Modelling geosmin concentrations in three sources of raw water in Quebec, Canada

The presence of off-flavour compounds such as geosmin, often found in raw water, significantly reduces the organoleptic quality of distributed water and diverts the consumer from its use. To adapt water treatment processes to eliminate these compounds, it is necessary to be able to identify them quickly. Routine analysis could be considered a solution, but it is expensive and delays associated with obtaining the results of analysis are often important, thereby constituting a serious disadvantage. The development of decision-making tools such as predictive models seems to be an economic and feasible solution to counterbalance the limitations of analytical methods. Among these tools, multi-linear regression and principal component regression are easy to implement. However, due to certain disadvantages inherent in these methods (multicollinearity or non-linearity of the processes), the use of emergent models involving artificial neurons networks such as multi-layer perceptron could prove to be an interesting alternative. In a previous paper (Parinet et al., Water Res 44: 5847-5856, 2010), the possible parameters that affect the variability of taste and odour compounds were investigated using principal component analysis. In the present study, we expand the research by comparing the performance of three tools using different modelling scenarios (multi-linear regression, principal component regression and multi-layer perceptron) to model geosmin in drinking water sources using 38 microbiological and physicochemical parameters. Three very different sources of water, in terms of quality, were selected for the study. These sources supply drinking water to the Québec City area (Canada) and its vicinity, and were monitored three times per month over a 1-year period. Seven different modelling methods were tested for predicting geosmin in these sources. The comparison of the seven different models showed that simple models based on multi-linear regression provide sufficient predictive capacity with performance levels comparable to those obtained with artificial neural networks. The multi-linear regression model (R ²?=?0.657, <0.001) used only four variables (phaeophytin, sum of green algae, chlorophyll-a and potential Redox) in comparison with ten variables (potassium, heterotrophic bacteria, organic nitrogen, total nitrogen, phaeophytin, total organic carbon, sum of green algae, potential Redox, UV absorbance at 254 nm and atypical bacteria) for the best model obtained with artificial neural networks (R ²?=?0.843).     

Developing a fuzzy neural network-based support vector regression (FNN-SVR) for regionalizing nitrate concentration in groundwater

The aim of this study is to develop a fuzzy neural network-based support vector regression model (FNN-SVR) for mapping crisp-input and fuzzy-output variables. In this model, an artificial neural network (ANN) estimator based on multilayer perceptron (MLP) is considered as the kernel function of the SVR, whereas asymmetric triangular fuzzy H-level sets are assumed for model parameters including weight and biases of the ANN model. A genetic algorithm (GA) with real coding is implemented to optimize the model parameters during the training phase. To evaluate the efficiency and applicability of the proposed model, it is applied for simulating and regionalizing nitrate concentration in Karaj Aquifer in Iran. The goodness-of-fit criteria indicate a better performance of the FNN-SVR compared to some benchmark models such as geostatistic techniques as well as traditional SVR models with linear, quadratic, polynomial, and Gaussian kernel functions for modeling nitrate concentrations in groundwater.     

Development of GP and GEP models to estimate an environmental issue induced by blasting operation

Air overpressure (AOp) is one of the most adverse effects induced by blasting in the surface mines and civil projects. So, proper evaluation and estimation of the AOp is important for minimizing the environmental problems resulting from blasting. The main aim of this study is to estimate AOp produced by blasting operation in Miduk copper mine, Iran, developing two artificial intelligence models, i.e., genetic programming (GP) and gene expression programming (GEP). Then, the accuracy of the GP and GEP models has been compared to multiple linear regression (MLR) and three empirical models. For this purpose, 92 blasting events were investigated, and subsequently, the AOp values were carefully measured. Moreover, in each operation, the values of maximum charge per delay and distance from blast points, as two effective parameters on the AOp, were measured. After predicting by the predictive models, their performance prediction was checked in terms of variance account for (VAF), coefficient of determination (CoD), and root mean square error (RMSE). Finally, it was found that the GEP with VAF of 94.12%, CoD of 0.941, and RMSE of 0.06 is a more precise model than other predictive models for the AOp prediction in the Miduk copper mine, and it can be introduced as a new powerful tool for estimating the AOp resulting from blasting.     

