T. Rajaee; R. Rahimi Benmaran
Abstract
IntroductionThe water quality is an issue of ongoing concern. Evaluation of the quantity and quality of running waters is considerable in hydro-environmental management.The prediction and control of the quality of Karaj river water, as one of the important needed water supply sources of Tehran, possesses ...
Read More
IntroductionThe water quality is an issue of ongoing concern. Evaluation of the quantity and quality of running waters is considerable in hydro-environmental management.The prediction and control of the quality of Karaj river water, as one of the important needed water supply sources of Tehran, possesses great importance. In this study, Performance of Artificial Neural Network (ANN), Wavelet Neural Network combination (WANN) and multi linear regression (MLR) models, to predict next month the Nitrate (NO3) and Chloride (CL) ions of "gate ofBylaqan sluice" station located in Karaj River has been evaluated.
Materials and MethodsIn this research two separate ANN models for prediction of NO3 and CL has been expanded. Each one of the parameters for prediction (NO3 / CL) has been put related to the past amounts of the same time series (NO3 / CL) and its amounts of Q in past months.From astatisticalperiod of10yearswas usedforthe input of the models. Hence 80% of entire data from (96 initial months of data) as training set, next 10% of data (12 months) and 10% of the end of time series (terminal 12 months) were considered as for validation and test of the models, respectively. In WANNcombination model, the real monthly observed time series of river discharge (Q) and mentioned qualityparameters(NO3 / CL) were decomposed to some sub-time series at different levels by wavelet analysis.Then the decomposed quality parameters to predict and Q time series were used at different levels as inputs to the ANN technique for predicting one-step-ahead Nitrate and Chloride. These time series play various roles in the original time series and the behavior of each is distinct, so the contribution to the original time series varies from each other. In addition, prediction of high NO3 and CL values greater than mean of data that have great importancewere investigated by the models. The capability of the models was evaluated by Coefficient of Efficiency (E) and the Root Mean Square Error (RMSE).An efficiency of one corresponds to an accurate match of forecasted data to the observed data. RMSE indicates the discrepancy between the observed and predicted values
Results Discussion The results indicates that the accuracy and the ability of hybrid model of wavelet neural network had been better than the other two modes; so that hybrid model of Wavelet artificial neural network was able the improve the rate of RMSE for Nitrate ions in comparison with ANN and MLR models respectively, amounting to 30.13% and 71.89%, for chloride ion as much as 31.3% and 57.1%. In the WANN model increasing the decomposition level, in level 1 to Level 3, increases the model’s performance, but increasing the decomposition level, in levels over Level 3, decreases the model’s efficiency, because high decomposition levels lead to a large number of parameters with complex nonlinear relationships in the ANN technique.The WANN model needed 1 to 7 neurons in the hidden layer for the best performance result. In prediction of high NO3 values the amount RMSE for ANN, MLR and WANN models are 1.487, 2.645 and 0.834 ppm, respectively. Also, for CL values the mentioned statistical parameter is 0.990, 3.003 and 0.188 ppm, respectively for models.The results exhibits that the combined model of WANN the forecast was better than the other two models.
Conclusion Wavelet transforms provide useful decompositions of original time series, so that wavelet-transformed data improve the ability of a predicting model by capturing useful information on various resolution levels. The main advantage of this study is that only from the Q and slightly quality of parameter time series are used until the same quality of parameter in one month ahead is predicted. The purpose of entering Q time series with quality of parameter as inputs of models is analysis the efficacy of Q in the accuracy of prediction. owing of the high capability wavelet neural network in the prediction of quality parameters of river's water, this model can be convenient and fast way to be proposed for management of water quality resources and assurance from water quality monitoring results and reduction its costs.
S. Babaei Hessar; R. Ghazavi
Abstract
Introduction: Precipitation is one of the most important and sensitive parameters of the tropical climate that influence the catchments hydrological regime. The prediction of rainfall is vital for strategic planning and water resources management. Despite its importance, statistical rainfall forecasting, ...
Read More
Introduction: Precipitation is one of the most important and sensitive parameters of the tropical climate that influence the catchments hydrological regime. The prediction of rainfall is vital for strategic planning and water resources management. Despite its importance, statistical rainfall forecasting, especially for long-term, has been proven to be a great challenge due to the dynamic nature of climate phenomena and random fluctuations involved in the process. Various methods, such as time series and artificial neural network models, have been proposed to predict the level of rainfall. But there is not enough attention to global warming and climate change issues. The main aim of this study is to investigate the conformity of artificial neural network and time series models with climate scenarios.
