support vector regression

Measuring the flow of rivers is one of the most important issues in river management, that's why it is always tried to use accurate methods for its measurement. The aim of this study is to enhance the performance of Support Vector Regression (SVR) model using the Gray Wolf Optimization (GWO) algorithm for monthly river flow modeling. For this purpose, the monthly data of river flow, precipitation and temperature during 15 years (from 1400 to 2015) are used. The trial and error procedure is used to select the best input variables to the SVR and GWO-SVR models. Based on the results of this method, Q(t-1), R(t-1), T(t-1) are the best independent variables for simulating the variable Q_t. 80% of all data are used for training and 20% for validating the SVR and GWO-SVR models. Also, R^2, RMS and NSE indices are utilized to evaluate the efficiency of the models, linear (LKF), polynomial (PKF), radial basis function (RBF), and sigmoid (SKF) activation functions are used to develop the models. First, the trial and error procedure is used to determine the parameters of the activation functions. Based on the results of this study, the SVR model with the polynomial activation function has the best performance in the training and validation stage, and the worst performance with the linear activation function in the training and verification stages. Then, the GWO algorithm is used to determine the parameters of the activation functions. Based on the results, the SVR model performs better with the GWO algorithm. Therefore, to simulate the monthly flow of river using this model, it is better to use the GWO algorithm instead of the trial and error procedure.

Timely and accurate crop yield estimation is important to regulate crop management and ensure agricultural sustainability. This research investigates the application of the support vector regression method in simulating the yield of barley in ten selected farms in the Qazvin plain from 2019 to 2020 using the drought indices NDVI, MSAVI, and EVI. Sentinel 2 satellite was used for drought indices. The results were evaluated using R2, RMSE, and ME statistics. To investigate the application of the support vector regression method in simulating the yield of barley using the mentioned drought indicators, seven scenarios were defined. The results of evaluating the relationship between NDVI, MSAVI, and EVI indices with yield barley showed that the support vector regression method in all scenarios except the first scenario in the test phase with a significant probability of 95% (P-value=0.0) and an explanatory coefficient of more than 0.7 and the low value of the RMSE index has a good estimate of the yield of the barley product. The results of this research will be useful in managing and planning the development of barley cultivation based on drought indices in the future.

In this study, a hybrid Support Vector Regression (SVR) model, combined with optimization algorithms including Wavelet, Firefly, and Grey Wolf, was employed to estimate groundwater hardness at the Naservand well in the Khorramabad Plain of Lorestan Province. For modeling, monthly quality data of the well during the water year (1382-1402) were used. This included parameters such as bicarbonate (HCO3), chloride (Cl), sulfate (SO4), magnesium (Mg), calcium (Ca), and total hardness (TH), all measured in ppm, which were used as inputs. Groundwater hardness was selected as the model output. To evaluate the performance of the models, assessment metrics such as the correlation coefficient, root mean square error (RMSE), mean absolute error (MAE), and Nash-Sutcliffe efficiency coefficient (NSE) were used. Time series, box plots, and Taylor diagrams were also used for analyzing the model results. The results indicated that hybrid scenarios in the studied models improve the model’s performance. Comparing results showed that the Support Vector Regression-Wavelet model performed better than the Support Vector Regression-Grey Wolf model, with the Support Vector Regression-Wavelet model achieving a correlation coefficient of 0.917, the lowest RMSE of 0.190 ppm, the lowest MAE of 0.115 ppm, and the highest NSE of 0.920 in the validation stage. Overall, the results indicate that using intelligent models based on the Support Vector Regression approach can be an effective approach for managing groundwater quality.