Selecting optimal hyperspectral bands to discriminate nitrogen status in durum wheat: a comparison of statistical approaches

Hyperspectral data can provide prediction of physical and chemical vegetation properties, but data handling, analysis, and interpretation still limit their use. In this study, different methods for selecting variables were compared for the analysis of on-the-ground hyperspectral signatures of wheat grown under a wide range of nitrogen supplies. Spectral signatures were recorded at the end of stem elongation, booting, and heading stages in 100 georeferenced locations, using a 512-channel portable spectroradiometer operating in the 325–1075-nm range. The following procedures were compared: (i) a heuristic combined approach including lambda-lambda R² (LL R²) model, principal component analysis (PCA), and stepwise discriminant analysis (SDA); (ii) variable importance for projection (VIP) statistics derived from partial least square (PLS) regression (PLS-VIP); and (iii) multiple linear regression (MLR) analysis through maximum R-square improvement (MAXR) and stepwise algorithms. The discriminating capability of selected wavelengths was evaluated by canonical discriminant analysis. Leaf-nitrogen concentration was quantified on samples collected at the same locations and dates and used as response variable in regressive methods. The different methods resulted in differences in the number and position of the selected wavebands. Bands extracted through regressive methods were mostly related to response variable, as shown by the importance of the visible region for PLS and stepwise. Band selection techniques can be extremely useful not only to improve the power of predictive models but also for data interpretation or sensor design.     

A Comparison Between Regression Models and Genetic Programming for Predictions of Chlorophyll-a Concentrations in Northern Lakes

Chlorophyll-a (chl-a) concentrations are often used as a proxy for water quality problems as well as phytoplankton blooms. Available chl-a models range from simple phosphorus loading models to complex regression and dynamic models. A comparison of multiple regression models was made with genetic programming (GP) techniques to predict chl-a concentrations over a large range of 104 Swedish lakes. Independent variables used were lake area, mean depth, iron, latitude, ammonium, nitrogen + nitrate, pH, phosphate, secchi depth, silicon, temperature, total phosphorus, total nitrogen and total organic carbon. GP is a method based on the Darwinian evolution theory. This implies that a program will be able to test different mathematical equations, iterating and improving each equation using fundamental ideas from evolution theory to increase the predictive power. A good correspondence was found between the multiple regression and the GP modelling approach. No significant improvement of the predictive power was found using GP, and it is therefore recommended that multiple regression methods should be preferred when predicting chl-a concentrations as these models tend to be less complex and the modelling approach is easier to use. Results from GP were in some cases more accurate compared to multiple regressions; however, the best model was created by multiple regressions which used concentrations of total phosphorus, total nitrogen and latitude as independent variables. These findings will be an important note for limnologists and modelling managers when developing future models of chl-a concentrations in lakes.     

Modeling of SO<Subscript>2</Subscript> distribution in Istanbul using artificial neural networks

Urban air pollution is a growing problem in developing countries. Some compounds especially sulphur dioxide (SO₂) is considered as typical indicators of the urban air quality. Air pollution modeling and prediction have great importance in preventing the occurrence of air pollution episodes and provide sufficient time to take the necessary precautions. Recently, various stochastic image-processing algorithms such as Artificial Neural Network (ANN) are applied to environmental engineering. ANN structure employs input, hidden and output layers. Due to the complexity of the problem, as the number of input–output parameters differs, ANN model settings such as the number of neurons of these layers changes. The ability of ANN models to learn, particularly capability of handling large amounts (or sets) of data simultaneously as well as their fast response time, are invariably the characteristics desired for predictive and forecasting purposes. In this paper, ANN models have been used to predict air pollutant parameter in meteorological considerations. We have especially focused on modeling of SO₂ distribution and predicting its future concentration in Istanbul, Turkey. We have obtained data sets including meteorological variables and SO₂ concentrations from Istanbul-Florya meteorological station and Istanbul-Yenibosna air pollution station. We have preferred three-layer perceptron type of ANN which consists of 10, 22 and 1 neurons for input, hidden and output layers, respectively. All considered parameters are measured as daily mean. The input parameters are: SO₂ concentration, pressure, temperature, humidity, wind direction, wind speed, strength of sunshine, sunshine, cloudy, rainfall and output parameter is the future prediction of SO₂. To evaluate the performance of ANN model, our results are compared to classical nonlinear regression methods. The over all system finds an optimum correlation between input–output variables. Here, the correlation parameter, r is 0.999 and 0.528 for training and test data. Thus in our model, the trend of SO₂ is well estimated and seasonal effects are well represented. As a result, we conclude that ANN is one of the compromising methods in estimation of environmental complex air pollution problems.     

