Latest Pleistocene and Holocene glacial events in the Colonia valley, Northern Patagonia Icefield, southern Chile

The Northern Patagonia Icefield (NPI) is the primary glaciated terrain worldwide at its latitude (46.5–47.5°S), and constraining its glacial history provides unique information for reconstructing Southern Hemisphere paleoclimate. The Colonia Glacier is the largest outlet glacier draining the eastern NPI. Ages were determined using dendrochronology, lichenometry, radiocarbon, cosmogenic 10Be and optically stimulated luminescence. Dated moraines in the Colonia valley defined advances at 13.2 ± 0.95, 11.0 ± 0.47 and 4.96 ± 0.21 ka, with the last being the first constraint on the onset of Neoglaciation for the eastern NPI from a directly dated landform. Dating in the tributary Cachet valley, which contains an ice‐dammed lake during periods of Colonia Glacier expansion, defined an advance at ca. 2.95 ± 0.21 ka, periods of advancement at 810 ± 49 cal a BP and 245 ± 13 cal a BP, and retreat during the intervening periods. Recent Colonia Glacier thinning, which began in the late 1800s, opened a lower‐elevation outlet channel for Lago Cachet Dos in ca. 1960. Our data provide the most comprehensive set of Latest Pleistocene and Holocene ages for a single NPI outlet glacier and expand previously developed NPI glacial chronologies.


Introduction
The Northern and Southern Patagonia Icefields (Fig. 1) form the largest continental ice mass in the world outside of Antarctica and Greenland (Loriaux and Casassa, 2013) and occupy the only terrain (except for very southern New Zealand) glaciated at their latitude during the Late Pleistocene and Holocene. Changes in the volume and extent of these icefields are tied closely to changes in climate (Glasser et al., 2004), and thus mapping and dating these fluctuations offers important and unique temporal information on past climatic changes in the Southern Hemisphere.
Most research on the glacial geology of the Northern Patagonia Icefield (NPI) has focused on the modern ca. 150-year period. This research has inventoried and dated the substantial reductions in volume and areal extent of ice as the NPI and its outlet glaciers retreated from late-1800s maximum positions (e.g. Harrison et al., 2007;Rivera et al., 2007;Davies and Glasser, 2012;Loriaux and Casassa, 2013). Other studies have examined the eastward extension of the icefield by almost 200 km into Argentina during the Last Glacial Maximum (LGM) (e.g. Kaplan et al., 2004;Hein et al., 2010), a regional advance of NPI outlet glaciers during the Latest Pleistocene and early Holocene Harrison et al., 2012), and Neoglacial advances during the mid to late Holocene (summarized by Aniya, 2013). However, the history of NPI advance and retreat between LGM deglaciation and the late-1800s maximum generally is poorly known (Glasser et al., 2004;Masiokas et al., 2009;Aniya, 2013). This lack of knowledge limits studies which compare either different parts of the NPI and its neighboring glaciers (e.g. Douglass et al., 2005) or the NPI with the better studied Southern Patagonia Icefield (SPI) (e.g. Strelin et al., 2014). More generally, this void limits our understanding of Southern Hemisphere glacial history and our ability to corroborate reconstructed temperature records.
The Colonia valley (Figs 1 and 2) on the eastern side of the NPI contains glacial features that date back ca. 13 ka and thus provide an important constraint on Patagonia Icefield glacial history for the period between the LGM and the late 1800s. The Colonia Glacier, with an ice area of 288 km 2 in 2001, is the fifth largest of 24 main outlet glaciers ( Fig. 1) draining the NPI (Rivera et al., 2007). Because it is the largest outlet glacier on the east side of the NPI and drains the central part of the icefield, the Colonia Glacier is probably a sensitive proxy of icefield changes. Glacial landforms in the valley have been mapped in the field (Tanaka, 1980;Harrison and Winchester, 2000) and from visible satellite imagery (Glasser et al., 2009, but dating has been restricted to landforms younger than ca. 150 years . The rapid retreat of NPI outlet glaciers during the past century Davies and Glasser, 2012) has increased the area of proglacial lakes surrounding the NPI (Loriaux and Casassa, 2013) and, more importantly, the frequency of catastrophic glacial lake outburst floods (GLOFs) Dussaillant et al., 2010). Most of the known GLOFs originating from the NPI during the past century have occurred in the Colonia valley (Tanaka, 1980;Friesen et al., 2015). GLOFs result from the sudden and catastrophic drainage of supraglacial or proglacial lakes and present a significant hazard to human populations and infrastructure (Richardson and Reynolds, 2000). Thus, understanding the history and mechanisms of the GLOFs in the Colonia valley has both local and global implications. A better resolved chronology of Colonia Glacier advance and retreat is essential to future studies of Colonia valley GLOFs.