Predicting Total Dissolved Solids (TDS) in water bodies such as rivers and lakes is challenging but essential for effective water resource management in agriculture, industry, and drinking water sectors. This study develops a hybrid intelligent model based on the Support Vector Regression (SVR) approach to predict the TDS levels in river water. Three optimization algorithms—Wavelet, Firefly, and Grey Wolf—were employed to model the TDS in river water. The study utilized hydrometric station data from the Kashkan River in Lorestan Province as a case study, using eight scenarios combining input parameters from 2003 to 2023. Model performance was evaluated using correlation coefficient, root mean square error (RMSE), mean absolute error (MAE), and Nash-Sutcliffe efficiency coefficient. Time series plots, box plots, and Taylor diagrams were used to analyze the model results. The findings indicated that the combined scenarios improved model performance. The evaluation criteria showed that the Wavelet-SVR model achieved a correlation coefficient of 0.980, RMSE of 0.881 mg/l, MAE of 0.473 mg/l, and a Nash-Sutcliffe efficiency of 0.990 during the validation phase. Overall, the results suggest that intelligent models based on the SVR approach can be an effective strategy for maintaining river quality for aquatic health.

Soil temperature is one of the important factors in agriculture and hydrology, and its accurate measurement is very important to ensure the optimal growth and development of the plants. Soil temperature is a factor that affects many processes such as seed germination, soil moisture level, aeration, nitrification and availability of plant nutrients. Because the soil temperature data is measured in some synoptic stations, most of the data have limitations or are incomplete. However, choosing the best method to estimate soil temperature with other available meteorological data is an efficient approach in many fields. Soil temperature depends on several factors including color, slope, vegetation, density, humidity and amount of sunlight. Currently, some physical models are available that are intrinsically related to the state of soil heat flow and energy balance in underlying soils to estimate soil temperature. The importance of soil temperature in agricultural sciences and hydrology, on the one hand, and the existence of many difficulties in recording this vital parameter, have led researchers to seek a relationship between soil temperature and other parameters in order to be able to estimate soil temperature with optimal accuracy.

In this research, daily soil temperature values were collected during the time period of 1990-2022 in Ramsar, Arak and Shiraz stations. On the other hand, the parameters of minimum temperature (Tmin), maximum temperature (Tmax), average temperature (Tm), maximum relative humidity (Umax), minimum relative humidity (Umin), average relative humidity (Um), average wind speed (FFM) and Sunshine hours (SSHN) was considered as the input parameters and soil temperature (T-soil) as the target parameter. It is worth mentioning that the way of choosing different input compounds to estimate the value of soil temperature in the studied models is based on having a higher correlation with soil temperature based on the thermal map. Moreover, the ability of data-driven models of Gaussian process regression (GPR), support vector regression (SVR), M5P algorithm, linear regression (LR), and multilayer perceptron (MLP) neural network in estimating soil temperature was evaluated using different statistical parameters of correlation coefficient (R), root mean square error (RMSE), Nash Sutcliffe coefficient (NS), average absolute value of percentage error (MAPE) and Wilmot agreement index (WI).

The evaluation of five GPR, SVR, M5P, LR and MLP models for three stations of Arak, Ramsar and Shiraz shows that the 8th M5P scenario and the 8th LR scenario with lower root mean square error respectively (0.899 and 0.889) for Ramsar station, (0.958 and 0.949) for Arak station and (0.966 and 0.953) for Shiraz station have better performance than other studied models. Also, the evaluation of the impact of the input parameters in creating the scenario for the models shows that the parameters of relative humidity and air temperature had more important role than other input parameters. So that by adding parameters of relative humidity and air temperature, the accuracy of the model has increased. Therefore, these parameters are among the most key and important parameters of soil temperature.

The analysis and evaluation of soil characteristics has an important impact in the fields of hydrology, agriculture and climate. On the other hand, soil temperature has a direct relationship with the amount of moisture available to the plant, so that an increase in soil temperature can increase the transpiration rate of plants, and as a result, soil moisture decreases. Soil temperature is also an essential factor in agriculture because it determines whether plants can grow, and controls soil chemistry and biology and atmosphere-land gas exchange. Therefore, predicting soil temperature is very important for successful crop management and yield optimization. So, In this research, five data-driven methods of GPR, SVR, M5P, LR and MLP were used to predict soil temperature in Arak, Ramsar and Shiraz stations during the time period of 1990-2022. The obtained results were compared using statistical parameters and it was concluded that the 8th M5P scenario and the 8th LR scenario have shown the best performance in three stations with the lowest error compared to all scenarios. Therefore, the application of the mentioned models to predict the soil temperature has proper accuracy and is recommended for management and evaluation in terms of environmental and civil aspects.

