Evaluation of artificial neural networks for fine particulate pollution (PM10 and PM2.5) forecasting

Multi-layer perceptron (MLP) artificial neural network (ANN) models are compared with traditional multiple regression (MLR) models for daily maximum and average O3 and particulate matter (PM10 and PM2.5) forecasting. MLP particulate forecasting models show little if any improvement over MLR models and exhibit less skill than do O3 forecasting models. Meteorological variables (precipitation, wind, and temperature), persistence, and co-pollutant data are shown to be useful PM predictors. If MLP approaches are adopted for PM forecasting, training methods that improve extreme value prediction are recommended.     

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.     

Multilayer perceptron neural network-based approach for modeling phycocyanin pigment concentrations: case study from lower Charles River buoy,USA

Environmental Science and Pollution Research - This paper proposes multilayer perceptron neural network (MLPNN) to predict phycocyanin (PC) pigment using water quality variables as predictor. In...     

Modelling of dissolved oxygen content using artificial neural networks: Danube River, North Serbia, case study

The aims of this study are to create an artificial neural network (ANN) model using non-specific water quality parameters and to examine the accuracy of three different ANN architectures: General Regression Neural Network (GRNN), Backpropagation Neural Network (BPNN) and Recurrent Neural Network (RNN), for prediction of dissolved oxygen (DO) concentration in the Danube River. The neural network model has been developed using measured data collected from the Bezdan monitoring station on the Danube River. The input variables used for the ANN model are water flow, temperature, pH and electrical conductivity. The model was trained and validated using available data from 2004 to 2008 and tested using the data from 2009. The order of performance for the created architectures based on their comparison with the test data is RNN > GRNN > BPNN. The ANN results are compared with multiple linear regression (MLR) model using multiple statistical indicators. The comparison of the RNN model with the MLR model indicates that the RNN model performs much better, since all predictions of the RNN model for the test data were within the error of less than ±10 %. In case of the MLR, only 55 % of predictions were within the error of less than ±10 %. The developed RNN model can be used as a tool for the prediction of DO in river waters.     

Impacts of Meteorological Factors on PM10: Artificial Neural Networks (ANN) and Multiple Linear Regression (MLR) Approaches

In this study, prediction capacities of multi-linear regression (MLR) and artificial neural networks (ANN) onto coarse particulate matter (PM₁₀) concentrations were investigated. Different meteorological factors on particulate pollution were chosen for operating variables in the model analyses. Two different regions (urban and industrial) were identified in the region of Kocaeli, Turkey. All data sets were obtained from air quality monitoring network of the Ministry of Environment and Urban Planning, and 120 data sets were used in the MLR and ANN models. Regression equations explained the effects of the meteorological factors in MLR analyses. In the ANN model, backpropagation network with two hidden layers has achieved the best prediction efficiency. Determination coefficients and error values were examined for each model. ANN models displayed more accurate results compared to MLR.     

Performance of ANFIS versus MLP-NN dissolved oxygen prediction models in water quality monitoring

We discuss the accuracy and performance of the adaptive neuro-fuzzy inference system (ANFIS) in training and prediction of dissolved oxygen (DO) concentrations. The model was used to analyze historical data generated through continuous monitoring of water quality parameters at several stations on the Johor River to predict DO concentrations. Four water quality parameters were selected for ANFIS modeling, including temperature, pH, nitrate (NO₃) concentration, and ammoniacal nitrogen concentration (NH₃-NL). Sensitivity analysis was performed to evaluate the effects of the input parameters. The inputs with the greatest effect were those related to oxygen content (NO₃) or oxygen demand (NH₃-NL). Temperature was the parameter with the least effect, whereas pH provided the lowest contribution to the proposed model. To evaluate the performance of the model, three statistical indices were used: the coefficient of determination (R ²), the mean absolute prediction error, and the correlation coefficient. The performance of the ANFIS model was compared with an artificial neural network model. The ANFIS model was capable of providing greater accuracy, particularly in the case of extreme events.     

