Location: Watershed Physical Processes Research
2018 Annual Report
Objectives
1. Provide accurate, efficient and user friendly multi-dimensional numerical models for studying (1) water driven soil erosion and sediment transport, (2) embankment breaching processes and associated flooding problems, and (3) agro-pollutant transport and water quality problems. (NP211 2016-2020 Action Plan: C2: PS 2.1; 2.2; 2.4; 2.5; C3: PS 3.1)
2. Develop a web-based Agricultural Integrated Management System (AIMS) that disseminates seamless geospatial data for modeling purposes and sustainable watershed management, and provides automated simulations of runoff, sediment, and agro-pollutant loadings for any watershed in the U.S. (NP211 2016-2020 Action Plan: C2: PS 2.4; C3: PS 3.1; C4: PS 4.1; 4.2; 4.3)
Approach
Sediment and agro-chemicals from agricultural watersheds and streams are transported into lakes and rivers where they degrade the aquatic eco-system and general water quality. The University of Mississippi, National Center for Computational Hydroscience and Engineering (NCCHE) has developed a number of computational models capable of simulating free-surface flows, soil erosion, embankment, levee and dam breaching, flood flows, sediment/contaminant transport, optimization analysis and decisions support systems for watershed management. These models have been rigorously verified and validated in previous research and are continuously being improved and upgraded. Computational modeling and effective decision support systems are needed in order to study problems and find solutions for soil erosion, gully erosion, sediment transport, embankment breaching and consequential flood inundation. NCCHE staff will work closely with the research scientists of the USDA to utilize these reliable and efficient models to study, understand and resolve the soil and water related problems in agriculture watersheds. At the same time, existing models need to be improved and enhanced by adopting new methods and merging technologies in order to better serve the needs of the agriculture research. The main focus of this project are issues of embankment breaching and flood inundation, detailed watershed runoff, erosion and pollutant transport, local scouring around instream structures, water quality and eco-systems affected by watershed, sediment transport optimal control, software efficiency improvements and decision support for watershed management. This research will help to achieve the goals of Water Availability and Watershed Management.
Progress Report
Soil erosion from farmlands results in soil and nutrient loss, and forms gullies on land surface downgrading the productivity and cultivability. A physically based hydrodynamic model, CCHE2D, is enhanced and applied to simulate the overland flow processes in physical experiments for soil erosion and gully development on Loess and Mollisol. The simulated physical parameters, flow velocity and water depth, were compared with the observed data of 19 experiments. The numerical simulations are in good agreement to the experiments. This study validated this physically based numerical model and confirmed its applicability to overland flows on hill-slopes with soil erosion and gully development.
The simulated overland flow distributions of water depth and velocities in the watersheds of the Loess and Mollisol tests revealed detailed dendritic structures of micro-channel networks formed by soil erosion from the watershed dividing to the developed gullies. The time series of the micro-channel network development revealed in the simulated runoff distributions represent the erosion effects of the sequential flow conditions of the physical experiments. The trends of interrill flows concentrating at the headcuts, and moving toward the gully from side slopes can be mimicked by the numerical model.
The study shows the kinematic wave method works well only in the areas of interrills where the sheet flow prevails. In the rills and gullies, the measured velocities have no relationship to the Manning’s equation. The concentrated flows in gullies with a complex topography can be predicted only by solving general shallow water equations. The consistency of the physically-based model and the 19 sets of experimental data measured in a variety of conditions of flow, rainfall, topography and soil indicates the application potential of the numerical model to overland flow and soil erosion studies.
Sediment transport in rivers, streams and lakes are three-dimensional phenomena. CCHE3D is a numerical model capable of simulating 3D turbulent flows and the associated sediment transport. The bed-load computation of this model was updated from a none-conservation scheme to a conservation scheme. The former has good stability and applicable to many situations, it may result in problems of inaccuracy particularly under complex flow conditions. The improved 3D model with the conservation bed-load transport scheme is tested rigorously using experimental and field cases with strong fluvial dynamics and complex topography.