Daily reference evapotranspiration prediction is a decisive and useful tool in sustainable agriculture and hydrological issues, especially in the design and management of water resources systems. The use of hybrid models with the help of climatic factors is an effective method in the daily reference evapotranspiration forecasting process. Therefore, in this study, the ability of the support vector regression model (SVR) and the combined model of support vector regression with the fruit fly algorithm (SVR-FOA) in estimating daily reference evapotranspiration in Ahvaz station during the period of 2000-2022 using four statistical criteria was evaluated. The inputs used included parameters of average temperature, minimum temperature, maximum temperature, average relative humidity, minimum relative humidity, maximum relative humidity, wind speed, and sunshine hours. The sensitivity analysis of the input parameters using Pearson's correlation coefficient also showed that among the input parameters, the parameters of sunshine hours and relative humidity were effective components in the prediction of evapotranspiration, thus reducing the error in all models. The obtained results showed that the sixth scenario of the SVR-FOA model provided the best performance with the lowest error (1.24 mm/day) compared to all models. Among the scenarios of the SVR model, the third scenario of the SVR model showed the lowest error (1.45 mm/day) compared to other SVR combinations. The results of this research showed that the sixth scenario of the SVR-FOA model had the best performance, and the fruit fly hybrid algorithm improved the performance of the support vector regression in estimating daily reference evapotranspiration.

Indiscriminate use of water resources and the occurrence of drought in recent years have caused many problems in the country's water resources. The increasing shortage of water resources and high irrigation costs require developing new irrigation methods for optimal water consumption, which can minimize the amount of water used to produce yields. Evapotranspiration is one of the most important parameters needed to estimate the water balance in any ecosystem. Evapotranspiration is an essential parameter in the hydrological cycle process in natural ecosystems, which links the water and energy balance of the earth's surface with the atmosphere. Reference evapotranspiration (ET0) plays an important role in the availability of water resources and stimulating the hydrological effect of climate change. Accurate estimation of ET0 is necessary for forecasting climate changes, predicting and monitoring droughts, assessing the lack of availability of water resources, assessing crop water needs, and planning irrigation. FAO's Penman-Monteith method is known as a standard reference method for estimating ET0. However, this model and, in general, water balance-based assessment methods require accurate and long-term meteorological data, which are not always and everywhere available. Therefore, alternative methods for predicting ET0 at different temporal and spatial scales should be developed, which are easily applied and require fewer input data without compromising the estimation accuracy. Also, due to the high rate of evapotranspiration in the coastal and central stations of the country, so far, few studies have predicted the ET0 parameter. Therefore, this study was carried out to predict daily reference evapotranspiration in Isfahan and Astara stations. The current study is forecasting daily reference evapotranspiration in two stations of Astara and Isfahan using Gaussian Process Regression (GPR), Support Vector Regression (SVR), M5P tree model, and M5Rules linear regression. For this purpose, the daily meteorological data of the stations including average temperature, minimum temperature, maximum temperature, average relative humidity, minimum relative humidity, maximum relative humidity, wind speed, and sunshine hours during the period of 1990-2021 as inputs to the models was used. Also, to evaluate the effectiveness of the models, the evaluation criteria of determination coefficient (R2), root mean square error (RMSE), Nash-Sutcliffe coefficient (NS), and Wilmott's index of agreement (WI) were used. The evaluation of the results of different scenarios of the GPR model in Astara station showed that the fifth scenario was recognized as the best scenario of this model due to having a lower error value (RMSE=1.52 mm day-1). For the M5Rules model, the fifth scenario has performed better than the other scenarios of the M5Rules model due to having fewer inputs and similar errors compared to the sixth to eighth scenarios (RMSE=1.42 mm day-1). In the M5P model, the fifth scenario has a higher accuracy than the other scenarios due to having a lower error value (RMSE=1.42 mm day-1). For the SVR model, the sixth scenario with the least error (RMSE=1.58 mm day-1) was selected as the best scenario compared to other scenarios of the SVR model. For the Isfahan station, for the GPR model, the fifth scenario has performed better than the other scenarios due to having fewer inputs. The comparison of M5Rules model scenarios also showed that the eighth scenario with RMSE=1.85 (mm day-1), had higher accuracy than other scenarios. The seventh scenario of the M5P model has performed better than other scenarios due to its RMSE=1.86 (mm day-1). Finally, the evaluation of SVR model scenarios showed that the eighth scenario with RMSE=1.88 (mm day-1) had a better performance than other scenarios. The comparison of the models used to predict daily reference evapotranspiration in Astara station showed that the fifth scenario of M5P and M5Rules models having evaluation criteria of R2=0.76, RMSE=1.42 (mm day-1), NS=0.7 and WI=0.89 had the highest accuracy compared to other models and showed the best performance. Also, the evaluation of the results of the models in Isfahan station showed that the eighth scenario of the M5Rules model, having the evaluation criteria of R2=0.8, RMSE=1.85 (mm day-1), NS=0.8 and WI=0.94 had the best performance compared to other models and the M5Rules model was selected as the best model. Also, the seventh scenario of the M5P model had almost the same performance as the eighth scenario of the M5Rules model and showed a good performance. Therefore, M5P and M5Rules models successfully predicted reference evapotranspiration. One of the limitations of the present study is the lack of access to dew point temperature and solar radiation data. Therefore, the use of these parameters is suggested for further studies.