Spatial scale and seasonal dependence of land use impacts on riverine water quality in the Huai River basin,China

Land use pattern is an effective reflection of anthropic activities, which are primarily responsible for water quality deterioration. A detailed understanding of relationship between water quality and land use is critical for effective land use management to improve water quality. Linear mixed effects and multiple regression models were applied to water quality data collected from 2003 to 2010 from 36 stations in the Huai River basin together with topography and climate data, to characterize the land use impacts on water quality and their spatial scale and seasonal dependence. The results indicated that the influence of land use categories on specific water quality parameter was multiple and varied with spatial scales and seasons. Land use exhibited strongest association with dissolved oxygen (DO) and ammonia nitrogen (NH₃-N) concentrations at entire watershed scale and with total phosphorus (TP) and fluoride concentrations at finer scales. However, the spatial scale, at which land use exerted strongest influence on instream chemical oxygen demand (COD) and biochemical oxygen demand (BOD) levels, varied with seasons. In addition, land use composition was responsible for the seasonal pattern observed in contaminant concentrations. COD, NH₃-N, and fluoride generally peaked during dry seasons in highly urbanized regions and during rainy seasons in less urbanized regions. High proportion of agricultural and rural areas was associated with high nutrient contamination risk during spring. The results highlight the spatial scale and seasonal dependence of land use impacts on water quality and can provide scientific basis for scale-specific land management and seasonal contamination control.     

Application of soft computing to predict water quality in wetland

Prediction of water quality is a critical issue because of its significant impact on human and ecosystem health. This research aims to predict water quality index (WQI) for the free surface wetland using three soft computing techniques namely, adaptive neuro-fuzzy system (ANFIS), artificial neural networks (ANNs), and group method of data handling (GMDH). Seventeen wetland points for a period of 14 months were considered for monitoring water quality parameters including conductivity, suspended solid (SS), biochemical oxygen demand (BOD), ammoniacal nitrogen (AN), chemical oxygen demand (COD), dissolved oxygen (DO), temperature, pH, phosphate nitrite, and nitrate. The sensitivity analysis performed by ANFIS indicates that the significant parameters to predict WQI are pH, COD, AN, and SS. The results indicated that ANFIS with Nash-Sutcliffe Efficiency (NSE = 0.9634) and mean absolute error (MAE = 0.0219) has better performance to predict the WQI comparing with ANNs (NSE = 0.9617 and MAE = 0.0222) and GMDH (NSE = 0.9594 and MAE = 0.0245) models. However, ANNs provided a comparable prediction and the GMDH can be considered as a technique with an acceptable prediction for practical purposes. The findings of this study could be used as an effective reference for policy makers in the field of water resource management. Decreasing variables, reduction of running time, and high speed of these approaches are the most important reasons to employ them in any aquatic environment worldwide.

     

Organic pollution of the Büyük Menderes River,Turkey and effects on aquaculture

Water quality was measured at eight stations on the Buyuk Menderes River in Turkey (Ad?güzel dam, Yenice regülator, Sarayköy bridge, Feslek regülator, Yenipazar bridge, Ayd?n bridge, Koçarl? bridge, Söke regülator) between 2000 and 2013 in February, April, June, August, October and December. The resulting data were evaluated in terms of biochemical oxygen demand (BOD), chemical oxygen demand (COD), dissolved oxygen (DO), ammoniac–nitrogen (NH₃–N), nitrite–nitrogen (NO₂–N), nitrate–nitrogen (NO₃–N) and orthophosphate (o-PO₄) aquaculture. According to the analysis, while river water pollution generally varied during each year, samples from certain measurement points demonstrated high pollution levels throughout the year. In this study, water parameters were classified according to the “Turkish water pollution control regulation.” The studied parameters are also compared with TS 266 and WHO guidelines. While levels at Sarayköy station were generally higher than other stations, values at Ad?güzel dam were the lowest, giving it the best water quality of the eight stations. The highest values on a yearly basis were obtained in 2007 due to the severe drought in the Menderes basin within which irrigation water levels fell to 4255 m³/ha. The BOD, COD levels are the lowest in 2009 and highest in 2007; the DO level is lowest in 2007 and highest in 2009; NH₃–N, NO₂–N, and NO₃–N parameters are the lowest in 2007 and highest in 2009; and the o-PO₄ are at the lowest level in 2004 and seen as the highest in 2007. Analysis of the data was performed by SPSS 21 statistics program. One direction ANOVA was applied to the data, which were also subject to Tukey multiple comparison tests. Differences between groups were evaluated at p?<?0.05. Box–plot graphs were used to demonstrate the data distribution. In the study, it was analyzed, the effect of fish species and pollution involved in the Büyük Menderes River so far on fish species.     

