artificial neural network

Iron is an essential element in the growth process of plants and plays a crucial role in chlorophyll production. Iron deficiency is a serious limitation in vineyards that can significantly affect both the yield and the quality of the crop. The use of modern methods such as digital image processing not only increases precision but also reduces the need for costly and time-consuming laboratory testing, thereby lowering costs and speeding up data-driven decision-making processes in orchard management. The aim of this study is to develop a system based on image processing and neural networks to estimate the active iron content in grape leaves. For this purpose, 55 leaf samples with different levels of iron deficiency were collected and analyzed from vineyards around Urmia. The total and active iron content in the samples was measured using atomic absorption spectroscopy and the leaves were photographed and processed under controlled light conditions. Statistical features were extracted from the images and their correlation with active and total iron content was analyzed. Finally, the best features were used to predict iron content using a multilayer artificial neural network. The results of the linear regression show that active iron correlates with the R, G, H, and S color channels with coefficients of 0.64, 0.58, 0.54, and 0.45, respectively, and that total iron does not correlate with the changes in leaf color. The neural network with an optimized structure of 8-9-1 was able to predict the data from the atomic absorption device with an accuracy of 0.83, 0.88, and 0.84 for training, test, and all data, respectively. In summary, image processing can be effectively and reliably used as a tool for optimal plant nutrition management and rapid diagnosis of iron deficiency.

The use of geospatial techniques for mapping soils is broadly covered by the term digital soil mapping (DSM). Soil maps have considerable significance as basic maps in many environmental and natural resources studies. Digital soil maps are based on the relationship between environmental variables and soil properties. With the development of computers and technology, digital and quantitative approaches have been developed. Continuous utilization of agricultural lands regardless of the land suitability caused soil destruction. Also, incompetency in custom methods, invention geographic information system (GIS), and remote sensing (RS) techniques cause erupt and use of digital soil mapping.

The study area is approximately 5000 ha which is located in the west of Heris region of East Azerbaijan province, Iran. In the first study, the potential of different models to predict soil classes at different taxonomic levels was investigated. According to semi-detailed soil, survey and using stratified random sampling method, 50 pedons and 50 augers with an approximate distance of 1000 m were excavated, described and soil samples were taken from different genetic horizons. Based on the pedon descriptions and soil analytical data, pedons were classified up to the family level. Different machine learning techniques, namely boosted regression tree (BRT), random forest (RF), artificial neural networks (ANNs), and multinomial logistic regression (MLR) were used to test the predictive power for mapping the soil classes. After preparing the soil properties maps and checking their accuracy, these maps were used along with auxiliary parameters for estimating soil classes using an artificial neural network model in the R software. Finally, the accuracy and uncertainty of the model were evaluated by overall accuracy and confusion index, respectively.

Results showed that the different models had the same ability for prediction of the soil classes across all taxonomic levels but a considerable decreasing trend was observed for their accuracy at subgroup and family levels. The terrain attributes were the most important auxiliary information to predict the soil classes up to the family level. The main goal of the second study was to predict soil surface properties (pH, electrical conductivity, gypsum, organic carbon, calcium carbonate equivalent, coarse fragments, and particle size distribution) using ANNs, BRT, generalized linear model (GLM), and multiple linear regression (MLR). Among the studied models, GLM showed the highest performance to predict most soil properties whereas the best model is not necessarily able to make an accurate estimation. Also, the terrain attributes were the most important environmental covariates to predict the soil classes in all taxonomic levels, but they could not display the soil variation entirely. This shows that the unexplained variations are controlled by unobserved variations in the environment, which can be due to the management over time. Results suggested that the DSM approaches have not enough prediction accuracy for the soil classes at lower taxonomic levels that focus on the soil properties affecting land use and management. Results showed that the entry of more details in the soil classification at the lower levels of the Soil Taxonomy system while increasing the number of classes, leads to decreasing the overall accuracy and increasing uncertainty. It is noticeable that the ANNs model has a good accuracy up to the great group level through the acceptable level of overall accuracy (i.e., 75 %), hence it has a high degree of uncertainty. Therefore, the accuracy of the model could not be effective in its selection through the modeling process; however, paying attention to its uncertainty is also very important along with the model error.
 

Terrain attributes were the main predictors among different studied auxiliary information. The accuracy of the estimations with more observations is recommended to give a better understanding about the performance of DSM approach over low-relief areas. Further studies may still be required to distinguish new environmental covariates and introduce new tools to capture the complex nature of soils. Accordingly, we suggest using the other methods of soft computing for modeling in plain areas or low relief regions. Finally, the use of DSM methods is increasing over time and will eventually be considered as distinct and novel techniques.

Today, the use of intelligent models in simulating runoff has been widely used in water resources management. In this study, in order to predict the daily flow time series of the Morghak hydrometric station in Karun basin, an intelligent model of artificial neural network combined with wavelet analysis has been used. For this purpose, the ERA-INTRIM observational and analytical precipitation time series for 16 years (1378-1382) was decomposed by wavelet transform into frequency subsets, then each subset separately as input data to the artificial neural network model was introduced. The results showed that the analytical data have a high ability to simulate runoff precipitation models and can be a good alternative to observation data of rainfall stations. Also, according to the results of the wavelet transform technique, it can be effective in improving the performance of the simple ANN model for the Bazoft basin by 38% on a daily scale and 72% on a monthly scale.