Studies in the Colonia valley were undertaken to identify glacial landforms and deposits and to collect data and samples suitable for improved constraint on the post-LGM glacial history of the NPI. Field studies focused on moraines in the Colonia valley proper (Fig. 2) and deposits and landforms in the tributary Cachet valley containing Lago Cachet Dos (Fig. 3), a lake currently dammed by the Colonia Glacier.

Description of study area
The Colonia valley extends east and north-east from the NPI to the Río Baker (Fig. 1). The eastern 19-km reach of the valley is drained by the Río de la Colonia (henceforth abbreviated to Río Colonia), a braided glacial outwash river with a wide ( 3 km), largely unvegetated floodplain composed of fluvio-glacial sediment. The center reach of the valley contains Lago Colonia, an 8-km-long morainedammed lake (Fig. 4A). The 5-km reach upstream of the lake contains a sparsely vegetated outwash plain with the eroded remains of moraines (Fig. 4B) formed during the past ca. 150 years  and a proglacial lake at the terminus of the Colonia Glacier (Fig. 5G). The Colonia Glacier extends north-west 18 km from its terminus (elevation of ca. 200 m) to the base of an icefall (elevation of ca. 950 m), where this outlet glacier flows from the NPI.
The Colonia Glacier today and in the past created icedammed lakes in two valleys tributary to the Colonia valley. Both lakes have been intermittent through time depending on the position of the Colonia Glacier and its effectiveness as a dam. GLOFs are (or were) a feature of both ice-dammed lakes (Tanaka, 1980;Dussaillant et al., 2010). One of these icedammed lakes is the current Lago Cachet Dos in the Cachet valley (Fig. 2). The second lake formed in the Arco valley on the south side of the Colonia valley (Fig. 5G) when the Colonia Glacier abutted the north flank of Cerro Colonia. This paleo Lago Arco existed during much of the 20th century and was larger than the present-day moraine-dammed Lago Arco shown in Fig. 2 (Tanaka, 1980). GLOFs from paleo Lago Arco occurred from before 1930 to 1968 (Tanaka, 1980), while GLOFs from Lago Cachet Dos started in 2008 (Dussaillant et al., 2010). Our field studies included the Cachet valley because determining previous fluctuations in the size and  (Fig. 2) and Cachet ( Fig. 3) valleys, and cosmogenic 10 Be sampling sites (triangles) and site names from Glasser et al. (2012). NPI outline and mask are from the Global Land Ice Measurements from Space (GLIMS) glacier database (Davies, 2012

Methods
Field studies were conducted during austral summers 2011/12 and 2012/13. Sample sites were located using a hand-held GPS with stated accuracy of <15 m using Universal Transverse Mercator projection, zone18S World Geodetic System of 1984 (WGS84). Dendrochronology and lichenometry ages are reported as calendar ages. Terrestrial cosmogenic nuclide ( 10 Be) and optically stimulated luminescence (OSL) ages are reported as calendar years before publication date. Radiocarbon ages are reported as calendar-calibrated years before present (BP, with present ¼ 1950) calculated from the conventional radiocarbon age, the SHCAL13 database (Hogg et al., 2013), and the 1-sigma error band.
Dendrochronological ages (Table 1) were determined from cores collected from live trees using an increment borer. Cores were mounted in the laboratory, and tree rings were counted using a stereoscopic microscope. Missing rings for cores not containing the tree center were estimated using the methodology of Duncan (1989). The growth rate for each tree for the section below coring height and ecesis were assumed to be 11.7 cm a À1 and 26 a, respectively, based on previous work in the Colonia valley .
