Theme for 2020:
Connecting Rain-on-Snow Events, Atmospheric Rivers, and Floods
|On this page:|
Special Recognition Award
The 2020 Special Recognition Award was presented to Michael D. Dettinger, Ph.D.. See the award language, biographical information, and hear the presentation on the Symposium's Mike Dettinger award page.
Moderator for the day was Adele Igel, Ph.D., Assistant Professor Atmospheric Science, UC Davis.
Rain-on-Snow: The Basic Science
Rain on snow in the Sierra Nevada is not well understood, but can pose significant flood risks to downstream areas. This presentation will explore basic snowpack formation, structure, and response to melt and rain on its surface.
Water transmission through snow is complex because the snowpack structure depends on the characteristics of the storms that created it, and no two years ever produce a snowpack with the same characteristics. The Sierran snowpack above 2100 m ranges from 2-5 m in depth at peak accumulation, and by spring, it is composed of 8-12 major layers or horizons that correspond to the prior major storm events. Melt and/or rain water transmission through snowpack is complex due to the layers having varying temperature, densities, and the presence of crustal boundary layers that form during sunny, inter-storm periods. Rain on snow is transmitted both vertically and laterally along layers and interfaces, and steeper slope increases the horizontal speed and amount that may be delivered directly to river channels. Vertical flow channels are often created that allow relatively rapid delivery of water to the soil surface. Rain can melt snow in the mid-elevation "transition zone" that is alternately snow covered and then snow free, but the contributions of that melted snowpack are typically small compared to the amount of rain that fell, some of which create immediate runoff.
Rain on snow frequency and magnitude are increasing in the central Sierra Nevada, and the rain is reaching higher elevations more often. Because runoff magnitude and timing are directly linked to the extent of the basin that receives rain as well as snowpack condition, this issue demands increased awareness and study of the snowpack as well as improved forecasting of the rain-snow elevation of an incoming storm.
Rain-on-Snow, Atmospheric Rivers, and California Floods — What Do We Need to Know?
Rain-on-snow (ROS) events and resulting major floods are a concern during most California winters. ROSs bring rainfall and rainfall-runoff generation to much larger than normal areas in Sierra Nevada catchments, can directly melt snowpack by heat carried in the rain, can be accompanied by vapor condensation (with vast amounts of latent heat release) and deposition of sensible heat onto the snowpacks, or combinations of all four.
ROS occurs with the arrival of warm wet storms over preexisting snowpacks. For example, the peak flow series for the Merced River in Yosemite Valley includes a handful of very large floods that stand head and shoulders above the rest of the 104-yr record, and those largest floods have all occurred with the arrival of atmospheric rivers (ARs). Farther north, on average, AR arrivals at Tahoe City are 2 ℃ warmer and 85% wetter than wet days in general. Thus ARs are key players in the occurrence (and consequences) of Sierra Nevadan ROS.
Under global warming, the fate of ROS depends upon a competition between increasing precipitation falling as rain rather than snow, and less snowpack present to be rained upon. In a 10-model ensemble of climate change projections, by century's end, ROSs in the northern Sierra are projected to all but cease, while 75 to >300% more ROSs are projected for the broad high-altitude areas of the southern Sierra. At the same time, AR intensities are projected to generally increase so that rainfall during ROSs is projected to increase by about 25 to 50% along most of the western Sierra between the Feather and Kings Rivers. With ROSs declining in numbers in some basins, increasing in others, and increasing in rainfall amount in most, the future of ROS floods from the Sierra is an uncertain and heterogeneous problem.
Footprint of Atmospheric Rivers on Land and Implications for Managing Water Resources
Atmospheric rivers (ARs) are one of the most prominent large-scale circulation patterns affecting the U.S. west coast. They contribute significantly to major hydrological extremes (precipitation, snowmelt, floods) in the region. Limited by the coarse spatial/temporal coverage of ground observations, a comprehensive analysis of the hydrological impact of ARs is still lacking. Convection-permitting climate simulations at a grid spacing of several kilometers allow us to capture the topography of the coastal mountains, and accurately construct the hydroclimate in the western U.S.