Life on Earth is influenced by precipitation. Precipitation is one of the most significant factors that affect the hydrological cycle. Considering that precipitation is non-linear, complex, and can be changed according to spatial and temporal, estimating the amount of this important atmospheric factor in each month or year for each region and watershed is particularly important in managing and optimizing water resources. Various optimization models and algorithms have been proposed for modeling hydrological systems in recent decades. These algorithms have significantly reduced errors and increased accuracy. Still, since hydrological systems rely on random events, none of the methods can be completely and accurately selected as a superior model for modeling and estimating. The honey badger algorithm is an innovative algorithm that requires a few iterations to achieve an optimal solution, and this increases the speed of reaching the desired results. In current study investigates the performance of three models, including multiple linear regression (MLR), artificial neural network (ANN), and hybrid artificial neural network with honey badger optimization algorithm (HBA-ANN) for modeling the temporal and spatial precipitation in East Azarbaijan province. The best-developed model was selected by evaluation criteria such as R, RMSE, NRMSE, MBE, and NSE, the best model is selected.

The MLR model is one of the methods to analyze and investigate several variables. In this method, the model has one dependent variable and several independent variables, so that a linear equation is generated between the independent variables called X1, X2, ..., Xn and the dependent variable Y. ANN is a black box model of neural networks in the human brain. One of the most used methods is the BP method, which includes two stages. In the first stage, which is entitled feed-forward, the error value is calculated, after comparing output and objective values. In the second stage, which is labeled the back-propagation, the error value calculated in the previous step is corrected. The mentioned two stages continue until the output of the model approaches the desired output. The HBA is a new algorithm that simulates the honey-seeking behavior of a creature called the honey badger. The HBA includes two stages. In the first phase, the locations of this creature are calculated, and in the second phase, the exact distance between the HBA and the prey (dj) is calculated based on the honey intensity (S) and the honey smell intensity (Ij), as well as its new and optimal location for the prey Xnew. In the HBA-ANN model, the HBA algorithm is used to determine the most optimal output value in the ANN and increase performance in modeling. Therefore, the developed hybrid model can have the characteristics of both ANN and HBA methods.

In this study, in the first stage, the temporal modeling, and in the second stage, the spatial modeling of the monthly precipitation of 18 stations in East Azarbaijan province during the period of 1996-2022 using MLP, ANN, and HBA-ANN models has been paid. For temporal modeling of precipitation, one and two-month precipitation delay steps of the stations were used as input parameters. The first 70% of the dataset was selected for the training phase and the last 30% of the dataset was selected for the testing phase. Based on the results obtained from evaluation criteria and graphic diagrams, it can be concluded that the HBA-ANN model indicated significant accuracy compared to other models in the temporal modeling of precipitation. Also, by comparing the results of the stations in the HBA-ANN model, the Heris station with R =0.94, RMSE=2.25, NSE=0.79, NRMSE=0.04, and MBE=1.06 in the testing stage performed better compared with other stations. For spatial modeling of precipitation, the geographic coordinates of the stations, which include longitude, latitude, and altitude, are used as input parameters, and average monthly precipitation is used as the output parameter. From eighteen stations, 70% of the stations were selected for the training phase and 30% of the stations were selected for the testing phase. Based on the results obtained from R=0.95, RMSE=1.03, NSE =0.92, NRMSE = 0.03, and MBE = -0.81 and graphical diagrams, it can be concluded that the HBA-ANN model revealed significant accuracy compared to other models in spatial modeling of precipitation.

Precipitation is one of the most important factors that significantly change the hydrological cycle. Therefore, modeling and estimating this parameter is vital. In this study, the performance of multiple linear regression (MLR), artificial neural network (ANN), and hybrid ANN using honey badger algorithm (HBA-ANN) models were used for the spatial and temporal modeling of precipitation in East Azarbaijan province. For spatial modeling, the time delay steps of one and two months of station precipitation were selected as input parameters. Also, for temporal modeling, the longitude, latitude, and altitude parameters were used. The mentioned models were evaluated by R, RMSE, NSE, NRMSE, and MBE assessment criteria. According to the results of temporal modeling, the HBA-ANN model for all stations, especially Heris station with R equal to 0.94, RMSE equal to 2.25, NSE equal to 0.79, NRMSE equal to 0.04, and MBEequal to 1.06 is selected as the superior model. Also, based on the results obtained from spatial modeling, the HBA-ANN model with R equal to 0.95, RMSE equal to 1.03, NSE equal to 0.92, NRMSE equal to 0.03, and MBEequal to -0.81 was selected as the best model. The MLR and ANN models, respectively, presented a relatively poor performance compared to the developed hybrid model.

Fish are cold blooded animals and their metabolism, growth and feeding are strongly dependent on water temperature. Temperature changes in fish breeding pools cause stress and disease outbreaks occur especially above the tolerance thresholds. The aim of this study is predicting pool water temperature from observed air temperate using several machine learning approaches, namely artificial neural network, gradient boosting and random forest in Gilan province.Maximum and minimum air temperature data of Rasht Agrometorological station for the period of June 2016 to November 2018 were collected and used for prediction of corresponding data of fish breeding pond .The obtained results showed that for prediction of the minimum temperature, the neural network model (with a root mean square of 1.93 and a correlation of 0.92) and for the pool water maximum temperature, the random forest model (with a root mean square of 1.61 and a correlation of 0.95) did a better job comparing to other two approaches. These selected models can be applied for prediction of water temperature using air Tmax and Tmin for improved management options under changing conditions.

