Yiping Wu

Automating calibration, sensitivity and uncertainty analysis of complex models using the R package Flexible Modeling Environment (FME): SWAT as an example

Parameter optimization and uncertainty issues are a great challenge for the application of large environmental models like the Soil and Water Assessment Tool (SWAT), which is a physically-based hydrological model for simulating water and nutrient cycles at the watershed scale. In this study, we present a comprehensive modeling environment for SWAT, including automated calibration, and sensitivity and uncertainty analysis capabilities through integration with the R package Flexible Modeling Environment (FME). To address challenges (e.g., calling the model in R and transferring variables between Fortran and R) in developing such a two-language coupling framework, 1) we converted the Fortran-based SWAT model to an R function (R-SWAT) using the RFortran platform, and alternatively 2) we compiled SWAT as a Dynamic Link Library (DLL). We then wrapped SWAT (via R-SWAT) with FME to perform complex applications including parameter identifiability, inverse modeling, and sensitivity and uncertainty analysis in the R environment. The final R-SWAT-FME framework has the following key functionalities: automatic initialization of R, running Fortran-based SWAT and R commands in parallel, transferring parameters and model output between SWAT and R, and inverse modeling with visualization. To examine this framework and demonstrate how it works, a case study simulating streamflow in the Cedar River Basin in Iowa in the United Sates was used, and we compared it with the built-in auto-calibration tool of SWAT in parameter optimization. Results indicate that both methods performed well and similarly in searching a set of optimal parameters. Nonetheless, the R-SWAT-FME is more attractive due to its instant visualization, and potential to take advantage of other R packages (e.g., inverse modeling and statistical graphics). The methods presented in the paper are readily adaptable to other model applications that require capability for automated calibration, and sensitivity and uncertainty analysis.

Predicting impacts of increased CO2 and climate change on the water cycle and water quality in the semiarid James River Basin of the Midwestern USA

Emissions of greenhouse gases and aerosols from human activities continue to alter the climate and likely will have significant impacts on the terrestrial hydrological cycle and water quality, especially in arid and semiarid regions. We applied an improved Soil and Water Assessment Tool (SWAT) to evaluate impacts of increased atmospheric CO(2) concentration and potential climate change on the water cycle and nitrogen loads in the semiarid James River Basin (JRB) in the Midwestern United States. We assessed responses of water yield, soil water content, groundwater recharge, and nitrate nitrogen (NO(3)-N) load under hypothetical climate-sensitivity scenarios in terms of CO(2), precipitation, and air temperature. We extended our predictions of the dynamics of these hydrological variables into the mid-21st century with downscaled climate projections integrated across output from six General Circulation Models. Our simulation results compared against the baseline period 1980 to 2009 suggest the JRB hydrological system is highly responsive to rising levels of CO(2) concentration and potential climate change. Under our scenarios, substantial decrease in precipitation and increase in air temperature by the mid-21st century could result in significant reduction in water yield, soil water content, and groundwater recharge. Our model also estimated decreased NO(3)-N load to streams, which could be beneficial, but a concomitant increase in NO(3)-N concentration due to a decrease in streamflow likely would degrade stream water and threaten aquatic ecosystems. These results highlight possible risks of drought, water supply shortage, and water quality degradation in this basin.

Identifying potential areas for biofuel production and evaluating the environmental effects: a case study of the James River Basin in the Midwestern United States

Biofuels are now an important resource in the United States because of the Energy Independence and Security Act of 2007. Both increased corn growth for ethanol production and perennial dedicated energy crop growth for cellulosic feedstocks are potential sources to meet the rising demand for biofuels. However, these measures may cause adverse environmental consequences that are not yet fully understood. This study 1) evaluates the long‐term impacts of increased frequency of corn in the crop rotation system on water quantity and quality as well as soil fertility in the James River Basin and 2) identifies potential grasslands for cultivating bioenergy crops (e.g. switchgrass), estimating the water quality impacts. We selected the soil and water assessment tool, a physically based multidisciplinary model, as the modeling approach to simulate a series of biofuel production scenarios involving crop rotation and land cover changes. The model simulations with different crop rotation scenarios indicate that decreases in water yield and soil nitrate nitrogen (NO3‐N) concentration along with an increase in NO3‐N load to stream water could justify serious concerns regarding increased corn rotations in this basin. Simulations with land cover change scenarios helped us spatially classify the grasslands in terms of biomass productivity and nitrogen loads, and we further derived the relationship of biomass production targets and the resulting nitrogen loads against switchgrass planting acreages. The suggested economically efficient (planting acreage) and environmentally friendly (water quality) planting locations and acreages can be a valuable guide for cultivating switchgrass in this basin. This information, along with the projected environmental costs (i.e. reduced water yield and increased nitrogen load), can contribute to decision support tools for land managers to seek the sustainability of biofuel development in this region.