Development of a Standardised Approach to River Habitat Assessment in Australia

Despite the demonstrated utility of the Australian River Assessment Scheme (AUSRIVAS) to provide national-scale information on the biological condition of rivers, there is no commensurate scheme that can provide standardised information on physical habitat. Existing habitat assessment methods are not suitable for implementation on a national scale, so we present a new habitat assessment protocol that incorporates favorable elements of existing methods. Habitat Predictive Modelling forms the basis for the protocol because it can predict the occurrence of local-scale features from large-scale data, uses the reference condition concept, can be modified to incorporate a range of biologically and geomorphologically relevant variables, and employs a rapid survey approach. However, the protocol has been augmented with geomorphological variables and incorporates principles of hierarchy and geomorphological river zonation. There are four sequential components to the implementation of the protocol: reference site selection, data collection, predictive model construction and assessment of test sites using the predictive models. Once implemented, the habitat assessment protocol will provide a standardised tool for the assessment of river habitat condition at a variety of governance levels.     

The effect of autocorrelated errors on change-detection statistics

In this paper, regression models with error terms generated by lower order ARMA schemes are analyzed. Methods are discussed for estimating the parameters of the regression coefficients and the ARMA processes. The problem of detecting changes in the regression parameters is considered. A change-detection statistic proposed by MacNeill (1978) for regression problems is modified for application to ARMA processes. The effect of autocorrelated errors on this statistic is briefly discussed.     

NARX神经网络模型在昆明市环境空气质量预测中的应用

神经网络模型作为一种重要的手段被广泛应用于数学计算、物理建模、水文模拟、环境预测、人工智能等研究领域。为验证神经网络模型在高原山地城市环境空气质量预测中的作用,以昆明市环境空气自动监测站气象因子和污染物浓度数据为基础,构建NARX神经网络模型,对污染物浓度进行预测。结果表明,基于NARX神经网络建立的预测模型具有很强的非线性动态描述能力,能够对环境空气6参数做出较为准确的预测,其预测浓度相对误差显著低于CMAQ、NAQPMS空气质量数值模式以及LSTM统计模型预测结果。优化后的NARX神经网络对污染物浓度变化趋势的预测较其他几个模式更为敏感。     

Minimax Statistical Models for Air Pollution Time Series. Application to Ozone Time Series Data Measured in Bordeaux

This paper deals with the application of l(infinity) (or minimax) optimization techniques to statistical modelling of high frequency air pollution data. The method was applied to ground-level ozone time-series data measured in Bordeaux over 4 years from 1998 to 2001. The aim of model building was to develop predictive models in order to provide forecasts of the maximal daily ground-level ozone concentration. Experimental results from this case study indicate that such techniques could be more appropriate than the commonly used l2 setting if only good estimation of high levels is of interest. When the free parameters are fitted by means of l(infinity) optimization techniques, the forecasting errors are more evenly distributed amongst the data points, resulting in a better estimation of high values. The paper compares the quality of forecasts produced by both a linear and a nonlinear model, using l2 and l(infinity) parameter optimization.     