Considering the importance of fresh water for human life and the vulnerability of groundwater sources to all kinds of pollution and the possibility of transferring pollutants to other surface and groundwater sources, as well as the location of Iran in the arid and semi-arid belt, protecting this valuable and rare element is imperative and its continuous monitoring must be one of the priorities of water resources managers. Therefore, in the current research, nitrate pollution in the eastern plains of Mazandaran province was discussed and relevant issues were investigated, and an efficient and optimal model for predicting nitrate concentration was presented. In this research, three machine learning models including decision tree, logistic regression and artificial neural network were compared. The physical and chemical data measured during the years 1985 to 2020 were used and entered as the input variables of the models. The variables include temperature, water level, pH, EC, HCO3-, CL-, SO24-, Na+, K+, Mg2+, Ca2+, TH and TDS; The amount of nitrate contamination of the groundwater, was predicted by dividing 70% of the dataset as training and 30% as testing data. The R2, RMSE, NSE and PBIAS indexes were applied for model evaluation. The results indicated that the Decision Tree model had the best performance with a large difference compared to the other two models (R2 = 0.957 and RMSE = 0.297, NSE = 0.95 and testing acc = 0.907). After That, logistic regression and artificial neural network had much weaker performances than the lead model. It is suggested to conduct another research with other machine learning models by changing the input variables and add some extra ones such as land-use and compare the results with the current research.

Predicting river flow is one of the most crucial aspects in water resources management. Improving forecasting methods can lead to a reduction in damages caused by hydrological phenomena. Studies indicate that artificial neural network models provide better predictions for river flow compared to physical and conceptual models. However, since these models may not offer reliable performance in estimating unstable data, using preprocessing techniques is necessary to enhance the accuracy and performance of artificial neural networks in estimating hydrological time series with nonlinear relationships. One of these methods is wavelet transformation, which utilizes signal processing techniques.

In this study, to evaluate the efficiency of discrete and continuous wavelet types in the Wavelet-Artificial Neural Network (WANN) hybrid model for monthly flow prediction, a case study was conducted on the Kardeh Dam watershed in the northeast of Iran, serving as a water source for part of Mashhad city and irrigation downstream agricultural lands. Monthly streamflow estimates for the upstream sub-basin of the Kardeh Dam were obtained from the meteorological and hydrometric stations' monthly statistics over a 30-year period (1991-2020). The WANN model is a hybrid time series model where the output of the wavelet transform serves as a data preprocessing method entering an artificial neural network as the predictive model. The combination of wavelet analysis and artificial neural network implies using wavelet capabilities for feature extraction, followed by the neural network to learn patterns and predict data, potentially enhancing the models' performance by leveraging both methods. The 4-fold cross-validation method was employed for the artificial neural network model validation, where the model underwent validation and accuracy assessment four times, each time using 75% of the data for training and the remaining 25% for model validation. The final results were presented by averaging the validation and accuracy results obtained from each of the four model runs. To evaluate and compare the performance of the models used in this study, three evaluation indices, Nash-Sutcliffe Efficiency (NSE), Root Mean Square Error (RMSE), and Pearson correlation coefficient (R), were employed.

The analysis of meteorological and hydrometric data in this study revealed that monthly streamflow in two time steps, T-1 and T-2, were the most effective predictive variables. Each of the two runoff variables of the previous month (Qt-1) and the previous two months (Qt-2) were analyzed by each of the Haar and Fejer-Korovkin2 discrete wavelet transforms and the two continuous Symlet3 and Daubechies2 wavelets at three levels. The results of each level of decomposition was given as input to the ANN model. The presented results at each decomposition level indicated that hybrid models could accurately predict lower flows compared to the single ANN model, and the estimation of maximum values also significantly improved in the hybrid models. Among the wavelets used, Haar wavelets exhibited the weakest performance, and the less commonly employed Kf2 wavelet showed a moderate performance. Since the Haar and Fk2 wavelets, with their discrete structure, did not perform well in decomposing continuous monthly streamflow data, continuous wavelet models outperformed discrete wavelet models. The hybrid models, combining wavelet analysis and artificial neural networks, demonstrated up to an 11% improvement over the performance of the single neural network model.

Streamflow is a crucial element in the hydrological cycle, and predicting it is vital for purposes such as flood prediction and providing water for consumption. The objective of this research was to evaluate the performance of different types of discrete and continuous wavelet models at various decomposition levels in enhancing the efficiency of artificial neural network (ANN) models for streamflow prediction. Since climate and watershed characteristics can influence the nature of data fluctuations and, consequently, the results of the wavelet model decomposition, choosing an appropriate wavelet model is essential for obtaining the best results. Considering the existing variations in the results of different studies regarding the selection of the best wavelet type, it is suggested to use both continuous and discrete wavelet types in modeling to achieve the best predictions and select the optimal results. Given that a lower number of input variables in neural network models lead to higher accuracy in modeling results, it is recommended to perform decomposition at a two-level depth to reduce input components to the neural network model, thereby reducing the model execution time.

