Addressing snow model uncertainty for hydrologic prediction

Franz, Kristie J.; Butcher, Phil; Ajami, Newsha K.

doi:10.1016/j.advwatres.2010.05.004

Cited by 49 publications

(34 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Comparisons of the earlier VIC dataset of Maurer et al (2002) with measurements from Surface Radiation (SURFRAD) stations showed that VIC daily downward solar radiation and net radiation were within 2% averaged over all stations. However, Franz et al (2010) found that the NARR downward shortwave radiation was biased high by about 40 W m 22 at the Reynolds Creek Experimental Watershed, Idaho. Zhu and Lettenmaier (2007) compared fields of similarly predicted downward shortwave radiation with NARR over Mexico and found the predicted data to be about 20 W m 22 lower than NARR (and up to 60 W m 22 lower in the summer regionally), although net shortwave (downward minus reflected) and net radiation was higher for the VIC-predicted dataset.…”

Section: Discussionmentioning

confidence: 95%

Representation of Terrestrial Hydrology and Large-Scale Drought of the Continental United States from the North American Regional Reanalysis

Sheffield

Livneh

Wood

2012

Journal of Hydrometeorology

View full text Add to dashboard Cite

The North American Regional Reanalysis (NARR) is a state-of-the-art land-atmosphere reanalysis product that provides improved representation of the terrestrial hydrologic cycle compared to previous global reanalyses, having the potential to provide an enhanced picture of hydrologic extremes such as floods and droughts and their driving mechanisms. This is partly because of the novel assimilation of observed precipitation, state-of-the-art land surface scheme, and higher spatial resolution. NARR is evaluated in terms of the terrestrial water budget and its depiction of drought at monthly to annual time scales against two offline land surface model [Noah v2.7.1 and Variable Infiltration Capacity (VIC)] simulations and observation-based runoff estimates over the continental United States for 1979-2003. An earlier version of the Noah model forms the land component of NARR and so the offline simulation provides an opportunity to diagnose NARR land surface variables independently of atmospheric feedbacks. The VIC model has been calibrated against measured streamflow and so provides a reasonable estimate of large-scale evapotranspiration. Despite similar precipitation, there are large differences in the partitioning of precipitation into evapotranspiration and runoff. Relative to VIC, NARR and Noah annual evapotranspiration is biased high by 28% and 24%, respectively, and the runoff ratios are 50% and 40% lower. This is confirmed by comparison with observationbased runoff estimates from 1130 small, relatively unmanaged basins across the continental United States. The overestimation of evapotranspiration by NARR is largely attributed to the evapotranspiration component of the Noah model, whereas other factors such as atmospheric forcings or biases induced by precipitation assimilation into NARR play only a minor role. A combination of differences in the parameterization of evapotranspiration and in particular low stomatal resistance values in NARR, the seasonality of vegetation characteristics, the near-surface radiation and meteorology, and the representation of soil moisture dynamics, including high infiltration rates and the relative coupling of soil moisture with baseflow in NARR, are responsible for the differences in the water budgets. Large-scale drought as quantified by soil moisture percentiles covaries closely over the continental United States between the three datasets, despite large differences in the seasonal water budgets. However, there are large regional differences, especially in the eastern United States where the VIC model shows higher variability in drought dynamics. This is mostly due to increased frequency of completely dry conditions in NARR that result from differences in soil depth, higher evapotranspiration, early snowmelt, and early peak runoff. In the western United States, differences in the precipitation forcing contribute to large discrepancies between NARR and Noah/VIC simulations in the representation of the early 2000s drought.

show abstract

Section: Discussionmentioning

confidence: 95%

Representation of Terrestrial Hydrology and Large-Scale Drought of the Continental United States from the North American Regional Reanalysis

Sheffield

Livneh

Wood

2012

Journal of Hydrometeorology

View full text Add to dashboard Cite

show abstract

“…WMO, 1986;Valéry, 2010;Franz et al, 2010) reported large uncertainties on the simulated discharge. It is thus expected that this variability remains at least as important under changing climate.…”

Section: G Seiller and F Anctil: Climate Change Impacts On The Hydrmentioning

confidence: 99%

Climate change impacts on the hydrologic regime of a Canadian river: comparing uncertainties arising from climate natural variability and lumped hydrological model structures

Seiller¹,

Anctil²

2014

Hydrol. Earth Syst. Sci.