Lichenometric ages (Table 1) were determined at six sites near a cored live tree. The longest diameter of 3-5 of the largest individual lichens (Placopsis perrugosa) on boulders at each site were measured (Fig. 5E). A lichen growth rate of 4.7 mm a À1 (which incorporates the colonization period) was assumed based on previous work in the Colonia valley . The mean age for the three largest lichens at each live-tree site was used for the site.
10 Be surface-exposure ages (Table 2) were determined for samples collected from the top of upward-facing surfaces of boulders (with a b-axis >1 m where possible) on lightly vegetated moraine crests using hammer and chisel ( Fig. 4C) according to techniques from Gosse and Phillips (2001). Some boulders had a thin layer of moss or lichen, which was removed before sampling. Topographic shielding was estimated from skyline measurements made with an Abney level. Physical preparation, quartz purification, dissolution and conversion into beryllium oxide, and 10 Be analyses were performed by the Purdue Rare Isotope Measurement Laboratory (West Lafayette, IN, USA). All 10 Be/ 9 Be ratios were measured against the revised ICN standard of Nishiizumi et al. (2007), which assumes a half-life of 1.36 Ma. The 10 Be/ 9 Be ratio (2.84 Â 10 À15 AE 0.57 Â 10 À15 ) of the processing blank prepared with the samples was subtracted from the 10 Be/ 9 Be ratios of the samples. 10 Be exposure ages were calculated with the CRONUS-Earth calculator  using a quartz density of 2.73 À g cm À3 , topographic shielding listed in Table 2, the Dunai timedependent scaling scheme and production rates from Lago Argentino in Patagonia . We applied no corrections for erosion rate or snow shielding (Gosse and Phillips, 2001), and these assumptions should not affect our conclusions. Similarly, use of alternative scaling schemes  resulted in age differences of <3% and therefore would not affect our conclusions. To facilitate comparisons, we recalculated the ages reported by Glasser et al. (2012) for nearby sites (Fig. 1) using the same scaling scheme, production rates and erosion rate listed above.  (Table 1). Trimlines and 2007 Lago Cachet Dos boundary are from Friesen et al. (2015). Elevation of lacustrine trimline is ca. 500 m. Image produced from Digital-Globe WorldView-2 data collected on 13 February 2014, 12 days after a GLOF emptied Lago Cachet Dos.
Burial ages based on optical dating (Table 3) were determined for sediment samples collected in 5-cm-diameter by 15-cm-long PVC core tubes inserted horizontally into shaded and freshly excavated vertical surfaces. Quartz and potassium feldspar grains (180-250 mm) were analysed by single aliquot regeneration Wintle, 2000, 2003) using continuous-wave OSL and continuous-wave infrared stimulated luminescence (IRSL), respectively. Luminescence data were subject to community standard quality-control tests including the recycling-ratio and dose-recovery tests (Rhodes, 2011). Additional information on OSL dating is given in supplementary Appendix S1.
Radiocarbon ages (Table 4) were determined for samples of outer tree rings collected using a handsaw from eight in situ dead trees exposed on the Cachet valley floor during post-GLOF periods. In addition, one sample was collected from the center of one of the trees and another from a paleo-soil excavated beneath an in situ tree. Radiocarbon analyses were performed by Beta Analytical (Miami, FL, USA) using gas proportional counting for tree-ring samples and accelerator mass spectrometry for the soil sample. Samples were pre-treated using sequential acid/alkali/acid washes. Calendar-calibrated ages were calculated using OxCal 4.2.4 software (Bronk Ramsey, 2009) and SHCAL13 database (Hogg et al., 2013).