Assessment and rationalization of water quality monitoring network: a multivariate statistical approach to the Kabbini River (India)

The establishment of an efficient surface water quality monitoring (WQM) network is a critical component in the assessment, restoration and protection of river water quality. A periodic evaluation of monitoring network is mandatory to ensure effective data collection and possible redesigning of existing network in a river catchment. In this study, the efficacy and appropriateness of existing water quality monitoring network in the Kabbini River basin of Kerala, India is presented. Significant multivariate statistical techniques like principal component analysis (PCA) and principal factor analysis (PFA) have been employed to evaluate the efficiency of the surface water quality monitoring network with monitoring stations as the evaluated variables for the interpretation of complex data matrix of the river basin. The main objective is to identify significant monitoring stations that must essentially be included in assessing annual and seasonal variations of river water quality. Moreover, the significance of seasonal redesign of the monitoring network was also investigated to capture valuable information on water quality from the network. Results identified few monitoring stations as insignificant in explaining the annual variance of the dataset. Moreover, the seasonal redesign of the monitoring network through a multivariate statistical framework was found to capture valuable information from the system, thus making the network more efficient. Cluster analysis (CA) classified the sampling sites into different groups based on similarity in water quality characteristics. The PCA/PFA identified significant latent factors standing for different pollution sources such as organic pollution, industrial pollution, diffuse pollution and faecal contamination. Thus, the present study illustrates that various multivariate statistical techniques can be effectively employed in sustainable management of water resources. Highlights ? The effectiveness of existing river water quality monitoring network is assessed ? Significance of seasonal redesign of the monitoring network is demonstrated ? Rationalization of water quality parameters is performed in a statistical framework     

Macroinvertebrate community as a biological indicator of ecological and toxicological factors in Lake Saint-François (Québec)

To assess the potential of the macroinvertebrate community for monitoring variation in the environmental quality of large rivers, the response of littoral macrobenthos in Lake Saint-Fran?ois, a fluvial lake of the St Lawrence River (Québec) are described. First, the composition of total macroinvertebrate communities and important taxonomic groups as well as the biotic ICI-SL index in 16 littoral stations varying in sedimentology, water chemistry and contamination are described to define indicator species groups and environmental quality ranks. Thereafter, the relative contribution of ecological and toxicological factors in explaining the variation observed in macroinvertebrate assemblages and biotic index were quantified using partial canonical correspondence analysis. Cluster analyses based on taxonomic composition separated five groups of stations where macroinvertebrate assemblages varied in density, composition and tolerance to pollution. The ICI-SL biotic index varied from 7.2 to 27.2 with a mean value of 19 +/- 6. The ICI-SL values determined for the macroinvertebrate communities in Lake Saint-Fran?ois did not reflect an important deterioration in environmental quality, and there was some agreement between the environmental quality ranking of the stations expressed either by the ICI-SL index or the community cluster analysis. Water conductivity and phosphorus concentration, followed by macrophyte types (Chara, Ceratophyllum) and sediment grain size, were the most significant ecological variables to explain variation in macroinvertebrate communities and derived ICI-SL index in Lake Saint-Fran?ois. Among the toxicological factors, metals in water (Fe, Cr, Pb, Mn, Zn) and sediment (Mn, Pb, Se), as well as the composite indices of metal and organic contamination (water CI, sediment CI, sediment total PAHs) were the most important factors. The contamination factors selected in our models represented contaminant sorption processes rather than direct toxicological effects. The lack of clear relationships between contaminants and macroinvertebrate variables reflected the relative low level of contamination in the stations sampled in Lake Saint-Fran?ois. There were some interactions between toxicological and ecological variables that should be considered in the planning of sampling and interpretation of biomonitoring studies. However, the large amount of unexplained variance (49.2-86.6%) in the CCA models underlined the limitations of the use of the indices of macroinvertebrate community structure that were assessed in this study for biomonitoring purposes in the absence of a contrasting pollution gradient.     

