Eigenvector selection with stepwise regression techniques to construct eigenvector spatial filters

Because eigenvector spatial filtering (ESF) provides a relatively simple and successful method to account for spatial autocorrelation in regression, increasingly it has been adopted in various fields. Although ESF can be easily implemented with a stepwise procedure, such as traditional stepwise regression, its computational efficiency can be further improved. Two major computational components in ESF are extracting eigenvectors and identifying a subset of these eigenvectors. This paper focuses on how a subset of eigenvectors can be efficiently and effectively identified. A simulation experiment summarized in this paper shows that, with a well-prepared candidate eigenvector set, ESF can effectively account for spatial autocorrelation and achieve computational efficiency. This paper further proposes a nonlinear equation for constructing an ideal candidate eigenvector set based on the results of the simulation experiment.     

The role of rainfall temporal and spatial averaging in seasonal simulations of the terrestrial water balance

The partitioning of rainfall into surface runoff and infiltration influences many other aspects of the hydrologic cycle including evapotranspiration, deep drainage and soil moisture. This partitioning is an instantaneous non-linear process that is strongly dependent on rainfall rate, soil moisture and soil hydraulic properties. Though all rainfall datasets involve some degree of spatial or temporal averaging, it is not understood how this averaging affects simulated partitioning and the land surface water balance across a wide range of soil and climate types. We used a one-dimensional physics-based model of the near-surface unsaturated zone to compare the effects of different rainfall discretization (5-min point-scale; hourly point-scale; hourly 0.125° gridded) on the simulated partitioning of rainfall for many locations across the United States. Coarser temporal resolution rainfall data underpredicted seasonal surface runoff for all soil types except those with very high infiltration capacities (i.e., sand, loamy sand). Soils with intermediate infiltration capacities (i.e., loam, sandy loam) were the most affected, with less than half of the expected surface runoff produced in most soil types when the gridded rainfall dataset was used as input. The impact of averaging on the water balance was less extreme but non-negligible, with the hourly point-scale predictions exhibiting median evapotranspiration, drainage and soil moisture values within 10% of those predicted using the higher resolution 5-min rainfall. Water balance impacts were greater using the gridded hourly dataset, with average underpredictions of ET up to 27% in fine-grained soils. The results suggest that “hyperresolution” modelling at continental to global scales may produce inaccurate predictions if there is not parallel effort to produce higher resolution precipitation inputs or sub-grid precipitation parameterizations.     

Predicting evaporation from mountain streams

Evaporation can be an important control on stream temperature, particularly in summer when it acts to limit daily maximum stream temperature. Evaporation from streams is usually modelled with the use of a wind function that includes empirically derived coefficients. A small number of studies derived wind functions for individual streams; the fitted parameters varied substantially among sites. In this study, stream evaporation and above-stream meteorological conditions (at 0.5 and 1.5 m above the water surface) were measured at nine mountain streams in southwestern British Columbia, Canada, covering a range of stream widths, temperatures, and riparian vegetation. Evaporation was measured on 20 site-days in total, at approximately hourly intervals, using nine floating evaporation pans distributed across the channels. The wind function was fit using mixed-effects models to account for among-stream variability in the parameters. The fixed-effects parameters were tested using leave-one-site-out cross-validation. The model based on 0.5 m measurements provided improved model performance compared to that based on 1.5 m values, with RMSE of 0.0162 and 0.0187 mm h⁻¹, respectively, relative to a mean evaporation rate of 0.06 mm h⁻¹. Inclusion of atmospheric stability and canopy openness as predictors improved model performance when using the 1.5 m meteorological measurements, with minimal improvement when based on 0.5 m measurements. Of the wind functions reported in the literature, two performed reasonably while five others exhibited substantial bias.     

Incorporating landscape features to obtain an object‐oriented landscape drainage network representing the connectivity of surface flow pathways over rural catchments

A topological representation of a rural catchment is proposed here in addition to the generally used topographic drainage network. This is an object‐oriented representation based on the identification of the inlets and outlets for surface water flow on each farmer's field (or plot) and their respective contributing areas and relationships. It represents the catchment as a set of independent plot outlet trees reaching the stream, while a given plot outlet tree represents the pattern of surface flow relationships between individual plots. In the present study, we propose to implement functions related to linear and surface elements of the landscape, such as hedges or road networks, or land use, to obtain what we call a landscape drainage network which delineates the effective contributing area to the stream, thus characterizing its topological structure. Landscape elements modify flow pathways and/or favour water infiltration, thus reducing the area contributing to the surface yield and modifying the structure of the plot outlet trees. This method is applied to a 4·4‐km² catchment area comprising 43 955 pixels and 312 plots. While the full set of 164 plot outlet trees, with an average of 7 plots per tree, covers 100% of the total surface area of the catchment, the landscape drainage network comprises no more than 37 plot outlet trees with an average of 2 plots per tree, accounting for 52 and 7% of the catchment surface area, when taking account of linear elements and land use, respectively. This topological representation can be easily adapted to changes in land use and land infrastructure, and provides a simple and functional display for intercomparison of catchments and decision support regarding landscape and water management. Copyright © 2011 John Wiley & Sons, Ltd.     

Landscape structure,groundwater dynamics,and soil water content influence soil respiration across riparian–hillslope transitions in the Tenderfoot Creek Experimental Forest,Montana

Variability in soil respiration at various spatial and temporal scales has been the focus of much research over the last decade aimed to improve our understanding and parameterization of physical and environmental controls on this flux. However, few studies have assessed the control of landscape position and groundwater table dynamics on the spatiotemporal variability of soil respiration. We investigated growing season soil respiration in a ～393 ha subalpine watershed in Montana across eight riparian–hillslope transitions that differed in slope, upslope accumulated area (UAA), aspect, and groundwater table dynamics. We collected daily‐to‐weekly measurements of soil water content (SWC), soil temperature, soil CO₂ concentrations, surface CO₂ efflux, and groundwater table depth, as well as soil C and N concentrations at 32 locations from June to August 2005. Instantaneous soil surface CO₂ efflux was not significantly different within or among riparian and hillslope zones at monthly timescales. However, cumulative integration of CO₂ efflux during the 83‐day growing season showed that efflux in the wetter riparian zones was ～25% greater than in the adjacent drier hillslopes. Furthermore, greater cumulative growing season efflux occurred in areas with high UAA and gentle slopes, where groundwater tables were higher and more persistent. Our findings reveal the influence of landscape position and groundwater table dynamics on riparian versus hillslope soil CO₂ efflux and the importance of time integration for assessment of soil CO₂ dynamics, which is critical for landscape‐scale simulation and modelling of soil CO₂ efflux in complex landscapes. Copyright © 2010 John Wiley & Sons, Ltd.