Lago Colonia terminal moraine and outwash terraces
The Lago Colonia terminal moraine that dams Lago Colonia (Figs 2 and 4A) was first identified by Tanaka (1980), who named it Colonia Moraine No. 1 and estimated it to be 2.8 km long, 1.5 km wide and as much as 85 m in height above Lago Colonia. The moraine crest ranges in elevation between ca. 190 and 210 m. Numerous granitic boulders 2-5 m in length litter the moraine surface ( Fig. 4C). Based on observation of the expansive north-facing slope cut through the moraine by the Río Colonia ( Fig. 4E), the moraine is underlain by massive diamicton. Several outwash terraces, which are at successively lower elevations and underlain primarily by coarse-grained gravel and sand, extend downstream from the moraine (Figs 2 and 4E). OSL samples were collected from a loess deposit of silt and fine-to-coarse sand near the top of the moraine (Fig. 4D) and from a medium sand lens near the top of the uppermost terrace (Fig. 4F).

Río Claro lateral moraine
The lateral moraine at the mouth of the Río Claro valley (Figs 2 and 4G) was first identified by Tanaka (1980), who named it Colonia Moraine No. 2. The moraine is ca. 200 m high, and its crest, which is more rounded than the crest of the Lago Colonia terminal moraine, ranges in elevation between ca. 350 and 380 m. Semi-rounded granitic boulders on the moraine crest are as large as 2 m in diameter (Fig. 4H).

Cachet valley
The Cachet valley is a hanging valley on the Colonia valley's north-east side, and its topographic configuration has allowed preservation of dateable materials that record Colonia Glacier advances and retreats. A lower outlet at the south-east corner of the lake (Fig. 3) limits the lake's maximum level to an elevation of ca. 420 m. A trimline at an elevation of ca. 500 m on both sides of the Cachet valley (Figs 3 and 5C) is demarcated by a dense and mature forest above the trimline and an immature and openly spaced forest of smaller diameter trees below the trimline (Friesen et al., 2015). The lateral extent and horizontal nature of this trimline indicate that it was formed during a high stand of Lago Cachet Dos and that, at that time, the lake extended from the Colonia Glacier to the modern boundary of Lago Cachet Uno. Downstream and near the south-east corner of Lago Cachet Dos, this lacustrine trimline extends to an abandoned upper outlet channel (ca. 500 m) that controlled the former 500-m level of Lago Cachet Dos. Near the south-west end of the lake, the lacustrine trimline grades into a glacial trimline that extends to the west up the Colonia valley (Figs 2 and 3). In the upper Cachet valley, the lacustrine trimline grades into another glacial trimline (Fig. 3), which rises up-valley on both sides of Lago Cachet Uno (Friesen et al., 2015). Based on age-dating of similar trimlines in and near the Arco valley ( Fig. 5G; Harrison and Winchester, 2000), these glacial trimlines probably demarcate the maximum late-1800s extent and thickness of the Colonia and Cachet Glaciers, respectively.
Before the onset of GLOFs in 2008, the Cachet valley between Lago Cachet Uno and the Colonia Glacier contained a flat valley floor and the braid plain of the Río Cachet over its upstream half while Lago Cachet Dos, at its 420-m level, filled the downstream half (Friesen et al., 2015). Large-scale  . Abutment of Colonia Glacier against the lower flank of Cerro Colonia created historical ice-dammed Lago Arco from which GLOFs emanated during the 20th century (Tanaka, 1980). Cerro Colonia lateral moraine is indicated. (H) Boulder located on crest of Cerro Colonia lateral moraine and sampled for 10 Be dating (ARCO3).  (Friesen et al., 2015). This erosion exposed the stratigraphy of the valley fill (e.g. Fig. 5A) and unearthed hundreds of in situ, upright trees (Fig. 5C, D) that presumably grew on the valley floor during periods when the Colonia Glacier was not large enough to dam the Cachet valley and create a Lago Cachet Dos. Trees in the mid portion of the valley tend to be short (<2 m) and show signs of abrasion (Fig. 5D), while trees in the upper half of the valley are taller ( 8 m) and much less abraded (Fig. 5C). A paleo-soil 14 C sample excavated directly beneath an in situ tree (Fig. 5D) came from a 5-cm-thick organic-rich ('A' horizon) layer located below a 10-cm-thick brown clayey coarse sand and above a >20-cm-thick medium brown, pebbly, fine-to-coarse sand (Fig. 5F). The valley fill in the Cachet valley is a combination of fluvial deposits consisting of cross-bedded, poorly to moderately sorted silt to sandy gravel and cobbles, lakebed deposits consisting of horizontally bedded gray clayey silts, and deltaic lake deposits. Much of this valley fill probably is outwash sediment that first came from the Cachet Glacier itself and then from streams that eroded the moraine that now dams Lago Cachet Uno. This drainage fed an extensive sandy delta, including fluvial topset beds of the delta plain that prograded into Lago Cachet Dos. These deltaic deposits (Fig. 5A, B) occur most prominently near the northern end of the 2007 Lago Cachet Dos (Fig. 3) and consist of thinly bedded and steeply dipping delta foreset strata composed of well-sorted, fine-to-coarse sand. Samples were collected at three locations for OSL dating.