Modeling of soil erosion and sediment transport in the East River Basin in southern China

Soil erosion is a major global environmental problem that has caused many issues involving land degradation, sedimentation of waterways, ecological degradation, and nonpoint source pollution. Therefore, it is significant to understand the processes of soil erosion and sediment transport along rivers, and this can help identify the erosion prone areas and find potential measures to alleviate the environmental effects. In this study, we investigated soil erosion and identified the most seriously eroded areas in the East River Basin in southern China using a physically-based model, Soil and Water Assessment Tool (SWAT). We also introduced a classical sediment transport method (Zhang) into SWAT and compared it with the built-in Bagnold method in simulating sediment transport process along the river. The derived spatial soil erosion map and land use based erosion levels can explicitly illustrate the identification and prioritization of the critical soil erosion areas in this basin. Our results also indicate that erosion is quite sensitive to soil properties and slope. Comparison of Bagnold and Zhang methods shows that the latter can give an overall better performance especially in tracking the peak and low sediment concentrations along the river. We also found that the East River is mainly characterized by sediment deposition in most of the segments and at most times of a year. Overall, the results presented in this paper can provide decision support for watershed managers about where the best management practices (conservation measures) can be implemented effectively and at low cost. The methods we used in this study can also be of interest in sediment modeling for other basins worldwide.

Parallelization of a hydrological model using the message passing interface

With the increasing knowledge about the natural processes, hydrological models such as the Soil and Water Assessment Tool (SWAT) are becoming larger and more complex with increasing computation time. Additionally, other procedures such as model calibration, which may require thousands of model iterations, can increase running time and thus further reduce rapid modeling and analysis. Using the widely-applied SWAT as an example, this study demonstrates how to parallelize a serial hydrological model in a Windows^(R) environment using a parallel programing technology-Message Passing Interface (MPI). With a case study, we derived the optimal values for the two parameters (the number of processes and the corresponding percentage of work to be distributed to the master process) of the parallel SWAT (P-SWAT) on an ordinary personal computer and a work station. Our study indicates that model execution time can be reduced by 42%-70% (or a speedup of 1.74-3.36) using multiple processes (two to five) with a proper task-distribution scheme (between the master and slave processes). Although the computation time cost becomes lower with an increasing number of processes (from two to five), this enhancement becomes less due to the accompanied increase in demand for message passing procedures between the master and all slave processes. Our case study demonstrates that the P-SWAT with a five-process run may reach the maximum speedup, and the performance can be quite stable (fairly independent of a project size). Overall, the P-SWAT can help reduce the computation time substantially for an individual model run, manual and automatic calibration procedures, and optimization of best management practices. In particular, the parallelization method we used and the scheme for deriving the optimal parameters in this study can be valuable and easily applied to other hydrological or environmental models.

Projecting the land cover change and its environmental impacts in the Cedar River Basin in the Midwestern United States

The physical surface of the Earth is in constant change due to climate forcing and human activities. In the Midwestern United States, urban area, farmland, and dedicated energy crop (e.g., switchgrass) cultivation are predicted to expand in the coming decades, which will lead to changes in hydrological processes. This study is designed to (1) project the land use and land cover (LULC) by mid-century using the FORecasting SCEnarios of future land-use (FORE-SCE) model under the A1B greenhouse gas emission scenario (future condition) and (2) assess its potential impacts on the water cycle and water quality against the 2001 baseline condition in the Cedar River Basin using the physically based soil and water assessment tool (SWAT). We compared the baseline LULC (National Land Cover data 2001) and 2050 projection, indicating substantial expansions of urban area and pastureland (including the cultivation of bioenergy crops) and a decrease in rangeland. We then used the above two LULC maps as the input data to drive the SWAT model, keeping other input data (e.g., climate) unchanged to isolate the LULC change impacts. The modeling results indicate that quick-response surface runoff would increase significantly (about 10.5%) due to the projected urban expansion (i.e., increase in impervious areas), and the baseflow would decrease substantially (about 7.3%) because of the reduced infiltration. Although the net effect may cause an increase in water yield, the increased variability may impede its use for public supply. Additionally, the cultivation of bioenergy crops such as switchgrass in the newly added pasture lands may further reduce the soil water content and lead to an increase in nitrogen loading (about 2.5% increase) due to intensified fertilizer application. These study results will be informative to decision makers for sustainable water resource management when facing LULC change and an increasing demand for biofuel production in this area.

An Operation-Based Scheme for a Multiyear and Multipurpose Reservoir to Enhance Macroscale Hydrologic Models

AbstractThis paper develops an operation-based numerical scheme for simulating storage in and outflow from a multiyear and multipurpose reservoir at a daily time step in order to enhance the simulation capacity of macroscale land surface hydrologic models. In the new scheme, besides the purpose of flood control, three other operational purposes—hydropower generation, downstream water supply, and water impoundment—are considered, and accordingly three related decision-based parameters are introduced. The new scheme is then integrated into the Soil and Water Assessment Tool (SWAT), which is a macroscale hydrologic model. The observed water storage and outflow from a multiyear and multipurpose reservoir, the Xinfengjiang Reservoir in southern China, are used to examine the new scheme. Compared with two other reservoir operation schemes—namely, a modified existing reservoir operation scheme in SWAT (i.e., the target release scheme) and a multilinear regression scheme—the new scheme can give a consistently bet...

