Location: Water Management and Systems ResearchTitle: Statistical analysis of soil-moisture patterns for probabilistic downscaling
|DESHON, JORDAN - Colorado State University|
|NIEMANN, JEFFREY - Colorado State University|
|JONES, ANDREW - Colorado State University|
Submitted to: Annual Hydrology Days Conference Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 2/20/2018
Publication Date: 3/19/2018
Citation: Deshon, J.P., Niemann, J.D., Green, T.R., Jones, A.S. 2018. Statistical analysis of soil-moisture patterns for probabilistic downscaling. Annual Hydrology Days Conference Proceedings. HTTP://HYDROLOGYDAYS.COLOSTATE.EDU..
Interpretive Summary: No summary needed for abstract
Technical Abstract: Soil moisture is a key variable for a variety of applications ranging from vector-borne disease-outbreak prediction to off-road vehicle trafficability. These applications require not only accurate, fine-resolution soil-moisture estimates across regions but also soil-moisture patterns that exhibit realistic statistical properties (e.g., variance and spatial correlation structure). Many existing downscaling models provide deterministic soil-moisture estimates using soil moisture’s dependence on topographic, vegetation, and soil characteristics. However, observed soil-moisture patterns also contain stochastic variations around such deterministic estimates. The primary objective of this research is to analyze these stochastic variations in soil moisture and include them when downscaling to produce more realistic spatial patterns and statistical properties. Extensive soil-moisture observations from two catchments (Tarrawarra in Victoria, Australia and Cache la Poudre in Colorado, USA) are used for the analysis and model development. The Equilibrium Moisture from Topography, Vegetation, and Soil (EMT+VS) model is used to obtain deterministic soil-moisture estimates from the catchment attributes, and the resulting residuals are considered to be the stochastic variations. Using semivariogram analysis, the stochastic variations are found to contain substantial spatial variance and correlation (i.e., they are not simply white noise). These patterns also include both temporally stable and unstable components. Moreover, the spatial variance of the stochastic variations increases with the spatial-average soil moisture. The EMT+VS model can reproduce these features if it is generalized to include stochastic deviations from the equilibrium state, stochastic variations in porosity and precipitation, and stochastic variations to account for measurement error. The generalized model produces realistic spatial patterns and extreme values of soil moisture, which are beneficial for the aforementioned applications.