Ages 10 Be surface-exposure ages for moraines 10 Be ages and associated uncertainties for individual boulders are listed in Table 2. Ages for the sampled boulders for each moraine were grouped fairly closely with no apparent outliers. Ages for the four boulders from the Lago Colonia terminal moraine overlap within 1-sigma uncertainties and ranged from 4.65 AE 0.34 to 5.13 AE 0.22 ka. The four ages for the Río Claro lateral moraine were grouped almost as closely and ranged from 10.3 AE 0.53 to 11.4 AE 0.51 ka. For the Cerro Colonia lateral moraine, the three ages were within 2-sigma uncertainties and ranged from 12.4 AE 0.54 to 14.1 AE 0.63 ka. Overall, the mean age and standard deviation for each site were 4.96 AE 0.21 ka for the Lago Colonia terminal moraine, 11.0 AE 0.47 ka for the Río Claro lateral moraine and 13.2 AE 0.95 ka for the Cerro Colonia lateral moraine.

OSL burial ages
The burial age for the deltaic sand sample (OSL1) from the Cachet valley (2.95 AE 0.21 ka; Table 3) suggests an earlier episode of delta formation in a paleo Lago Cachet Dos. IRSL ages were 4.70 AE 0.23 ka for the loess sample (OSL5) from the Lago Colonia terminal moraine and 2.79 AE 0.34 ka for the fluvial sample (OSL6) from the outwash plain.

Radiocarbon ages from Lago Cachet Dos lake bed
Ages (Table 4) for the paleo-soil (995 AE 38 cal a BP; sample Soil11) and outer rings from Tree 18 and Tree 11 (830 AE 50 and 790 AE 47 cal a BP, respectively) indicate that the Cachet valley had terrestrial vegetation and no lake over a multicentury period ca. 0.85-1.1 ka ago. Potential causes for the death for these two trees include inundation by lake water or burial by fluvial or colluvial sediment. If a paleo Lago Cachet Dos had formed and killed the trees by inundation, the lake had disappeared by 410 AE 55 cal a BP, the age of sample Tree 10B collected from the center of Tree 10.
Outer tree-ring samples from six in situ trees provided conventional radiocarbon ages ranging from 90 to 210 14 C a BP (Table 4) and, due to the non-linearity of the radiocarbon calibration curve, a wide but largely overlapping range of  Fig. 3), these trees were probably killed nearly simultaneously by inundation when Lago Cachet Dos last formed, probably as the Colonia Glacier thickened en route to its late-1800s maximum. The possible age of these tress can be narrowed to the earliest part of the 1650-1950 period by assuming that Lago Cachet Dos was probably created >100 years before the late-1800s maximum. This assumption is reasonable considering that the lake has continued to exist for >100 years after the late-1800s maximum even in the face of substantial post-late-1800s retreat of the Colonia Glacier Davies and Glasser, 2012). Five of the six trees have a distinct calibrated age range before 1750 (!100 years before the late-1800s maximum); the mean of these ranges (245 AE 13 cal a BP) is considered the most plausible age for the formation of Lago Cachet Dos.

