[PDF] Calibration Of Watershed Models Using Moments And Scaling Properties Of Streamflows eBook

Author : Qingyun Duan
Publisher : John Wiley & Sons
Page : 356 pages
File Size : 41,19 MB
Release : 2003-01-10
Category : Science
ISBN : 087590355X

GET BOOK

Published by the American Geophysical Union as part of the Water Science and Application Series, Volume 6. During the past four decades, computer-based mathematical models of watershed hydrology have been widely used for a variety of applications including hydrologic forecasting, hydrologic design, and water resources management. These models are based on general mathematical descriptions of the watershed processes that transform natural forcing (e.g., rainfall over the landscape) into response (e.g., runoff in the rivers). The user of a watershed hydrology model must specify the model parameters before the model is able to properly simulate the watershed behavior.

Author : Xuan Yu
Publisher :
Page : pages
File Size : 30,89 MB
Release : 2014
Category :
ISBN :

GET BOOK

Integrated watershed models describe the land-phase of hydrologic cycles by coupling processes such as canopy interception, infiltration, recharge, evapotranspiration, overland flow, vadose zone flow, groundwater flow, and channel routing. This modeling scheme serves as an important strategy for understanding the moisture redistribution processes across the watershed and river-basin landscape. For example, the Penn State Integrated Hydrologic Model (PIHM) has successfully been applied to explain the impacts of antecedent soil moisture on peak streamflow and timing. However, due to the heavy computational cost of solving integrated models with complex model structure, efficient parameter estimation for PIHM is a major computational and algorithmic challenge. The focus of this dissertation has four main themes: (1) develop an efficient calibration strategy for PIHM; (2) develop a weighted-objective calibration scheme for multi-variable distributed parameters (e.g., streamflow, water table depth, and eddy flux data); (3) test the parameter-estimation process for spatial shallow groundwater calibration of PIHM using national wetland geospatial data (National Wetland Inventory: NWI); (4) extend the capabilities of PIHM for linking vegetation dynamics from an ecosystem model and evaluating the importance of vegetation growth in water balance studies.The temporal and geospatial complexity of the data requirements for integrated and fully coupled catchment models increases the difficulty of applying parameter optimization in real watershed applications. In this research, a new strategy known as partition calibration was proposed to enable the automatic calibration of PIHM. The concept can be thought of as a "divide-and-conquer algorithm," where the parameter space is divided into two or more sub-problems that can be solved sequentially. The first partition of the parameter vector is divided according to the two dominant time-scales of catchment hydrological processes: 1) event-scale hydrologic response parameters; and 2) seasonal-scale response parameters. Once divided, the event-scale group parameters and seasonal-scale group parameters are then calibrated sequentially. The second partition focused on the separation of the total calibration objective onto multiple targets to predict each observed hydrological variable. The "informativeness" of each calibration target was defined in terms of a weighted objective function. Application of the scheme suggested the use of an informativeness-based partitioning of streamflow, groundwater, and ET parameters and demonstrated that partition calibration was superior to both single-objective calibration and un-weighted average multi-objective calibration.Applications of the PIHM were found to be efficient with the partition calibration strategy. The first PIHM application involves characterization of the freshwater wetland response to climate change at seven catchments within the Susquehanna River Basin. In this case, streamflow time series and geospatial mapping of wetlands in the National Wetland Inventory (NWI) were used to calibrate the model to capture the distributed groundwater and streamflow dynamics. After calibration, the model was applied to an IPCC climate change scenario (2046-2065), and the modeling results suggested that upland groundwater levels were more sensitive to climate change than water levels of wetlands in lower parts of the catchment, as expected. In the final part of this research, long-term modeling of PIHM compared the role of fixed seasonal variation in LAI (Leaf Area Index) and a fully dynamic vegetation growth model. The community ecosystem model BIOME-BGC was linked to PIHM to test the hypothesis that default monthly LAI values are sufficient to represent long-term water balances in a catchment. By comparing model results for fixed LAI and dynamic LAI, it was demonstrated that fixed LAI is not sufficient for capturing interannual variability of forest vegetation and water flow dynamics, especially as it relates to the onset and growth of forest.

