Estimating Snowfall From Lake Water Pressure: Results from Three Lakes in Humid Continental Climate

Snowfall is a fundamental winter weather observation that is used for benchmarking weather models, forcing snowpack models, and tracking climate change. However, accurate measurement of snowfall is challenging; it is vulnerable to several sources of measurement error, including precipitation gauge undercatch, wind redistribution and other local effects. Snowfall is typically measured at the point scale and thus is subject to spatial representativeness uncertainty. Here, we use a recently published method (Pritchard et al., 2021, J. Hydrometeorology) to infer the mass of snowfall falling onto a given lake from the increase in water pressure measured in the lake. The rationale behind this method is simple: if snow falls on to the lake surface, the water pressure in the lake will increase, either by additional mass on top of the ice (if the lake surface is frozen) or by adding mass directly to lake (if it is not frozen). This method has the potential to increase the footprint over which snow is measured by multiple orders of magnitude, because the entire lake surface acts as a giant snow pillow, reducing measurement errors that can arise in conventional, point-scale snowfall measurements. This large measurement footprint has the added benefit of reducing the spatial scale discrepancy between ground-based snowfall observations and the resolution of weather and climate models. Water pressure sensors are relatively inexpensive and robust, and lakes are ubiquitous across Earth’s cold regions, including in locations that are not easily accessible and experience harsh weather conditions (e.g. alpine and Arctic regions), highlighting the transformative potential of this method for snowfall monitoring. We evaluate the viability of the lake pressure-snowfall method three lakes in the Northeastern USA and southern Canada: Arbutus Lake (Newcomb, New York, USA), Mirror Lake (Woodstock, NH, USA), and Sunfish Lake (Waterloo, ON, Canada). Lake-estimated snowfall is compared to nearby snow and precipitation gauge observations at these sites in order to evaluate whether the method can be applied in regions with temperate winter climates. Agreement between the lake-estimated snowfall and a nearby snowfall observations vary by site, but can reach root mean squared error (RMSE) values of approximately 2 mm snowfall water equivalent and R² values above 0.95 on a snowfall event basis, supporting the viability and accuracy of lake pressure as a method to infer snowfall.