Correlations between fine particulate matter (PM2.5) and meteorological variables in the United States: Implications for the sensitivity of PM2.5 to climate change

We applied a multiple linear regression (MLR) model to study the correlations of total PM_2.5 and its components with meteorological variables using an 11-year (1998–2008) observational record over the contiguous US. The data were deseasonalized and detrended to focus on synoptic-scale correlations. We find that daily variation in meteorology as described by the MLR can explain up to 50% of PM_2.5 variability with temperature, relative humidity (RH), precipitation, and circulation all being important predictors. Temperature is positively correlated with sulfate, organic carbon (OC) and elemental carbon (EC) almost everywhere. The correlation of nitrate with temperature is negative in the Southeast but positive in California and the Great Plains. RH is positively correlated with sulfate and nitrate, but negatively with OC and EC. Precipitation is strongly negatively correlated with all PM_2.5 components. We find that PM_2.5 concentrations are on average 2.6 μg m^?3 higher on stagnant vs. non-stagnant days. Our observed correlations provide a test for chemical transport models used to simulate the sensitivity of PM_2.5 to climate change. They point to the importance of adequately representing the temperature dependence of agricultural, biogenic and wildfire emissions in these models.     

Effects of seasonal succession and water pollution on the protozoan community structure in an eutrophic lake

The purpose of the research is to study the seasonal succession of protozoa community and the effect of water quality on the protozoa community to characterize biochemical processes occurring at a eutrophic Lake Donghu, a large shallow lake in Wuhan City, China. Samples of protozoa communities were obtained monthly at three stations by PFU (polyurethane foam unit) method over a year. Synchronously, water samples also were taken from the stations for the water chemical quality analysis. Six major variables were examined in a principal component analysis (PCA), which indicate the fast changes of water quality in this station I and less within-year variation and a comparatively stable water quality in stations II and III. The community data were analyzed using multivariate techniques, and we show that clusters are rather mixed and poorly separated, suggesting that the community structure is changing gradually, giving a slight merging of clusters form the summer to the autumn and the autumn to the winter. Canonical correspondence analysis (CCA) was used to infer the relationship between water quality variables and phytoplankton community structure, which changed substantially over the survey period. From the analysis of cluster and CCA, coupled by community pollution value (CPV), it is concluded that the key factors driving the change in protozoa community composition in Lake Donghu was water qualities rather than seasons.     

Forecasting summertime surface-level ozone concentrations in the Lower Fraser Valley of British Columbia: an ensemble neural network approach

Empirical models for predicting daily maximum hourly average ozone concentrations were developed for 10 monitoring stations in the Lower Fraser Valley (LFV) of British Columbia. According to data from 1991 to 1996, ensemble neural network models increased explained variance an average of 7% over multiple linear regression models using the same input variables. Without modification, all models performed poorly on days when the observed peak ozone concentration exceeded 82 parts per billion, the National Ambient Air Quality Objective. When numbers of extreme events in training data were increased using a histogram equalization process, models were able to forecast exceedances with improved accuracy. Modified generalized additive model (GAM) plots and associated measures of input variable importance and interaction were generated for a subset of the trained models and used to investigate relationships between input variables and ozone levels. The neural network models displayed a high degree of interaction among inputs, and it is likely the ability of these model types to account for interactions, rather than the nonlinearity of individual input variables, that explains their improved forecast skill. Inspection of GAM-style plots indicated that the relative importance of input variables in the ensemble neural network models varied with geographic location within the LFV. Four distinct groups of stations were identified, and rankings of inputs within the groups were generally consistent with physical intuition and results of prior studies.     

