We use cosmogenic radionuclide (CRN) exposure ages to constrain numerical simulations of deglaciation histories in the Middle Boulder Creek drainage, Colorado Front Range, and the Animas River valley, San Juan Mountains, Colorado. We present 18 new ¹⁰Be exposure ages from glacially polished bedrock sampled in the Middle Boulder Creek valley (MBCV). All of these ages are younger than the 20.6±1.2 ka terminal moraine age based on ²⁶Al and ³⁶Cl measurements by Schildgen (2000) and Benson et al. (2005), and a new ¹⁰Be measurement from this study. Exposure ages decrease with distance upvalley from the moraine, and the youngest ages in the uppermost valley are uniformly ~13 ka. We include 4 ¹⁰Be ages in a cross section across the mid-valley, which show a pattern of post-Last Glacial Maximum (LGM) ages (12-14 ka) within the glacial footprint, and older exposure ages (~40 ka) near the trim lines. A similar age trend is seen in the Animas River valley in southwestern Colorado, which was occupied by a lobe of the LGM ice sheet that capped the San Juan mountains. Deglaciation began here ca. 19.4 ka, based on a 10Be depth profile in a proglacial terrace. A longitudinal transect of exposure ages from glacially polished samples indicates that terminus retreat proceeded at ~15 m/yr until complete deglaciation ca. 12.3 ka (Guido et al., 2007). Neither valley has major recessional deposits within the LGM glacial footprint.

The first-order trend in each valley is a monotonic glacial retreat, but there are other possible retreat scenarios, with different implications for regional climate change after the LGM. For instance, we would like to test whether the same trend in ¹⁰Be concentrations could be generated by episodic retreat punctuated by periods of readvance or stillstand. To investigate these scenarios, we modified the GC2D numerical glacier simulation (see Kessler et al., 2006) to incorporate a CRN accumulation layer. This layer can contain any starting value of CRN concentration. Production over each timestep is scaled to DEM latitude and altitude. Production is taken to be zero in areas covered by more than 10 m of ice. The CRN inventory can also decline due to glacial erosion. We incorporate a selectable erosion rule based on basal sliding velocity, and calculate the reduction in CRN inventory by the depth stripped in each timestep. We then simulate a glacier responding to equilibrium line altitude (ELA) changes imposed stepwise, gradually, or including short periods of lowering during an overall rise. Each scenario generates a pattern of ages in the CRN layer that can be compared with the map pattern of measured ¹⁰Be concentrations.

Because different tributaries of a glacial valley have differing hypsometries, they respond at different rates to the same forcing signal. This helps to constrain the range of ELA rise scenarios following the LGM that explain the CRN concentrations. For instance, a slow ELA rise from 21 ka until ~14 ka, followed by rapid deglaciation, can fit well the cosmogenic ages from the larger, northern tributary of MBCV, but underpredicts ages in the southern tributary by 2000 years. To fit most data from both tributaries, the MBCV glacier’s retreat must have included a stillstand or partial readvance between 16 and 14 ka. This is consistent with other cosmogenic deglaciation records from the U.S. Rocky Mountains, which indicate readvance ca. 16-17 ka (e.g. Licciardi et al., 2004) and rapid retreat beginning at about 14 ka. The Animas River valley’s record does not require such an event, and the data there are well-explained by a linear, monotonic ELA rise from 20 to 12 ka.