Incorporating layer- and local-scale heterogeneities in numerical simulation of unsaturated flow and tracer transport

This study characterizes layer- and local-scale heterogeneities in hydraulic parameters (i.e., matrix permeability and porosity) and investigates the relative effect of layer- and local-scale heterogeneities on the uncertainty assessment of unsaturated flow and tracer transport in the unsaturated zone of Yucca Mountain, USA. The layer-scale heterogeneity is specific to hydrogeologic layers with layerwise properties, while the local-scale heterogeneity refers to the spatial variation of hydraulic properties within a layer. A Monte Carlo method is used to estimate mean, variance, and 5th, and 95th percentiles for the quantities of interest (e.g., matrix saturation and normalized cumulative mass arrival). Model simulations of unsaturated flow are evaluated by comparing the simulated and observed matrix saturations. Local-scale heterogeneity is examined by comparing the results of this study with those of the previous study that only considers layer-scale heterogeneity. We find that local-scale heterogeneity significantly increases predictive uncertainty in the percolation fluxes and tracer plumes, whereas the mean predictions are only slightly affected by the local-scale heterogeneity. The mean travel time of the conservative and reactive tracers to the water table in the early stage increases significantly due to the local-scale heterogeneity, while the influence of local-scale heterogeneity on travel time gradually decreases over time. Layer-scale heterogeneity is more important than local-scale heterogeneity for simulating overall tracer travel time, suggesting that it would be more cost-effective to reduce the layer-scale parameter uncertainty in order to reduce predictive uncertainty in tracer transport.     

Predictions of U.K. regulated power station contributions to regional air pollution and deposition: a model comparison exercise

Contributions of the emissions from a U.K. regulated fossil-fuel power station to regional air pollution and deposition are estimated using four air quality modeling systems for the year 2003. The modeling systems vary in complexity and emphasis in the way they treat atmospheric and chemical processes, and include the Community Multiscale Air Quality (CMAQ) modeling system in its versions 4.6 and 4.7, a nested modeling system that combines long- and short-range impacts (referred to as TRACK-ADMS [Trajectory Model with Atmospheric Chemical Kinetics-Atmospheric Dispersion Modelling System]), and the Fine Resolution Atmospheric Multi-pollutant Exchange (FRAME) model. An evaluation of the baseline calculations against U.K. monitoring network data is performed. The CMAQ modeling system version 4.6 data set is selected as the reference data set for the model footprint comparison. The annual mean air concentration and total deposition footprints are summarized for each modeling system. The footprints of the power station emissions can account for a significant fraction of the local impacts for some species (e.g., more than 50% for SO2 air concentration and non-sea-salt sulfur deposition close to the source) for 2003. The spatial correlation and the coefficient of variation of the root mean square error (CVRMSE) are calculated between each model footprint and that calculated by the CMAQ modeling system version 4.6. The correlation coefficient quantifies model agreement in terms of spatial patterns, and the CVRMSE measures the magnitude of the difference between model footprints. Possible reasons for the differences between model results are discussed. Finally, implications and recommendations for the regulatory assessment of the impact of major industrial sources using regional air quality modeling systems are discussed in the light of results from this case study.     

Meteorological factors of ozone predictability at Houston, Texas

Several ozone modeling approaches were investigated to determine if uncertainties in the meteorological data would be sufficiently large to limit the application of physically realistic ozone (O3) forecast models. Three diagnostic schemes were evaluated for the period of May through September 1997 for Houston, TX. Correlations between measured daily maximum and model calculated O3 air concentrations were found to be 0.70 using a linear regression model, 0.65 using a non-advective box model, and 0.49 using a three-dimensional (3-D) transport and dispersion model. Although the regression model had the highest correlation, it showed substantial underestimates of the highest concentrations. The box model results were the most similar to the regression model and did not show as much underestimation. The more complex 3-D modeling approach yielded the worst results, likely resulting from O3 maxima that were driven by local factors rather than by the transport of pollutants from outside of the Houston domain. The highest O3 concentrations at Houston were associated with light winds and meandering trajectories. A comparison of the gridded meteorological data used by the 3-D model to the observations showed that the wind direction and speed values at Houston differed most on those days on which the O3 underestimations were the greatest. These periods also tended to correspond with poor precipitation and temperature estimates. It is concluded that better results are not just obtained through additional modeling complexity, but there needs to be a comparable increase in the accuracy of the meteorological data.     

