Artificial neural network modeling of dissolved oxygen in reservoir

The water quality of reservoirs is one of the key factors in the operation and water quality management of reservoirs. Dissolved oxygen (DO) in water column is essential for microorganisms and a significant indicator of the state of aquatic ecosystems. In this study, two artificial neural network (ANN) models including back propagation neural network (BPNN) and adaptive neural-based fuzzy inference system (ANFIS) approaches and multilinear regression (MLR) model were developed to estimate the DO concentration in the Feitsui Reservoir of northern Taiwan. The input variables of the neural network are determined as water temperature, pH, conductivity, turbidity, suspended solids, total hardness, total alkalinity, and ammonium nitrogen. The performance of the ANN models and MLR model was assessed through the mean absolute error, root mean square error, and correlation coefficient computed from the measured and model-simulated DO values. The results reveal that ANN estimation performances were superior to those of MLR. Comparing to the BPNN and ANFIS models through the performance criteria, the ANFIS model is better than the BPNN model for predicting the DO values. Study results show that the neural network particularly using ANFIS model is able to predict the DO concentrations with reasonable accuracy, suggesting that the neural network is a valuable tool for reservoir management in Taiwan.     

Soil water content forecasting by ANN and SVM hybrid architecture

Soil water content prediction is essential to the development of advanced agriculture information systems. Because soil water content series are inherently noise and non-stationary, it is difficult to get an accurate forecasting result. Considering the problems, in this paper, a novel hybrid learning architecture is proposed according to divide-and-conquer principle, the forecasting accuracy is improved. This novel hierarchical architecture is composed of ANN (Kohonen neural network) and SVM (support vector machine). The Kohonen network is used as a classifier, which partitions the whole input space into several distinct feature regions. Then, the best SVM predictor combined with an appropriate kernel function can be achieved for correspondence regions. The experimental results based on the hybrid model exhibit good agreement with actual soil water content measurements and outperform ANN and SVM single-stage models.     

Artificial neural network modeling of dissolved oxygen in the Heihe River, Northwestern China

Identification and quantification of dissolved oxygen (DO) profiles of river is one of the primary concerns for water resources managers. In this research, an artificial neural network (ANN) was developed to simulate the DO concentrations in the Heihe River, Northwestern China. A three-layer back-propagation ANN was used with the Bayesian regularization training algorithm. The input variables of the neural network were pH, electrical conductivity, chloride (Cl^?), calcium (Ca²⁺), total alkalinity, total hardness, nitrate nitrogen (NO₃-N), and ammonical nitrogen (NH₄-N). The ANN structure with 14 hidden neurons obtained the best selection. By making comparison between the results of the ANN model and the measured data on the basis of correlation coefficient (r) and root mean square error (RMSE), a good model-fitting DO values indicated the effectiveness of neural network model. It is found that the coefficient of correlation (r) values for the training, validation, and test sets were 0.9654, 0.9841, and 0.9680, respectively, and the respective values of RMSE for the training, validation, and test sets were 0.4272, 0.3667, and 0.4570, respectively. Sensitivity analysis was used to determine the influence of input variables on the dependent variable. The most effective inputs were determined as pH, NO₃-N, NH₄-N, and Ca²⁺. Cl^? was found to be least effective variables on the proposed model. The identified ANN model can be used to simulate the water quality parameters.     

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.     

Transition index maps for urban growth simulation: application of artificial neural networks,weight of evidence and fuzzy multi-criteria evaluation

Transition index maps (TIMs) are key products in urban growth simulation models. However, their operationalization is still conflicting. Our aim was to compare the prediction accuracy of three TIM-based spatially explicit land cover change (LCC) models in the mega city of Mumbai, India. These LCC models include two data-driven approaches, namely artificial neural networks (ANNs) and weight of evidence (WOE), and one knowledge-based approach which integrates an analytical hierarchical process with fuzzy membership functions (FAHP). Using the relative operating characteristics (ROC), the performance of these three LCC models were evaluated. The results showed 85%, 75%, and 73% accuracy for the ANN, FAHP, and WOE. The ANN was clearly superior compared to the other LCC models when simulating urban growth for the year 2010; hence, ANN was used to predict urban growth for 2020 and 2030. Projected urban growth maps were assessed using statistical measures, including figure of merit, average spatial distance deviation, producer accuracy, and overall accuracy. Based on our findings, we recomend ANNs as an and accurate method for simulating future patterns of urban growth.     

