W. Locke - Paleoclimate Modeling by Multiple Regression

Loewe (1971, Arctic and Alpine Research, 3, 331-344) observed that, while glaciers occurs as a generally predictable function of winter precipitation and summer temperature, the variability in that correlation suggested the presence of another significant variable (perhaps continentality). Continentality per se is not significant, but it serves as a proxy for cloudiness/humidity, thus for the distribution of energy inputs other than simple conductive heat transfer. I (Locke, 1990, Annals of Glaciology, 14, 164-166) concluded that the addition of continentality to the simple regression of summer temperature against winter precipitation at glacial ELAs improved the explanation by 15% (from 69% to 81%; 12/81=0.15), thus significantly reducing predictive error.

Late-Pleistocene Equilibrium Line Altitudes in Montana

I applied this linear predictive model to the Late Pleistocene glaciers of Montana (USA)(Locke, 1990, Arctic and Alpine Research , 22 , 1-13). Late Pleistocene equilibrium line altitudes, interpreted largely, but not entirely, from N-NE-facing cirque floor elevations, varied from above 2800 m in the SE to below 1700 m in the NW, with significant local complexity. NOTE: Triangles show glacier locations used in the synthesis - dashed lines show the maximum extent of the Laurentide (long) and Cordilleran/Northern Rocky Mountain (short) ice sheets.

The trend in ELAs is clearly dominated by NW-SE moisture transport driven by prevailing westerly winds. This trend is statistically indistinguishable from that of ELAs on the scattered mountain glaciers in the region

Assuming a uniform regional cooling of 10^oC (based on the regional occurrence of ice-wedge polygons where mean modern annual temperatures are well above freezing), we can infer the regional precipitation at glacier ELAs at the last glacial maximum.

Inferred Late Pleistocene precipitation at glacial ELA's

Estimated winter snow accumulation at glacier ELAs varied from less than 40 cm water equivalent in the dry interior mountain masses to more than 100 cm along the NW margin of the study area, where relatively moist maritime Pacific air masses enter the region along the prevailing storm track. Another area of relatively high winter accumulation occurred in the center of the study area, where the lowest pass across the Continental Divide funneled airflow to the SE. The presence of glacial lakes Missoula (west of the divide) and Great Falls (to the east) may have provided some recharge as well.

From the inferred late Pleistocene winter accumulation and modern measurements, the difference between modern and Late Pleistocene snowfall at glacial elevations can be calculated.

Difference between Modern and Late Pleistocene Precipitation

Pleistocene snowfall was apparently less than present across most of the region, except in areas where major proglacial lakes increased local air saturation. There may also have been convergent airflow (prevailing westerlies and catabatic drainage between the ice sheets) increasing local air uplift, thus cooling and precipitation. The conclusion of drier conditions at peak glaciation is supported by independent glacier flow modeling.

I applied the same technique to reconstruct late Pleistocene precipitation across both Montana and adjacent northern Idaho (Locke, Murray and Slaughter, 1989, CANQUA Program and Abstracts, p. 37). Using the same cooling assumption and linear model (r²=0.81) as applied in Montana alone, inferred precipitation in the interior ranges of Idaho was negative! A slightly weaker logarithmic model (r²=0.75) yields positive, but very low precipitation. A stronger model still would use an inferred gradient of late Pleistocene cooling across the region, from a maximum (>12^oC) in the continental interior to a minimum (6^oC?) in maritime regions. However, such a model still awaits testing, as by regional palynology.

Multiple Regression Modeling