Climatic mapping to identify high-risk areas for Cylindrocladium quinqueseptatum leaf blight on eucalypts in mainland South East Asia and around the world

Cylindrocladium quinqueseptatum is a pathogen on a wide range of hosts. It affects at least 20 species of eucalypts and is an important causal agent of leaf blight of Eucalyptus camaldulensis in central and southern Vietnam. Results from previous studies and observations of broadscale infection patterns in mainland South East Asia were used to derive simple rules (i.e. mean minimum temperature of coldest month > or =16 degrees C and mean annual precipitation > or =1400 mm) to identify locations which are likely to have a high risk of C. quinqueseptatum leaf blight (CqLB). Climatic mapping programs, which include interpolated climatic data estimated for numerous locations, were used to map these high risk areas in Africa, Australia and Latin America as well as in South East Asia. The predicted high-risk areas included several regions where CqLB has already been reported and the maps generated suggested other areas which may be at risk under present climatic conditions given the presence of C. quinqueseptatum and susceptible hosts. Some simple climate change scenarios were also used to suggest areas in mainland South East Asia which may become vulnerable to CqLB over the next 50 years. It is concluded that climatic mapping programs can assist the broadscale evaluation of risk of CqLB infections, although it is recognised that more detailed models and survey information are also needed.     

Single- and dual-porosity modelling of flow in reclaimed mine soil cores with embedded lignitic fragments

Lignitic mine soils represent a typical two-scale dual-porosity medium consisting of a technogenic mixture of overburden sediments that include lignitic components as dust and as porous fragments embedded within a mostly coarse-textured matrix. Flow and transport processes in such soils are not sufficiently understood to predict the course of soil reclamation or of mine drainage. The objective of this contribution is to identify the most appropriate conceptual model for describing small-scale heterogeneity effects on flow on the basis of the physical structure of the system. Multistep flow experiments on soil cores are analyzed using either mobile–immobile or mobile–mobile type 1D dual-porosity models, and a 3D numerical model that considers a local-scale distribution of fragments. Simulations are compared with time series' of upward infiltration and matric potential heads measured at two depths using miniature tensiometers. The 3D and the 1D dual-permeability models yielded comparable results as long as pressure heads are in local equilibrium; however, could describe either the upward infiltration or the matric potential curves but not both at the same time. The mobile–immobile type dual-porosity model failed to describe the data. A simultaneous match with pressure heads and upward infiltration data could only be obtained with the 1D dual-permeability model (i.e., mobile–mobile) by assuming an additional restriction of the inter-domain water transfer. These results indicate that for unsaturated flow conditions at higher matric potential heads (i.e., here >− 40 hPa), water in a restricted part of the fragment domain must be more mobile as compared to water in the sandy matrix domain. Closer inspections of the pore system and first neutron radiographic imaging support the hypothesis that a more continuous pore region exists at these pressure heads in the vicinity of the lignitic fragments possibly formed by fragment contacts and a lignitic dust interface-region between the two domains. The results suggest that the small-scale structure is too complex as to be represented by weighted contributions of individual components alone.     

Stochastic modeling of atmospheric pollution: a spatial time-series framework. Part I: methodology

A geostatistical framework for joint spatiotemporal modeling of atmospheric pollution is presented. The spatiotemporal distribution of concentration levels is modeled as a joint realization of a collection of spatially correlated time series. Parametric temporal trend models, associated with long-term pollution variability are established from concentration profiles at monitoring stations. Such parameters, e.g., amplitude of seasonal variation, are then regionalized in space for determining trend models at any unmonitored location. The resulting spatiotemporal residual field, associated with short-term pollution variability, is also modeled as a collection of spatially correlated residual time series. Stochastic conditional simulation is proposed for generating alternative realizations of the concentration spatiotemporal distribution, which identify concentration measurements available at monitoring stations. Simulated realizations also reproduce the histogram of the sample data, and a model of their spatiotemporal correlation. Such alternative concentration fields can be used for risk analysis studies.     

