This project is one component of the larger Hydro-Kansas (HK) Project. It has the common goal to develop a robust, testable theory of coupled processes on ordered networks (COPON) in order to predict atmospheric, hydrologic, landscape and ecological responses at multiple space and time scales. This project addresses evapotranspiration (ET) in riparian zones; which make up only a small percentage of the total landscape, but contribute disproportionately to basin-wide ET. Riparian areas are generally narrow areas that do not meet ?fetch? requirements and so are not amenable to traditional methods of flux measurement. This project will apply the most modern technology available, from a scanning water vapor Raman lidar to a subsurface water budget approach that can be implemented for measurement of ET on streams of different orders. The proposed lidar method involves a novel two dimensional inversion method that estimates what the ET had to have been to generate lidar-measured two dimensional water vapor concentrations downwind of the riparian area. The information on riparian ET will provide a theoretical basis for testing COPON theory.
Riparian Evapotranspiration (RET) is an important but poorly understood watershed flux. Its importance ranges from negligible in low, heavily forested areas (northern Wisconsin, Minnesota) to being the only major river-loss mechanism (for example, the Rio Grande). Water consumption by riparian vegetation is a widespread concern in water management in the semi-arid regions of the U.S. and has been the subject of many previous studies. It is infeasible to employ the eddy correlation method, which is considered to be the most direct estimate of ET and a de facto standard, to measure RET in riparian zones, because they are too narrow. Alternative approaches based on water balance analysis of subsurface water storage are cost-effective and have a long history. They have been demonstrated to provide reasonably accurate estimates when the water table is shallow and groundwater consumption dominates RET. However, the accuracy of water balance estimates of reach-scale RET is compromised by the ubiquitous heterogeneity of subsurface environments. This project focused on the challenge of estimating RET in a system where it is dominated by soil moisture consumption, because of a deep water table and a very thick vadose zone. A network of soil moisture profilers, groundwater wells, stream gages and a weather station were installed to obtain continuous measurements of meteorological conditions and subsurface water storage over a period of three years, at a site in central Kansas (Figure 1). The riparian reach is shown in Figure 2, with a schematic description of the sensor network. The weather station was located in the open pasture to the south of the reach. Figure 3 (primary image) shows a detailed view of the Sentek EnviroSCAN soil moisture profiler system used in our research. To improve understanding of the accuracy of water balance estimates, these RET estimates were compared to a high-resolution measurement of RET using a Raman LIDAR (a technology that is very accurate, but too expensive for routine measurements) for the period from 7/7/11-7/10/11). To our knowledge, this is the first comparison between water balance estimates and LIDAR estimates in a riparian system where RET is dominated by extraction from soil moisture storage. Water balance estimates of RET were obtained at the location of each soil moisture profile (and associated water table measurements). Figures 4 and 5 show observations of well water levels and depth-integrated soil moisture storage. In Figure 4, the shallow water table wells show diurnal variations characteristic of water consumption by riparian vegetation (RET occurs primarily during the day). However, the deep wells are not influenced by the diurnal variations. The steep rise and fall in the heads in the deep wells is due to the pumping of a nearby lawn irrigation well. The influence of pumping is not observed in the shallow wells. Occasionally, high streamflow events produce a hydraulic connection between the stream and the shallow water table system as indicated following the high precipitation event in early June 2012. Figure 5 shows total soil moisture variations at the locations of two soil moisture profilers. At profiler F, there is a net inflow (rise in soil moisture storage) during the night time periods in the absence of RET, whereas at profiler location E, there is a net outflow even during the night. Both profilers show the depletion of soil moisture storage during the day, which serves as the basis for estimating RET. Figure 6 shows the comparison between the RET estimates from different profilers, LIDAR estimates and a canopy level eddy covariance measurement. The RET estimates from individual soil moisture profilers show significant variability, even within a relatively small reach (approximately 300' x 100' area). These results highlight the caution that must be exercised in estimating RET from single soil moisture profilers under complex natural vegetation in heterogeneous natural soil environments. RET is often estimated from single profilers, and our observations clarify that these estimates can be very misleading. On the other hand, averaged RET estimates obtained using soil moisture measurements from all profilers agreed very well with the high-resolution accurate LIDAR based estimates of RET. However, we acknowledge that the good agreement may not have been achieved but for one of the soil moisture profiler locations (F) where the RET estimate was much higher than at other locations. This fact highlights the significant uncertainty that may be expected in estimates of RET based on single soil moisture profile water balance analyses. The underlying source of this uncertainty is the significant heterogeneity in subsurface soil properties, and in vegetation density. Our results suggest that arrays of soil moisture profilers should be installed for reliable estimation of RET. However, the effort and cost of installation may limit the number of profilers that can be used for routine applications. Furthermore, the number of sensors and profilers needed to obtain accurate estimates may vary from site to site.