Evaporation is one of the main components of hydrological cycle and one of the effective climatic variables in arid areas such as Iran. Accurate estimate of evaporation rate plays an important role in sustainable development and optimal management of water resources. Evaporation is one of the essential processes, because it depends on meteorological variables such as solar radiation, air temperature, wind speed, relative humidity and atmospheric pressure, which are related to the topography and the climate of the region. Class A pan-evaporation is one of the standard and direct tools for measuring evaporation, which is used all over the world due to its ease of application in determining evaporation. However, in most stations accurate evaporation recording is not practical due to instrument limitations and maintenance problems. On the other hand, the temporal and spatial distribution of evaporation stations compared to meteorological stations is limited, so according to the problems mentioned, the use of meteorological variables in estimating the rate of evaporation from the pan will be useful. In different regions, the impact of different climatic factors on changes evaporation from the pan has not be fully understood, so the relatively accurate estimation and prediction of this phenomenon is an effective step in the relevant fields. In recent years, for estimating the amount of evaporation from the pan, a variety of intelligent systems and software calculations such as data mining methods have been developed.

In this study, meteorological data of Tabriz station in the period of 2003 to 2018 have been used to estimate the evaporation values from the class A pan. For this purpose, a simple correlation between meteorological variables and evaporation from class A pan was created and based on the result of this correlation, in the studied station the minimum temperature and relative humidity were inversely and the maximum and average temperature were directly affected by evaporation. Thus, ten combined scenarios were defined and modeling was performed using Support vector regression (SVR), Gaussian process regression (GPR), M5tree, Random forest (RF) and Linear regression (LR) methods. It should be noted that in this study, 70% of the data were selected for training and 30% for testing. Finally, the performance of each method in estimating evaporation values was evaluated using root mean squared error (RMSE), mean absolute error (MAE), Nash- Sutcliffe coefficient (NS) and Akaike information criterion (AIC).

