Calibration Of Water Quality Model Research Articles

Calibration, validation and uncertainty analysis of a SWAT water quality model

Sediment and nutrient pollution in water bodies is threatening human health and the ecosystem, due to rapid land use changes and improper agricultural practices. The impact of the nonpoint source pollution needs to be evaluated for the sustainable use of water resources. An ideal tool like the soil and water assessment tool (SWAT) can assess the impact of pollutant loads on the drainage area, which could be beneficial for developing a water quality management model. This study aims to evaluate the SWAT model’s multi-objective and multivariable calibration, validation, and uncertainty analysis at three different sites of the Yarra River drainage area in Victoria, Australia. The drainage area is split into 51 subdrainage areas in the SWAT model. The model is calibrated and validated for streamflow from 1990 to 2008 and sediment and nutrients from 1998 to 2008. The results show that most of the monthly and annual calibration and validation for streamflow, nutrients, and sediment at the three selected sites are found with Nash–Sutcliffe efficiency values greater than 0.50. Furthermore, the uncertainty analysis of the model shows satisfactory results where the p-factor value is reliable by considering 95% prediction uncertainty and the d-factor value is close to zero. The model's results indicate that the model performs well in the river's watershed, which helps construct a water quality management model. Finally, the model application in the cost-effective management of water quality might reduce pollution in water bodies due to land use and agricultural activities, which would be beneficial to water management managers.

Calibration of QUAL2K water quality model in Pattipul stream (Saran) with Site-specific Parameters

Calibration of QUAL2K water quality model in Pattipul stream (Saran) with Site-specific Parameters

Quantifying input uncertainty in the calibration of water quality models: reordering errors via the secant method

Abstract. Uncertainty in input can significantly impair parameter estimation in water quality modeling, necessitating the accurate quantification of input errors. However, decomposing the input error from the model residual error is still challenging. This study develops a new algorithm, referred to as the Bayesian Error Analysis with Reordering (BEAR), to address this problem. The basic approach requires sampling errors from a pre-estimated error distribution and then reordering them with their inferred ranks via the secant method. This approach is demonstrated in the case of total suspended solids (TSSs) simulation via a conceptual water quality model. Based on case studies using synthetic data, the BEAR method successfully improves the input error identification and parameter estimation by introducing the error rank estimation and the error position reordering. The results of a real case study demonstrate that, even with the presence of model structural error and output data error, the BEAR method can approximate the true input and bring a better model fit through an effective input modification. However, its effectiveness depends on the accuracy and selection of the input error model. The application of the BEAR method in TSS simulation can be extended to other water quality models.

Incorporating multiple observational uncertainties in water quality model calibration

AbstractThe presence of error in water quality and hydrologic variables can significantly impair the calibration of water quality models. Precise and reliable identification of observational errors can have a significant impact on improving model parameter estimation. This study develops the Bayesian error analysis with reordering (BEAR) method to accommodate multiple sources of observational errors in the calibration of a water quality model. It realizes this goal by sampling the errors for input and output data from their respective error distributions and reordering them with inferred ranks via the secant method. This approach is demonstrated in the case of total suspended solids (TSS) simulated via a conceptual water quality model. Based on case studies using synthetic data, the new algorithm successfully quantifies one source of observational error when the error model of another source of observational error can be estimated accurately in advance. The results of a real case study also illustrate that considering observational errors in both model inputs and outputs, rather than in just the inputs or outputs, can improve the parameter calibration and error characterization. The improvements depend on the precision of the prior information for the error model. The application of this new algorithm in TSS simulation can be an example to understand how the BEAR algorithm works in other water quality models. The core idea of the BEAR algorithm is flexible and can be extended to the calibration of other environmental models.

