THE EFFECT OF DATA INDEPENDENCE IN MODEL CALIBRATION AND MODEL TESTING1

ABSTRACT: With the increased use of models in hydrologic design, there is an immediate need for a comprehensive comparison of hydrologic models, especially those intended for use at ungaged locations (i.e., where measured data are either not available or inadequate for model calibration). But some past comparisons of hydrologic models have used the same data base for both calibration and testing of the different models or implied that the results of model calibration are indicative of the accuracy at ungaged locations. This practice was examined using both the regression equation approach to peak discharge estimation and a unit hydrograph model that was intended for use in urban areas. The results suggested that the lack of data independence in the calibration and testing of regression equations may lead to both biased results and misleading statements about prediction accuracy. Additionally, although split-sample testing is recognized as desirable, the split-samples should be selected using a systematic-random sampling scheme, rather than random sampling, because random sampling with small samples may lead to a testing sample that is not representative of the population. A systematic-random sampling technique should lead to more valid conclusions about model reliability. For models like a unit hydrograph model, which are more complex and for which calibration is a more involved process, data independence is not as critical because the data fitting error variation is not as dominant as the error variation due to the calibration process and the inability of the model structure to conform with data variability.     

A SURVEY OF HYDROLOGY COURSE CONTENT IN NORTH AMERICAN UNIVERSITIES1

ABSTRACT: This report presents the results of a survey of hydrology faculties of colleges and universities in the United States and Canada. Information is presented on topics covered in classes, allocation of class periods to individual topics, textbooks, prerequisites, computer use, and accreditation categories for hydrology courses offered by engineering departments. Hydrology courses generally require courses in fluid mechanics, mathematics, statistics, and computer science as prerequisites. Topics that receive the largest allocation of time in both introductory and advanced courses include rainfall-runoff relations, the hydrologic cycle, routing and open channel flow, and statistics. Advanced courses place greater emphasis on watershed models than do the introductory courses. Hydrology courses at both levels allocate the smallest amounts of time to snow hydrology ground-water hydrology, and “other topics.” Very few courses include field or experimental work. In a discipline where computer modeling is a major tool, this lack of field and data-collection experience may lead students to underestimate the uncertainties associated with data used to calibrate models and the modeling results themselves. Survey responses on hydrology courses taught in departments other than civil engineering were too few to permit detailed analysis. Most of these courses spend approximately two-thirds of available class time on the same topics as presented in engineering hydrology courses. The balance of class time is spent on topics that emphasize the specialized interest of the particular discipline, such as soil physics and soil moisture in agricultural engineering.     

ASSESSMENT OF REMOTE SENSING INPUT TO HYDROLOGIC MODELS1

Remotely sensed variables such as land cover type and snow-cover extent can currently be used directly and effectively in a few specific hydrologic models. Regression models can also be developed using physiographic and snow-cover data to permit estimation of discharge characteristics over extended periods such as a season or year. Most models, however, are not of an appropriate design to readily accept as input the various types of remote sensing parameters that can be obtained now or in the future. Because this new technology has the potential for producing hydrologic data that has significant information content on an areal basis, both inexpensively and repetitively, effort should be devoted now to either modifying existing models or developing new models that can use these data. Minor modifications would at least allow the remote sensing data to be used in an ancillary way to update the model state variables, whereas major structural modifications or new models would permit direct input of the data through remote sensing compatible algorithms. Although current remote sensing inputs to hydrologic models employ only visible and near infrared data, model modification or development should accommodate microwave and thermal infrared data that will be more widely available in the future.     

On the Design of Computer-Based Models for Integrated Environmental Science

The current research agenda in environmental science is dominated by calls to integrate science and policy to better understand and manage links between social (human) and natural (nonhuman) processes. Freshwater resource management is one area where such calls can be heard. Designing computer-based models for integrated environmental science poses special challenges to the research community. At present it is not clear whether such tools, or their outputs, receive much practical policy or planning application. It is argued that this is a result of (1) a lack of appreciation within the research modeling community of the characteristics of different decision-making processes including policy, planning, and (2) participation, (3) a lack of appreciation of the characteristics of different decision-making contexts, (4) the technical difficulties in implementing the necessary support tool functionality, and (5) the socio-technical demands of designing tools to be of practical use. This article presents a critical synthesis of ideas from each of these areas and interprets them in terms of design requirements for computer-based models being developed to provide scientific information support for policy and planning. Illustrative examples are given from the field of freshwater resources management. Although computer-based diagramming and modeling tools can facilitate processes of dialogue, they lack adequate simulation capabilities. Component-based models and modeling frameworks provide such functionality and may be suited to supporting problematic or messy decision contexts. However, significant technical (implementation) and socio-technical (use) challenges need to be addressed before such ambition can be realized.     