Materials and Methods: For this study, 50 years of daily rainfall data (1961 to 2010) of the synoptic station of Urmia, Tabriz and Khoy was investigated. Data was obtained from Meteorological Organization of Iran. In the present study, the results of two Artificial Neural Network (ANN) and Time Seri (TS) methods were compared with the result of the Emission Scenarios (A2 & B1). HadCM3 model in LARS-WG software was used to generate rainfall for the next 18 years (2011-2029). The results of models were compared with climate scenarios over the next 18 years in the three synoptic stations located in the basin of the Lake Urmia. At the first stage, the best model of time series method was selected. The precipitation was estimated for the next 18 years using these models. For the same period, precipitation was forecast using artificial neural networks. Finally, the results of two models were compared with data generated under two scenarios (B1 and A2) in LARS-WG.
Results and Discussion: Different order of AR, MA and ARMA was examined to select the best model of TS The results show that AR(1) was suitable for Tabriz and Khoy stations .In the Urmia station MA(1) was the best performance. Multiple Layer Perceptron with a 10 neurons in hidden layer and the output layer consists of five neurons had the lowest MSE and the highest correlation coefficient in modeling the values of annual precipitation. So MLP was determined as the best structure of neural network for rainfall prediction. According to results, precipitation predicted by the ANN model was very close to the results of A2 and B1 scenario, whereas TS has a significant difference with these scenarios. Average rainfall predicted by two A2 and B1 scenarios in Urmia station has more difference than other stations. Based on the B1 scenario, precipitation will increase 11 percent over the next two decades. It will decrease 10.7 percent according to A2 emissions scenario. According to ANN models and two A2 and B1 scenarios, the rates of rainfall will increase in Tabriz and Khoy stations. However, according to TS model, rainfall will decline 5.94 and 3.63 percent for these two stations, respectively.
Conclusion: Global warming and climate change should have adverse effects on groundwater and surface water resources. Different models are used for simulating of thes effects. But, conformity of these models with the results of climate scenarios is an issue that has not been addressed. In the present research coincidence of TS model, ANN model and climate change scenarios was investigated. Results show under emissions scenarios, during the next two decades in Tabriz and Khoy stations, precipitation will increase. In Urmia station B1 and A2 scenario percent increase by 11 percent and 10.5 percent decline predicted, respectively. The results of Roshan and et al (4) and Golmohammad and et al, (7) investigations show increasing trend in the rainfall rate and confirming the results of this study According to results, the performance of ANN model is better than TS model for rainfall prediction and its result is similar to climate change scenarios. Similar results have been reported by Wang et al (29) and the Norani et al (20). Due to the significant difference between the TS and climate scenarios used in the study area, is not recommended, though it can be used as a plausible climate scenario to predict the precipitation of stations in the future studied. At the end, it is suggested that the similar studies carried out in a larger number of stations in the country with respect to global warming and climate change, to determine the validity of the methods used to the predicted rainfall.
Farzaneh Nazarieh; H. Ansari
Abstract
Introduction: Rainfall is affected by changes in the global sea level change, especially changes in sea surface temperature SST Sea Surface Temperature and sea level pressure SLP Sea level Pressure. Climate anomalies being related to each other at large distance is called teleconnection. As physical ...
Read More
Introduction: Rainfall is affected by changes in the global sea level change, especially changes in sea surface temperature SST Sea Surface Temperature and sea level pressure SLP Sea level Pressure. Climate anomalies being related to each other at large distance is called teleconnection. As physical relationships between rainfall and teleconnection patterns are not defined clearly, we used intelligent models for forecasting rainfall. The intelligent models used in this study included Fuzzy Inference Systems, neural network and Neuro-fuzzy. In this study, first the teleconnection indices that could affect rainfall in the study area were identified. Then intelligent models were trained for rainfall forecasting. Finally, the most capable model for forecasting rainfall was presented. The study area for this research is the Khorasan Razavi Province. In order to present a model for rainfall forecasting, rainfall data of seven synoptic stations including Mashhad, Golmakan, Nishapur, Sabzevar, Kashmar, Torbate and Sharks since 1991 to 2010 were used.