Empirical modeling of hypolimnion and sediment oxygen demand in temperate Canadian lakes

ABSTRACT Sediment oxygen demand (SOD) and hypolimnetic oxygen demand (HOD) drive deep-water dissolved oxygen (DO) depletion in lakes, yet these parameters can be difficult to be measure routinely. To address this issue, we present an empirical DO depletion model from time-series measurements of hypolimnetic DO and water-column temperature profiles to estimate hypolimnion thickness (H). The model is based on a dataset that includes 3 temperate lakes (Lake Erie, Lake Simcoe, and Eagle Lake) in Ontario, Canada, of varying size and trophic state. We report SOD (mean [standard deviation]; 0.30 [0.07] g m−2 d−1) and HOD (0.08 [0.03] g m−3 d−1) values based on regression fits of (where t is time) and H from these lakes. The model shows that when vertical (through thermocline) and horizontal fluxes can be neglected in the DO budget during summer, SOD and HOD are the first-order parameters driving . The empirical model predicted hypolimnetic DO in the 3 lakes (root-mean-square error [RMSE] of DO < 3.58 g m−3) and was subsequently validated against observations from a fourth lake (Little Silver Lake, Ontario; RMSE of DO = 1.07 g m−3). The model provides insight into the importance of physical characteristics (i.e., 1/H) in the hypolimnetic DO budget and the relative impact of physical transport versus biogeochemical sources and sinks in the DO budget. When the spring turnover DO concentration is known, the model can be used to simulate depletion of DO in the hypolimnion, including the onset of hypoxia, using routinely collected temperature profile data. We suggest that the proposed values for SOD and HOD can serve as estimates for water quality model calibration when no information is available.

Benchmarking the efficiency of a metamodeling-enabled algorithm for the calibration of surface water quality models

Abstract Computational efficiency is a major obstacle imposed in the automatic calibration of numerical, high-fidelity surface water quality models. To surpass this obstacle, the present work formulated a metamodeling-enabled algorithm for the calibration of surface water quality models and assessed the computational gains from this approach compared to a benchmark alternative (a derivative-free optimization algorithm). A radial basis function was trained over multiple snapshots of the original high-fidelity model to emulate the latter's behavior. This data-driven proxy of the original model was subsequently employed in the automatic calibration of the water quality models of two water reservoirs and, finally, the computational gains over the benchmark alternative were estimated. The benchmark analysis revealed that the metamodeling-enabled optimizer reached a solution with the same quality compared to its benchmark alternative in 20–38% lower process times. Thereby, this work manifests tangible evidence of the potential of metamodeling-enabled strategies and sets out a discussion on how to maximize computational gains deriving from such strategies in surface water quality modeling.

Robustness analysis of storm water quality modelling with LID infrastructures from natural event-based field monitoring

Sponge city construction (SCC) in China, as a new concept and a practical application of low-impact development (LID), is gaining wide popularity. Modelling tools are widely used to evaluate the ecological benefits of SCC in stormwater pollution mitigation. However, the understanding of the robustness of water quality modelling with different LID design options is still limited due to the paucity of water quality data as well as the high cost of water quality data collection and model calibration. This study develops a new concept of ‘robustness’ measured by model calibration performances. It combines an automatic calibration technique with intensive field monitoring data to perform the robustness analysis of storm water quality modelling using the SWMM (Storm Water Management Model). One of the national pilot areas of SCC, Fenghuang Cheng, in Shenzhen, China, is selected as the study area. Five water quality variables (COD, NH3-N, TN, TP, and SS) and 13 types of LID/non-LID infrastructures are simulated using 37 rainfall events. The results show that the model performance is satisfactory for different water quality variables and LID types. Water quality modelling of greenbelts and rain gardens has the best performance, while the models of barrels and green roofs are not as robust as those of the other LID types. In urban runoff, three water quality parameters, namely, SS, TN and COD, are better captured by the SWMM models than NH3-N and TP. The modelling performance tends to be better under heavy rain and significant pollutant concentrations, denoting a potentially more stable and reliable design of infrastructures. This study helps to improve the current understanding of the feasibility and robustness of using the SWMM model in sponge city design.

Direct assimilation of measured soil water content in Root Zone Water Quality Model calibration for deficit‐irrigated maize