Using GA-Ridge regression to select hydro-geological parameters influencing groundwater pollution vulnerability

For groundwater conservation and management, it is important to accurately assess groundwater pollution vulnerability. This study proposed an integrated model using ridge regression and a genetic algorithm (GA) to effectively select the major hydro-geological parameters influencing groundwater pollution vulnerability in an aquifer. The GA-Ridge regression method determined that depth to water, net recharge, topography, and the impact of vadose zone media were the hydro-geological parameters that influenced trichloroethene pollution vulnerability in a Korean aquifer. When using these selected hydro-geological parameters, the accuracy was improved for various statistical nonlinear and artificial intelligence (AI) techniques, such as multinomial logistic regression, decision trees, artificial neural networks, and case-based reasoning. These results provide a proof of concept that the GA-Ridge regression is effective at determining influential hydro-geological parameters for the pollution vulnerability of an aquifer, and in turn, improves the AI performance in assessing groundwater pollution vulnerability.     

Precipitation Modeling by Polyhedral RCMARS and Comparison with MARS and CMARS

Climate change is becoming an ever important issue due to the possibility that it may result in extreme weather events such as floods or droughts. Consequently, precipitation forecasting has similarly gained in significance as it is a useful tool in meeting the increasing need for the efficient management of water resources as well as in preventing disasters before they happen. In the literature, there are various statistical and computational methods used for this purpose, including linear and nonlinear regression, kriging, time series models, neural networks, and multivariate adaptive regression splines (MARS). Among them, MARS stands out as the better performing precipitation modeling method. In this article, we used a recently developed method called robust conic mars (RCMARS), based on MARS (also on CMARS), to forecast precipitation owing to its ability to model complex uncertain data. In CMARS, which was developed as a powerful alternative to MARS, the model complexity is penalized in the form of Tikhonov regularization and studied as a conic quadratic programming. In RCMARS, on the other hand, CMARS is refined further by including the existence of uncertainty in the future scenarios and robustifying it with a robust optimization technique. To evaluate the performance of the RCMARS method, it was applied to build a precipitation model constructed as an early warning system for the continental Central Anatolia Region of Turkey, where drought has been a recurrent phenomenon for the last few decades. Furthermore, the performance of the RCMARS precipitation model was also compared to that of MARS and CMARS. The results indicated that RCMARS builds more accurate, precise, and stable precipitation models compared to those of MARS and CMARS. In addition to these advantageous features of the RCMARS precipitation model, it also provided a good fit to the data. As a result, we propose its use in precipitation forecasting for the region studied.     

Wasteload Allocation Using Wastewater Treatment and Flow Augmentation

A Wasteload allocation model, named Cost-Flow-Augmentation Model involving wastewater treatment and flow augmentation as a method of pollution abatement has been developed. The cost functions for wastewater treatment were developed as power functions of biochemical oxygen demand (BOD) removal using the regression module of the SPSS10 software. The cost function for flow augmentation was also developed using a regression between cost of dam/barrage and corresponding flow released from upstream reservoir for downstream water quality improvement. The response of wasteloads and flow augmentation on the water quality was quantified in terms of transfer coefficient calculated using the QUAL2E water quality simulation model. The performance of these models is demonstrated on the 22-km-long Delhi stretch of river Yamuna, India. Optimal solutions of the formulated models were obtained using the Web-based interactive non-differentiable interactive multiobjective bundle-based optimization system software. The optimal solutions obtained reveal that flow augmentation is not an economically feasible pollution abatement option for the Delhi stretch of river Yamuna.     

Hybrid Neural Networks and Boosted Regression Tree Models for Predicting Roadside Particulate Matter

This paper examines the application of artificial neural network (ANN) and boosted regression tree (BRT) methods in air quality modelling. The methods were applied to developing air quality models for predicting roadside particle mass concentration (PM₁₀, PM_2.5) and particle number counts (PNC) based on air pollution, traffic and meteorological data from Marylebone Road in London. Elastic net, Lasso and principal components analysis were used as feature selection methods for the ANN models to reduce the number of predictor variables and improve their generalisation. The performance of the ANN with feature selection (ANN hybrid) and the BRT models was evaluated and compared using statistical performance metrics. The performance parameters include root mean square error (RMSE), fraction of prediction within a factor of two of the observation (FAC2), mean bias (MB), mean gross error (MGE), the coefficient of correlation (R) and coefficient of efficiency (CoE) values. The input variables selected by the elastic net produced the best performing ANN models. The ANN hybrid produced models performed only slightly better than the BRT models. The R values of the ANN elastic net and BRT models were 0.96 and 0.95 for PM₁₀, 0.96 and 0.96 for PM_2.5 and 0.89 and 0.87 for PNC, respectively. Their corresponding CoE values were 0.72 and 0.70 for PM₁₀, 0.74 and 0.76 for PM_2.5 and 0.81 and 0.71 for PNC respectively. About 80–99% of all the model predictions are within a factor of two of the observed particle concentrations. The BRT models offer more advantages regarding model interpretation and permit feature selection. Therefore, the study recommends the use of BRT over ANN where the model interpretation is a priority.     