Dendrochronology and lichenometry ages for Lago Cachet Dos trimline
The age of the lacustrine trimline that encircles much of Lago Cachet Dos at an elevation of ca. 500 m was bracketed from its appearance in aerial photographs and estimated using dendrochronology and lichenometry. The 500-m lake was at this trimline elevation in 1945, based on a map compiled by Lliboutry (1998, fig. 27) from aerial photographs. By 1975, Lago Cachet Dos had lowered to the 420-m level (Fig. 2) based on the 1: 50 000 Cord on Soler quadrangle (published in 1982 by the Instituto Geogr afico Militar de Chile and based on 1975 aerial photographs). Based on our dendrochronological and lichenometric ages, the lake level dropped to its current 420-m position in ca. 1960. Tree-ring cores from 13 live trees located below the lacustrine trimline provided ages between 1945 and 1964, with a mean of 1955 (Table 1). Ages based on the mean of the three largest lichens at the five live-tree sites where lichens were measured ranged from 1957 to 1965, with a mean of 1961 (Table 1).

Chronology of Colonia Glacier advance and retreat
During the LGM, the Colonia Glacier filled the Colonia valley, coalesced with other NPI outlet glaciers, flowed eastward and formed moraine systems in Argentina (Kaplan et al., 2004;Singer et al., 2004). The LGM occurred at 27-25 ka with subsequent advances at 23-22, 20-18 and ca. 18-17 ka; rapid deglaciation from the LGM moraines began after 18-17 ka (Hein et al., 2010;Boex et al., 2013). During deglaciation but while the Río Baker's path to the Pacific Ocean was still dammed by the retreating outlet glaciers, paleo lakes formed with surface elevations of ca. 489-512 m and later at ca. 375-397 m near the Nef and Colonia valleys; the Río Baker finally drained to the Pacific Ocean at ca. 12.8 ka (Turner et al., 2005). The earliest evidence for a post-LGM ice position of the Colonia Glacier is the Cerro Colonia lateral moraine high on Table 4. Radiocarbon ages for wood and soil samples collected in October 2011 or February 2012. Data for samples Tree 10 to Tree 1 are shown in downstream-to-upstream order in the Cachet valley. Tree samples were collected from outer rings except sample Tree 10B, which was collected from the center of Tree 10.  AD 1720-1815AD 1835-1850AD 1855-1880AD 1925-1950 Ã Radiocarbon years before present with 'present' ¼ 1950. The conventional radiocarbon age represents the measured radiocarbon age corrected for isotopic fractionation calculated using the measured 13 C/ 12 C ratio. †Calculated using OxCal 4.2.4 software (Bronk Ramsey, 2009) and SHCAL13 database (Hogg et al., 2013) from the conventional radiocarbon age. ‡Arithmetic mean and 1-sigma error in calendar-calibrated years before 1950. §Mean age calculated for only the earliest calendar-calibrated age range. ¶Mean age not calculated.
the north flank of this mountain. Dated to 13.2 AE 0.95 ka (Table 2), this moraine records the most extensive Colonia Glacier position that has been found within the Colonia valley proper and the position held just before the initiation of westward Río Baker drainage at ca. 12.8 ka (Turner et al., 2005). The Río Claro lateral moraine also records a distant downvalley advance or perhaps stabilization during post-LGM retreat of the Colonia Glacier at 11.0 AE 0.47 ka. No terminal moraine associated with either the Cerro Colonia or the Río Claro lateral moraines has been identified, but the moraine mounds at Lago Esmeralda (site LE, Fig. 1) dated by 10 Be at 12.0 AE 0.75 ka ) may indicate an approximately contemporaneous downstream limit. Although we found no evidence for early Holocene activity for the Colonia Glacier, evidence for advance and retreat during the Neoglacial (Porter and Denton, 1967) is relatively abundant. In the early Neoglacial, the Colonia Glacier advanced and created the terminal moraine at Lago Colonia at 4.96 AE 0.21 ka. The IRSL age (4.70 AE 0.23 ka, sample OSL5, Table 3) for the loess from the top of the Lago Colonia terminal moraine is a minimum age for formation of this moraine and is consistent with the 10 Be age.
