Time series modeling of the spatial distribution of soil moisture in a mountainous hillslope headwater

Abstract

The spatial distribution of soil moisture depth profiles on a hillslope at the headwater of a mountainous forest watershed in Korea was obtained from time series measurements performed using a refined soil moisture monitoring system. Digital terrain analysis based on intensive field surveying was applied to configure the spatial distribution of soil moisture along the steep regolith. The upslope contributing area and topographic wetness index were employed to determine the optimal locations of sensors for time domain reflectometry (TDR) monitoring along two transects. In the early autumn of 2007, 37 time series were recorded over three weeks and used as data for modeling, accounting for the known reliability of the data and the preliminary evaluation of measured soil moisture histories in the context of time series modeling assumptions. A systematic modeling procedure involving pretreatment, investigation of stochastic structure, modeling practice with heuristic repetition, and diagnostic checking was applied to selected soil moisture time series. The two transects showed differences in the distribution of the explanatory models. Transect A, covering a region of relatively fine soil texture, showed a systematic model distribution in the context of soil depth, whereas transect B across coarser soil occasionally produced inconsistent depth profile modeling results. The predictability of the models tends to reduce substantially at the outlet of the hillslope region, possibly due to active subsurface runoff generation near the channel.