Simultaneous spectrophotometric determination of trace amount of polycyclic aromatic hydrocarbons in water samples after magnetic solid-phase extraction by using projection pursuit regression

Magnetic solid-phase extraction based on coated nano-magnets Fe₃O₄ was applied for the preconcentration of four polycyclic aromatic hydrocarbons (PAHs; anthracene, phenanthrene, fluorine, and pyrene) in environmental water samples prior to simultaneous spectrophotometric determination using multivariate calibration method. Magnetic nanoparticles, carrying target metals, were easily separated from the aqueous solution by applying an external magnetic field so, no filtration or centrifugation was necessary. After elution of the adsorbed PAHs, the concentration of PAHs was determined spectrophotometrically with the aid of a new and efficient multivariate spectral analysis base on principal component analysis-projection pursuit regression, without separation of analytes. The obtained results revealed that using projection pursuit regression as a flexible modeling approach improves the predictive quality of the developed models compared with partial least squares and least squares support vector machine methods. The method was used to determine four PAHs in environmental water samples.     

Prediction of lichen diversity in an UNESCO biosphere reserve – correlation of high resolution remote sensing data with field samples

The present study focuses on developing models to predict lichen species richness in a UNESCO Biosphere Reserve of the Swiss Pre-Alps following a gradient of land-use intensity combining remote sensing data and regression models. The predictive power of the models and the obtained r ranging from 0.5 for lichens on soil to 0.8 for lichens on trees can be regarded as satisfactory to good, respectively. The study revealed that a combination of airborne and spaceborne remote sensing data produced a variety of ecological meaningful variables.     

A comparison of radiometric correction techniques in the evaluation of the relationship between LST and NDVI in Landsat imagery

Atmospheric corrections for multi-temporal optical satellite images are necessary, especially in change detection analyses, such as normalized difference vegetation index (NDVI) rationing. Abrupt change detection analysis using remote-sensing techniques requires radiometric congruity and atmospheric correction to monitor terrestrial surfaces over time. Two atmospheric correction methods were used for this study: relative radiometric normalization and the simplified method for atmospheric correction (SMAC) in the solar spectrum. A multi-temporal data set consisting of two sets of Landsat images from the period between 1991 and 2002 of Penang Island, Malaysia, was used to compare NDVI maps, which were generated using the proposed atmospheric correction methods. Land surface temperature (LST) was retrieved using ATCOR3_T in PCI Geomatica 10.1 image processing software. Linear regression analysis was utilized to analyze the relationship between NDVI and LST. This study reveals that both of the proposed atmospheric correction methods yielded high accuracy through examination of the linear correlation coefficients. To check for the accuracy of the equation obtained through linear regression analysis for every single satellite image, 20 points were randomly chosen. The results showed that the SMAC method yielded a constant value (in terms of error) to predict the NDVI value from linear regression analysis-derived equation. The errors (average) from both proposed atmospheric correction methods were less than 10%.     

样本优化核主元分析及其在水质监测中的应用

核主元分析(KPCA)方法通过核变换将输入空间映射到高维特征空间,在特征空间进行主元分析。由于KPCA不适合大样本数据建模与分析,因此建模数据的选取非常重要,合理的数据样本可以简化运算,提高核主元分析的诊断准确度。文章提出一种优化数据样本的KPCA方法,利用相似度函数的方法实现样本优化,再建立核主元分析模型,提取数据特征信息,并将该方法应用到水环境监测的传感器故障诊断中,通过试验分析,验证了该方法的有效性。