In natural rivers and streams, the surface and ground water interact actively in a sediment layer called hyporhiec zone. The thickness of the hyporhiec zone depends on the sediment size, sediment layer formation, channel morphology and the flow conditions. The water exchange is important to river eco-system and ground water recharging process. A 3D ground water model based on the CCHE3D general model is developed and verified using multiple analytical solutions for simulating the hyporhiec flows. Excellent agreements were found between the numerical simulation results and the analytical solutions. The numerical model is also used to conduct numerical experiments to study the hyporheic flow distributions in sediment layers and to extract empirical relationships between the hyporheic flow and sediment properties. This model has a good potential to be used to study ground water recharging and stream ecosystem.
Meandering is a fluvial process that dynamically changes rivers’ course over an alluvial plain or in a river valley. In a collaboration research, CCHE2D flow and sediment transport models have been applied to simulate sediment transport in a large scale experimental channel. Loose sediments were used to form the channel bed which is actively changed by sediment transport. The simulated change of the channel bed agreed well with that of the measured. The numerical simulation results were then used to track the sediment particle movement in the longitudinal and transverse channel directions. The meander river bed evolution processes can be interpreted by the simulated information which can hardly be obtained even from physical experiments.
Parallel computing generally greatly improve computation efficiency utilizing more hardware resources at the same time. In this study, the ADI (Alternating Direction Implicit) solver and the parallel cyclic reduction (PCR) algorithm, implemented in CCHE2D for graphics processing unit (GPU) parallel computing, have been refined. When large and complex geometric domains are simulated using a computational fluid dynamics (CFD) method, there are multiple problems involved in the fast ADI solvers parallelized with PCR and run on GPU hardware. We have identified and improved three main problems: (1) the numerical stability due to the semi-implicitness of ADI method; (2) the accuracy of GPU computing using single-precision variables and the fast math library; and, (3) the limitations of on-board shared memory of GPU to PCR for large-scale computation domains.
Both the serial version and parallel version of the ADI solver were evaluated. Several mapping methods (one-block-without-iteration, one-block-with-iterations, multi-block-without-iteration, and one-block-with-partitions) were evaluated and compared in numerical simulations of water flow with complex domain geometries. A new mapping method, one-block-with-partition mapping, was proposed to resolve the size limitations problems of the PCR on GPU.
Velocity correction methods are widely used in numerical models simulating fluid flows. In this method, the velocity correction coefficient (ECC) dominates the model’s stability and efficiency. Existing methods compute ECC locally using information at each element. A more robust and stable method to evaluate ECC is developed that a linear equation is constructed by the coefficient matrix of the momentum equations and it is simultaneously solved for ECC for the whole solution domain. The ECC method was integrated into CCHE2D Hybrid model based on a hybrid mesh system (triangle + quadrilateral). The updated model was validated using several sets of experimental data and applied to a field case as well. According to the numerical tests, the proposed ECC method has demonstrated an enhanced numerical stability and improved efficiency.
In many closed water bodies, the water is well-mixed dynamically: the net flow velocity is negligible, the concentrations of dissolved chemicals and suspended sediments are almost uniformly distributed horizontally. However, the solar intensity, temperature and suspended sediment have vertical distributions affecting the water quality processes. The water quality (WQ) constituents may also have vertical distributions. Taking advantage of these conditions, a vertical distribution (VD) model has been developed to simulate the WQ processes. Because all of the concerned properties are well mixed in the horizontal direction, the model can be executed very fast.
In this model the water level is affected by the inflow, outflow, evaporation, and precipitation. The solar intensity, temperature and suspended sediment are vertically distributed according to empirical formulations. Similar to the current CCHE water quality models, the VD model simulates eight WQ constituents: ammonia nitrogen (NH4), nitrate nitrogen (NO3), inorganic phosphorus (PO4), phytoplankton (PHYTO), carbonaceous biochemical oxygen demand (CBOD), dissolved oxygen (DO), organic nitrogen (ON), and organic phosphorus (OP). CCHE surface water quality model and the USDA watershed model are connected to study the effect of pollutant loads from upland watershed on the water quality of the receiving surface water body. The AnnAGNPS watershed model, developed at the USDA ARS, National Sedimentation Laboratory (NSL), was applied to simulate the loads of water, sediment and nutrients from upland watersheds. The effects of land use/land cover, soil properties, climate, agriculture management, etc. on the watershed loads are considered. The computed results from AnnAGNPS are used as boundary conditions for CCHE water quality model to simulate the water quality concentration in water bodies. In this model, the effects of suspended sediment on the water quality constituents are considered, and the distributions of nutrients, chlorophyll and dissolved oxygen in the water body can be obtained efficiently. The Beasley Lake watershed in the Mississippi Delta, a lake surrounded by agricultural lands, was used as a study site. Sediment concentration is relatively high due to farming activities. The sediment-associated water quality processes need to be taken into account. The lake water quality is monitored by NSL, and some of the measured data is used to calibrate and validate the numerical model. The water quality model is shown to be a useful tool to assess long term impacts of watershed nutrient and sediment loads on the water quality of the receiving water bodies.