EFFICIENCY OF HYDROLOGIC DATA COLLECTION SYSTEMS ROLE OF TYPE I AND II ERRORS1

ABSTRACT The efficiency of hydrologic data collection systems is relevant to solution of environmental problems, scientific understanding of hydrologic processes, model-building and management of water resources. Because these goals may be overlapping and non-commensurate, design of data networks is not simple. Identified are four elements of error or risk in such networks: (a) choice of variables and mathematical model for the same process, (b) accuracy of model parameter estimates, (c) acceptance of wrong hypothesis or rejection of correct hypothesis and (d) economic losses associated with error. Of these four, the classical hypothesis testing problem is specifically evaluated in terms of costs of type I and II errors for simple and composite hypotheses; mathematical models for these economic analyses also include costs of sample data and costs of waiting while new data is obtained. An illustrative computational example focuses on the hypothesis that natural recharge might be augmented by a system of pumping wells along an ephemeral channel. The relationship of the hypothesis testing problem to Bayesian decision theory is discussed; it is felt that the latter theory offers a more comprehensive framework for design and use of hydrologic data networks.     

A conceptual framework for assessing cumulative impacts on the hydrology of nontidal wetlands

Wetlands occur in geologic and hydrologic settings that enhance the accumulation or retention of water. Regional slope, local relief, and permeability of the land surface are major controls on the formation of wetlands by surface-water sources. However, these landscape features also have significant control over groundwater flow systems, which commonly play a role in the formation of wetlands. Because the hydrologic system is a continuum, any modification of one component will have an effect on contiguous components. Disturbances commonly affecting the hydrologic system as it relates to wetlands include weather modification, alteration of plant communities, storage of surface water, road construction, drainage of surface water and soil water, alteration of groundwater recharge and discharge areas, and pumping of groundwater. Assessments of the cumulative effects of one or more of these disturbances on the hydrologic system as related to wetlands must take into account uncertainty in the measurements and in the assumptions that are made in hydrologic studies. For example, it may be appropriate to assume that regional groundwater flow systems are recharged in uplands and discharged in lowlands. However, a similar assumption commonly does not apply on a local scale, because of the spatial and temporal dynamics of groundwater recharge. Lack of appreciation of such hydrologic factors can lead to misunderstanding of the hydrologic function of wetlands within various parts of the landscape and mismanagement of wetland ecosystems.     

EFFECTIVE USE OF LANDSAT DATA IN HYDROLOGIC MODELS1

ABSTRACT: In a cooperative demonstration project, NASA and the U.S. Army Corps of Engineers (Corps) compared conventional and Landsat-derived land-use data for use in hydrologic models, and the resulting discharge frequency curves were analyzed. When a grid-based data-management system was used on a cell-by-cell basis (size about 1.1 acres or 0.45 hectare), Landsat classification accuracy was only 64 percent, but, when the grid cells were aggregated into watersheds, the classification accuracy increased to about 95 percent. When both conventional and Landsat land-use data were input to the HEC-1 model for generating discharge frequency curves, the differences in calculated discharge were judged insignificant for subbasins as small as 1.0mi² (2.59 km²). For basins larger than 10mi² (25.9km²), use of the Landsat approach is more cost-effective than use of conventional methods. Digital Landsat data can also be used effectively by local and regional agencies for hydrologic analysis by incorporating the data into grid-based data-management systems. The transfer of this new technology is well under way through inclusion in some Corps training courses and through use by both county government personnel and private consultants.     

SUBWATERSHED SPATIAL ANALYSIS TOOL: DISCRETIZATION OF A DISTRIBUTED HYDROLOGIC MODEL BY STATISTICAL CRITERIA1

ABSTRACT: Complex hydrologic models, designed for simulating larger watersheds, require a huge amount of input data. Most of these models use spatially distributed data as inputs. Spatial data can be aggregated or disaggregated for use as input to a model, which can impact model outputs. Although, it is efficient to minimize data redundancy by aggregating the spatial data, upscaling reduces the detail/resolution of input information and increases model uncertainty. On the other hand, a large number of model inputs with high degrees of disaggregation take more computer time and space to process. Hence, a balance between striving for a maximum level of aggregation and a minimum level of information loss has to be achieved. This study presents a definition of an appropriate level of discretization, derived by establishing a relationship between a model's efficiency and the number of subwater‐sheds modeled. An entropy based statistical approach/tool called Subwatershed Spatial Analysis Tool (SUSAT) was developed to find an objective choice of an appropriate level of discretization. The new approach/tool was applied to three watersheds, each representing different hydrologic conditions, using a hydrologic model. Coefficients of efficiency and entropy estimated at different levels of discretization were used to validate the success of the new approach.     

