentina. Mar Chiquita lake-level measurement began in 1968, but paleolimnological studies allow reconstruction to c.189015. All main peaks of Mar Chiquita levels coincide with well-defined increased glacial retreat periods (C, D, E, F and G, Fig. 4). As at Titicaca, the e1 peak also occurs in Mar Chiquita (although as a single peak), while the peak centered in 1994, does not occur in either the Lake Titicaca or the Tarija Glacier records, suggesting local controls. Lago Mar Chiquita is not meltwater fed and follows almost the same evolution of other Pampean lakes (Fig. 3), suggesting a connection between temperature, precipitation and lake level. We may summarize that in one mode, both precipitation and temperature increase and in a secondary mode precipitation increases without a significant change of temperature. The covariance of temperature and precipitation over longer time periods in southern South America is also suggested from analysis of other Holocene geoarchives16.
Temperature and precipitation can be linked by the available environmental energy: a temperature rise may fuel the water cycle by fostering evaporation. In the Bolivian Andes, this relation is logical as most snowfall over glaciers depends on Amazonian evaporation. Given that the primary source of energy is the sun, we compare fluctuations in solar irradiance with the Tarija Glacier retreat record.
Solar energy and terrestrial cycles
Decadal differences notwithstanding, post-LIA glacial retreat in Bolivia began almost synchronously with well-dated northern hemisphere glaciers 17, 18. The fact that the LIA glacial maximum coincides roughly with the Maunder sunspot minimum suggests that solar irradiance played some role in glacial behavior. We alluded above to the relation between annual retreat D and summer temperature6, and its potential connection with precipitation anomalies through the regional ambient energy. It is beyond the scope of this work to discuss the dynamic connections and potential buffering of solar energy in different regions by different processes. We simply suggest that there is a potential connection between solar irradiance fluctuations and environmental effects.
The Tarija Glacier retreat curve was compared with two established, solar-related curves: the Heliomagnetic Field (HMF) strength calculated for the last 3 centuries19 and the open solar magnetic flux (Fo20), exhibiting a diachronous match (Fig. 4). The HMF curve does not provide an unambiguous measure of heliospheric activity and is not a physical index, but rather a formal fitting parameter describing the cosmic rays affecting the Earth’s outer ionosphere and thus, it cannot be linked directly to a universal solar activity index21. The long-term scale of HMF potential (beyond the 11-year solar cycle), however, is closely related to Fo, which is a physical quantity describing the solar magnetic variability21, 22. Both are shown in Fig. 4, averaged over 11- and 50-years to remove the effect of the sunspot cycle. The atmospheric CO2 trend for the same period was added23, revealing a negative correlation with the Tarija Glacier ice loss history.
We noted peaks of higher HMF strength and faster glacial mass loss (A to G, Fig.4), with a phase lag of 30 -35 years, which could result, in part, from the potential time gaps involved by larger debris ridges. The last 4 decades, however, can be confidently regarded to be continuous; yet the lag between the HMF and the retreat rate peaks F and G, remains in spite of shape similarity: fast initial climb, a higher peak F, a short trough, and a second minor peak G. This delay may suggest the existence of buffering mechanisms delaying the effect of external energy over natural systems.