Information on more recent Colonia Glacier activity comes from the Cachet valley, which has a multi-millennia record of alternating periods of the valley either containing a lake during periods of an advanced and thickened Colonia Glacier or being forested during periods of stable fluvial drainage, probably when the glacier was smaller than it is today. The oldest deltaic-sand sample (OSL1 , Table 3) indicates a Colonia Glacier advance that created a paleo Lago Cachet Dos at or before 2.95 AE 0.21 ka. No terminal moraine associated with this advance has been identified. However, the fluvial sand from the upper outwash terrace of the Lago Colonia terminal moraine (2.79 AE 0.34 ka, sample OSL6, Table 3) also dates to this period and suggests that the Colonia Glacier may have readvanced as far down valley as this moraine. Sometime after 2.95 AE 0.21 ka, but before 995 AE 38 cal a BP, the glacier retreated and the paleo Lago Cachet Dos disappeared, allowing soil development (sample Soil11) on the valley floor. However, the lake may have been re-dammed at 810 AE 49 cal a BP (mean of ages for outer-ring samples from Tree 18 and Tree 11, Table 4) assuming these two trees were killed by inundation. If a paleo Lago Cachet Dos did form at 810 AE 49 cal a BP, it had disappeared by 410 AE 55 cal a BP based on the radiocarbon sample from the center of Tree 10, and the valley floor was again forested.
At 245 AE 13 cal a BP (Fig. 5C), thickening of the Colonia Glacier dammed the Cachet valley forming a lake, presumably with a water level of ca. 420 m controlled by the current outlet channel for Lago Cachet Dos. Sometime later but probably before the late-1800s maximum, continued advance of the Colonia Glacier into the Cachet valley sealed the lower outlet channel. This event allowed the lake level to rise to the elevation of the upper outlet channel at ca. 500 m. This continued advancement of the Colonia Glacier reached a maximum, recorded by glacial trimlines (Fig. 5G) and remnants of a terminal moraine (Fig. 4B) ca. 5 km downstream from the current terminus, dated to 1850-1880 .
Between the late-1880s maximum and 1996, the Colonia Glacier terminus retreated ca. 1.5 km ; Fig. 2). Only small and isolated remnants of the late-1800s terminal moraine and post-late-1800s recessional moraines remain downstream of the current Colonia Glacier terminus, probably because of erosion by numerous GLOFs during the past century from the Arco and Cachet valleys. GLOFs from the Arco valley stopped occurring in 1968 presumably because retreat of the Colonia Glacier away from the northern flank of Cerro Colonia removed the dam.
By ca. 1960, retreat of the Colonia Glacier from the Cachet valley uncovered the lower outlet channel for Lago Cachet Dos at ca. 420 m, and the lake abandoned its previous 500-m level. Between 1996 and 2013, the Colonia Glacier terminus retreated another ca. 1.5 km, and the individual proglacial lakes in the Arco and Colonia valleys (Fig. 5G) joined to form a single larger proglacial lake in 2014 (Fig. 2). GLOFs from Lago Cachet Dos started in 2008 probably due to thinning and weakening of the Colonia Glacier, thus allowing the lake to drain catastrophically via meltwater channel(s) within or beneath the 8-km terminal reach of the glacier (Dussaillant et al., 2010).

Post-LGM glacial chronology for the NPI
Combining the Colonia record reported here with previous work on other NPI outlet glaciers (Fig. 6) provides a clearer post-LGM glacial chronology for the NPI as a whole and allows a better comparison with the more thoroughly studied SPI. However, as recently noted by Strelin et al. (2014) and Aniya (2013), who reviewed Neoglacial advances for the SPI and the combined icefields (Table 5), respectively, attempts to find regional patterns in Holocene glacial chronology for the Patagonia Icefields can be elusive with existing data. As discussed in this section, the most regionally consistent data for the NPI document advances at the beginning of the Holocene and in the late-1880s (Fig. 6). The questions of what happened before and between these two periods and if advances in this intervening period were synchronous have not yet been completely resolved.