This study addresses the challenge of measuring soil cation exchange capacity (CEC), a vital factor influencing soil fertility, by exploring the impact of grouping soil samples based on different characteristics on the performance of estimation models. Recognizing the difficulties associated with traditional CEC measurement methods, the study employs a cost-effective and rapid approach using various models and equations. The research, conducted at Bu Ali Sina University in Hamedan, utilizes a substantial dataset of 45,948 soil samples from the standardized database of world soils. Soil samples are initially categorized into different groups, and nine estimator variables are examined across 11 models for the entire dataset and specific classes within each group. These variables include soil texture components, organic carbon, calcium sulfate, calcium carbonate, bulk density, base saturation percentage, total exchangeable base cations, and soil reaction. The results demonstrate that grouping soil samples, especially based on texture classes, significantly improves the performance of artificial neural network models, with a remarkable 87% relative improvement coefficient in the test section. The study reveals that data grouping enhances the model's estimation capabilities, as evidenced by reduced root mean square error (RMSE) values in the test sections for different texture classes. In conclusion, the findings suggest that utilizing functions derived from grouped data offers an effective and cost-efficient method for estimating soil cation exchange capacity. This approach provides valuable insights for soil fertility management, offering a simplified yet accurate means of assessing this critical soil parameter.

River flow forecasting has been one of the important challenges of water resources management in recent decades, so many researchers have proposed different methods to improve the performance of forecasting models. In the last decade, artificial intelligence methods have been most widely used in the simulation of various processes, including hydrological processes, due to their flexibility and high accuracy in modeling. However, the results of these methods have always been associated with uncertainty due to several factors such as structure, algorithm, input data, and the type of method chosen for data calibration. One of the methods that can somewhat solve this problem is the uncertainty analysis of the predictions made by these models.

In this study, the uncertainty of the results of artificial neural network (ANN) and support vector machine (SVM) models in predicting the monthly flow of the river has been evaluated. In this research, the time series of the monthly flow of the Ghezelozan River using the data of the Bianlu-Yasaul Stream gauging station in 39 years from 1976 to 2014 was used, where 75% and 25% of the data was used for training and testing the models, respectively. In these models, to estimate the monthly flow of the Ghezelozan River, six different input combinations including the flow of one, two, and three months before and the number of months of the flow were used. Then, the accuracy and performance of the models were compared using the coefficient of determination (R) and root mean square of errors (RMSE). Finally, the uncertainty of the results of ANN and SVM models in predicting the monthly flow of the river was evaluated by the Monte-Carlo method using d-factor and 95PPU values.

The evaluation of the performance of the ANN model shows that the best performance is related to the combination where the flow of the previous two months and the number of the month of the flow are the inputs of the model so that R and RMSE indexes were obtained as 0.757 and 9.45, respectively. In the SVM model for the monthly river flow series, the best performance is related to the combination where the flow of one, two, and three months ago and the number of months of the flow were the inputs of the model, and the R and RMSE indexes for this input pattern were 0.729 and 10.946, respectively. After studying the uncertainty of the models, the results showed that the ANN model has more uncertainty in the output values compared to the SVM model, and this is while the d-factor of the ANN model, both in the training and test phase, it was more than the SVM model. The comparison of the uncertainty analysis of the results of the ANN and SVM models showed that the SVM model with d-factor and 95PPU values equal to 0.155 and 17.241, respectively, compared to the ANN model with d-factor and 95PPU values equal to 0.993 and 85.470, respectively, has less uncertainty in the output values. So the number of observation data placed in the 95% confidence range (95PPU) of the ANN model compared to the SVM model has increased significantly in both the training and testing phases. Also, the results showed that both models have more uncertainty in the months with a large volume of water, which can be due to the complexity of the process and the involvement of uncertain factors in these months, as well as the effect of factors that are not considered in the structure of predictive models.

The results of ANN and SVM models in predicting the monthly flow of the Ghezelozan River showed that although the ANN model with R-value equal to 0.757 and RMSE value equal to 9.45 has a good performance compared to the SVM model with R-value equal to 0.729 and RMSE value equal to 10.946 in predicting the river flow, the results of this model are associated with high uncertainty. The comparison of the uncertainty analysis of the results of ANN and SVM models by Monte-Carlo method showed that the SVM model with d-factor and 95PPU values equal to 0.155 and 17.241, respectively, compared to the ANN model with d-factor and 95PPU values equal to 0.993 and 85.470, respectively, has less uncertainty in predicting the monthly flow of the Ghezelozan River and it is better than ANN model. According to the results of this research, taking into account the fact that advanced artificial intelligence models also have uncertainty, it is necessary to apply these methods in the fields of risk management and future planning to obtain the best performance.

Due to global warming, accurately estimating evaporation has become a key challenge in water resource management, and due to the important role it plays in the withdrawal of water from human reach, it has always attracted the attention of researchers. Therefore, modeling and awareness of the value of evaporation as one of the hydrological variables is of great importance in agricultural research and soil and water conservation. Gorgan was chosen for the study due to its proximity to the Caspian Sea with a humid climate and a higher rate of evaporation than other cities. On the other hand, Shiraz has a hot and dry climate, is located in central Iran far from water resources such as the sea, and has a lower evaporation rate. Kish also has a warm and humid climate due to its proximity to the sea, with a lower evaporation rate than Shiraz but higher than Gorgan. Several meteorological variables affect the process of evaporation and transpiration, and due to the complexity of the evaporation parameter, a method with high accuracy should be used to determine them. Recently, artificial neural network methods have become very popular among researchers due to their common use and the ease with which they can solve complex problems. Therefore, many intelligent algorithms have been suggested to find the best solution for complex engineering problems, as they can find optimal answers faster and more accurately.