Meteorological Factors of Ozone Predictability at Houston,Texas

ABSTRACT

Several ozone modeling approaches were investigated to determine if uncertainties in the meteorological data would be sufficiently large to limit the application of physically realistic ozone (O₃) forecast models. Three diagnostic schemes were evaluated for the period of May through September 1997 for Houston, TX. Correlations between measured daily maximum and model calculated O₃ air concentrations were found to be 0.70 using a linear regression model, 0.65 using a non-advective box model, and 0.49 using a three-dimensional (3-D) transport and dispersion model. Although the regression model had the highest correlation, it showed substantial underestimates of the highest concentrations. The box model results were the most similar to the regression model and did not show as much underestimation. The more complex 3-D modeling approach yielded the worst results, likely resulting from O₃ maxima that were driven by local factors rather than by the transport of pollutants from outside of the Houston domain. The highest O₃ concentrations at Houston were associated with light winds and meandering trajectories. A comparison of the gridded meteorological data used by the 3-D model to the observations showed that the wind direction and speed values at Houston differed most on those days on which the O₃ underestima-tions were the greatest. These periods also tended to correspond with poor precipitation and temperature estimates. It is concluded that better results are not just obtained through additional modeling complexity, but there needs to be a comparable increase in the accuracy of the meteorological data.     

Determination of nitrate and other water quality parameters in groundwater from UV/Vis spectra employing partial least squares regression

The use of UV/Vis spectroscopy in combination with partial least squares (PLS) regression for the simultaneous prediction of nitrate and non-purgeable organic carbon (NPOC) in groundwaters was evaluated. A model of high quality was obtained using first order derivative spectra in the range 200-300 nm. Inclusion of non-UV-absorbing constituents in the modeling procedure, i.e., chloride, sulfate, fluoride, total carbon (TC), inorganic carbon (IC), alkalinity, pH and conductivity was also evaluated. This model seemed to be useful for prediction of chloride, TC, IC, alkalinity and conductivity, while its ability to predict sulfate, fluoride and pH was poor. In conclusion, application of PLS regression, which requires neither filtration of samples nor addition of chemicals, is a promising alternative for fast interpretation of geochemical patterns of groundwater quality.     

Temporal evolution of DNAPL source and contaminant flux distribution: impacts of source mass depletion

We investigated, using model simulations, the changes occurring in the distribution of dense non-aqueous phase liquid (DNAPL) mass (Sn) within the source zone during depletion through dissolution, and the resulting changes in the contaminant flux distribution (J) at the source control plane (CP). Two numerical codes (ISCO3D and T2VOC) were used to simulate selected scenarios of DNAPL dissolution and transport in three-dimensional, heterogeneous, spatially correlated, random permeability fields with emplaced sources. Data from the model simulations were interpreted based on population statistics (mean, standard deviation, coefficient of variation) and spatial statistics (centroid, second moments, variograms). The mean and standard deviation of the Sn and J distributions decreased with source mass depletion by dissolution. The decrease in mean and standard deviation was proportional for the J distribution resulting in a constant coefficient of variation (CV), while for the Sn distribution, the mean decreased faster than the standard deviation. The spatial distributions exhibited similar behavior as the population distribution, i.e., the CP flux distribution was more stable (defined by temporally constant second moments and range of variograms) than the Sn distribution. These observations appeared to be independent of the heterogeneity of the permeability (k) field (variance of the log permeability field=1 and 2.45), correlation structure (positive vs. negative correlation between the k and Sn domains) and the DNAPL dissolution model (equilibrium vs. rate-limited), for the cases studied. Analysis of data from a flux monitoring field study (Hill Air Force Base, Utah) at a DNAPL source CP before and after source remediation also revealed temporal invariance of the contaminant flux distribution. These modeling and field observations suggest that the temporal evolution of the contaminant flux distribution can be estimated if the initial distribution is known. However, the findings are preliminary and broader implications to sampling strategies for remediation performance assessment need to be evaluated in additional modeling and experimental studies.     

Spatial and temporal variability in urban fine particulate matter concentrations

Identification of hot spots for urban fine particulate matter (PM(2.5)) concentrations is complicated by the significant contributions from regional atmospheric transport and the dependence of spatial and temporal variability on averaging time. We focus on PM(2.5) patterns in New York City, which includes significant local sources, street canyons, and upwind contributions to concentrations. A literature synthesis demonstrates that long-term (e.g., one-year) average PM(2.5) concentrations at a small number of widely-distributed monitoring sites would not show substantial variability, whereas short-term (e.g., 1-h) average measurements with high spatial density would show significant variability. Statistical analyses of ambient monitoring data as a function of wind speed and direction reinforce the significance of regional transport but show evidence of local contributions. We conclude that current monitor siting may not adequately capture PM(2.5) variability in an urban area, especially in a mega-city, reinforcing the necessity of dispersion modeling and methods for analyzing high-resolution monitoring observations.     

