The Value of Accurate High-Resolution and Spatially Continuous Snow Information to Streamflow Forecasts

Abstract

Previous studies have shown limited success in improving streamflow forecasting for snow-dominated watersheds using physically based models, primarily due to the lack of reliable snow water equivalent (SWE) information. Here we use a hindcasting approach to evaluate the potential benefit that a high-resolution, spatiotemporally continuous, and accurate SWE reanalysis product would have on the seasonal streamflow forecast in the snow-dominated Sierra Nevada mountains of California if such an SWE product were available in real time. We tested the efficacy of a physically based ensemble streamflow prediction (ESP) framework when initialized with the reanalysis SWE. We reinitialized the SWE over the Sierra Nevada at the time when the Sierra Nevada had domain-wide annual maximum SWE for each year in 1985–2015, and on 1 February of the driest years within the same period. The early season forecasts on 1 February provide valuable lead time for mitigating the impact of drought. In both experiments, initializing the ESP with the reanalysis SWE reduced the seasonal streamflow forecast errors; compared with existing operational statistical forecasts, the peak-annual SWE insertion and the 1 February SWE insertion reduced the overall root-mean-square error of the seasonal streamflow forecasts by 13% and 23%, respectively, over the 13 major rivers draining the Sierra Nevada. The benefits of the reanalysis SWE insertion are more pronounced in areas with greater snow accumulation, while the complex snow and runoff-generation processes in low-elevation areas impede the forecasting skill improvement through SWE reinitialization alone.

Monitoring Spatial and Temporal Differences in Andean Snow Depth Derived From Satellite Tri-Stereo Photogrammetry

“…Further years of data, even at single "well-timed" end-of-winter dates could have benefits for modelling seasonal streamflow (Li et al, 2019;Margulis and Fang, 2019;Shaw et al, 2020b), calibration of suitable parameterisations for snow accumulation (e.g., solid precipitation threshold) or elevation gradients of forcing variables, namely precipitation (Ragettli et al, 2014;Ayala et al, 2016;Vögeli et al, 2016;Burger et al, 2019;Ayala et al, 2020). In combination with the increasing resolution of (merged) snow cover products at short revisit periods (e.g., Gascoin et al, 2019), leveraging more, "welltimed" Pléiades imagery may improve our ability to characterize the snow cover distribution and evolution in mountain environments, in particular for data-scarce regions such as the central Andes.…”

“…Our understanding of large scale snow processes has been greatly aided by decades of global satellite observations of snow cover extent (e.g., Hall et al, 2010) and more recently of snow volume by leveraging highly detailed airborne LiDAR surveys (Painter et al, 2016;Hedrick et al, 2018;Margulis and Fang, 2019). Studies have found that improving representation of snow water equivalent (SWE) from such detailed snow depth observations can increase the predictive capability of seasonal streamflow forecast models (Li et al, 2019;Margulis and Fang, 2019) and help to quantify the inter-annual variability and persistence of the snow depth in space (Hedrick et al, 2018). Nevertheless, the spatial extent and cost of such intensive campaigns, despite their invaluable contribution to process understanding, limits the wider applicability to catchments in many parts of the world where snow water resources are highly important, yet more poorly understood (e.g., Cortés et al, 2016;Fayad et al, 2017;Smith and Bookhagen, 2018;Baba et al, 2018).…”

Data Assimilation Improves Estimates of Climate-Sensitive Seasonal Snow

“…There exists a wide variety of data assimilation techniques spanning degrees of complexity and the way in which modeling and observation errors are treated. They vary from the simple direct insertion of observations into the model (e.g., Rodell and Houser 2004;Li et al, 2019), where observation are treated as perfect (i.e., zero observation errors), to more mathematical Bayesian methods such as ensemble Kalman filter and particle filter approaches which are designed to account for the uncertainties of the model and observations using error statistics and an ensemble of possible model realizations. While modeling and observation errors are assumed to be of Gaussian shape in the ensemble Kalman filters, particle filters relax this assumption.…”

“…A simple direct insertion application is provided by Li et al, (2019). They directly insert a blended satellite-and model-based SWE product (Margulis et al, 2016) for the initialization of a seasonal streamflow forecast model applied over the snow-dominated Sierra Nevada.…”

Verification of Regional Deterministic Precipitation Analysis Products Using Snow Data Assimilation for Application in Meteorological Network Assessment in Sparsely Gauged Nordic Basins

“…In many forecasting centers around the globe where streamflow simulation is performed in basins with a hydrology dominated by snowpack melt during spring freshet, in the absence of a high-density precipitation observation network, assimilation of in situ and remotely sensed measurements of snowpack state variables has become increasingly important for accurate flow estimation (Helmert et al, 2018). Li et al (2019) have shown that in snow dominated basins, where the meteorological uncertainty during the forecast period is significant (which is the case for sparsely gauged networks), reinitializing the model based on observed snow water equivalent (SWE) information can significantly improve streamflow forecasts. Similarly, in the absence of a high-density precipitation observation network, assimilation of snowpack state variables can provide the possibility to handle different sources of uncertainty by merging the value of observed information into the model in order to correct the effects of model errors and improve forecasting capabilities (Turcotte et al, 2010).…”