RELATIONS BETWEEN TOPOGRAPHY AND WATER TABLE DEPTH IN A SHALLOW FOREST SOIL

Abstract

The relations between water-table depth in a shallow forest soil and topographic characteristics derived from a raster digital elevation model (DEM) were studied. Water-table depths at times of observation, and peak levels between times of observation, were measured weekly during an autumn-winter rainy season at 59 wells in a catchment (ca 0.04 km 2 ) drained by an ephemeral stream, located 80 km east of Vancouver, Canada. Digital elevation models with grid spacings of 4, 8 and 16 m were derived from a ground survey and values of a (upslope contributing area per unit contour length), tanβ (surface slope) and x (a measure of surface curvature) were computed from the DEMs. To avoid problems with data censoring when the water-table dropped below the bottom of some wells, the depths were discretized into classes and discriminant analysis was used to relate water-table depth classes to the topographic indices. The relations were derived using data from 30 wells and were tested on data from the other 29 wells to provide an independent test of predictive capability. The index In(a/tanβ) provided more reliable classifications than a multiple discriminant analysis using In(a), tanβ and x as separate predictor variables. Water-table depths at some wells were consistently misclassified (e.g. 0-3% correct classifications), whereas at others they were classified correctly in up to 100% of cases. Consistent misclassification at individual wells could reflect either the effects of errors in the derived topographic indices (through errors in the DEM or the algorithms used to calculate the indices) or phenomena not related to surface topography, such as variability in soil properties or differences between the surface topography and the topography of the base of the soil profile. The predictive accuracy and statistical significance of derived relations decreased as the DEM grid size increased from 4 to 8 to 16 m. These results are valid only for predictions of water-table depths at points. Further research should explore the effect of grid scale on relations between grid-averaged water-table depths and topographic indices and on methods to parameterize subgrid-scale variability.