During the Late Pleistocene, the NPI remained close to LGM positions until ca. 18 ka, when the extensive ice lobes of the ice sheet quickly shrank and separated into the discrete outlet glaciers found today in valleys draining all sides of the icefield (Boex et al., 2013). Erratic boulders on the Río Bayo valley floor (Fig. 1)  At the beginning of the Holocene (10 000 14 C a BP or 11 360 cal a BP), outlet glaciers in four valleys (Fig. 6) extended as much as 50 km eastward forming terminal moraines in the Nef, Soler and Leones valleys (sites RC, LPB and MLV, respectively, in Fig. 1) and lateral moraines in the Colonia and Nef valleys (sites NEF in Fig. 1 and CLARO in Fig. 2), all dated by 10 Be at 10.4-11.2 ka ( Table 2; Glasser et al., 2012). Whether these moraines represent stabilization of the NPI during ongoing retreat from LGM positions or regional expansion of NPI outlet glaciers is not clear , but overall this advance is one of the best documented for the NPI.
Evidence for NPI advances during the first half of the Holocene is scant. Harrison et al. (2012) dated advances at 9.7-9.3 AE 1.2 and 7.7 AE 1.1 ka (Fig. 6) for the San Rafael Glacier (Fig. 1), and Douglass et al. (2005) dated advances at >10 and ca. 7 AE 1 ka [recalculated by Strelin et al. (2014) using Kaplan et al. (2011) production rates] in the Río Avil es valley ca. 25 km east of the NPI. As noted by Aniya (2013), more data are necessary to determine whether any of these advances was regionally synchronous.
The onset of Neoglaciation for the NPI appears to be indicated by terminal moraines of the San Rafael Glacier dated at 5.7 AE 0.6 ka  and the Colonia Glacier dated at 4.96 AE 0.21 ka (Table 2) and 4.70 AE 0.23 ka (Table 3). No other glacial features on either side of the NPI have been dated to this period. Their timing coincides with the Strelin et al. (2014) age for an SPI advance at 5000-6000 cal a BP and is similar to the Aniya (2013) age for Neoglacial I (5130-4430 cal a BP, Table 5).
The record compiled to date for the main part of the Neoglacial (Fig. 6) does not indicate broad regional patterns for NPI and vicinity. The oldest advances include the terminal moraine in the Leones valley dated to ca. 3.3-2.4 ka by Harrison et al. (2008) and deltaic sediment in the Cachet valley indicating a Colonia Glacier advance at ca. 2.95 AE 0.21 ka (Table 3). These events broadly coincide with advances dated to 2500-2000 cal a BP for the SPI (Strelin et al., 2014) and the Aniya (2013) age for Neoglacial III (2770-1910 cal a BP, Table 5). Similarly, reported ages for advances in a few eastside NPI valleys coincide with SPI advances dated to 1500-1100 and ca. 700 cal a BP (Strelin et al., 2014) and the Aniya (2013) age for Neoglacial IV (1450-750 cal a BP, Table 5). These advances of the NPI are recorded by moraines dated at 814-657 cal a BP for the Exploradores Glacier , 1210 cal a BP (Aniya and Naruse, 1999) and 721-507 cal a BP  for the Soler Glacier, and before 580 cal a BP for the Nef Glacier . In addition, radiocarbon data indicate possible inundation of trees in the Cachet valley caused by an advancing Colonia Glacier at 810 AE 49 cal a BP (Table 2). Lastly, sedimentological and geochemical analysis of fjord sediment indicates three advance/retreat cycles of the Gualas Glacier at 4180-850 cal a BP .
The best documented regional advance of NPI outlet glaciers occurred at 350-50 cal a BP (Aniya, 2013;Strelin et al., 2014), a period commonly referred to as the Little Ice Age. Although advances early in this period have been reported for several SPI outlet glaciers (Strelin et al., 2014), only our 245 AE 13 cal a BP radiocarbon age indicating an advancing Colonia Glacier provides similar temporal evidence. Conversely, contemporaneous advances to late-1800s maximum positions are a common feature of almost all studied NPI outlet glaciers: seven shown in Fig. 6 and five others listed in Masiokas et al. (2009). After the late-1800s, retreat of and volume loss from all NPI outlet glaciers has been comparable (Rivera et al., 2007;Masiokas et al., 2009;Davies and Glasser, 2012). mentoring of Stephen Porter and completed with the very able field