Development and Comparison of Multiple Regression Models to Predict Bankfull Channel Dimensions for Use in Hydrologic Models

Channel dimensions are important input variables for many hydrologic models. As measurements of channel geometry are not available in most watersheds, they are often predicted using bankfull hydraulic geometry relationships. This study aims at improving existing equations that relate bankfull width, depth, and cross‐sectional area to drainage area (DA) without limiting their use to well‐gauged watersheds. We included seven additional variables in the equations that can be derived from data that are generally required by hydrologic models anyway and conducted several multiple regression analyses to identify the ideal combination of additional variables for nationwide and regional models for each Physiographic Division of the United States (U.S.). Results indicate that including the additional variables in the regression equations generally improves predictions considerably. The selection of relevant variables varies by Physiographic Division, but average annual precipitation (PCP) and temperature (TMP) were generally found to improve the models the most. Therefore, we recommend using regression equations with three independent variables (DA, PCP, and TMP) to predict bankfull channel dimensions for hydrologic models. Furthermore, we recommend using the regional equations for watersheds within regions from which data were used for model development, whereas in all other parts of the U.S. and the rest of the world, the nationwide equations should be given preference.     

KALMAN FILTER ESTIMATION AND PREDICTION OF DAILY STREAM FLOWS: I. REVIEW,ALGORITHM, AND SIMULATION EXPERIMENTS1

An important class of models, frequently used in hydrology for the forecasting of hydrologic variables one or more time periods ahead, or for the generation of synthetic data sequences, is the class of autoregressive(AR) models. As the AR models belong to the family of linear stochastic difference equations, they have both a deterministic and a stochastic component. The stochastic component is often assumed to have a Gaussian distribution. It is well known that hydrologic observations (e.g., stream flows) are heavily affected by noise. To account explicitly for the observation noise, the linear stochastic difference equation is expressed in state variable form and an observation model is introduced. The discrete Kalman filter algorithm can then be used to obtain estimates of the state variable vector. Typically, in hydrologic systems, model parameters, system noise statistics and measurement noise statistics are unknown, and have to be estimated. In this study an adaptive algorithm is discussed which estimates these quantities simultaneously with the state variables. The performance of the algorithm is evaluated by using simulated data.     

COMPARISON OF PROCESS‐BASED AND ARTIFICIAL NEURAL NETWORK APPROACHES FOR STREAMFLOW MODELING IN AN AGRICULTURAL WATERSHED1

ABSTRACT: The performance of the Soil and Water Assessment Tool (SWAT) and artificial neural network (ANN) models in simulating hydrologic response was assessed in an agricultural watershed in southeastern Pennsylvania. All of the performance evaluation measures including Nash‐Sutcliffe coefficient of efficiency (E) and coefficient of determination (R²) suggest that the ANN monthly predictions were closer to the observed flows than the monthly predictions from the SWAT model. More specifically, monthly streamflow E and R² were 0.54 and 0.57, respectively, for the SWAT model calibration period, and 0.71 and 0.75, respectively, for the ANN model training period. For the validation period, these values were ?0.17 and 0.34 for the SWAT and 0.43 and 0.45 for the ANN model. SWAT model performance was affected by snowmelt events during winter months and by the model's inability to adequately simulate base flows. Even though this and other studies using ANN models suggest that these models provide a viable alternative approach for hydrologic and water quality modeling, ANN models in their current form are not spatially distributed watershed modeling systems. However, considering the promising performance of the simple ANN model, this study suggests that the ANN approach warrants further development to explicitly address the spatial distribution of hydrologic/water quality processes within watersheds.     