With such a high-resolution (6-km) dataset over the western U.S., we examine the surface hydrological responses to AR and non-AR induced precipitations. Most of the contrasting responses can be attributed to the unique meteorological conditions (heavy precipitation, modulated temperature and radiation) during ARs, as well as the rain-on-snow effects. We also show that the number of ARs in winter is well correlated with the winter precipitation and April 1st snowpack and hence summer runoff, indicating the possibility of making seasonal water resource prediction based on ARs. These results clarify the role of ARs in surface hydrology and provide new insights on the water resources management at local-to-regional scales.
In the second part, we explore the predictability of extreme precipitation based on ARs in the western U.S., which has important contributions to the predictability of hydrologic extremes. Using machine learning techniques, we classify ARs into three categories, two of which are found to be more closely related to the occurrences and magnitudes of extreme precipitation events. By analyzing the AR characteristics from the AR Tracking Method Intercomparison Project (ARTMIP) data archive, we reveal a significant variation of predictability of AR extreme precipitation based on ARs detected using different methods. Our results provide guidance on the AR tracking methods that best predict extreme precipitation in the Pacific Northwest and California regions.
Operational Forecasting Perspectives of Rain-On-Snow During Flood Events
All hydrologic forecasting entails uncertainty. Rain-on-Snow (ROS) events contribute additional elements of uncertainty to the forecasting process. Forecasting the Rain-Snow Elevation (RSEL), with its inherent uncertainty, becomes more important during ROS events in determining how much of the watershed will experience rain-on-snow. Uncertainty in snowpack conditions at the onset of ROS events presents another challenge. Finally, the modeling of ROS events includes assumptions and parameterization that add to forecast uncertainty.
Looking more closely at the Merced River, a couple of ROS events in 2018 demonstrate these three elements of uncertainty.
Folsom Dam Flood Operations During Rain-on-Snow Events
Recent changes to Folsom Dam flood management focus on the new forecast informed reservoir operations (FIRO). Although new rules codify a five-day inflow forecast into threshold-action relationships for flood management decisions (and this operation is a significant shift from the prior century's mindset), past flood control actions also exercised this concept. The U.S. Bureau of Reclamation's (Reclamation) reservoir operations office will review past and current flood operation practice, rain-on-snow storm event operations, and reservoir management changes for snow runoff timing. Reclamation will share the hallmarks of large snowpack year successes and highlight balancing multi-objective reservoir purposes for water supply and flood management. The new flood management framework offers the potential for a wide range of benefits including features for flood preparation and potential to recover storage refill after storm events. We view decision making through the reservoir operations lens while harnessing forecasting knowledge for a multi-use reservoir during rain, snow, or both conditions.
Past and Future Rain-on-Snow Floods at Don Pedro Dam on Tuolumne River
The most notable rain-on-snow event to occur in the Tuolumne River Watershed occurred on New Years of 1997. With limited weather and runoff information operators had to act in a reactive fashion rather than proactive. Since 1997 the hydrology within the Tuolumne River Watershed has continued to increase in volatility and brought a record-breaking 4-year drought followed by one of the wettest years on record. Water operators have had to adapt quickly and follow the development of new tools to integrate into their daily operations. Most important are improvements to weather and water model accuracy, lead time, and resolution. Having these new tools gives operators the confidence to make proactive decisions before a rain-on-snow or large precipitation event occurs and then better communicate information to individuals internally and externally. This talk takes a deeper look into how new tools could be used to operate Don Pedro Reservoir in a rain-on-snow event to reduce peak flood flows and monitor the changes to the watershed during and after the event.
Observing Rain on Snow — Using Existing Networks and New Technologies to Improve Forecasts
91 years ago, the California State Legislature started the California Cooperative Snow Survey Program. Today there are over 265 snow courses and 130 snow sensors monitoring this precious resource through winter. These networks were designed to support planning for the seasonal drought and forecast the timing of spring runoff. The data collection started monthly, and with the development of snow pillows in the 1960's, the measurements were able to be made daily and even hourly. 40 years later we mainly study and operate off daily data, but seasonal runoff forecasting for many is done with the wider snow course network monthly data.