Author : Huicheng Chien
Publisher :
Page : 299 pages
File Size : 25,6 MB
Release : 2011
Category :
ISBN :

GET BOOK

Complex hydrological models are widely used to predict overall watershed responses by incorporating knowledge acquired on field or plot scales. However, processes or complexity critical at smaller scales may not necessarily be important at larger scales. Consequently, it is unnecessary, and could potentially be problematic, to predict hydrological responses at the watershed scale using driving variables acquired at the field scale. With long-term detailed water quality data and a comprehensive set of forcing, state, and flux variables at the watershed scale, dominant variables were able to be successfully identified for watershed scale streamflow, suspended sediment (SS), particulate phosphorus (PP), and soluble reactive phosphorus (SRP). The identification of dominant variables and their relative importance was conducted through the establishment of time series seasonal autoregressive integrated moving average (SARIMA) models.^I found that catchment scale hydrological responses including streamflow, SS, PP, and SRP had different dominant variables. The results showed that models based on the dominant variables were capable of replicating watershed scale hydrological responses. As such, simple models were sufficient for watershed hydrological response simulations and it appeared that identification of dominant variables was the first step to achieve simple models. The application of predefined model complexity and model structure developed for one watershed may not guarantee successful predictions in another watershed. To address this problem, I tested how model complexity, as expressed through differences in the number and configuration of flux and state equations, affects hydrological processes, and to evaluate the validity of current water quality models' assumption that driving variables must include those implied by the plot or field scale empirical studies.^Four models with different model complexity were used to generate runoff and test the needs of model complexity. By removing the assumption, dominant variables of water quality models were identified based entirely on their statistical significance as determined from the SARIMA analysis. The results suggested that the more complex models did not generate better predictions. Simple models were sufficient to generate total runoff at different time scales for water quality modeling purpose. Without runoff flux variable, water quality models with identified forcing and state variables still presented reasonable predictions of hydrological responses. It is difficult to transfer all model details in terms of model structure and parameters from where a model was developed to another watershed especially ungauged ones.^When essential and important features of the watershed hydrological dynamics could be reliably represented using only a few dominant variables, it may be sufficient to transfer dominant variables among watersheds. Model transferability was compared using models based on flux variables and based on dominant variables. The results showed more credible model transferability for streamflow, SS, PP, and SRP across watersheds when models were based on dominant variables. It suggested that simple models that simulate one flux may be easier to move among watersheds without a lot of calibration.

Author : Amin Haghnegahdar
Publisher :
Page : 102 pages
File Size : 28,37 MB
Release : 2015
Category :
ISBN :