Accomplishments
1. Improved CCHE2D (Center for Computational Hydroscience and Engineering) model has been successfully validated and applied to watershed process analyses. Watershed hydrologic processes are modeled using simplified and parametric models for a long time, which simulate general watershed characteristics and processes. However, the predictions of many of these models are not strictly based on physics. A physically based hydrodynamic model for watershed overland flows, CCHE2D, has been developed and validated using multiple sets of experimental data by ARS researchers at Oxford, Mississippi. Data sets of measured velocities and water depth are rarely available for watershed studies. This is the first numerical simulation study utilizing measured overland flow velocity and water depth in 19 sets of experiments. The comparison of the simulation and the data indicated that the physically based numerical model produced good results: all the data measured over watershed slopes, in rills, gullies and tillage furrows, under a wide variety of experimental conditions: rainfall intensity, side in-flow, upstream in-flow, and soil types, are consistent with the model’s prediction. Four commercial licenses were granted to apply the model for individual use in their simulation projects.
2. An automatic mesh generation method has been developed and used in water resource software. Natural water bodies such as rivers and lakes have complex geometric forms which require to use irregular shaped meshes for computational models to solve water resource problems. Generating such meshes is, however, often time consuming. The faculty of the University of Mississippi, in collaboration with ARS researchers at Oxford, Mississippi, developed an automatic mesh generation method for applications with complex geometric domains. A complex shaped domain is decomposed into multiple blocks with a simpler shape to alleviate the generation difficulties; the overall mesh is then generated based on these identified blocks effectively. Studies have showed that the proposed method can successfully generate complex shaped meshes automatically and save time for overall numerical simulations.
3. An efficient mapping method for projecting topographic data to computational meshes has been developed. In CFD (Computational Fluids Dynamics) analysis in water resource related projects, topography elevation of land, riverbed, land-use mapping and other geometric distributions, have to be mapped by ARS researchers at Oxford, Mississippi, to a computational grid or mesh from a database of topography or other property distributions. This mapping is called interpolation. It is a crucial step of mesh generation that enables the computational model to recognize a specific physical problem and be used for numerical simulations. For any practice large research project, an accurate and reasonable interpolation takes a lot of calculation time. In this research, we have developed a fast interpolation method based on the idea of quick sorting optimization. According to large computation example cases, the computing time of the optimized interpolation method is reduced to less than 0.1% of the original method.
4. An update of the Center for Computational Hydroscience and Engineering (CCHE) pollutant transport and water quality model have been completed and made available to users. Water quality is an important environmental concern of human society. Water quality processes can be effectively simulated and evaluated using numerical models. The numerical models for simulating water quality and contaminant transport processes of the NCCHE, CCHE-WQ and CCHE-Chem models, have been thoroughly revised and connected to the user-friendly Graphic User Interface for easy simulation management and control. These updated models are currently being applied to simulate the water quality processes of the Pelahatchie Bay of the Rose Barnett Reservoir in Mississippi. The major objective is to study the process of the excessive sediment and nutrients in the bay induced by human activities. This is a research collaboration between ARS researchers at Oxford, Mississippi and researchers at the University of Mississippi at Oxford, Mississippi.
Review Publications
Zhang, Y., Jia, Y., Bingner, R.L. 2017. 2D automatic body-fitted structured mesh generation using advancing extraction method. Journal of Computational Physics 2. 353:316-335.
Jia, Y., Altinakar, M., Guney, M. 2017. Three-dimensional numerical simulations of local scouring around bridge piers. Journal of Hydraulic Research IAHR. DOI: https://doi.org/10.1080/00221686.2017.1356389.
Zhang, Y., Jia, Y. 2017. 2D Automatic body-fitted structured mesh generation using advancing extraction method. Journal of Computational Physics 2. 353(1)316-335 https://doi.org/10.1016/j.jcp.2017.10.018.
