Exploring the Potential of Long Short-Term Memory Networks for Improving Understanding of Continental- and Regional-Scale Snowpack Dynamics

Abstract

Introduction The Problem of Continental-Scale Estimation of Snow Water EquivalentAccurate monitoring of the large-scale dynamics of snowpack is essential for understanding the details of climate dynamics and climate change (Robinson et al., 1993). Warming under a changing climate is expected to cause snowpack to melt earlier in the year (Xiao, 2021; and to reduce the amount of water stored as snow (Musselman et al., 2021;Nijssen et al., 2001). This is expected to have broad and potentially severe impacts to ecosystem productivity (Boisvenue & Running, 2006), winter flood risk (Musselman et al., 2018), groundwater recharge (Ford et al., 2020), agriculture and food security (Qin et al., 2020;Shindell et al., 2012), wildfire hazard (Westerling, 2016), and frequency and severity of drought (Arevalo et al., 2021). In western North America, snow is the primary source of water and streamflow (Li et al., 2017), while globally it supports the water supply needs for more than 1 billion people (Barnett et al., 2005). Therefore, having accurate estimates of the quantity of water stored in snowpack, called snow water equivalent (SWE), is critical for the forecasting and management of water supply and hydropower (Bales et al., 2006;Mankin et al., 2015).