The results showed that GPR10 method with RMSE = 1.90 mm/day, MAE = 1.48, NS = 0.81 and SVR10 method with RMSE = 1.92 mm/day, MAE = 1.51, NS = 0.8 had reasonable performance in estimating the values of daily evaporation from class A pan. The GPR method showed its higher capability to estimate daily evaporation values in all definition scenarios with the least error and the most accuracy. The SVR model with appropriate results was in the second place. The results of statistical parameters for random forest model were even weaker than the results of linear regression. In general, scenario number 10 with all meteorological variables and scenario number 1 with only the input minimum temperature variable had the best and weakest results among all defined scenarios, respectively. Scenarios 6 to 10 have more accuracy and less error and modeling structures with the least number of variables has the least accuracy. Also, wind speed and solar radiation variables were introduced as the most effective factors in estimating the evaporation rate from class A pan.

Evaporation is one of the important processes that cause the losses of half of precipitation in arid and semi- arid regions. Accordingly, knowledge of the amount of evaporation and its modeling as one of the most important hydrological variables in agricultural research and factors related to water and soil of great importance. So, accurate estimation of this phenomenon is essential. In this study, meteorological data from Tabriz station were utilized to assessment capability of machine learning methods. Evaporation values were estimated using five data mining methods including SVR, GPR, M5, RF and LR. Conclusively, the results of evaluation criteria indicated that GPR and SVR models using all variable of meteorological data performed more accurate than others. Finally, both of them are recommended to estimate the amount of evaporation from class A pan.

Wind erosion is a key process in land degradation worldwide, especially in arid and semi-arid regions of Iran. This phenomenon is affected by many soil characteristics. The main objective of this study was to estimate the wind erosion threshold velocity using easily measurable soil characteristics along with data mining methods. For this purpose, wind erosion threshold velocity was measured in 100 areas in Fars province using a portable wind tunnel. Wind erosion threshold velocity was predicted by a support vector regression algorithm using easily measurable soil properties. In this regard, a genetic algorithm was used in order to obtain a set of parameters effective in estimating wind erosion threshold velocity. The results showed that the characteristics of soil moisture (r = 0.77), the size distribution of soil particles including the mean weight diameter of aggregate (r = 0.87) and the wind-erodible fraction of soils (r = -0.81), penetration resistance (r = 0.75), and organic matter (r = 0.33) have a high and significant correlation with wind erosion threshold velocity and play a key role in determining the threshold velocity of wind erosion in the region. According to the evaluation criteria, the combined support vector regression model with the genetic algorithm had the best performance and the most accurate estimate for wind erosion threshold velocity (RMSE = 0.53 and R2 = 0.92) and can be a promising method for estimation of wind erosion threshold velocity.

Drought is one of the most important natural disasters with devastating and harmful effects in various economic, social, and environmental fields. Due to the repetitive behavior of this phenomenon, if the appropriate solutions are not implemented, its destructive effects can remain in the region for years after its occurrence. Most natural disasters, such as floods, earthquakes, hurricanes, and landslides in the short term, can cause severe financial and human damage to society, but droughts are slow-moving and creepy in nature, and their devastating effects appear gradually and over a longer period of time. Therefore, by modeling drought, it is possible to provide plans for drought preparation and reduce the damage caused by it. In this study, computational intelligence algorithms of Multi-Layer Perceptron neural network, Generalized Regression Neural Network, Support Vector Regression with support kernel, and Support Vector regression with the proposed kernel (Support Vector) Regression New kernel has been used to model the drought using the Standardized Precipitation Index. The modeling results, in most cases, showed better performance of the proposed SVR_N model than other models. The values of RMSE and R2 were 0.093 and 0.991, respectively, and the GRNN, MLP, and SVR models performed better in modeling after SVR_N, respectively. Modeling of drought phenomenon in modeling is supported by vector regression method.