AbstractCorrect soil water simulation is critical for water balance and plant growth in agricultural systems. Crop production simulation errors have often been attributed to a lack of accuracy in soil water content (SWC) estimates. However, only a few studies have quantified the effects of SWC estimate errors on crop production and evapotranspiration (ET), especially under different irrigation treatments. The objective of this study was to investigate the impacts of direct assimilation of measured SWC during model calibration for deficit irrigated maize (Zea mays L.) on simulated ET, leaf area index (LAI), biomass, and yield. The CERES‐Maize model within the Root Zone Water Quality Model (RZWQM) was calibrated using the automatic parameter estimation (PEST) software. Simulation results showed that, using PEST‐optimized crop parameters, RZWQM was able to adequately predict crop yield (relative root mean squared error, rRMSE, of 4.8%) and biomass (rRMSE of 7.1%) in response to irrigation levels, in spite of the bias in SWC and ET simulation. However, with the same crop parameters but replacing simulated SWC with measured data, simulations of crop yield and biomass became worse, with higher rRMSE values (14.5% for yield and 21.5% for biomass). This unexpected model performance with SWC assimilation was mainly associated with the water addition and removal from the soil, which was improved only by recalibration of both soil and crop parameters. This study suggested compensating effects between soil and crop parameters during model calibration. Caution should be applied when using measured SWC as model inputs, especially under water stress conditions.

Ensemble data assimilation methods for improving river water quality forecasting accuracy

River water quality is one of the main challenges that societies face during the 21st century. Accurate and reliable real-time prediction of water quality is an effective adaptation measure to counteract water quality issues such as accidental spill and harmful algae blooms. To improve accuracy and skill of water quality forecasts along the Yeongsan River in South Korea three different ensemble data assimilation (DA) methods have been investigated: the traditional Ensemble Kalman Filter (EnKF) and two related algorithms (Dud-EnKF and EnKF-GS) that offer either possibilities to improve initial conditions for non-linear models or reduce computation time (important for real-time forecasting) by using a (smaller) time-lagged ensemble to estimate the Kalman gain. Twin experiments, assimilating synthetic observations of three algae species and phosphate concentrations, with relatively small ensemble sizes showed that all three DA methods improved forecast accuracy and skill with only subtle difference between the methods. They all improved the model accuracy at downstream locations with very similar performances but due to spurious correlation, the accuracy at upstream locations was somewhat deteriorated. The experiments also showed no clear trend of improvement by increasing the ensemble size from 8 to 64. The real world experiments, assimilating real observations of three algae species and phosphate concentrations, showed that less improvement was achieved compared to the twin experiments. Further improvement of the model accuracy may be achieved with different state variable definitions, use of different perturbation and error modelling settings and/or better calibration of the deterministic water quality model.

The effect of sampling frequency and strategy on water quality modelling driven by high-frequency monitoring data in a boreal catchment

With conventional water quality monitoring based on grab sampling, usually carried out once a month, it is quite challenging to obtain reliable sediment and nutrient loading estimates, which is a pre-requisite for testing of water quality models. Despite the growing availability of high-frequency water quality monitoring data obtained by using modern sensors, their use in the calibration of water quality models is still limited. The main objective of this study is to evaluate six years of continuous, high-frequency water quality monitoring data set from a case study carried out in a medium-sized, boreal catchment in southern Finland to be used as a source for the calibration and testing of the Soil & Water Assessment Tool (SWAT). This objective is accomplished by developing a set of ‘sub-sampling’ scenarios focused on quantifying the effect of the sampling frequency (12 vs. 26 samples per year) and strategy (regular vs. flow-proportional sampling) on the model performance and output uncertainty. The goodness-of-fit of the model calibrated against high-frequency data (‘benchmark’), evaluated using the Kling-Gupta Efficiency index for total suspended solids (TSS), nitrate and total nitrogen (NO3-N and TN) and total phosphorus (TP) daily loads ranged between 0.76 and 0.83 (0.69 and 0.73 in validation period). Overall, the model calibrated against low-frequency data performed worse when evaluated against high-frequency data and compared with the benchmark model for each of the studied water quality parameters. The results show that increasing the sampling frequency from 12 to 26 samples per year leads to an improvement in the model performance, particularly for TSS and TP loads. In contrast to the sampling frequency, the evidence for the effect of the sampling strategy on the model performance was much weaker. The model output uncertainty due to different realizations of sub-sampling scenarios was high for four analysed output indices: basin-averaged, long-term average annual values of soil erosion rate, denitrification rate, NO3-N yield and soluble P yield. This study demonstrates that more attention should be paid to the calibration data as a source of uncertainty in model predictions. It also provides evidence for either a wider adoption of modern sensors providing high-frequency data, or for increasing the grab sampling frequency to reduce the uncertainty of loading estimates, particularly in rivers with flashy flow regime such as the Vantaanjoki River.