View full text Add to dashboard Cite

Abstract. Diagnosing the impacts of climate change on water resources is a difficult task pertaining to the uncertainties arising from the different modelling steps. Lumped hydrological model structures contribute to this uncertainty as well as the natural climate variability, illustrated by several members from the same Global Circulation Model. In this paper, the hydroclimatic modelling chain consists of twenty-four potential evapotranspiration formulations, twenty lumped conceptual hydrological models, and seven snowmelt modules. These structures are applied on a natural Canadian subcatchment to address related uncertainties and compare them to the natural internal variability of simulated climate system as depicted by five climatic members. Uncertainty in simulated streamflow under current and projected climates is assessed. They rely on interannual hydrographs and hydrological indicators analysis. Results show that natural climate variability is the major source of uncertainty, followed by potential evapotranspiration formulations and hydrological models. The selected snowmelt modules, however, do not contribute much to the uncertainty. The analysis also illustrates that the streamflow simulation over the current climate period is already conditioned by the tools' selection. This uncertainty is propagated to reference simulations and future projections, amplified by climatic members. These findings demonstrate the importance of opting for several climatic members to encompass the important uncertainty related to the climate natural variability, but also of selecting multiple modelling tools to provide a trustworthy diagnosis of the impacts of climate change on water resources.

show abstract

“…Surface albedo and snow conductivity are thus key parameters for modeling the surface energy balance in order to understand the impact of snow cover on global and regional climate dynamics (Armstrong and Brun, 2008). They also have a major impact on the prediction of the snow water equivalent as well as the timing of melt onset (Franz et al, 2010).…”

Section: Introductionmentioning

confidence: 99%

Snow specific surface area simulation using the one-layer snow model in the Canadian LAnd Surface Scheme (CLASS)

et al. 2013

View full text Add to dashboard Cite

Abstract. Snow grain size is a key parameter for modeling microwave snow emission properties and the surface energy balance because of its influence on the snow albedo, thermal conductivity and diffusivity. A model of the specific surface area (SSA) of snow was implemented in the one-layer snow model in the Canadian LAnd Surface Scheme (CLASS) version 3.4. This offline multilayer model (CLASS-SSA) simulates the decrease of SSA based on snow age, snow temperature and the temperature gradient under dry snow conditions, while it considers the liquid water content of the snowpack for wet snow metamorphism. We compare the model with ground-based measurements from several sites (alpine, arctic and subarctic) with different types of snow. The model provides simulated SSA in good agreement with measurements with an overall point-to-point comparison RMSE of 8.0 m 2 kg −1 , and a root mean square error (RMSE) of 5.1 m 2 kg −1 for the snowpack average SSA. The model, however, is limited under wet conditions due to the single-layer nature of the CLASS model, leading to a single liquid water content value for the whole snowpack. The SSA simulations are of great interest for satellite passive microwave brightness temperature assimilations, snow mass balance retrievals and surface energy balance calculations with associated climate feedbacks.

show abstract

Addressing snow model uncertainty for hydrologic prediction

Cited by 49 publications

References 69 publications

Representation of Terrestrial Hydrology and Large-Scale Drought of the Continental United States from the North American Regional Reanalysis

Representation of Terrestrial Hydrology and Large-Scale Drought of the Continental United States from the North American Regional Reanalysis

Climate change impacts on the hydrologic regime of a Canadian river: comparing uncertainties arising from climate natural variability and lumped hydrological model structures

Snow specific surface area simulation using the one-layer snow model in the Canadian LAnd Surface Scheme (CLASS)

Contact Info

Product

Resources

About