GIS‐Based Predictive Models of Hillslope Runoff Generation Processes1

Abstract: Successful nonpoint source pollution control using best management practice placement is a complex process that requires in‐depth knowledge of the locations of runoff source areas in a watershed. Currently, very few simulation tools are capable of identifying critical runoff source areas on hillslopes and those available are not directly applicable under all runoff conditions. In this paper, a comparison of two geographic information system (GIS)‐based approaches: a topographic index model and a likelihood indicator model is presented, in predicting likely locations of saturation excess and infiltration excess runoff source areas in a hillslope of the Savoy Experimental Watershed located in northwest Arkansas. Based on intensive data collected from a two‐year field study, the spatial distributions of hydrologic variables were processed using GIS software to develop the models. The likelihood indicator model was used to produce probability surfaces that indicated the likelihood of location of both saturation and infiltration excess runoff mechanisms on the hillslope. Overall accuracies of the likelihood indicator model predictions varied between 81 and 87% for the infiltration excess and saturation excess runoff locations respectively. On the basis of accuracy of prediction, the likelihood indicator models were found to be superior (accuracy 81‐87%) to the predications made by the topographic index model (accuracy 69.5%). By combining statistics with GIS, runoff source areas on a hillslope can be identified by incorporating easily determined hydrologic measurements (such as bulk density, porosity, slope, depth to bed rock, depth to water table) and could serve as a watershed management tool for identifying critical runoff source areas in locations where the topographic index or other similar methods do not provide reliable results.     

An imprecise fuzzy risk approach for water quality management of a river system

Uncertainty plays an important role in water quality management problems. The major sources of uncertainty in a water quality management problem are the random nature of hydrologic variables and imprecision (fuzziness) associated with goals of the dischargers and pollution control agencies (PCA). Many Waste Load Allocation (WLA) problems are solved by considering these two sources of uncertainty. Apart from randomness and fuzziness, missing data in the time series of a hydrologic variable may result in additional uncertainty due to partial ignorance. These uncertainties render the input parameters as imprecise parameters in water quality decision making. In this paper an Imprecise Fuzzy Waste Load Allocation Model (IFWLAM) is developed for water quality management of a river system subject to uncertainty arising from partial ignorance. In a WLA problem, both randomness and imprecision can be addressed simultaneously by fuzzy risk of low water quality. A methodology is developed for the computation of imprecise fuzzy risk of low water quality, when the parameters are characterized by uncertainty due to partial ignorance. A Monte-Carlo simulation is performed to evaluate the imprecise fuzzy risk of low water quality by considering the input variables as imprecise. Fuzzy multiobjective optimization is used to formulate the multiobjective model. The model developed is based on a fuzzy multiobjective optimization problem with max–min as the operator. This usually does not result in a unique solution but gives multiple solutions. Two optimization models are developed to capture all the decision alternatives or multiple solutions. The objective of the two optimization models is to obtain a range of fractional removal levels for the dischargers, such that the resultant fuzzy risk will be within acceptable limits. Specification of a range for fractional removal levels enhances flexibility in decision making. The methodology is demonstrated with a case study of the Tunga–Bhadra river system in India.     

WATERSHED RESPONSES TO CLIMATE CHANGE AT GLACIER NATIONAL PARK1

ABSTRACT: We have developed an approach which examines ecosystem function and the potential effects of climatic shifts. The Lake McDonald watershed of Glacier National Park was the focus for two linked research activities: acquisition of baseline data on hydrologic, chemical and aquatic organism attributes that characterize this pristine northern rocky mountain watershed, and further developing the Regional Hydro-Ecosystem Simulation System (RHESSys), a collection of integrated models which collectively provide spatially explicit, mechanistically-derived outputs of ecosystem processes, including hydrologic outflow, soil moisture, and snow-pack water equivalence. In this unique setting field validation of RHESSys, outputs demonstrated that reasonable estimates of SWE and streamflow are being produced. RHESSys was used to predict annual stream discharge and temperature. The predictions, in conjunction with the field data, indicated that aquatic resources of the park may be significantly affected. Utilizing RHESSys to predict potential climate scenarios and response of other key ecosystem components can provide scientific insights as well as proactive guidelines for national park management.     

