Spatial and temporal patterns of soil moisture: A study on soil moisture observation and modeling

Abstract

This dissertation addressed three research questions related to observing and modeling soil moisture spatial and temporal patterns. 1) What is the spatial pattern like for soil moisture at the mesoscale? A mobile device was used to measure soil moisture along a 150-km transect 18 times over 13 months. Spatial structures of soil moisture, sand content, and precipitation were characterized and compared. Soil texture turned out to exert a stronger influence than precipitation on mesoscale soil moisture patterns. 2) How can we effectively upscale in situ soil moisture measurements to a larger scale? A phase space analysis method was applied to upscale soil moisture from the point-scale to field-scale, and its performance was evaluated and compared to traditional scaling methods - linear regression and CDF matching. The phase space method was not able to improve the prediction accuracy compared to the two traditional scaling methods. 3) How can we discover and model hidden structures in soil moisture temporal dynamics? An approach from computational mechanics, called epsilon-machines, was applied to symbolized soil moisture time series and their first and second order derivatives. Based on the reconstructed epsilon-machines, soil moisture time series exhibit a degree of complexity hidden structures, and unpredictability. Statistical complexities tend to increase with the orders of derivatives for soil moisture processes.