Soil and Water Assessment Tool (SWAT) Model: A Systemic Review

Wang, Y.; Jiang, R.; Xie, J.; Zhao, Y.; Yan, D., and Yang, S., 2019. Soil and water assessment tool (SWAT) model: A systemic review. In: Guido-Aldana, P.A. and Mulahasan, S. (eds.), Advances in Water Resources and Exploration. Journal of Coastal Research, Special Issue No. 93, pp. 22–30. Coconut Creek (Florida), ISSN 0749-0208.Under the influences of climate change and human activities, the water cycle under changing environment and distributed hydrological model based on physical mechanism have become the focus of hydrology research. Soil & Water Assessment Tool (SWAT) model is one of typical distributed models, and it has been widely used in testing and forecasting the water and sediment circulation and agriculture production with chemicals. The paper used CiteSpace to analyze the research progress based on previous literatures. The results of high frequency keywords and co-word analysis show that the research hotspots of SWAT model include water quality problem, sediment yield, model calibration, uncertainty analysis and sensitivity analysis. The literatures related to SWAT were divided into model application, model accuracy and model coupling. The paper mainly focused on the literature related to model application. The domestic and abroad research processes were reviewed from three hot application fields such as runoff simulation, hydrological impacts under changing environment (HIUCE) and non-point source (NPS) pollution. The results show that: (1) the research content of runoff simulation mainly focused on the adaptability and accuracy of the model; (2) the research content of HIUCE mainly concentrated on analysis of historical changes in hydrological effects and prediction of future changes using the scenario analysis; (3) the research content of NPS pollution mainly focused on the spatial and temporal distribution of nutrients, the influencing factors and degree of nutrients, and the prediction of future to select Best Management Practices (BMPs). The paper pointed out some existing problems, prospected the future research, and proposed research idea of model improvement to adapt to the domestic changing environment. This study should be beneficial to systematically understand the structure, principle, application fields and future development trends of SWAT model, and thus provide references for users of SWAT model in related subjects.

The Value of Intensive Sampling—A Comparison of Fluvial Loads

Most long-term sampling regimes are calendar based, collecting one or two samples per month regardless of the stream conditions. Loads estimated with calendar-based sampling are often used for expensive water quality mitigation measures. In this paper, we have tested the differences between the calendar-based and extensive sampling methods for two watersheds of different sizes, and three parameters—total nitrogen, total phosphorus, and total suspended solids. Based on the results obtained and the costs associated with the remediation, a simple decision-making framework is proposed for watershed managers to decide on the applicability of a calendar-based sampling method. Direct loads (DL) were computed using a method based on an intensive sampling of flow and other water quality parameters. Weighted regression loads (WL) were estimated using the WRTDS model designed for modified calendar-based sampling. The results suggest that for trend analysis and planning on a larger scale, long-term loads obtained from a modified calendar-based sampling regime may be used as a reasonable substitute for loads obtained from intensive sampling. However, for purposes where accurate daily loads are needed (e.g., water quality model calibration) WL may not be an effective substitute for DL. Finally, we recommend that the costs of control measures should be assessed when deciding on a sampling regime.

Calibration of mathematical water quality modeling in a river basin under critical conditions

In order to assist with the management of water resources, this study aimed to calibrate the water quality model "Ottocoded Critical Basins Analysis" (ABaCO) in a section of the Lambari River, located in the city of Poços de Caldas, Minas Gerais, Brazil, and it is considered qualitatively critical by the "National Water Agency" (ANA). The calibration results, obtained automatically by the Microsoft Excel® tool Solver, showed good adjustments between the simulated and observed concentrations for the following parameters: total nitrogen, organic phosphorus, inorganic phosphorus, and total phosphorus, considering a visual interpretation of the graphics. The other calibrated parameters (BOD, DO, organic nitrogen, ammoniacal nitrogen, and nitrate) obtained satisfactory adjustments. When evaluating the results by coefficient of determination, it was observed that all parameters showed negative values to one or more field campaigns. However, the calibrated values of the parameters can be incorporated into the model to obtain a first view of the river's qualitative situation, aiming to plan, control, and protect water resources to assist in decision-making and to favor performance of intervention analyses of environmental control measures. PRACTITIONER POINTS: Calibration of water quality model in a section of a river located in Brazil considered qualitatively critical; The calibration occurred automatically and manually, when it was necessary, in an Excel® spreadsheet, generated from the water quality model; The automation process was performed by the Excel® Solver tool through the objective function; Through visual interpretation, it was noticed that the calibrated parameters were well adjusted to the data observed in the field; Sensitivity analysis showed that the studied parameters did not show significant differences in relation to the values calibrated for parameters.