Results of a Receptor Modeling Feasibility Study

This paper summarizes the results of an Electric Power Research Institute funded research effort to determine the feasibility of using receptor models for the apportionment of power plant contributions to air quality, deposition quality, and light extinction on local and regional scales. Sufficient information currently exists (or was developed during the course of this study) to establish feasibility for the apportionment of power plant contributions to local-scale air quality and to strongly suggest the usefulness of receptor modeling for regional-scale air quality, light extinction, and dry deposition quality apportionment. Insufficient information existed to determine whether or not receptor modeling can be useful for the allocation of wet deposition quality.

A series of seven future research recommendations were prepared for the purpose of advancing receptor modeling from theoretical feasibility to practical utility. Two recommendations address model evaluation: (1) prepare a user-oriented receptor model application and testing package, and (2) perform computer simulation testing of past, current, and proposed receptor model applications. Two recommendations address model development: (1) develop a receptor model requiring minimal information about source profiles, and (2) develop a “hybrid” model (i.e., a model that combines source and receptor oriented methods). Finally, three recommendations address source characterization: (1) develop procedures for individual particle characterization, (2) develop sampling and analysis methods for source profile determination, and (3) measure source profiles of coal- and oil-fired power plants, and other sources that confound identification of emissions from coal- and oilfired power plants.     

On the use of mean groundwater age, life expectancy and capture probability for defining aquifer vulnerability and time-of-travel zones for source water protection

Protection and sustainability of water supply wells requires the assessment of vulnerability to contamination and the delineation of well capture zones. Capture zones, or more generally, time-of-travel zones corresponding to specific contaminant travel times, are most commonly delineated using advective particle tracking. More recently, the capture probability approach has been used in which a probability of capture of P=1 is assigned to the well and the growth of a probability-of-capture plume is tracked backward in time using an advective-dispersive transport model. This approach accounts for uncertainty due to local-scale heterogeneities through the use of macrodispersion. In this paper, we develop an alternative approach to capture zone delineation by applying the concept of mean life expectancy E (time remaining before being captured by the well), and we show how life expectancy E is related to capture probability P. Either approach can be used to delineate time-of-travel zones corresponding to specific travel times, as well as the ultimate capture zone. The related concept of mean groundwater age A (time since recharge) can also be applied in the context of defining the vulnerability of a pumped aquifer. In the same way as capture probability, mean life expectancy and groundwater age account for local-scale uncertainty or unresolved heterogeneities through macrodispersion, which standard particle tracking neglects. The approach is tested on 2D and 3D idealized systems, as well as on several watershed-scale well fields within the Regional Municipality of Waterloo, Ontario, Canada.     

A feasibility study on the predictive emission monitoring system applied to the Hsinta power plant of Taiwan Power Company

The continuous emission monitoring system (CEMS) can monitor flue gas emissions continuously and instantaneously. However, it has the disadvantages of enormous cost, easily producing errors in sampling periods of bad weather, lagging response in variable ambient environments, and missing data in daily zero and span tests and maintenance. The concept of a predictive emission monitoring system (PEMS) is to use the operating parameters of combustion equipment through thermodynamic or statistical methods to construct a mathematic model that can predict emissions by a computer program. The goal of this study is to set up a PEMS in a gas-fired combined cycle power generation unit at the Hsinta station of Taiwan Power Co. The emissions to be monitored include nitrogen oxides (NOx) and oxygen (O2) in flue gas. The major variables of the predictive model were determined based on the combustion theory. The data of these variables then were analyzed to establish a regression model. From the regression results, the influences of these variables are discussed and the predicted values are compared with the CEMS data for accuracy. In addition, according to the cost information, the capital and operation and maintenance costs for a PEMS can be much lower than those for a CEMS.     

Receptor Model Applied to Patterns in Space (RMAPS)

Abstract

The physical and mathematical bases of a new receptor model based on spatially intensive data are presented. The model apportions the average concentration of a species as measured at many sites among several spatially distinct sources and can be applied to primary or secondary species. In the latter case, no assumptions concerning transformation or deposition rates are required. The methodology is a combination of the empirical orthogonal function approach that is well known in meteorology and the self-modeling multivariate modeling approach that has long been applied in chemometrics and multivariate receptor modeling of air quality data. A simple, geometrical example of the modeling approach is given.     

