A Framework for Testing Distributed Hydrologic Models at the Hillslope Scale

点击翻译按钮获取中文摘要

0537410BurgesSpatially distributed hydrologic models are tools with great potential in applications such as non-point source pollutant transport, erosion and sediment transport, or hydrologic responses to landscape change. The predictive skill in these models hinges on accurate simulation of flows along the many pathways through a watershed. Numerous sophisticated computer models exist for simulating these flow pathways, yet due to lack of data, natural system complexity, and model complexity, links between distributed models and reality remain weak and poorly understood. To test, improve, and/or develop effective and practical distributed modeling tools, we need to forge stronger links between field measurements and process simulations to learn more about the strengths and limitations of distributed modeling. In forging these links, however, we face inherent limitations stemming from lack of information and system complexity. Our research aims to design a distributed model test bed whereby, as much as possible, we capture the system complexity and reduce the uncertainties. We do this by combining one of the most detailed available field measurement sets with a state-of-the-art numerical model to create a measurement-driven hypothetical reality. of hydrologic response. The basis for this test bed is the 10 ha (25 acre) Tarrawarra catchment (Australia), which has a simple but representative structure, with 0.5-1.5 m deep soils overlaying bedrock, gently undulating convergent and divergent topography, and a relatively uniform grass cover. Under a subcontract, Dr. Keith Loague, Stanford University, will generate Tarrawarra hypothetical realities. with the detailed Integrated Hydrology Model (InHM), which is fully capable of representing the physical mechanisms of water movement through the subsurface and subsurface with high resolution in space and time. When carefully calibrated to match Tarrawarra observations, InHM simulations come as close as is currently possible to simulating Tarrawarra water flow pathways accurately. These simulations will provide a useful and comprehensive benchmark against which to test simplified representations of surface and subsurface flow. Calibrated InHM simulations become the .hypothetical realities., which we treat as a comprehensive error-free database for evaluation of the types of simpler distributed models that could be used for hydrologic decision making. Although it would be preferable to evaluate simpler models against a range of actual realities, even the Tarrawarra data set is not sufficiently comprehensive in space and time to conduct rigorous distributed model testing. Our .hypothetical reality. approach has the advantage of minimizing constraints on measurement collection and model evaluation. We will test two distributed hydrologic models against the .hypothetical realities. following a combinatorial scheme in which we vary the simulation time periods, spatial and temporal scales, input information, and observation data. Tests within this framework will be used to inform selection of model structure, field measurement types and locations, model element scale, and model parameter values. At this stage, we have a long path ahead of us in the attempt to identify appropriate techniques for measuring and simulating distributed water pathways. This research represents an initial step along that path; the particular goal is to develop a framework and a tool-base for examining complex, multidimensional, heterogeneous systems. Although the small scale, simple structure, and synthetic nature of the Tarrawarra .hypothetical reality. creates challenges in transferring the particular model test results to other areas, we hope that the testing approach will be a useful first step in addressing this challenge of spatially explicit hydrologic simulation. The hypothetical realities, framework database, and distributed model testing results will all be made available through a website, offering the opportunity for others to test different distributed models or to use the framework to address a science question of interest. The work will provide educational opportunities for doctoral and undergraduate students, and we will develop movie representations of parts of the continuous hydrological data sequences for use in high school and university courses.