Optimal experimental design for calibration of a new sewer water quality model

Water quality data in the sewer system are indispensable for modelling, but rarely available, as measurements in sewers are challenging to conduct. Optimal experimental design (OED) is a powerful tool to identify and maximize the information content of measurement data. This paper adopts a model-based OED methodology to efficiently plan a measurement campaign for final model calibration and validation of a new sewer water quality model. To do so, a preliminary calibrated model of the case study is used to evaluate the information content of different potential measurement locations and scenarios for suspended solids as measured variable. The case study first demonstrates how OED can identify the best measurement location within a complex sewer network. It secondly demonstrates that measuring the beginning of a big rain event results in the most information-rich data among all scenarios evaluated. Thirdly, it analyses in detail the information content of dry weather flow (DWF) data. In comparison to previous studies the methodology is improved by considering the actual measurement error characteristics when calculating the information content of measurement data.

Calibration of water quality model for distribution networks using genetic algorithm, particle swarm optimization, and hybrid methods

Chlorine reacts with both organic and inorganic matters in water. That is why water quality modeling has received great attention in recent years. The serious issue in municipal water quality modeling is gathering the essential input parameters of the model, particularly bulk decay (kb) and wall decay (kw) coefficients as well as their calibrations. Therefore, this study first thoroughly formulates the problem in the form of a heuristic optimization and then utilizes Genetic Algorithm, Particle Swarm Optimization, and Hybrid GA-PSO as the model optimizers in order to best calibrate kw for minimizing the difference of residual chlorine concentrations that exist between the simulated and observed values. These three algorithms are linked to EPANET, the hydraulic and water quality simulator. The method is then applied to a real-world water distribution network. Here, kw is considered as a decision variable. The objective function is to minimize both the Sum of Square Error and Root Mean Square Error between the observed and simulated chlorine concentrations. According to the simulation results obtained, the optimal value of wall decay coefficient is 1.233 m/day during the calibration process while the minimum and maximum differences between the measured and simulated chlorine rates were 0 and 0.18, respectively.•The method presented in this article can be useful for managers of water and wastewater companies, water resources facilities and operators and operation manager of water distribution system to manage chlorine dosing rate.•Due to adverse health effect of disinfection by product and poor microbial water quality as results of inefficient chlorination, control chlorine concentration in water distribution networks and its consequence on human health effect is necessary.•Hybrid PSO and GA methods are used to cope with their falling in local optimum and requiring highly computational effort.

A regionalized river water quality model calibration method based on watershed physical characteristics: application to the Cau River in Vietnam

A methodology is proposed for the calibration of river water quality models on large watersheds, in the absence of intensive measurements for water quality and quantity. This methodology is based on: 1) the use of the results from a hydrological model to provide the required hydrological variables to the water quality model; 2) five assumptions for the definition of initial and boundary conditions; 3) a three-step regionalized calibration method, in which the specific characteristics of the different subwatersheds are taken into account and 4) the adjustment of some parameters in order to reproduce processes that are not explicitly represented in the model. The regionalized calibration method relies on a comprehensive study of the land use and characteristics on each subwatershed and the definition of different sets of parameters values in distinct regions. Application to the Cau River, in Vietnam, with QUAL-GIBSI, an adaptation of the QUAL2E model, showed that: i) calibration and validation results were significantly improved by applying regionalized calibration as compared to an initial calibration for which a single set of parameters values was used for the whole simulated river stretch and ii) use of a hydrological model to provide discharge at various points in the watershed allowed to overcome the lack of detailed measurements of discharge at locations other than the watershed outlet.

