No abstract

Abstract. Spatial patterns of soil moisture cannot be adequately characterized by direct measurement for most practical applications, so interpolation between observations is required. Interpolation of soil moisture is complicated because multiple hydrologic processes can affect soil moisture and these processes can introduce distinct modes of variation into the soil moisture patterns. In this paper, a new method to interpolate soil moisture data is presented. This method accepts a dataset of soil moisture at widely-spaced locations on multiple dates and produces fine-scale patterns of soil moisture on the same dates. The method first uses Empirical Orthogonal Function (EOF) analysis to decompose the dataset into a set of time-invariant patterns of covariation (EOFs) and a set of associated time series (called expansion coefficients or ECs) that indicate the importance of the patterns on each date. The method then uses a statistical test to retain only the most important EOFs, and these EOFs are interpolated to the desired resolution using a standard estimation or interpolation method. The interpolated EOFs are finally combined with the spatial averages and the ECs to construct the finescale soil moisture patterns. Using the Tarrawarra dataset, the EOF-based interpolation method is shown to outperform analogous direct interpolation methods, and this improved performance is observed when as few as two observation dates are available. The improved performance occurs because EOF analysis decomposes soil moisture roughly according to the controlling processes and the most important EOFs exhibit distinct but more consistent spatial structures than soil moisture itself. Less predictable variation is also separated into higher order EOFs, which are discarded by the method.

Plant mixtures were established that differed in both proportion and density of loblolly pine (Pinustaeda L.), sweetgum (Liquidambarstyraciflua L.), and broomsedge (Andropogonvirginicus L.). Soil moisture availability to the pine seedlings was quantified every 2 weeks by measuring predawn xylem pressure potentials. Temporal variation in pine water potential was accounted for by a water stress integral approach. Cumulative water stress integral values were calculated over four overlapping periods, from May to June, May to July, May to August, and May to September and compared with the mean seedling stem volume index at each period to determine competitive responses at the whole plant scale. Diurnal measures of stomatal conductance were taken each month to compare competitive responses at the leaf scale. In addition, environmental and plant responses that may control stomatal behavior were quantified. The pine water stress integral was strongly influenced by competing vegetation after the onset of a period of drought in early summer. The correlation between the water stress integral and pine growth increased after a significant drying period, accounting for more than half of the variation in stem volume index at the end of the first growing season. Stomatal conductance was also influenced by competition, with competitive effects more evident during times of drought. Conductance was most often related to bulk leaf water potential, which in turn was related to competitive effects on soil moisture availability. Vapor pressure deficit also influenced stomatal conductance, but this was largely unrelated to competitive effects.

The effects of different plant life-forms, including a bunch grass species, Andropogon virginicus L. (broomsedge), and a sprouting deciduous tree species, Liquidambar styraciflua L. (sweetgum), on soil moisture and competitive responses of a transplanted coniferous tree seedling, Pinus taeda L. (loblolly pine), were investigated. Addition of the bunch grass and/or hardwood sprouts either had no effect or increased soil moisture in the surface soil (0-14 cm) depending on the time, while addition of sweetgum and/or broomsedge (greatest density alone) decreased soil moisture in deeper portions of the solum during the summer months. Soil moisture available to pine seedlings at various points in time was assessed by measuring predawn xylem pressure potential. Temporal variation in predawn xylem pressure potential was accounted for through a water stress integral approach. More than half of the variation in pine size after one growing season could be accounted for by the water stress integral.

Abstract. Spatial patterns of soil moisture cannot be adequately characterized by direct measurement for most practical applications, so interpolation between observations is required. Interpolation of soil moisture is complicated because multiple hydrologic processes can affect soil moisture and these processes can introduce distinct modes of variation into the soil moisture patterns. In this paper, a new method to interpolate soil moisture data is presented. This method accepts a dataset of soil moisture at widely-spaced locations on multiple dates and produces fine-scale patterns of soil moisture on the same dates. The method first uses Empirical Orthogonal Function (EOF) analysis to decompose the dataset into a set of time-invariant patterns of covariation (EOFs) and a set of associated time series (called expansion coefficients or ECs) that indicate the importance of the patterns on each date. The method then uses a statistical test to retain only the most important EOFs, and these EOFs are interpolated to the desired resolution using a standard estimation or interpolation method. The interpolated EOFs are finally combined with the spatial averages and the ECs to construct the fine-scale soil moisture patterns. Using the Tarrawarra dataset, the EOF-based interpolation method is shown to outperform analogous direct interpolation methods, and this improved performance is observed when as few as two observation dates are available. The improved performance occurs because EOF analysis decomposes soil moisture roughly according to the controlling processes and the most important EOFs exhibit distinct but more consistent spatial structures than soil moisture itself. Less predictable variation is also separated into higher order EOFs, which are discarded by the method.