Forest cover change and fragmentation using Landsat data in Maçka State Forest Enterprise in Turkey

In this study, Grey model (GM) and artificial neural network (ANN) were employed to predict suspended solids (SSeff) and chemical oxygen demand (CODeff) in the effluent from a wastewater treatment plant in industrial park of Taiwan. When constructing model or predicting, the influent quality or online monitoring parameters were adopted as the input variables. ANN was also adopted for comparison. The results indicated that the minimum MAPEs of 16.13 and 9.85% for SSeff and CODeff could be achieved using GMs when online monitoring parameters were taken as the input variables. Although a good fitness could be achieved using ANN, they required a large quantity of data. Contrarily, GM only required a small amount of data (at least four data) and the prediction results were even better than those of ANN. Therefore, GM could be applied successfully in predicting effluent when the information was not sufficient. The results also indicated that these simple online monitoring parameters could be applied on prediction of effluent quality well.     

WEPP and ANN models for simulating soil loss and runoff in a semi-arid Mediterranean region

This paper presents the use of both the Water Erosion Prediction Project (WEPP) and the artificial neural network (ANN) for the prediction of runoff and soil loss in the central highland mountainous of the Palestinian territories. Analyses show that the soil erosion is highly dependent on both the rainfall depth and the rainfall event duration rather than on the rainfall intensity as mostly mentioned in the literature. The results obtained from the WEPP model for the soil loss and runoff disagree with the field data. The WEPP underestimates both the runoff and soil loss. Analyses conducted with the ANN agree well with the observation. In addition, the global network models developed using the data of all the land use type show a relatively unbiased estimation for both runoff and soil loss. The study showed that the ANN model could be used as a management tool for predicting runoff and soil loss.     

Predicting copper concentrations in acid mine drainage: a comparative analysis of five machine learning techniques

Acid mine drainage (AMD) is a global problem that may have serious human health and environmental implications. Laboratory and field tests are commonly used for predicting AMD, however, this is challenging since its formation varies from site-to-site for a number of reasons. Furthermore, these tests are often conducted at small-scale over a short period of time. Subsequently, extrapolation of these results into large-scale setting of mine sites introduce huge uncertainties for decision-makers. This study presents machine learning techniques to develop models to predict AMD quality using historical monitoring data of a mine site. The machine learning techniques explored in this study include artificial neural networks (ANN), support vector machine with polynomial (SVM-Poly) and radial base function (SVM-RBF) kernels, model tree (M5P), and K-nearest neighbors (K-NN). Input variables (physico-chemical parameters) that influence drainage dynamics are identified and used to develop models to predict copper concentrations. For these selected techniques, the predictive accuracy and uncertainty were evaluated based on different statistical measures. The results showed that SVM-Poly performed best, followed by the SVM-RBF, ANN, M5P, and KNN techniques. Overall, this study demonstrates that the machine learning techniques are promising tools for predicting AMD quality.     

Improving Estimation of Specific Surface Area by Artificial Neural Network Ensembles Using Fractal and Particle Size Distribution Curve Parameters as Predictors

Specific surface area (SSA) is one of the principal soil properties used in modeling soil processes. In this study, artificial neural network (ANN) ensembles were evaluated to predict SSA. Complete soil particle-size distribution was estimated from sand, silt, and clay fractions using the model by Skaggs et al. and then the particle-size distribution curve parameters (PSDCPs) and fractal parameters were calculated. The PSDCPs were used to predict 20 particle-size classes for a soil sample’s particle size distribution. Fractal parameters were calculated by the model of Bird et al. In addition, total soil-specific surface area (TSS) was calculated using the above 20 size classes. Pedotransfer functions were developed for SSA and TSS using ANN ensembles from 63 pieces of SSA data taken from the literature. Fractal parameters, PSDCPs, and some other soil properties were used to predict SSA and TSS. Introducing fractal parameters and PSDCPs improved the SSA estimations by 12.5 and 11.1 %, respectively. The improvements were even better for TSS estimations (27.7 and 27.0 %, respectively). The use of fractal parameters as estimators described 44 and 92.8 % of the variation in SSA and TSS, respectively, while PSDCPs explained 42 and 6.6 % of the variation in SSA and TSS, respectively. The results suggested that fractal parameters and PSDCPs could be successfully used as predictors in ANN ensembles to predict SSA and TSS.     

