Location: Watershed Management ResearchTitle: Modeling time-series wind fields over a semi-arid mountain catchment) Author
|Marks, Daniel - Danny|
Submitted to: European Geosciences Union General Assembly Proceedings
Publication Type: Abstract only
Publication Acceptance Date: 11/27/2007
Publication Date: 12/8/2007
Citation: Winstral, A.H., Marks, D.G., Gurney, R. 2007. Modeling time-series wind fields over a semi-arid mountain catchment. Presented in IAHS Symposium HS1003: Hydrolgy in Mountain Regions: Observations, Processes and Dynamics, The 24th General Assembly of teh IUGG, Perugia, Italy, July 2-13, 2007. Interpretive Summary:
Technical Abstract: The spatial variability of winds is considerable over mountain landscapes producing great spatial variability in mass and energy fluxes. Variable winds are often cited for the strong heterogeneity of snow distribution in non-forested mountain locations. Distributed models capable of capturing the variability of these mass and energy fluxes require time-series of distributed wind data at a comparably fine spatial scale. Atmospheric and surface wind flow models in these regions have been limited by our abilities to represent the inherent complexities of the processes being modeled in a computationally efficient manner. Simplified parameterized models, such as those based on terrain and vegetation, are not as explicit as a model of fluid flow, but are computationally efficient for operational use, including in real time. This study applied a simplified wind model based on digital analysis of upwind terrain to predict wind speeds at three sites with diverse topographic exposures in the 0.27 km2 Upper Sheep Creek headwater catchment in the Reynolds Creek Experimental Watershed (RCEW) in southwestern Idaho, USA. Differences in upwind topographic structure were significantly related to wind speed differences between sites. Statistical measures of goodness-of-fit between model-predicted and measured wind speeds at the three sites yielded a coefficient of determination (R2) of 0.54 and a mean absolute error of 1.21 m s-1 normalized to 25% of the mean measured wind speeds. Forcing the model with data from either of the two moderate wind speed sites to predict wind speeds at the other two sites produced the best results (R2: 0.64 – 0.83; normalized mean absolute errors of 16 – 26%) whereas results were not as strong forcing the model with data from the high wind site (R2: 0.13 – 0.29; 32%). These results were very encouraging given the inherent process complexities and the profound variability present in the system.