In California it is not uncommon to receive more than 50% of the annual precipitation from 3-5 large storms. Atmospheric Rivers (AR's) can cause natural disasters but they are also the state's largest opportunity to capture water. Thinking about water supply should also mean thinking about extreme events as just one storm can contribute up to 25% of the total annual rainfall (Lamjiri et al., 2018). Some of the state's historic and most damaging floods occurred from rain on snow, but do we fully understand the contribution of the snowpack to the watershed during these events? With advancements in technology and data acquisition systems, we can begin to use these networks as our early warning system to understand how the snowpack is responding to the rain during the event. This discussion covers the progress in snowpack monitoring and new approaches in observations for rain-on-snow from leveraging existing technologies to exploring new ones and how this can create tools for better decision making.
Detecting Rain-Snow Transition Elevations in Mountain Basins Using Wireless-Sensor Network
We used two independent methods to estimate on-the-ground rain-snow transition elevations, based on continuous data from 17 spatially distributed wireless-sensor clusters in the American and Feather River basins of California's Sierra Nevada. One method used wet-bulb temperature (Twet), and the other was based on snow accumulation and ablation. We analyzed 68 winter events in water years 2014-2018. Each of the 17 clusters had 10-12 measurement nodes strategically placed to capture differences in landscape attributes. Each node measures air temperature, relative humidity, and snow depth. A 0.5 ℃Twet threshold for rain versus snow was selected based on snow-depth-derived transition elevation.
We also found the difference between the on-the-ground transition elevation using Twet was on the order of 100 m lower than the atmospheric snow level from radars; and the root-mean-square difference between the two measurements, over all hourly data, was about 350 m. Compared to the American, transition-elevation estimates in the Feather had higher uncertainties, attributed to its more-complex terrain and fewer sensor clusters (4 in the Feather, 13 in the American). The American tended to exhibit a 100 m higher transition elevation than did the Feather. Yet, the Feather is more vulnerable to rain-on-snow risk since its transition elevation often covers a substantially larger area. For intense-precipitation events related to atmospheric rivers, the transition elevation averaged 150 m higher than for non-atmospheric-river events. Meanwhile, the difference between atmospheric snow level and ground transition elevation during atmospheric-river events was 90 m larger than non-atmospheric-river events.
Overall, the on-the-ground continuous observations from the wireless-sensor networks provide important value-added information on the transition elevation in mountain basins, and can be extended to near-real-time applications.
Modeling Extreme Rain-on-Snow Melt Responses in the Sierra Nevada Snowpack: When Do You Need a More Sophisticated Model?
How critical are physically-based models, with near-real time predictions of winter flooding to protecting built infrastructure and managing water supplies in California's northern Sierra Nevada?
Rain-on-snow (ROS) events are an important element for producing floods in the region and are expected to increase as a result of climate change. Because rainfall during ROS can be captured by freezing within the snowpack, can drain through (or off) the snowpack while melting snow, or some combination of the two, runoff prediction using hydrological models presents a challenge. Depending on model structure and model timing, some models predict liquid water drainage through the snowpack with more skill than others. Whereas physically-based models incorporate mass and energy balance equations to characterize snowmelt and rainfall percolation through the snowpack, conceptual models ignore the complexity of liquid water drainage, if not ignoring water drainage altogether.
To answer our research question, we perform a sensitivity analysis on three models of increasing complexity: a simple degree-day model, a conceptual model that operates similarly to the SNOW-17 model used by the California Nevada River Forecast Center, and the physically-based model, SNOWPACK, originally developed in Switzerland for avalanche forecasting. These models are used to analyze two ROS events in California's northern Sierra Nevada mountains: the Oroville Dam flood (2017) and the Valentines Day ROS (2019). We explore the sensitivity of model structure and time-stepping for these ROS storms, along an elevation gradient on the leeward side of the Sierras. Model forcing data incorporates field observations, the SNOTEL network, as well as remote sensing.