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.     

Modeling of nitrate concentration in groundwater using artificial intelligence approach—a case study of Gaza coastal aquifer

Nitrate concentration in groundwater is influenced by complex and interrelated variables, leading to great difficulty during the modeling process. The objectives of this study are (1) to evaluate the performance of two artificial intelligence (AI) techniques, namely artificial neural networks and support vector machine, in modeling groundwater nitrate concentration using scant input data, as well as (2) to assess the effect of data clustering as a pre-modeling technique on the developed models' performance. The AI models were developed using data from 22 municipal wells of the Gaza coastal aquifer in Palestine from 2000 to 2010. Results indicated high simulation performance, with the correlation coefficient and the mean average percentage error of the best model reaching 0.996 and 7 %, respectively. The variables that strongly influenced groundwater nitrate concentration were previous nitrate concentration, groundwater recharge, and on-ground nitrogen load of each land use land cover category in the well's vicinity. The results also demonstrated the merit of performing clustering of input data prior to the application of AI models. With their high performance and simplicity, the developed AI models can be effectively utilized to assess the effects of future management scenarios on groundwater nitrate concentration, leading to more reasonable groundwater resources management and decision-making     

Artificial neural network based carbon monoxide persistence models for episodic urban air quality management

This paper describes the development of artificial neural network (ANN) based carbon monoxide (CO) persistence (ANNCOP) models to forecast 8-h average CO concentration using 1-h maximum predicted CO data for the critical (winter) period (November–March). The models have been developed for three 8-h groupings of 10 p.m. to 6 a.m., 6 a.m. to 2 p.m. and 2–10 p.m., at two air quality control regions (AQCRs) in Delhi city, representing an urban intersection and an arterial road consisting heterogeneous traffic flows. The result indicates that time grouping of 2–10 pm is dominantly affected by inversion conditions and peak traffic flow. The ANNCOP model corresponding to this grouping predicts the 8-h average CO concentrations within the accuracy range of 68–71%. The CO persistence values derived from ANNCOP model are comparable with the persistence values as suggested by the Environmental Protection Agency (EPA), USA. This work demonstrates that ANN based model is capable of describing winter period CO persistence phenomena.     

Multilayer perceptron neural network for flow prediction

Artificial neural networks (ANNs) have proven to be a tool for characterizing, modeling and predicting many of the non-linear hydrological processes such as rainfall-runoff, groundwater evaluation or simulation of water quality. After proper training they are able to generate satisfactory predictive results for many of these processes. In this paper they have been used to predict 1 or 2 days ahead the average and maximum daily flow of a river in a small forest headwaters in northwestern Spain. The inputs used were the flow and climate data (precipitation, temperature, relative humidity, solar radiation and wind speed) as recorded in the basin between 2003 and 2008. Climatic data have been utilized in a disaggregated form by considering each one as an input variable in ANN(1), or in an aggregated form by its use in the calculation of evapotranspiration and using this as input variable in ANN(2). Both ANN(1) and ANN(2), after being trained with the data for the period 2003-2007, have provided a good fit between estimated and observed data, with R(2) values exceeding 0.95. Subsequently, its operation has been verified making use of the data for the year 2008. The correlation coefficients obtained between the data estimated by ANNs and those observed were in all cases superior to 0.85, confirming the capacity of ANNs as a model for predicting average and maximum daily flow 1 or 2 days in advance.     