SOIL MOISTURE PREDICTION USING SUPPORT VECTOR MACHINES1

ABSTRACT: Herein, a recently developed methodology, Support Vector Machines (SVMs), is presented and applied to the challenge of soil moisture prediction. Support Vector Machines are derived from statistical learning theory and can be used to predict a quantity forward in time based on training that uses past data, hence providing a statistically sound approach to solving inverse problems. The principal strength of SVMs lies in the fact that they employ Structural Risk Minimization (SRM) instead of Empirical Risk Minimization (ERM). The SVMs formulate a quadratic optimization problem that ensures a global optimum, which makes them superior to traditional learning algorithms such as Artificial Neural Networks (ANNs). The resulting model is sparse and not characterized by the “curse of dimensionality.” Soil moisture distribution and variation is helpful in predicting and understanding various hydrologic processes, including weather changes, energy and moisture fluxes, drought, irrigation scheduling, and rainfall/runoff generation. Soil moisture and meteorological data are used to generate SVM predictions for four and seven days ahead. Predictions show good agreement with actual soil moisture measurements. Results from the SVM modeling are compared with predictions obtained from ANN models and show that SVM models performed better for soil moisture forecasting than ANN models.     

AUTOMATED EXTRACTION OF DRAINAGE NETWORK AND WATERSHED DATA FROM DIGITAL ELEVATION MODELS1

ABSTRACT: This paper discusses a computer program which extracts a number of watershed and drainage network properties directly from digital elevation models (DEM) to assist in the rapid parameterization of hydrologic runoff models. The program integrates new and established algorithms to address problems inherent in the analysis low-relief terrain from raster DEMs similar to those distributed by the U.S. Geological Survey for 7.5-minute quadrangles. The program delineates the drainage network from a DEM, and determines the Strahler order, total and direct drainage area, length, slope, and upstream and downstream coordinates of each channel link. It also identifies the subwatershed of each channel source and of the left and right bank of each channel link, and assigns a unique number to each network node. The node numbers are used to associate each subwatershed with the channel link to which it drains, and can be used to control flow routing in cascade hydrologic models. Program output includes tabular data and raster maps of the drainage network and subwatersheds. The raster maps are intended for import to a Geographical Information System where they can be registered to other data layers and used as templates to extract additional network and subwatershed information.     

Distributed Hydrologic Modeling in the Suburban Landscape: Assessing Parameter Transferability from Gauged Reference Catchments1

Mittman, Tamara, Lawrence E. Band, Taehee Hwang, and Monica Lipscomb Smith, 2012. Distributed Hydrologic Modeling in the Suburban Landscape: Assessing Parameter Transferability from Gauged Reference Catchments. Journal of the American Water Resources Association (JAWRA) 48(3): 546-557. DOI: 10.1111/j.1752-1688.2011.00636.x Abstract: Distributed, process-based models of catchment hydrologic response are potentially useful tools for the assessment of Low Impact Development (LID) techniques in urbanized catchments. Their application is often limited, however, by the lack of continuous streamflow records to calibrate poorly constrained parameters. This article examines the transferability of soil and groundwater parameters from a forested reference catchment to a nearby suburban catchment. We use the Regional Hydro-Ecologic Simulation System (RHESSys) to develop hydrologic models of one gauged forested and one ungauged suburban catchment within the Baltimore Ecosystem Study (BES) study area. We use a parameter uncertainty framework to calibrate soil and groundwater parameters for the forested catchment, and discrete measurements of streamflow from the suburban catchment to assess parameter transferability. Results indicate that the transfer of soil and groundwater parameters from forested reference to nearby suburban catchments is viable, with performance measures for the suburban catchment often exceeding those for the forested catchment. We propose that the simplification of hydrologic processes in urbanized catchments may account for the increase in model performance in the suburban catchment.     

GIS‐BASED HYIROLOGIC MODELING OF RIPARIAN AREAS: IMPLICATIONS FOR STREAM WATER QUALITY1

ABSTRACT: Riparian buffers have potential for reducing excess nutrient levels in surface water. Spatial variation in riparian buffer effectiveness is well recognized, yet researchers and managers still lack effective general tools for understanding the relevance of different hydrologic settings. We present several terrain‐based GIS models to predict spatial patterns of shallow, subsurface hydrologic flux and riparian hydrology. We then link predictions of riparian hydrology to patterns of nutrient export in order to demonstrate potential for augmenting the predictive power of land use/land cover (LU/LC) maps. Using predicted hydrology in addition to LUILC, we observed increases in the explained variation of nutrient exports from 290 sites across Lower Michigan. The results suggest that our hydrologic predictions relate more strongly to patterns of nutrient export than the presence or absence of wetland vegetation, and that in fact the influence of vegetative structure largely depends on its hydrologic context. Such GIS models are useful and complimentary tools for exploring the role of hydrologic routing in riparian ecosystem function and stream water quality. Modeling efforts that take a similar GIS approach to material transport might be used to further explore the causal implications of riparian buffers in heterogeneous watersheds.