Artificial neural networks are designed based on inspiration from the memory and learning mechanisms in the human brain. To train artificial neural networks, a set of valid input and output data is used based on the type of problem. The accuracy of the network output depends on the amount of training data and how the inputs and their features are processed. To design different scenarios for adjusting input data, the correlation values of the data with evaporation were used. In this study, three synoptic stations with different climates, including Gorgan, Shiraz, and Kish, were chosen. Three stations with different climates were used to better evaluate and repeat the steps of the method so that the efficiency of the method could be more accurately assessed. Considering the importance of the value of evaporation in nature, evaporation modeling with the ANN and its combination with the COOT algorithm, which mimics the natural life of a COOT bird, was performed using five meteorological parameters, including the minimum air temperature, maximum air temperature, wind speed, average relative humidity, and sunshine hours on a monthly between 2000 and 2022. The dataset was divided into two phases: training (70 % of the dataset) and testing (30% of the dataset). To evaluate the performance of developed models, statistical indices of these models such as correlation coefficient (R), root-mean-square error (RMSE), Nash-Sutcliffe coefficient (NS), and their graphical representations were compared with each other.

As mentioned, four models of ANN-COOT with varying input parameters were developed and compared to four conventional ANN models. Statistical performances were calculated, and comparison plots were made in the training and testing phases to find the most adequate model for the prediction of evaporation. Comparing of obtained results from statistical indices for the testing phase revealed that the COOT-ANN4 model had the best performance for Gorgan with the R, RMSE, and NS equal to 0.99, 8.19, and 0.99 respectively. Shiraz also obtained values of the R, RMSE, and NS equal to 0.99, 18.43, and 0.98 respectively. Similarly, for Kish, the values of the R, RMSE, and NS equal to 0.97, 19.36, and 0.93 respectively, have better performance than the other models. Compared with the results of different input combinations, the hybrid ANN-COOT model (ANN-COOT4) at three stations was found superior with input combinations of Tmin, Tmax, SSH, RH, and WS. Additionally, to evaluate the accuracy of developed models, Scatter plots, Violin plots, Relative error percent plots (RE %), Taylor diagrams, and Histograms were drawn. By comparing the graphical representations, it can be determined that the hybrid COOT with ANN4, namely the COOT-ANN4 model, has improved the artificial neural network at Gorgan, Shiraz, and Kish stations.

The algorithm of COOT is an optimization algorithm that is generally used to solve optimization problems. As observed from the overall performance of the results of the hybrid model in predicting evaporation, the objective function was minimized. The results indicated that scenario four of the COOT-ANN4 hybrid model with input parameters of minimum temperature, maximum temperature, sunshine hours, relative humidity, and wind speed has better accuracy and performance at all three stations. In general, the findings of this study revealed that the COOT algorithm can improve the artificial neural network (ANN) structure in any climate and provide a hybrid model with higher accuracy and less error for modeling the evaporation parameter. Considering that the COOT algorithm is powerful and efficient, it is better to use it in various fields to improve the performance and accuracy of models. The testing results revealed that the lowest Root Mean Square Error RMSE (18.43, 19.36 and 8.19) and highest coefficient of correlation R (0.99, 0.97, and 0.99), and the highest Nash–Sutcliffe Efficiency Coefficient (N-S) (0.98, 0.93 and 0.99) attained by the ANN-COOT4 hybrid model (relative to other ANN and ANN-COOT models) tested for three selected stations in Shiraz, Kish and Gorgan sites. Concerning the predictive efficiency, the developed ANN-COOT hybrid model, improved the modeling performance at extreme points, which outperforms the ANN model, indicating its capability in the prediction of monthly evaporation.

An artificial neural network (ANN) is a powerful data-driven tool capable of learning the linear and nonlinear relationships governing different systems. However, determining the best-performing algorithm in terms of convergence speed and accuracy for a particular problem is still a fundamental challenge for users of artificial neural networks.

We investigated the most effective algorithm among widely used processes to simulate and estimate nonlinear water quality parameters. For this purpose, we constructed 42 models combining artificial neural network topology (single or multilayer) and training processes. The quality parameters’ data acquired at 107 wells throughout the aquifer of Qorveh-Dehgolan plain were used for training (data from 1996 to 2013) and to test (data from 2014 to 2016) each model.

The results showed that artificial neural networks with a hidden layer that benefits from the optimal number of neurons could simulate the aquifer behavior with high accuracy and in less time. Also, increasing the number of hidden layers while increasing the response accuracy increases the number of optimal network neurons and the duration of the problem analysis. Finally, artificial neural networks based on the Broyden-Fletcher-Goldfarb (BFG) method had the highest efficiency in simulating aquifer behavior, although the performance of the Levenberg Marquart (LM) method is very close to BFG. BFG is more efficient than LM due to its lower Mean Square Error and Standard Deviation (3.46 and 3.09, respectively).