Materials and Methods: Based on previous studies about Teleconnection Patterns in the study area, effective Teleconnection indexes were identified. After calculating the correlation between the identified teleconnection indices and rainfall in one, two and three months ahead for all stations, fourteen teleconnection indices were chosen as inputs for intelligent models. These indices include, SLP Adriatic , SLP northern Red Sea, SLP Mediterranean Sea, SLP Aral sea, SST Sea surface temperature Labrador sea, SST Oman Sea, SST Caspian Sea, SST Persian Gulf, North Pacific pattern, SST Tropical Pacific in NINO12 and NINO3 regions, North Pacific Oscillation, Trans-Nino Index, Multivariable Enso Index. Inputs of the intelligent models include fourteen teleconnection indices, latitude and altitude of each station and their outputs are the prediction of rainfall for one, two and three months ahead. For calibration of the models, eighty percent of the data belonged to six stations. Mashhad, Golmakan, Sabzevar, Kashmar, Torbate and Sarakhs were used. Verification of the model was carried out in two parts. The first part of verification was done with twenty percent of the remaining data which belonged to the mentioned six stations. The second part of verification was done with data from the Nishapur station. Nishapur geographically is located between other stations and did not participate in the calibration. So, it provides a ondition for assessing models in location except for the calibration stations. To assess and compare the accuracy of the models, the following statistical criteria have been used: correlation coefficient (R), normal root mean square error (NRMSE), mean bias error (MBE), Jacovides criteria (t), and ratio (R2/t). To evaluate models in different rainfall depths, rainfall data based on standard precipitation index (SPI) was divided into seven classes, and the accuracy of each class was calculated separately.
Results and Discussion: By comparing the models' ability to predict rainfall according to the R2 /t criteria it can be concluded that the ranking of the models is Neuro-fuzzy model, Fuzzy Inference Systems, and Neural network, respectively. R2 /to criteria for prediction of rainfall one, two, and three month earlier in the Neuro-fuzzy model are 0.91, 0.4, 0.36, in Fuzzy Inference Systems are 0.76, 0.38, 0.31 and in the neural network model are 0.43,0.27, 0.2. The statistical criteria of Neuro-fuzzy model (R, MBE, NRMSE, t, R2/ t) for rainfall forecasting one month earlier are 0.8, -0.55,0.43, 0.7 , 0.91; two months earlier are 0.79, -1.32, 0.48, 1.56, 0.4; and three months earlier are 0.73,-1.37, 0.54, 1.47, 0.36 . Calculation of MBE criteria for Neuro-fuzzy models in all classes of SPI indicated that this model has a lower estimate in extremely wet and very wet classes. This is because of lack of data belonging to these classes for model training.
Conclusion: The results of this research showed that teleconnection indices are suitable inputs for intelligent models for rainfall prediction. Computing the best structure of fuzzy, neural network and Neuro-fuzzy models showed that Neuro-fuzzy can predict rainfall the most accurately. But, the results of these models in very wet and extremely wet condition are not reliable .So, these models should be used with more caution in these conditions.
H. Mir; Ahmad Gholamalizadeh Ahangar; A. Shabani
Abstract
Introduction: Phosphorus is important as an essential element in the production of agricultural products. On the other hand, its ability to induce essential micronutrient deficiency and its negative effects on the environment, have attracted more attention to this element. The knowledge of phosphorus ...
Read More
Introduction: Phosphorus is important as an essential element in the production of agricultural products. On the other hand, its ability to induce essential micronutrient deficiency and its negative effects on the environment, have attracted more attention to this element. The knowledge of phosphorus availability conditions in the soil and consequently the accurate management of fertilizer consumption has a key role in the environmental protection. The degree of phosphorus absorption in the soil depends on the environmental factors, soil characteristics and compositions, and phosphorus fertilizer which have been used. The amount of available phosphorus in the soil has relationship with some of the physical and chemical properties of the soil. Since, the soil characteristics are important factors in the reaction of phosphorus in the soil, the present study aimed to investigate and determine the most important soil characteristics affecting the availability of phosphorus using regression and artificial neural network techniques, in the soils of Sistan plain.