基于遥感影像的木兰溪水质参数反演

水质监测对水环境评价及污染预防至关重要,但地面监测成本高、监测面积有限等,难以满足实时、大范围监测的要求。为了更好地解决该问题,基于遥感影像的空中监测技术越来越得到研究人员的青睐。以木兰溪为研究区,利用和地面监测数据同步的Landsat-8卫星遥感影像数据,对木兰溪的典型水质参数总磷、总氮、溶解氧、高锰酸盐指数的反演问题进行研究。首先,根据Landsat-8的水体敏感波段,分别选取总磷、总氮、溶解氧、高锰酸盐指数的反演特征波段组合为(b1-b2)/(b2+b3),(b1-b2)/(b3-b4),b2/(b1+b4),b1/b2;其次,利用反演特征波段组合分别构建总磷、总氮、溶解氧、高锰酸盐指数浓度的SVR(Support Vector Regression)反演模型,通过IPSO算法对SVR模型的参数进行优选;然后,将IPSO-SVR反演模型和统计回归反演模型、广义回归神经网络(GRNN)反演模型在验证集上进行评估,以平均绝对误差和均方根误差作为评价指标进行对比分析,结果表明IPSO-SVR反演模型的平均绝对误差和均方根误差最小,说明IPSO-SVR反演模型具有较高的精度和较好的实用性...     

Use of a GIS-based hybrid artificial neural network to prioritize the order of pipe replacement in a water distribution network

A methodology based on the integration of a seismic-based artificial neural network (ANN) model and a geographic information system (GIS) to assess water leakage and to prioritize pipeline replacement is developed in this work. Qualified pipeline break-event data derived from the Taiwan Water Corporation Pipeline Leakage Repair Management System were analyzed. “Pipe diameter,” “pipe material,” and “the number of magnitude-3^?+? earthquakes” were employed as the input factors of ANN, while “the number of monthly breaks” was used for the prediction output. This study is the first attempt to manipulate earthquake data in the break-event ANN prediction model. Spatial distribution of the pipeline break-event data was analyzed and visualized by GIS. Through this, the users can swiftly figure out the hotspots of the leakage areas. A northeastern township in Taiwan, frequently affected by earthquakes, is chosen as the case study. Compared to the traditional processes for determining the priorities of pipeline replacement, the methodology developed is more effective and efficient. Likewise, the methodology can overcome the difficulty of prioritizing pipeline replacement even in situations where the break-event records are unavailable.     

Retrieval of total suspended matter (TSM) and chlorophyll-a (Chl-a) concentration from remote-sensing data for drinking water resources

The concentrations of chlorophyll-a (Chl-a) and total suspended matter (TSM) are major water quality parameters that can be retrieved using remotely sensed data. Water sampling works were conducted on 15 July 2007 and 13 September 2008 concurrent with the Indian Remote-Sensing Satellite (IRS-P6) overpass of the Shitoukoumen Reservoir. Both empirical regression and back-propagation artificial neural network (ANN) models were established to estimate Chl-a and TSM concentration with both in situ and satellite-received radiances signals. It was found that empirical models performed well on the TSM concentration estimation with better accuracy (R ² = 0.94, 0.91) than their performance on Chl-a concentration (R ² = 0.62, 0.75) with IRS-P6 imagery data, and the models accuracy marginally improved with in situ spectra data. Our results indicated that the ANN model performed better for both Chl-a (R ² = 0.91, 0.82) and TSM (R ² = 0.98, 0.94) concentration estimation through in situ collected spectra; the same trend followed for IRS-P6 imagery data (R ² = 0.75 and 0.90 for Chl-a; R ² = 0.97 and 0.95 for TSM). The relative root mean square errors (RMSEs) from the empirical model for TSM (Chl-a) were less than 15% (respectively 27.2%) with both in situ and IRS-P6 imagery data, while the RMSEs were less than 7.5% (respectively 18.4%) from the ANN model. Future work still needs to be undertaken to derive the dynamic characteristic of Shitoukoumen Reservoir water quality with remotely sensed IRS-P6 or Landsat-TM data. The algorithms developed in this study will also need to be tested and refined with more imagery data acquisitions combined with in situ spectra data.     

