Spatial Variability in Seepage from Unlined, Open Channels
AuthorShanafield, Margaret Almut
StatisticsView Usage Statistics
Recent research into the interactions between streams or lakes and the aquifers beneath them highlights the usefulness of methods for determining movement of water between surface and groundwater sources. Several methods have successfully been employed to estimate these interactions in a variety of environments. As temperature sensor technology and computer capability has improved in the past decades, an increasing number of studies have used temperature and pressure information collected in surface water, streambed porous media, and groundwater in one-dimensional analytical or one-, two-, or even three-dimensional (1D, 2D, or 3D) numerical models that combine heat transport and fluid flux equations. This work compared the advantages and limitations of one- vs. two-dimensional numerical models and the applicability of the Stallman analytical solution for use in applications of heat as a tracer. Although the use of temperature data for estimating seepage and hydraulic conductivity has been shown to have several advantages over many other methods, it shares a common disadvantage of being a point measurement. Because spatial variability in seepage is often an important factor, a new model using the diffusion analogy to the shallow-water (Saint-Venant) equations was developed as a step toward estimating seepage losses from unlined channels.Comparison of numerical models showed that a 1D model can be useful when a simple, comparatively low-cost estimate of seepage from streambed temperature data is desired and streambed heterogeneity is not a concern. However, the assumption of strictly vertical water movement inherent in 1D models highlights the importance of temperature sensor placement to satisfy this condition. Alternatively, the Stallman analytical solution precisely reproduced velocities for moderate rates of infiltration (approximately ~1.5-4 m/d) despite sensor noise and uncertainty in either sensor spacing or thermal diffusivity. When seepage was close to zero or flow was from the groundwater into the streambed, uncertainty in input parameters and noise due to sensor accuracy had an effect on the accuracy of predicted seepage rates, likely due to a reduced change in amplitude with depth. The diffusion-wave model, coupled to the U.S. Geological survey groundwater model MODFLOW, produced longitudinal seepage velocities for several reaches of two irrigation canals that were comparable to the estimates reported in previous studies. This model may be a useful tool for predicting surface water/groundwater interactions along comparatively long stretches of man-made or natural channels at a finer scale than was previously convenient.