Sampled Monte Carlo uncertainty analysis for photochemical grid models

This study reports on the development and testing of a method of quantifying the uncertainties in concentration predictions by a complex photochemical grid model (PGM), using a modification of the basic Monte Carlo method (MCM). The computationally intensive aspects of applying a full MCM to hundreds of PGM inputs and model parameters is replaced by a highly restricted sampling approach that exploits the spatial persistence found in predicted concentration fields. The sampling approach to the MCM is being explored as an efficient approach to assess the uncertainty in the differences in predicted maximum ozone concentration between base case and control scenarios. The MCM is applied to several dozen surface cells, with the goal of sampling the spatial pattern of uncertainty in the PGM-predicted differences in surface ozone concentration fields between a pair of base and control scenarios. The uncertainty in model inputs and parameters is simulated using several types of stochastic models. These stochastic models are driven using Latin hypercube sampling (LHS) to generate a non-redundant ensemble of alternative model inputs. Preliminary testing of the sampled MCM approach was conducted using the UAM-IV PGM on the New York ozone attainment modeling domain for the 6–8 July 1988 ozone episode. One hundred alternative concentration estimates were generated for a base scenario and for control scenarios representing 50%, 10% and 5% reduction of NO_x emissions. The upper and lower bounds of the concentration difference ensemble that define a 95% confidence range were spatially interpolated from 27 monitoring sites to the full (surface) modeling domain, using the field of zero uncertainty (ZU) concentration differences. For the 50% NO_x control scenario, predicted increases in peak ozone concentration smaller than 20 ppb were generally not significant from zero. By contrast, predicted decreases in peak ozone greater than 10 ppb were usually significant. For a control scenario with a small 5% NO_x reduction, predicted concentration differences and confidence intervals were much smaller, but predicted changes in peak ozone were significant at a number of sample cells.     

Comparative statistical study of hourly precipitation determined by radar-based Stage IV and ground-based methods in the North Central United States

Detailed hourly precipitation data are required for long-range modeling of dispersion and wet deposition of particulate matter and water-soluble pollutants using the CALPUFF model. In sparsely populated areas such as the north central United States, ground-based precipitation measurement stations may be too widely spaced to offer a complete and accurate spatial representation of hourly precipitation within a modeling domain. The availability of remotely sensed precipitation data by satellite and the National Weather Service array of next-generation radars (NEXRAD) deployed nationally provide an opportunity to improve on the paucity of data for these areas. Before adopting a new method of precipitation estimation in a modeling protocol, it should be compared with the ground-based precipitation measurements, which are currently relied upon for modeling purposes. This paper presents a statistical comparison between hourly precipitation measurements for the years 2006 through 2008 at 25 ground-based stations in the north central United States and radar-based precipitation measurements available from the National Center for Environmental Predictions (NCEP) as Stage IV data at the nearest grid cell to each selected precipitation station. It was found that the statistical agreement between the two methods depends strongly on whether the ground-based hourly precipitation is measured to within 0.1 in/hr or to within 0.01 in/hr. The results of the statistical comparison indicate that it would be more accurate to use gridded Stage IV precipitation data in a gridded dispersion model for a long-range simulation, than to rely on precipitation data interpolated between widely scattered rain gauges.

Implications:

The current reliance on ground-based rain gauges for precipitation events and hourly data for modeling of dispersion and wet deposition of particulate matter and water-soluble pollutants results in potentially large discontinuity in data coverage and the need to extrapolate data between monitoring stations. The use of radar-based precipitation data, which is available for the entire continental United States and nearby areas, would resolve these data gaps and provide a complete and accurate spatial representation of hourly precipitation within a large modeling domain.     

Flow and transport in unsaturated fractured rock: effects of multiscale heterogeneity of hydrogeologic properties

The heterogeneity of hydrogeologic properties at different scales may have different effects on flow and transport processes in a subsurface system. A model for the unsaturated zone of Yucca Mountain, Nevada, is developed to represent complex heterogeneity at two different scales: (1) layer scale corresponding to geologic layering and (2) local scale. The layer-scale hydrogeologic properties are obtained using inverse modeling, based on the available measurements collected from the Yucca Mountain site. Calibration results show a significant lateral and vertical variability in matrix and fracture properties. Hydrogeologic property distributions in a two-dimensional, vertical cross-section of the site are generated by combining the average layer-scale matrix and fracture properties with local-scale perturbations generated using a stochastic simulation method. The unsaturated water flow and conservative (nonsorbing) tracer transport through the cross-section are simulated for different sets of matrix and fracture property fields. Comparison of simulation results indicates that the local-scale heterogeneity of matrix and fracture properties has a considerable effect on unsaturated flow processes, leading to fast flow paths in fractures and the matrix. These paths shorten the travel time of a conservative tracer from the source (repository) horizon in the unsaturated zone to the water table for small fractions of total released tracer mass. As a result, the local-scale heterogeneity also has a noticeable effect on global tracer transport processes, characterized by an average breakthrough curve at the water table, especially at the early arrival time of tracer mass. However, the effect is not significant at the later time after 20% tracer mass reaches the water table. The simulation results also verify that matrix diffusion plays an important role in overall solute transport processes in the unsaturated zone at Yucca Mountain.     

