Search for: All records

Abstract The Ensemble Streamflow Prediction (ESP) framework combines a probabilistic forecast structure with process‐based models for water supply predictions. However, process‐based models require computationally intensive parameter estimation, increasing uncertainties and limiting usability. Motivated by the strong performance of deep learning models, we seek to assess whether the Long Short‐Term Memory (LSTM) model can provide skillful forecasts and replace process‐based models within the ESP framework. Given challenges inimplicitlycapturing snowpack dynamics within LSTMs for streamflow prediction, we also evaluated the added skill ofexplicitlyincorporating snowpack information to improve hydrologic memory representation. LSTM‐ESPs were evaluated under four different scenarios: one excluding snow and three including snow with varied snowpack representations. The LSTM models were trained using information from 664 GAGES‐II basins during WY1983–2000. During a testing period, WY2001–2010, 80% of basins exhibited Nash‐Sutcliffe Efficiency (NSE) above 0.5 with a median NSE of around 0.70, indicating satisfactory utility in simulating seasonal water supply. LSTM‐ESP forecasts were then tested during WY2011–2020 over 76 western US basins with operational Natural Resources Conservation Services (NRCS) forecasts. A key finding is that in high snow regions, LSTM‐ESP forecasts using simplified ablation assumptions performed worse than those excluding snow, highlighting that snow data do not consistently improve LSTM‐ESP performance. However, LSTM‐ESP forecasts that explicitly incorporated past years' snow accumulation and ablation performed comparably to NRCS forecasts and better than forecasts excluding snow entirely. Overall, integrating deep learning within an ESP framework shows promise and highlights important considerations for including snowpack information in forecasting. 

Abstract. Modeling the multidimensional flow of liquid waterthrough snow has been limited in spatial and temporal scales to date. Here,we present simulations using the inverse TOUGH2 (iTOUGH2) model informed by the modelSNOWPACK, referred to as SnowTOUGH. We use SnowTOUGH to simulate snowmetamorphism, melt/freeze processes, and liquid water movement intwo-dimensional snowpacks at the plot scale (20 m) on a sloping groundsurface during multi-day observation periods at three field sites innorthern Colorado, USA. Model results compare well with sites below the treelineand above the treeline but not at a site near the treeline. Results show theimportance of longitudinal intra-snowpack flow paths (i.e., parallel toground surface in the downslope direction and sometimes referred to aslateral flow), particularly during times when the snow surface (i.e.,snow–atmosphere interface) is not actively melting. At our above-treelinesite, simulations show that longitudinal flow can occur at rates orders ofmagnitude greater than vertically downward percolating water flow at a meanratio of 75:1 as a result of hydraulic barriers that divert flow. Our near-treeline site simulations resulted in slightly less longitudinal flow thanvertically percolating water, and the below-treeline site resulted innegligible longitudinal flow of liquid water. These results show theincreasing influence of longitudinal intra-snowpack flow paths withelevation, similar to field observations. Results of this study suggest thatintra-snowpack longitudinal flow may be an important process forconsideration in hydrologic modeling for higher-elevation headwatercatchments. 

Abstract With an increasing number of continental‐scale hydrologic models, the ability to evaluate performance is key to understanding uncertainty and making improvements to the model(s). We hypothesize that any model, running a single set of physics, cannot be “properly” calibrated for the range of hydroclimatic diversity as seen in the contenintal United States. Here, we evaluate the NOAA National Water Model (NWM) version 2.0 historical streamflow record in over 4,200 natural and controlled basins using the Nash‐Sutcliffe Efficiency metric decomposed into relative performance, and conditional, and unconditional bias. Each of these is evaluated in the contexts of meteorologic, landscape, and anthropogenic characteristics to better understand where the model does poorly, what potentially causes the poor performance, and what similarities systemically poor performing areas share. The primary objective is to pinpoint traits in places with good/bad performance and low/high bias. NWM relative performance is higher when there is high precipitation, snow coverage (depth and fraction), and barren area. Low relative skill is associated with high potential evapotranspiration, aridity, moisture‐and‐energy phase correlation, and forest, shrubland, grassland, and imperviousness area. We see less bias in locations with high precipitation, moisture‐and‐energy phase correlation, barren, and grassland areas and more bias in areas with high aridity, snow coverage/fraction, and urbanization. The insights gained can help identify key hydrological factors underpinning NWM predictive skill; enforce the need for regionalized parameterization and modeling; and help inform heterogenous modeling systems, like the NOAA Next Generation Water Resource Modeling Framework, to enhance ongoing development and evaluation. 