Study on the Effect of Sulfate on the Degradation of BTEX in Leakage Area of Gasoline by Using Numerical Simulation

In order to explore the migration of sulfate electron acceptors injected into aqueous media, and study the specific attenuation of BTEX with sulfate injecting. In this study, the laboratory sand tank was used for the injecting and natural attenuation monitoring experiments. The combination of MODFLOW and RT3D was used to simulate the attenuation process of sulfate and natural degradation of BTEX. The results showed that the concentration of sulfate in the source area was significantly lower than that of the surrounding area at the 600th day after injected on the 416th day and BTEX concentration also gradually tends to zero. The injection experiments were conducted for a total of 46 days with the injected sulfate quality of 1488.24 g, and the calculated sulfate output of 1320.52 g. It can be seen that the input was significantly higher than the output and the sulfate loss could be judged as acting on Redox reactions .Sulfate promoted the degradation of BTEX. The numerical simulation was carried out for the water flow model and water quality model calibration, the normalized RMS of water flow model and water quality model are 9.8% and 8% respectively, which are less than 10% of the conventional limit. The model can accurately simulate the migration of sulfate and BTEX attenuation and predict the further pollution of BTEX. And can be applied to the actual site contaminated by BTEX that has important practical significance.

Parameter Estimation of Water Quality Models Using an Improved Multi-Objective Particle Swarm Optimization

Water quality models are of great importance for developing policies to control water pollution, with the model parameters playing a decisive role in the simulation results. It is necessary to introduce estimation through multi-objective parameters, which is often affected by noise in the data, into water quality models. This paper presents a multi-objective particle swarm optimization algorithm, which is based on the Mahalanobis distance operation, mechanism of cardinality preference and advection-diffusion operator. The Mahalanobis distance operation can effectively reduce the influence of noise in the data on model calibration. The mechanism of cardinality preference and the use of the advection-diffusion operator can prevent non-dominated solutions from falling into the local optimum. Four cases were used to test the proposed approach. The first two cases with true Pareto fronts show that this approach can accurately estimate the true Pareto front with a good distribution, even in the presence of noise. Furthermore, the application of the approach was tested by the O’Connor model and Crops of Engineers Integrated Compartment Water Quality Model. We show that our approach can produce satisfactory results for the multi-objective calibration of complex water quality models. In general, the proposed approach can provide accurate and efficient parameter estimation in water quality models.

Environmental Flow for Sungai Johor Estuary

Sungai Johor estuary is a vital water body in the south of Johor and greatly affects the water quality in the Johor Straits. In the development of the hydrodynamic and water quality models for Sungai Johor estuary, the Environmental Fluid Dynamics Code (EFDC) model was selected. In this application, the EFDC hydrodynamic model was configured to simulate time varying surface elevation, velocity, salinity, and water temperature. The EFDC water quality model was configured to simulate dissolved oxygen (DO), dissolved organic carbon (DOC), chemical oxygen demand (COD), ammoniacal nitrogen (NH3-N), nitrate nitrogen (NO3-N), phosphate (PO4), and Chlorophyll a. The hydrodynamic and water quality model calibration was performed utilizing a set of site specific data acquired in January 2008. The simulated water temperature, salinity and DO showed good and fairly good agreement with observations. The calculated correlation coefficients between computed and observed temperature and salinity were lower compared with the water level. Sensitivity analysis was performed on hydrodynamic and water quality models input parameters to quantify their impact on modeling results such as water surface elevation, salinity and dissolved oxygen concentration. It is anticipated and recommended that the development of this model be continued to synthesize additional field data into the modeling process.

Impact of Length of Dataset on Streamflow Calibration Parameters and Performance of APEX Model

AbstractDue to resource constraints, long‐term monitoring data for calibration and validation of hydrologic and water quality models are rare. As a result, most models are calibrated and, if possible, validated using limited measured data. However, little research has been done to determine the impact of length of available calibration data on model parameterization and performance. The main objective of this study was to evaluate the impact of length of calibration data (LCD) on parameterization and performance of the Agricultural Policy Environmental eXtender model for predicting daily, monthly, and annual streamflow. Long‐term (1984‐2015) measured daily streamflow data from Rock Creek watershed, an agricultural watershed in northern Ohio, were used for this study. Data were divided into five Short (5‐year), two Medium (15‐year), and one Long (25‐year) streamflow calibration data scenarios. All LCD scenarios were calibrated and validated at three time steps: daily, monthly, and annual. Results showed LCD affected the ability of the model to accurately capture temporal variability in simulated streamflow. However, overall average streamflow, water budgets, and crop yields were simulated reasonably well for all LCD scenarios.