Digital Soil Mapping (DSM) encompasses a variety of methodologies that can yield precise spatial information about soil by establishing quantitative relationships between environmental covariates (predictors) and soil classes or properties. In this study, Artificial Neural Networks (ANNs), Decision Tree (DT), Multinomial Logistic Regression (MLR), and Random Forest (RF) algorithms were used to predict the soil map of downstream lands of Azad dam with an area of approximately 178.3 ha in the northwest of Sanandaj city in Kurdistan province. A random soil sampling method was used to determine the location and distribution of the 84 soil profiles in the study area. After recording soil morphological attributes, sampling of all horizons was conducted for required laboratory analysis. Afterward, the soil profiles were classified up to the family taxonomic level based on US classification system. Based on the soil taxonomy classification system, Inceptisols and Entisols order were observed by frequency, two Suborder, three Great groups, five Subgroups, and Family. To calculate the predictor variables, a digital elevation model (DEM) with a 10 m spatial resolution and Sentinel 2-B satellite images were used in the study area. To check the prediction accuracy of the models the Overall accuracy (OA), Kappa Index (K), and Brier Score (BS) were used. The best result was obtained by the ANN model (OA=0.65, K=0.53, and BS=0.16, respectively). The weakest predictions were found by DT model with OA, K, and BS of 0.38, 0.22, and 0.87, respectively.

Rainfall is one of the complex natural phenomena and one of the most crucial component of the water cycle, playing a significant role in assessing the climatic characteristics of each region. Understanding the amount and trends of rainfall changes is essential for effective management and more precise planning in agricultural, economic, and social sectors, as well as for studies related to runoff, droughts, groundwater status, and floods. Additionally, rainfall prediction in urban areas has a significant impact on traffic control, sewage flow, and construction activities.

The objective of this study is to compare the accuracy of classification models, including Chi-squared Automatic Interaction Detector (CHAID), C5 decision tree, Naive Bayes (NB), Quest tree, and Random Forest, k-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Artificial Neural Network (ANN) in predicting rainfall occurrence using 50 years of data from the synoptic station at Hamedan Airport. In this study, 80% of the data is used for training the models, and 20% for model validation and the results obtained from the model executions are compared using metrics such as confusion matrix, Receiver Operating Characteristic (ROC) curve, and the Area Under the Curve (AUC) index. To create the classification variable for rainfall and non-rainfall data, based on rainfall data, the days of the year are categorized into two classes: days with rainfall (y) and days without rainfall (n). Data preprocessing is performed using Automatic Data Preprocessing (ADP). Then, Principal Component Analysis (PCA) is employed to reduce the dimensions of the variables.

In this study, the PCA method reduces the dimensions of the variables to 5. Also, approximately 80% of the available data corresponds to rainless days, while 20% corresponds to rainy days. The research results indicated that the KNN model with an accuracy of 91.9% for training data and the SVM model with 89.13% for test data exhibit the best performance among the data mining models. The AUC index for the KNN model is 0.967 for training data and 0.935 for test data, while for the SVM algorithm, it is 0.967 for training data and 0.935 for test data. According to the ROC curve for Hamedan rainfall data, the KNN model outperforms other models. Considering the sensitivity index in the confusion matrix, the KNN and SVM models perform better in predicting non-rainfall occurrence for training data. In terms of the precipitation occurrence prediction, the RT and KNN models show better results according to the specificity index.

The results demonstrated that for the RT, C5, ANN, SVM, BN, KNN, CHAID, QUEST, accuracy metrics was obtained 86.82%, 89.78%, 89.55%, 89.96%, 88.06%, 91.9%, 88.29%, 87.46%, 91.9%, respectively for training data. Moreover, for test data, the accuracy metrics for this model was obtained 83.82%, 87.9%, 88.12%, 89.13%, 87.12%, 89.13%, 87.12%, 88.19%, 86.93%, 86.76%, respectively. The AUC index in the training data for RT, C5, ANN, SVM, BN, KNN, CHAID QUEST models was 0.94%, 0.99%, 0.94%, 0.94%, 0.93%, 0.97%, 0.93%, 0.89%, respectively. In addition, for the test data, this metric was evaluated 0.89%, 0.89%, 0.93%, 0.94%, 0.92%, 0.90%, 0.92%, 0.88% respectively. As observed, considering accuracy metric and AUC index for training data KNN model and for test data SVM model were more sufficient in rainfall prediction.

Improving the lifestyle of city residents is conditional on benefiting from high-quality urban infrastructure to satisfy daily demands. The urban water supply network is one of the most basic urban infrastructures, and its optimal design and service are essential during the planning period. Therefore, it is important to determine the actual amount of consumption and predict it for the future. For this purpose, in this research, a method based on artificial intelligence, i.e., genetic programming (GP), as well as Pearson's correlation coefficient data mining method, is proposed. The data mining method is applied here for the database, including daily data on temperature, precipitation, humidity, and the amount of daily water produced in Najafabad city (presenting the total water consumption) from the beginning of 2014 to the end of 2018, and the best set of input data vectors is selected. The selected data are used as input data vectors for the proposed. The obtained results are compared with the results of models based on artificial neural network (ANN). To investigate the performance of the models, R², RMSE, and NSE statistical indices are calculated. A comparison of the results indicates the acceptable performance of the proposed models based on the GP. In other words, the values of RMSE, NSE, R², and MAPE statistical indices for training data in the best GP model are equal to 3262.59 MCM, 0.80, 0.80, and 5.38%, respectively, and for test data equal to 3507.68 MCM, 0.78, 0.78, and 6.67%.