Abstract. A critical component of hydrologic modeling in cold andtemperate regions is partitioning precipitation into snow and rain, yetlittle is known about how uncertainty in precipitation phase propagates intovariability in simulated snow accumulation and melt. Given the wide varietyof methods for distinguishing between snow and rain, it is imperative toevaluate the sensitivity of snowpack model output to precipitation phasedetermination methods, especially considering the potential of snow-to-rainshifts associated with climate warming to fundamentally change the hydrologyof snow-dominated areas. To address these needs we quantified thesensitivity of simulated snow accumulation and melt to rain–snowpartitioning methods at sites in the western United States using theSNOWPACK model without the canopy module activated. The methods in thisstudy included different permutations of air, wet bulb and dew pointtemperature thresholds, air temperature ranges, and binary logisticregression models. Compared to observations of snow depth and snow water equivalent (SWE), thebinary logistic regression models produced the lowest mean biases, whilehigh and low air temperature thresholds tended to overpredict andunderpredict snow accumulation, respectively. Relative differences betweenthe minimum and maximum annual snowfall fractions predicted by the differentmethods sometimes exceeded 100 % at elevations less than 2000 m in theOregon Cascades and California's Sierra Nevada. This led to rangesin annual peak SWE typically greater than 200 mm,exceeding 400 mm in certain years. At the warmer sites, ranges in snowmelttiming predicted by the different methods were generally larger than 2 weeks, while ranges in snow cover duration approached 1 month and greater.Conversely, the three coldest sites in this work were relatively insensitiveto the choice of a precipitation phase method, with average ranges in annualsnowfall fraction, peak SWE, snowmelt timing, and snow cover duration of lessthan 18 %, 62 mm, 10 d, and 15 d, respectively. Average ranges in snowmeltrate were typically less than 4 mm d−1 and exhibited a smallrelationship to seasonal climate. Overall, sites with a greater proportionof precipitation falling at air temperatures between 0 and4 ∘C exhibited the greatest sensitivity to method selection,suggesting that the identification and use of an optimal precipitation phasemethod is most important at the warmer fringes of the seasonal snow zone. 

Abstract. Cold content is a measure of a snowpack's energy deficit and is a linear function of snowpack mass and temperature. Positive energy fluxes into a snowpack must first satisfy the remaining energy deficit before snowmelt runoff begins, making cold content a key component of the snowpack energy budget. Nevertheless, uncertainty surrounds cold content development and its relationship to snowmelt, likely because of a lack of direct observations. This work clarifies the controls exerted by air temperature, precipitation, and negative energy fluxes on cold content development and quantifies the relationship between cold content and snowmelt timing and rate at daily to seasonal timescales. The analysis presented herein leverages a unique long-term snow pit record along with validated output from the SNOWPACK model forced with 23 water years (1991–2013) of quality controlled, infilled hourly meteorological data from an alpine and subalpine site in the Colorado Rocky Mountains. The results indicated that precipitation exerted the primary control on cold content development at our two sites with snowfall responsible for 84.4 and 73.0% of simulated daily gains in the alpine and subalpine, respectively. A negative surface energy balance – primarily driven by sublimation and longwave radiation emission from the snowpack – during days without snowfall provided a secondary pathway for cold content development, and was responsible for the remaining 15.6 and 27.0% of cold content additions. Non-zero cold content values were associated with reduced snowmelt rates and delayed snowmelt onset at daily to sub-seasonal timescales, while peak cold content magnitude had no significant relationship to seasonal snowmelt timing. These results suggest that the information provided by cold content observations and/or simulations is most relevant to snowmelt processes at shorter timescales, and may help water resource managers to better predict melt onset and rate.