Stream biodiversity and monitoring in North Central,Nigeria: the use of macroinvertebrate indicator species as surrogates

Indicator species (IS) have been employed in modern aquatic research for monitoring of environmental changes and evaluating the efficiency of environmental management procedures. In this study, we evaluated the possibility of developing surrogate indicator groups as tools for the conservation and management of the biodiversity of Northern Nigeria streams by surveying 15 streams in Niger state for benthic macroinvertebrates and environmental variables as data sets, over a period of 24 months (2016 and 2017). Samples were collected in two locations of reference and impacted sites for each of the streams surveyed. The statistically significant (P < 0.05; based on 1000 permutations) indicator species for each of the status classes (reference versus impacted) was identified using the indicator species analysis/indicator value (Indval) method. Canonical correspondence analysis (CCA) was used to evaluate the IS-environment relationships. Indicator value found fifteen species for the reference streams including Ephemeropteran (Bugilliesia sp., Tricorythus sp., Thraulus sp., Crassabwa sp.) and the Tricopteran (Leptonema sp.). Opposite, the Indval found seven (7) indicator species for the impacted streams, which included the Dipteran (Pentaneura sp., Tabanus sp.). Multivariate analysis revealed that species assemblage had wide dispersal patterns in relation to the sites for both status classes. CCA revealed that the reference and impacted indicator species responded to entirely different environmental factors, indicating their preference to particular environmental variables along the ecological gradients. While the indicator species of reference sites were associated with environmental predictors of good water quality such as high DO, increased flow, low conductivity, and low BOD, the indicator species of impacted sites were strongly related to environmental predictors of anthropogenic pollution, including low DO, high BOD, and increased nutrients concentrations. This study has provided a reference point and effective tool to monitor environmental changes, community, and ecosystem dynamics across the Northern Nigeria streams.

     

A new hybrid system of QSAR models for predicting bioconcentration factors (BCF)

The aim was to develop a reliable and practical quantitative structure–activity relationship (QSAR) model validated by strict conditions for predicting bioconcentration factors (BCF). We built up several QSAR models starting from a large data set of 473 heterogeneous chemicals, based on multiple linear regression (MLR), radial basis function neural network (RBFNN) and support vector machine (SVM) methods. To improve the results, we also applied a hybrid model, which gave better prediction than single models. All models were statistically analysed using strict criteria, including an external test set. The outliers were also examined to understand better in which cases large errors were to be expected and to improve the predictive models. The models offer more robust tools for regulatory purposes, on the basis of the statistical results and the quality check on the input data.     

Evaluation of natural and anthropogenic influences on the Guadalquivir River (Spain) by dissolved heavy metals and nutrients

To evaluate both the natural and anthropogenic influences on surface waters of Guadalquivir River (SW-Spain), concentrations of dissolved trace metals (Mn, Co, Ni, Cu, Cd, Zn, and Pb), inorganic nutrients (N-NH(4)(+), N-NO(3)(-), N-NO(2)(-), and P-PO(4)(3-)) and other variables as conductivity, pH, dissolved oxygen (DO) and suspended solids (SS) were measured during a three-years period (2001-2003). Samples were taken at 26 stations twice a year, during rain and dry periods. The analysis of variance (ANOVA) suggested that temporal variations within the period of study were statistically negligible. Spatial distributions identified three different zones, mainly influenced by sewage (Eastern Zone), agriculture runoffs (Central Zone), and estuarine processes (Western Zone), respectively. Principal Component and Cluster Analysis allowed to identify the variables controlling the water quality of each zone as: N-NH(4)(+), N-NO(2)(-), Mn, and Co, (Eastern Zone), SS, and P-PO(4)(3-) (Central Zone), and Cd, pH and conductivity (Western Zone). Other variables such as Ni, Cu, Zn or N-NO(3)(-), influenced two different zones, while Pb presented a singular behavior.     

A new hybrid system of QSAR models for predicting bioconcentration factors (BCF)

The aim was to develop a reliable and practical quantitative structure-activity relationship (QSAR) model validated by strict conditions for predicting bioconcentration factors (BCF). We built up several QSAR models starting from a large data set of 473 heterogeneous chemicals, based on multiple linear regression (MLR), radial basis function neural network (RBFNN) and support vector machine (SVM) methods. To improve the results, we also applied a hybrid model, which gave better prediction than single models. All models were statistically analysed using strict criteria, including an external test set. The outliers were also examined to understand better in which cases large errors were to be expected and to improve the predictive models. The models offer more robust tools for regulatory purposes, on the basis of the statistical results and the quality check on the input data.