Dynamic regression modeling of daily nitrate-nitrogen concentrations in a large agricultural watershed

Nitrate-nitrogen concentrations in rivers represent challenges for water supplies that use surface water sources. Nitrate concentrations are often modeled using time-series approaches, but previous efforts have typically relied on monthly time steps. In this study, we developed a dynamic regression model of daily nitrate concentrations in the Raccoon River, Iowa, that incorporated contemporaneous and lags of precipitation and discharge occurring at several locations around the basin. Results suggested that 95 % of the variation in daily nitrate concentrations measured at the outlet of a large agricultural watershed can be explained by time-series patterns of precipitation and discharge occurring in the basin. Discharge was found to be a more important regression variable than precipitation in our model but both regression parameters were strongly correlated with nitrate concentrations. The time-series model was consistent with known patterns of nitrate behavior in the watershed, successfully identifying contemporaneous dilution mechanisms from higher relief and urban areas of the basin while incorporating the delayed contribution of nitrate from tile-drained regions in a lagged response. The first difference of the model errors were modeled as an AR(16) process and suggest that daily nitrate concentration changes remain temporally correlated for more than 2 weeks although temporal correlation was stronger in the first few days before tapering off. Consequently, daily nitrate concentrations are non-stationary, i.e. of strong memory. Using time-series models to reliably forecast daily nitrate concentrations in a river based on patterns of precipitation and discharge occurring in its basin may be of great interest to water suppliers.     

Exploring the Relationship Between Hydrologic Parameters and Nutrient Loads Using Digital Elevation Model and GIS – A Case Study from Sugarcreek Headwaters, Ohio, U.S.A.

Ohio is typical among the Midwestern and Eastern United States with high levels of water pollutants, the main sources being from agriculture. In this study, we used a digital elevation model in conjunction with hydrological indices to determine the role of landscape complexity affecting the spatial and temporal variation in pollutant levels, in one of the most impaired headwater streams in Ohio. More than eighty five percent of the study area is dominated by agriculture. Spatial distribution of slope (S), altitude and wetness index along with other watershed parameters such as flow direction, flow accumulation, stream networks, flow stream orders and erosion index were used within a Geographic Information Systems framework to quantify variation in nitrate and phosphate loads to headwater streams. Stream monitoring data for nutrient loads were used to correlate the observed spatial and temporal patterns with hydrological parameters using multiple linear regressions. Results from the wetness index calculated from a digital elevation model suggested a range of 0.10–16.39, with more than 35% having values less than 4.0. A Revised Universal Soil Loss Equation (RUSLE) predicted soil loss in the range of 0.01–4.0 t/ha/yr. Nitrate nitrogen levels in the study area paralleled precipitation patterns over time, with higher nitrate levels corresponding to high precipitation. Atmospheric deposition through precipitation could explain approximately 35% of total nitrate levels observed in streams. Among the different topographic variables and hydrological indices, results from the step-wise multiple regression suggested the following best predictors, (1) elevation range and upstream flow length for nitrate, (2) flow direction and upstream flow length for ammonia-nitrogen and slope, and (3) elevation range for phosphate levels. Differences in the landscape models observed for nitrate, phosphate and ammonia-nitrogen in the surface waters were attributed partly to differences in the chemical activity and source strengths of the different forms of these nutrients through agricultural management practices. The results identify geomorphologic and landscape characteristics that influence pollutant levels in the study area.     

Estimating soil temperature using neighboring station data via multi-nonlinear regression and artificial neural network models

The aim of this study is to estimate the soil temperatures of a target station using only the soil temperatures of neighboring stations without any consideration of the other variables or parameters related to soil properties. For this aim, the soil temperatures were measured at depths of 5, 10, 20, 50, and 100 cm below the earth surface at eight measuring stations in Turkey. Firstly, the multiple nonlinear regression analysis was performed with the “Enter” method to determine the relationship between the values of target station and neighboring stations. Then, the stepwise regression analysis was applied to determine the best independent variables. Finally, an artificial neural network (ANN) model was developed to estimate the soil temperature of a target station. According to the derived results for the training data set, the mean absolute percentage error and correlation coefficient ranged from 1.45% to 3.11% and from 0.9979 to 0.9986, respectively, while corresponding ranges of 1.685–3.65% and 0.9988–0.9991, respectively, were obtained based on the testing data set. The obtained results show that the developed ANN model provides a simple and accurate prediction to determine the soil temperature. In addition, the missing data at the target station could be determined within a high degree of accuracy.