Impact of Atmospheric Rivers on Snow Accumulation in the Sierra Nevada with Implications for Rain-on-Snow Flood Risk
Agricultural water supplies in California are heavily dependent on natural water storage in the Sierra Nevada snowpack which stores approximately 14 million acre feet (~17 km3) of water annually. Snowmelt from this natural storage fills reservoirs in the Sacramento and San Joaquin River Valleys equivalent to 13.5 and 11.5 million acre feet, respectively. Efficient management of snowmelt runoff relies heavily on water supply forecasts, which exhibit significant uncertainties associated with inadequate measurement of snowpack water storage.
To address this issue, we have developed a technique to utilize spatial patterns of reconstructed snow water equivalent (SWE) from previous years toward estimating real-time SWE distribution, blending both satellite measurements and in situ observations. The technique uses a generalized linear regression model which relates observed in situ SWE from 120 snow stations (dependent variable) to 13 physiographic variables such slope, aspect, elevation, distance to ocean, etc (independent variables).
Using these data, we assess the snowmelt contribution to the large flooding event of February 2017 that led to the Oroville spillway incident. The event generated exceptional runoff volumes (second-largest in a 30 year record) partially at odds with the event precipitation totals (ninth-largest). We explain the discrepancy with observed record melt of deep antecedent snowpack, heavy rainfall extending to unusually high elevations, and high water vapor transport during the atmospheric river storms. An analysis of distributed snow water equivalent indicates that snowmelt increased water available for runoff watershed-wide by 37% (25-52% at 90% confidence). The results highlight potential threats to public safety and infrastructure associated with warmer and more variable climate.
The Effect of Vertical Canopy Structure on Snow Processes
The precise modeling of snow and snowmelt processes in forested areas remains a challenge because of the complex interactions and feedbacks between the canopy and snow. Forest canopies alter both the horizontal snow distribution in the landscape and the vertical snow distribution in the canopy, changing the exposure of snow cover to radiation and wind. Yet, most snow studies simulate the canopy as a single layer and do not explicitly resolve vertical profiles of canopy biomass, intercepted snow, and complex energy balance processes.
We evaluated how different vertical canopy structures impact the seasonal water and energy budgets utilizing a vertically resolved plant canopy model, the Advanced Canopy-Atmosphere-Soil-Algorithm (ACASA). We additionally evaluated how sensitive the multilayer model is to variations in snow unloading formulations. Alterations to the vertical canopy structure yielded significant differences in seasonal water and energy budgets with a peak snow depth difference of up to 40% and snow season length up to 2.5 weeks. In contrast, the selected variations to the unloading formulations did not yield significant alterations to the modeled snow depth. The results of this work indicate that accounting for vertical canopy structure may be a critical step for improving estimates of annual water budgets from forested areas.
Drivers of Snowpack Response to Rain-on-Snow Events in Forested Terrain
Interactions between canopy and snowpack during rain-on-snow (ROS) events are poorly understood, particularly in complex, forested terrain. Seven wireless sensor networks with 74 measurement sites in the southern Sierra Nevada of California were used to investigate the response of snowpack to rain events. We identified a total of 496 ROS events that occurred across a range of topographic and canopy conditions over a ten-year period (2008-2017). Leveraging climate, vegetation, topographic, and timing data of the events we trained a Random Forest algorithm to predict change in snow depth during ROS events. The model performed well across all validation events (R2 = 0.87, RMSE = 3.91 cm, bias = 2.97 cm, Kling-Gupta Efficiency = 0.72).
Based on model results, we discuss insights into processes affecting snow depth change during ROS. First, the order of precipitation phase (rain before snow or snow before rain) in mixed rain-and-snow ROS events causes different responses in terms of the sign of snow-depth change. In addition, there is a nonlinear relationship between Leaf Area Index and snow-depth change during ROS, supporting a more-quantitative characterization for forest structure than simply "open" and "closed". This work more fully characterizes the impact of precipitation phase and vegetation patterns on ROS response of the snowpack, providing a foundation for improving predictive modeling of these events in forested regions.
Rain Sponsors — $500+
Shower Sponsors — $250+
|CEPSYM is a Floodplain Management Association project|