Improvement of the R-SWAT-FME framework to support multiple variables and multi-objective functions

Application of numerical models is a common practice in the environmental field for investigation and prediction of natural and anthropogenic processes. However, process knowledge, parameter identifiability, sensitivity, and uncertainty analyses are still a challenge for large and complex mathematical models such as the hydrological/water quality model, Soil and Water Assessment Tool (SWAT). In this study, the previously developed R program language-SWAT-Flexible Modeling Environment (R-SWAT-FME) was improved to support multiple model variables and objectives at multiple time steps (i.e., daily, monthly, and annually). This expansion is significant because there is usually more than one variable (e.g., water, nutrients, and pesticides) of interest for environmental models like SWAT. To further facilitate its easy use, we also simplified its application requirements without compromising its merits, such as the user-friendly interface. To evaluate the performance of the improved framework, we used a case study focusing on both streamflow and nitrate nitrogen in the Upper Iowa River Basin (above Marengo) in the United States. Results indicated that the R-SWAT-FME performs well and is comparable to the built-in auto-calibration tool in multi-objective model calibration. Overall, the enhanced R-SWAT-FME can be useful for the SWAT community, and the methods we used can also be valuable for wrapping potential R packages with other environmental models.

Development of a generic auto-calibration package for regional ecological modeling and application in the Central Plains of the United States

article i nfo Process-orientedecologicalmodels are frequentlyusedfor predicting potentialimpactsof global changes suchas climate and land-cover changes, which can be useful for policy making. It is critical but challenging to automat- ically derive optimal parametervaluesat different scales,especially at regional scale, and validate the model per- formance. In this study, we developed an automatic calibration (auto-calibration) function for a well-established biogeochemical model—the General Ensemble Biogeochemical Modeling System (GEMS)-Erosion Deposition Carbon Model (EDCM)—using data assimilation technique: the Shuffled Complex Evolution algorithm and a model-inversion R package—Flexible Modeling Environment (FME). The new functionality can support multi- parameter and multi-objectiveauto-calibration of EDCM atthe both pixel and regional levels. Wealsodeveloped a post-processing procedure for GEMS to provide options to save the pixel-based or aggregated county-land cover specific parameter values for subsequent simulations. In our case study, we successfully applied the up- dated model (EDCM-Auto) for a single crop pixel with a corn-wheat rotation and a large ecological region (Level II)—Central USA Plains. The evaluation results indicate that EDCM-Auto is applicable at multiple scales and is capable to handle land cover changes (e.g., crop rotations). The model also performs well in capturing the spatial pattern of grain yield production for crops and net primary production (NPP) for other ecosystems across the region, which is a good example for implementing calibration and validation of ecological models with readily available survey data (grain yield) and remote sensing data (NPP) at regional and national levels. The developed platform for auto-calibration can be readily expanded to incorporate other model inversion algorithms and potential R packages, and also be applied to other ecological models. Published by Elsevier B.V.

Parameter optimization, sensitivity, and uncertainty analysis of an ecosystem model at a forest flux tower site in the United States

Ecosystem models are useful tools for understanding ecological processes and for sustainable management of resources. In biogeochemical field, numerical models have been widely used for investigating carbon dynamics under global changes from site to regional and global scales. However, it is still challenging to optimize parameters and estimate parameterization uncertainty for complex process-based models such as the Erosion Deposition Carbon Model (EDCM), a modified version of CENTURY, that consider carbon, water, and nutrient cycles of ecosystems. This study was designed to conduct the parameter identifiability, optimization, sensitivity, and uncertainty analysis of EDCM using our developed EDCM-Auto, which incorporated a comprehensive R package—Flexible Modeling Framework (FME) and the Shuffled Complex Evolution (SCE) algorithm. Using a forest flux tower site as a case study, we implemented a comprehensive modeling analysis involving nine parameters and four target variables (carbon and water fluxes) with their corresponding measurements based on the eddy covariance technique. The local sensitivity analysis shows that the plant production-related parameters (e.g., PPDF1 and PRDX) are most sensitive to the model cost function. Both SCE and FME are comparable and performed well in deriving the optimal parameter set with satisfactory simulations of target variables. Global sensitivity and uncertainty analysis indicate that the parameter uncertainty and the resulting output uncertainty can be quantified, and that the magnitude of parameter-uncertainty effects depends on variables and seasons. This study also demonstrates that using the cutting-edge R functions such as FME can be feasible and attractive for conducting comprehensive parameter analysis for ecosystem modeling.