SNOWPACK MODELING TO SUPPORT WATER SUPPLY FORECASTING

Please see attached Statment of Work (SOW) for further details and deliverables. To outline the scope, objectives, deliverables, and timeline for a project aimed at improving the skill of the existing statistical and machine learning water supply forecasting methods in use at the USDA-NRCS National Water and Climate Center (NWCC).

This project investigates the potential use of long-term, historical, physically-based snowpack modeling to enhance existing NWCC water supply forecasting methods by providing additional spatial context of the snowpack state. The proposed project has two components: creation of the historical gridded data set and demonstration of near real-time capabilities.

The first component is long-term historical snow state modeling (including a minimum of SWE, cold content, surface water input), encompassing water year (WY) 1980 through WY 2024 for approximately 15 unique watershed-based modeling domains to be specified through a collaborative effort with NWCC hydrologists.

Modeling domains range in size from 1,000-3,000 square miles, encompassing different hydroclimatic and snowpack regimes in the mountainous Western United States where snowmelt sources a majority of available surface water. High resolution gridded modeled snow water equivalent (SWE) data will then be aggregated and summarized by subbasin and elevation bands within the modeling domain.

The summarized information will be passed to NWCC hydrologists for evaluation in the existing operational water supply forecasting system to determine viability of these input variables for improving water supply forecast skill. The second component is the near real-time running of the same modeling systems and framework to support operational water supply forecasting.

Thus, any products derived from gridded snow data must also be capable of supporting real-time modeling, which is defined here as the capability of delivering model results from the previous day by 8:00 am Pacific Time (24 hours latency) for all basins of interest within the modeling domains.