GET BOOK

Large-scale (~103-106 km2) physically-based distributed hydrological models have been used increasingly, due to advances in computational capabilities and data availability, in a variety of water and environmental resources management, such as assessing human impacts on regional water budget. These models inevitably contain a large number of parameters used in simulation of various physical processes. Many of these parameters are not measurable or nearly impossible to measure. These parameters are typically estimated using model calibration, defined as adjusting the parameters so that model simulations can reproduce the observed data as close as possible. Due to the large number of model parameters, it is essential to use a formal automated calibration approach in distributed hydrological modelling. The St. Lawrence River Basin in North America contains the largest body of surface fresh water, the Great Lakes, and is of paramount importance for United States and Canada. The Lakes' water levels have huge impact on the society, ecosystem, and economy of North America. A proper hydrological modelling and basin-wide water budget for the Great Lakes Basin is essential for addressing some of the challenges associated with this valuable water resource, such as a persistent extreme low water levels in the lakes. Environment Canada applied its Modélisation Environnementale-Surface et Hydrologie (MESH) modelling system to the Great Lakes watershed in 2007. MESH is a coupled semi-distributed land surface-hydrological model intended for various water management purposes including improved operational streamflow forecasts. In that application, model parameters were only slightly adjusted during a brief manual calibration process. Therefore, MESH streamflow simulations were not satisfactory and there was a considerable need to improve its performance for proper evaluation of the MESH modelling system. Collaborative studies between the United States and Canada also highlighted the need for inclusion of the prediction uncertainty in modelling results, for more effective management of the Great Lakes system. One of the primary goals of this study is to build an enhanced well-calibrated MESH model over the Great Lakes Basin, particularly in the context of streamflow predictions in ungauged basins. This major contribution is achieved in two steps. First, the MESH performance in predicting streamflows is benchmarked through a rather extensive formal calibration, for the first time, in the Great Lakes Basin. It is observed that a global calibration strategy using multiple sub-basins substantially improved MESH streamflow predictions, confirming the essential role of a formal model calibration. At the next step, benchmark results are enhanced by further refining the calibration approach and adding uncertainty assessment to the MESH streamflow predictions. This enhancement was mainly achieved by modifying the calibration parameters and increasing the number of sub-basins used in calibration. A rigorous multi-criteria comparison between the two experiments confirmed that the MESH model performance is indeed improved using the revised calibration approach. The prediction uncertainty bands for the MESH streamflow predictions were also estimated in the new experiment. The most influential parameters in MESH were also identified to be soil and channel roughness parameters based on a local sensitivity test. One of the main challenges in hydrological distributed modelling is how to represent the existing spatial heterogeneity in nature. This task is normally performed via watershed discretization, defined as the process of subdividing the basin into manageable “hydrologically similar” computational units. The model performance, and how well it can be calibrated using a limited budget, largely depends on how a basin is discretized. Discretization decisions in hydrologic modelling studies are, however, often insufficiently assessed prior to model simulation and parameter. Few studies explicitly present an organized and objective methodology for assessing discretization schemes, particularly with respect to the streamflow predictions in ungauged basins. Another major goal of this research is to quantitatively assess watershed discretization schemes for distributed hydrological models, with various level of spatial data aggregation, in terms of their skill to predict flows in ungauged basins. The methodology was demonstrated using the MESH model as applied to the Nottawasaga river basin in Ontario, Canada. The schemes differed from a simple lumped scheme to more complex ones by adding spatial land cover and then spatial soil information. Results reveal that calibration budget is an important factor in model performance. In other words, when constrained by calibration budget, using a more complex scheme did not necessarily lead to improved performance in validation. The proposed methodology was also implemented using a shorter sub-period for calibration, aiming at substantial computational saving. This strategy is shown to be promising in producing consistent results and has the potential to increase computational efficiency of this comparison framework. The outcome of this very computationally intensive research, i.e., the well-calibrated MESH model for the Great Lakes and all the final parameter sets found, are now available to be used by other research groups trying to study various aspects of the Great Lakes System. In fact, the benchmark results are already used in the Great Lakes Runoff Intercomparison Project (GRIP). The proposed comparison framework can also be applied to any distributed hydrological model to evaluate alternative discretization schemes, and identify one that is preferred for a certain case.

Author : Jasmine C. Vittori
Publisher :
Page : 156 pages
File Size : 38,53 MB
Release : 2011
Category : Thesis
ISBN :

GET BOOK

Measured streamflow data in a given basin are important for determining regional patterns of climate and streamflow trends, but are often unavailable or extend back less than 100 years. Observed streamflows can be regressed against tree-ring data that serve as proxies for streamflow to extend the measured record, however this empirical approach cannot account for factors that do not directly affect tree-ring growth, but which may influence streamflow. To reconstruct past streamflows in a more mechanistic way, seasonal water balance models were reviewed and developed for the upper West Walker River basin that can use proxy precipitation and air temperature data derived from treering records. The final model incorporates simplistic relationships between temperature, precipitation, and other components of the hydrologic cycle, and operates at a seasonal time scale. The model was able to reproduce streamflow with an r2 of 0.90, and a RMSE of 7.50 cm with average seasonal air temperature as input. Simulated streamflow was 0.66% greater than observed streamflow for WY 1940 through WY 2006. This model was subsequently used to simulate the effects of wildfire on streamflow in the upper West Walker River Basin. The earliest historical record of wildfire in this basin dates back to 1961, with the most recent recorded in 2005. Evapotranspiration and runoff coefficients were adjusted to simulate reduced vegetation cover as a result of fire, and were applied to the dry season when fire was recorded and the subsequent wet season to reflect time required for re-vegetation to occur. The resulting r2 value decreased to 0.85, with RMSE increasing to 9.02 cm, and the overall streamflow simulation increased to 1.57% greater than observed streamflow. Based on the results of this modeling exercise, the modeling approach with average seasonal air temperature would be appropriate for utilizing proxy tree-ring data as input. The model performed very well using only air temperature and precipitation as input and incorporated 6 parameters representing hydrologic processes influencing streamflow. However, simulating wildfire with this model did not improve streamflow simulations, indicating that the model was sensitive to modeling such landcover manipulations.