Indiscriminate exploitation of groundwater has caused a decrease in the groundwater level and as a result, has caused land subsidence in many areas. This problem is especially visible in arid and semi-arid regions like Iran, where water supply for agriculture, drinking, and industry is done from groundwater water sources, and in recent years, it has seriously threatened the aquifers of the plains as a serious danger. In this research, due to the importance of the problem, land subsidence in the Damghan Plain aquifer located in Semnan province was studied and investigated.In this research, the amount of subsidence was measured at the field, and then its spatial changes were investigated using conventional methods such as Kriging interpolation, Co-kriging, and inverse distance weighted interpolation (IDW). Also, the artificial neural network model was used to estimate and interpolate the amount of subsidence. Three statistical indices namely the coefficient of correlation (R2), the root mean square error (RMSE), and the mean absolute error (MAE) were used to compare the estimation of subsidence values using an artificial neural network model, kriging interpolation method, cokriging, and IDW. To perform interpolation using kriging, cokriging, and IDW interpolation methods, the variogram of land subsidence data was drawn. Also, in order to increase the accuracy of the mentioned models in predicting the amount of subsidence, the auxiliary variable of water level reduction was used.Using different functions such as circular, spherical, exponential, Gaussian, and linear models for plotting the semivariogram, results show that the Gaussian function with segment-to-threshold ratio (C0/(C0+C)) equal to 0.26 has better performance compared to other models. Also, the artificial neural network has a better performance compared to the kriging method and the inverse weighted distance method and has been able to reduce the RMSE error value in the validation stage by 17.6% and 31.3%, respectively. It has also increased the value of the R2 from 0.502 and 0.421 to 0.721.The use of the auxiliary variable of water level reduction has increased the accuracy of the models used in predicting the amount of subsidence. In this case, the comparison between the estimation of subsidence values using the artificial neural network model compared to the interpolation method of kriging, cokriging, and IDW shows that the artificial neural network model with a coefficient of determination (R2) of 0.860 and 0.751 and 0.015 and 0.017, has a better performance compared to the mentioned methods and reduce the amount of prediction error. Therefore, the artificial neural network model can be used with good accuracy as an alternative method instead of conventional interpolation methods to investigate the spatial changes of land subsidence.In recent years, the phenomenon of land subsidence has occurred in many plains of Iran due to excessive extraction of underground water, including Damghan Plain. In this research, by using the measured data of land subsidence in this plain, its spatial changes were investigated using kriging and cokriging methods (geostatistics method), inverse distance weighted method (IDW), and weighted IDW (deterministic method) in the whole plain. it placed. The research results showed:• Compared to circular, spherical, exponential, and linear functions, the Gaussian function can better estimate the spatial changes of land subsidence.• The results of the kriging method were better than the inverse distance weighted (IDW) method.• The artificial neural network model with Gauss function and two intermediate layers has better performance than other transfer models such as sigmoid, hyperbolic tangent, and hyperbolic Bade secant.• The use of artificial neural network model has increased the accuracy of land subsidence estimation compared to conventional methods such as the kriging method and inverse weighted distance method.

In Iran, the climatic conditions are such that even in the rainiest areas of the country, there is a need for groundwater resources, and this demand is increasing every year. Since, groundwater is one of the most valuable water resources in Iran, it is very necessary to predict its changes in order to use it optimally with the aim of sustainable development. One of the most complex hydrological processes in nature is the rainfall-groundwater level process, which is affected by various physical and hydrological parameters. Although, various models have been presented to predict the changes in the groundwater level using the rainfall patterns, but less attention has been paid to the transfer function model. Hence, the main objective of this research is to introduce and use the transfer function (TF) model to predict the monthly groundwater level using rainfall data and to compare its results with ANFIS and Artificial Neural Network (ANN) models.In the present study, 30-year data (1992-2021) of meteorological stations and observation wells in 3 watersheds of Galikesh, Ramian and Mohamadabad were used to model the rainfall and groundwater level. of the Gorganroud river basin.Then, considering that the years closer to the present time have more accurate information about the situation of this time, the years were considered as a forward process in artificial neural networks. The model fitting and prediction of the groundwater level values using rainfall data for the next 12 months was performed with applying three models: Artificial Neural Network (ANN), Adaptive neuro fuzzy inference system (ANFIS), and transfer function (TF). For this purpose, MINITAB SAS, SPSS, and R software were used. Next step, the validation of the values predicted by the models was evaluated using three indices Mean Absolute Distance (MAD), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE).The results of the autocorrelation plots of the groundwater level of the wells revealed that all time series have a seasonal trend with a period of 12 months. Based on the cross-correlation plots, it was also found that rainfall has direct effect on the groundwater level in the two watersheds of Galikesh and Mohamadabad with lag time of three months and in the Ramian watershed with a delay of one month. The validation results of the models using three indexes MAD, RMSE and MAPE revealed that the artificial neural network model for predicting the groundwater level using monthly rainfall data in all three investigated watersheds had the most appropriate performance (RMSE=0.0778, 0.0243, 0.0532m) and the ANFIS model is ranked second (RMSE=0.1841, 0.0832, 0.1012m). Although the transfer function model was less accurate than the other two methods (RMSE=0.5711, 0.5023, 0.3234m), but this model has performed well in fitting the monthly groundwater level values. This model is very effective in identifying the delay in the impact between the input and output variables, as well as expressing the model based on which the impact of rainfall can be expressed as a model.The results of this research show that all three models of artificial neural network, ANFIS and transfer function can be used to predict the groundwater level using monthly rainfall values. Consecutive overestimation and underestimation, which increases the error and decreases the performance of the models, was not observed for the three used models. Also, all three models perform well in detecting trends and data changes. However, the artificial neural network model is more accurate than the other models. In addition, when forward process is used in artificial neural network modeling, compared to the case where the complete series of data is used, the efficiency of the model is significantly improved.