Materials and Methods: Soil sampling was done in 1.5×1.5 km intervals, from 0-30 cm depth, and 200 soil samples were taken. The amounts of available phosphorus and the soil properties including the percentages of clay , organic matter, calcium carbonate and the amount of pH were measured. Then, stepwise multivariate linear regression analysis was performed to determine linear relation between available phosphorus and the soil properties. In order to model and validate the regression model, respectively 80 and 20% of data were selected and entered into SPSS software. To train the neural network, multilayer perceptron (MLP) network was used by MATLAB 7.6 package. In this type of network, 70% of data is selected for training, 15% for validation and 15% for testing the model. Levenberg-Marquardt algorithm and hyperbolic tangent (as a transfer function) were used to train the network. The numbers of neurons in the hidden layer were calculated based on the trial and error method and finally the best structure was selected according to the highest R2 and the lowest RMSE value. Moreover, quantifying the importance of variables in the neural network was done through employing connection weight approach. In this method, the connection weights of input-hidden and hidden-output neurons were used to indicate the significance of variables.
Results and Discussion: The values of the coefficient of variation for the soil properties were in the range of 5.66 for pH (the lowest) and 69.90 for available phosphorus (the highest). The high variation of the available phosphorus could be due to the different amounts of phosphorus fertilizers consumption and their diverse rate of conversion to less soluble forms. The validation results of regression and neural network methods showed that the latter technique was more accurate compared with the multivariate linear regression method, in the estimation of available phosphorus, as multi-layer perceptron neural network with 4-6-1 layout predicts nearly 90% of available phosphorus variability using soil properties (percentage of clay, organic matter, calcium carbonate and the amount of pH); however, the obtained regression equation could explain only 43% of phosphorus variances. The reasons for this could be: 1) considering nonlinear relations between the variables in the artificial neural network method, and 2) less sensitivity of this method to the existence of error in input data, comparing with the regression method. The values of R2 and RMSE were 0.43 and 11.23, respectively for training the multivariate linear regression method and 0.91 and 4.28, respectively for training the artificial neural network method. From the investigated soil properties in the current study, the percentage of organic matter and clay were entered in the regression model, and the values of standardized regression coefficient (beta) showed that the first variable is more important to explain the available phosphorus variability. The results of quantifying the importance of variables by the connection weight method showed that pH had the greatest contribution in the variability of phosphorus in the study area. In the other words, the high values of pH were the most important limiting factor for the availability of phosphorus in Sistan soils.
Conclusion: Considering nonlinear and complicated relations between variables, the artificial neural network model is an effective tool to assess the effect of soil properties on the availability of phosphorus in the study area. The results of quantifying the importance of variables by using the connection weight method showed that pH had the greatest contribution in the variability of phosphorus in the study area. In fact, the existence of lime in the soils of the study area, arid climate and lack of precipitation have resulted in the accumulation of basic cations in the soil and consequently increased pH values. Furthermore, the observed average values of pH that are more than 8.5 demonstrated the risk of soil sodicity in the study area. Thus, the management of this area by cultivating tolerant plants could be resulted in increasing organic matter content, which along with using chemical amendments such as sulfur will decrease pH values and increase the availability of phosphorus in Sistan plain. Applying such practices and through it modifying soil characteristics, decreasing the consumption of phosphate fertilizers and preventing their hazardous environmental effects would be expected in long run.
H. Ghorbani; A. Roohani; N. Hafezi Moghaddas
Abstract
In this research, a learning vector quantization neural network (LVQ) model was developed to predict and classify the spatial distribution of cadmium in soil in Golestan province. The cadmium data were obtained from soils measuring total Cd contents in soil samples. Some statistical tests, such as means ...
Read More
In this research, a learning vector quantization neural network (LVQ) model was developed to predict and classify the spatial distribution of cadmium in soil in Golestan province. The cadmium data were obtained from soils measuring total Cd contents in soil samples. Some statistical tests, such as means comparision, variance and statistical distribution were performed between the observed points samples data and the estimated cadmium values to evaluate the performance of the pattern recognition method. The Results showed that in training and test phase, there were no significant differences, with the confidence level of 95%, between the statsitcal parameters such as average, variance, statistical distribution and also coefficient of determination in the observed and the estimated cadmium concentrations. The results suggest that learning vector quantization (LVQ) neural network can learn cadmium cocentration model precisely. In addition the results also indicated that trained LVQ neural network had a high capability in predicting cadmium concentrations for non-sampled points. The technique showed that the LVQNN could predict and map the spatial cadmium concentrations variability. Our results indicated that it is possible to discriminate different cadmium levels in soil, using LVQNN.