Modeling porosity reductions caused by mineral fouling in continuous-wall permeable reactive barriers

A study was conducted to assess key factors to include when modeling porosity reductions caused by mineral fouling in permeable reactive barriers (PRBs) containing granular zero valent iron. The public domain codes MODFLOW and RT3D were used and a geochemical algorithm was developed for RT3D to simulate geochemical reactions occurring in PRBs. Results of simulations conducted with the model show that the largest porosity reductions occur between the entrance and mid-plane of the PRB as a result of precipitation of carbonate minerals and that smaller porosity reductions occur between the mid-plane and exit face due to precipitation of ferrous hydroxide. These findings are consistent with field and laboratory observations, as well as modeling predictions made by others. Parametric studies were conducted to identify the most important variables to include in a model evaluating porosity reduction. These studies showed that three minerals (CaCO3, FeCO3, and Fe(OH)2 (am)) account for more than 99% of the porosity reductions that were predicted. The porosity reduction is sensitive to influent concentrations of HCO3-, Ca2+, CO3(2-), and dissolved oxygen, the anaerobic iron corrosion rate, and the rates of CaCO3 and FeCO3 formation. The predictions also show that porosity reductions in PRBs can be spatially variable and mineral forming ions penetrate deeper into the PRB as a result of flow heterogeneities, which reflects the balance between the rate of mass transport and geochemical reaction rates. Level of aquifer heterogeneity and the contrast in hydraulic conductivity between the aquifer and PRB are the most important hydraulic variables affecting porosity reduction. Spatial continuity of aquifer hydraulic conductivity is less significant.     

Prediction of dissolved oxygen concentration in hypoxic river systems using support vector machine: a case study of Wen-Rui Tang River,China

Accurate quantification of dissolved oxygen (DO) is critically important for managing water resources and controlling pollution. Artificial intelligence (AI) models have been successfully applied for modeling DO content in aquatic ecosystems with limited data. However, the efficacy of these AI models in predicting DO levels in the hypoxic river systems having multiple pollution sources and complicated pollutants behaviors is unclear. Given this dilemma, we developed a promising AI model, known as support vector machine (SVM), to predict the DO concentration in a hypoxic river in southeastern China. Four different calibration models, specifically, multiple linear regression, back propagation neural network, general regression neural network, and SVM, were established, and their prediction accuracy was systemically investigated and compared. A total of 11 hydro-chemical variables were used as model inputs. These variables were measured bimonthly at eight sampling sites along the rural-suburban-urban portion of Wen-Rui Tang River from 2004 to 2008. The performances of the established models were assessed through the mean square error (MSE), determination coefficient (R ²), and Nash-Sutcliffe (NS) model efficiency. The results indicated that the SVM model was superior to other models in predicting DO concentration in Wen-Rui Tang River. For SVM, the MSE, R ², and NS values for the testing subset were 0.9416 mg/L, 0.8646, and 0.8763, respectively. Sensitivity analysis showed that ammonium-nitrogen was the most significant input variable of the proposal SVM model. Overall, these results demonstrated that the proposed SVM model can efficiently predict water quality, especially for highly impaired and hypoxic river systems.     

Validation of regression models for nitrate concentrations in the upper groundwater in sandy soils

For Dutch sandy regions, linear regression models have been developed that predict nitrate concentrations in the upper groundwater on the basis of residual nitrate contents in the soil in autumn. The objective of our study was to validate these regression models for one particular sandy region dominated by dairy farming. No data from this area were used for calibrating the regression models. The model was validated by additional probability sampling. This sample was used to estimate errors in 1) the predicted areal fractions where the EU standard of 50 mg l⁻¹ is exceeded for farms with low N surpluses (ALT) and farms with higher N surpluses (REF); 2) predicted cumulative frequency distributions of nitrate concentration for both groups of farms.Both the errors in the predicted areal fractions as well as the errors in the predicted cumulative frequency distributions indicate that the regression models are invalid for the sandy soils of this study area.