The present research performed to estimate soil texture using visible near-infrared spectrometry in Semirom, Isfahan. A total number of 200 soil samples (0-10 cm) were collected from the Semirom area (51º 17' - 52º 3' E; 30º 42' - 31º 51' N), Isfahan. The samples were air dried and passed through a 2 mm sieve, and soil particles percentage was determined in the laboratory using hydrometry method. Reflectance spectra of all samples were measured using an ASD field spectrometer. Different pre-processing methods i.e., First Derivatives and Savitzky-Golay Filter, Multiplicative Scatter Correction and Standard Normal Variable were applied and performed on spectral data. The Partial Least Squares Regression, Support Vector Machine Regression and Artificial Neural Network models were used to estimate soil texture. The best result was obtained for Silt estimation, with excellent values of RPD >2, R2 =0.98 and RMSE=1.08 using Artificial Neural Network model with MSC pre-processing technique. The results indicated the desirable capability of Artificial Neural Network model with MSC and SNV pre-processing techniques in estimating the Clay (RPD >2, R2=0.94 and RMSE=1.21) and Sand (RPD >2, R2=0.84 and RMSE=6.24) contents of the soils, respectively. In general, based on the results of this study, VNIR spectroscopy was successful in estimating soil particles percentage and showed its potential for substituting laboratory analyses.

Water infiltration to soil play an important role in irrigation management, storage moisture in soil especially in dry and semi-dry area and increasing agronomy yield. Understanding water infiltration to soil is of important in designing and applying water conservation methods, flood and runoff control and soil erosion management. Additionally, accurately measuring water infiltration to soil in different times is more important for predicting water storage in root zone, irrigation designing and planning and agronomy management. In other hand, water infiltration to soil is a base for precision agriculture, therefore, producing accurate maps for water infiltration to soil play an important role in land management and applying precision agriculture. Modeling soil water infiltration at the field scale with ruler of calcareous, saline and sodic conditions is important for a better understanding of infiltration processes in these soils and future of infiltration modeling. The present study aimed for estimating water infiltration to soil at different times using soil spatial prediction functions and spatial estimators.

in present study, 72 soil samples were collected using a random sampling method in Marvdasht plain, Fars Province. In selected points, soil bulk density, sand silt, clay, pH, electrical conductivity, calcium carbonate, solution sodium, solution calcium and magnesium and organic carbon contents. For measuring water infiltration to soil, the double ring method were used. Multiple linear regression (MLR), artificial neural network (ANN) and spatial estimators were used for deriving soil spatial prediction functions models between water infiltration to soil in different times including 5, 10, 20, 45, 90, 150, 210 and 270 min. In this study, the readily available soil properties and auxiliary variables such as remote sensing and topography data were used in soil spatial prediction functions.

The results of evaluating regression and artificial neural networks models based on the mean error (ME), coefficient of determination (R2) and root mean square error (RMSE) criteria in testing phase were showed that the developed artificial neural network models in present study performed better than multiple linear regression models in water infiltration to soil prediction at different times. Moreover, the results showed that the combined estimators (artificial neural network - Kriging) performed better than ordinary kriging model for estimating water infiltration to soil.

In totally, the results of this study showed that the applying soil spatial prediction functions (using auxiliary variables such as remote sensing data and topography data with the readily available soil properties) had a great potential to predict spatial estimation of water infiltration to soil at most considered times.

It is common to use different data mining methods in drought prediction. However, the selection of the best model is mainly based on the accuracy of the simulation, while most of the studies do not mention the features of the models. In this paper, the performance of the most common data mining models, including Multilayer Perceptron Artificial Neural Network (ANN-MLP), Radial Base Function Neural Network (ANN-RBF), Regression Decision Tree (CART), Model Tree (M5P), and Support Vector Machine (SVM) is evaluated in order to predict monthly one year ahead rainfall at Bandar Abbas synoptic station and then the characteristics of each of them are described. Calibration and validation of the models were done using raw data and a three-year moving average of climatic parameters from 1347 to 1396. The performance of the models has been evaluated using different statistical indices and comparative diagrams. The results showed that the SVM and M5P models have good prediction performance with RMSE of 7.93 and 8.31 mm, the MAE of 3.66 and 4.69 mm, and the CC of 0.83 and 0.82, respectively. Also, with the exception of the CART, the change in the data mining tool makes an eight to 11 percent difference in the accuracy of the estimates. Therefore, the most appropriate model should be selected based on other characteristics of the methods besides their accuracy. In addition, using the three-year moving average of the input parameters has increased the correlation coefficient by about 78 percent and reduced the RMSE by about 63 percent. The analysis of the long-term drought situation showed that with the increase in the period of the standard precipitation index, the separation of wet and dry years becomes more specific.