Application of a Physically-Based Distributed Snowmelt Model in Support of Reservoir Operations and Water Management - Phase 2

Project ID: 2157
Principal Investigator: Eric Rothwell
Research Topic: Managing Hydrologic Events
Funded Fiscal Years: 2016
Keywords: None

Research Question

This project will focus on Reclamation's need for hydrologic modeling tools that can optimize reservoir
management decisions to improve planning and management of water supplies.
We will apply the latest advances in snow accumulation and melt modeling to the Boise River Basin, which
contributes inflows to three reservoirs. In phase 1 a physically-based distributed snow model was applied in an
operational forecasting setting. Phase 2 is needed to investigate and further develop the physically based
modeling techniques for coupling snowpack models with hydrologic and weather forecasting models. Secondly,
what modifications are required to insure that the modeling tools are 1) appropriate to operational needs and 2)
capable of delivering products that will improve water supply forecast and reservoir management outcomes?
Phase 2 will continue to provide weekly snow state and reservoir inflow data to PN Region and the National
Weather Service Northwest River Forecast Center (NWRFC), and will further test and refine the applicability of these
modeling tools in an operational forecast setting.

Need and Benefit

Current operational snowmelt-driven streamflow forecasts are derived from statistical relationships largely based
on a combination of historic trends and calibrations to point observations of SWE or as-available satellite
observations of snow-covered-area. These models rarely contain a physical basis. It has been shown that such
models become unreliable when non-normal conditions (e.g. the 2015 water year) are encountered.
Physically-based, distributed models require little to no calibration and are based on current and predicted
conditions. The physical basis means that all mass and energy fluxes that affect the snowcover are numerically
calculated based on the governing physics. These models are robust to non-normal climate conditions and ideal
tools for evaluating streamflow responses to short-term extreme events such as rain-on-snow, the extended
effects of unseasonable wet, dry, warm, or cold periods, and the long-term effects of climate warming.
Up until now the reasoning has been that the computational demands of modeling over large basins required
simpler, parameterized models. A lack of driving data (i.e. mountain weather observations) for physically-based
modeling has also been seen as an impediment to more complex solutions. Today however, computational
capabilities have multiplied, efficient techniques for distributing limited observations have been developed, and
gridded weather forecasts are readily available.
The proposed work will provide immediate benefits to Reclamation. Current maps of basin-wide SWE will answer
the oft-asked questions, "How much snow is still up there and where is it?", 'Where is the snow located?', 'When
will the snow come off, with the next warm spell, rain storm, etc.?' The up-to-date maps of snowcover cold
content will provide managers data on how sensitive reservoir inflows will be to subsequent energy inputs. Upon
completion, the potential role of physically-based snow modeling in the realm of operational river forecasting will
be known. The physically-based foundations of these modeling tools will be directly applicable to basins
throughout the region – basin-specific calibrations are not necessary. If successful, reservoir managers will have
an advanced, modern tool for predicting inflows, optimizing water usage, and increasing flood protection.
The immediate benefit of this proposal is to the PN Region. However, a pilot study was completed for the
Tuolumne River (1400 km2) watershed in the Sierra Nevada, California for the 2013, 2014 & 2015 water years. For
this test, iSnobal was combined with weekly LiDAR over-flights so that a measured snow depth and distribution
could be used to update the snow model. The simulated snow parameters were used to convert the
LiDAR-measured depths to SWE, providing operational snow distribution and volume estimates to forecasters.
Weekly LiDAR over-flights were then used to update the iSnobal model, so that it could "hone-in" on the true
snow cover distribution and volume as the snow season progressed. This work was particularly important during
the very dry 2015 water year in the southern Sierra.
If techniques such as these can be effectively scaled for a large river basin like the Boise, they could be effective in
any snow dominated mountain basin in the Western US. USBR and ARS have reached out to the NWRFC to receive
feedback on the Boise modeling effort, and are pursuing a similar relationship with the California-Nevada River
Forecast Center (CNRFC). With the ARS application of iSnobal in the Tuolumne River Basin, above Hetch Hetchy
Reservoir, initial conversations with the CNRFC have been promising for potential collaboration, starting with a
webinar overview of the Tuolumne modeling effort. The unique coupling of the snow model with weekly
LiDAR-derived snow depths in the Tuolumne project make it particularly exciting.

Contributing Partners

Contact the Principal Investigator for information about partners.

Research Products

Bureau of Reclamation Review

The following documents were reviewed by experts in fields relating to this project's study and findings. The results were determined to be achieved using valid means.

Application of a Physically-Based Distributed Snowmelt Model in Support of Reservoir Operations and Water Management - Phase 2 (final, PDF, 4.6MB)
By Eric Rothwell
Publication completed on September 30, 2016

Current operational snowmelt models are derived from statistical relationships between historic point measurements of SWE, and rarely contain detailed physics-based model representations. The statistical models have been shown to be unreliable in non-normal conditions that have not been observed in the past or due to changing climatic conditions. In contrast, physically based, distributed models represent the actual physical processes and are as accurate as the forcing information, making them more robust to non-normal conditions. They have the potential to improve reservoir management decisions by providing distributed snowpack properties that are resilient to climate change. This project extended a previous project (S&T 2264) and focused on applying the physically based, distributed snow model iSnobal in an operational setting for water year 2016. Forcing data for iSnobal was derived from a short term weather forecast to provide a 3-day snowpack forecast in real time. The snowpack results, such as spatially distributed snow water equivalent (SWE), susceptibility to melt, the volume of liquid water delivered to the soil (snow melt or rain), and the 3-day forecast were provided on a weekly basis to local area water managers. The results show great agreement between model results and SNOTEL measurement locations, building confidence that the results can be used and trusted in an operational setting.

Application of a Physically-Based Distributed Snowmelt Model in Support of Reservoir Operations and Water Management - Phase 2 (final, PDF, 4.5MB)
By Scott Havens, Daniel Marks, Eric Rothwell, Jennifer M. Johnson
Research Product completed on September 30, 2017

This research product summarizes the research results and potential application to Reclamation's mission.

Application of a Physically-Based Distributed Snowmelt Model in Support of Reservoir Operations and Water Management - Phase 2 (final, PDF, 4.5MB)
By Scott Havens, Daniel Marks, Eric Rothwell, Jennifer M. Johnson
Research Product completed on September 30, 2017

This research product summarizes the research results and potential application to Reclamation's mission.


Return to Research Projects

Last Updated: 4/4/17