In this study, machine learning based models (linear regression, k-nearest neighbor, support vector regression, random forest) were evaluated to estimate carbon sequestration and soil respiration in the forests of eastern Mazandaran province. After identifying the sampling points, in each plot, the diameter and height of the trees were measured and the above-ground biomass of the trees was calculated using allometric models of Hyrcanian forest. Soil samples were taken from 0-20 cm depth and soil respiration was measured by CO2-port. Soil respiration was estimated using bulk density, moisture content, texture, soil temperature, total nitrogen, available phosphorus and potassium, organic carbon and above-ground trees biomass. Soil carbon sequestration was estimated using soil temperature and moisture and above-ground tree biomass. Random forest model (RMSE = 10.47 and R2 = 0.82) and support vector machine (RMSE = 0.77 and R2 = 0.90) had the highest performance in estimating carbon sequestration and soil respiration, respectively. Soil moisture had the highest relative importance in estimating carbon sequestration (random forest model) and soil respiration (support vector machine model). According to the obtained results, it is possible to estimate the amount of carbon sequestration and soil respiration in the forest with appropriate accuracy by having the above-ground biomass of trees and basic soil characteristics.

Accurate prediction of streamflow is crucial for water resources management, especially for the prediction of floods and soil erosion. In the current study, the capability of three machine learning (ML) methods, including Support Vector Regression (SVR), Artificial Neural Network with Backpropagation (ANN-BP), and Gradient Boosting Regression (GBR) was investigated using meteorological observations and MODIS snow cover data to forecast daily streamflow in two different basins, namely Latian and Navroud. For model development, four major predictors, including daily rainfall (P), maximum temperature (Tmax), minimum temperature (Tmin), and the Normalized Difference Snow Index (NDSI) from the MODIS satellite were used from 2000 to 2018. The performance of these models was evaluated using statistical indices. Simulation results revealed that all models presented satisfactory results in simulating daily streamflow using meteorological predictors, and the efficiency of all applied models was improved when the NDSI index was applied as an additional predictor. The best performance was observed in GBR among all studied models in both basins, whereas SVR revealed the lowest performance in forecasting streamflow for both validation and calibration steps in most cases. In general, the simulation results demonstrated higher accuracy in Latian basin than Navroud basin in both calibration and validation steps. The best performance among all models was observed using GBR with R = 0.85, NS=0.72, and RMSE = 3.43 m3/s using the NDSI index in Latian basin indicating the significant effect of snowmelt on streamflow generation in snowmelt-dominated regions.

Surface soil moisture is an important variable in nature's water cycle and can be affected by various factors, including temperature and soil characteristics. The use of ground sensors for measuring moisture can lead to spending time and expense and inappropriate distribution of samples on large scales. Therefore, Remote sensing observations can be an important tool in estimating soil moisture. The present study aims to use the TOTRAM model using Landsat 8 images and the SVR method using Sentile 1 images to estimate soil moisture.

In the present study, two TOTRAM methods based on pixel distribution in LST- VI space and the SVR method were used to extract soil moisture using the SAR technique and Sentinel 1 data. To implement the TOTRAM method, Landsat 8 images related to 4/29/1398 and 5/30/1398 are downloaded and after extracting NDVI and LST maps, The correlation between the dependent variable of moisture and independent temperature variables and vegetation variables has been investigated using Geographically weighted regression (GWR). To implement the SVR method after acquiring Sentinel 1 images related to 31/05/1398 and 27/04/1398, Soil Moisture Data Product FLDAS and 500 meters product of Modis Satellite (MCD12q1) were called to classify land cover in the Google Earth Engine system, and maps related to soil moisture were extracted. After extracting the moisture maps the distribution of moisture using the local Moran index has been investigated. By defining this index, positive values ​​for this index represent the cluster of distribution.

Examination of the soil moisture map obtained by the SVR method showed the concentration of moisture in areas with vegetation and water and the change in moisture status from July to August was visible. The humidity pattern has shown the reflection of the precipitation pattern so that maximum precipitation and humidity were observed in April and in summer both precipitation and humidity components decreased. Examination of the TOTRAM method and application of the GWR method has shown a complete correlation between NDVI LST and moisture. However, the correlation between LST and humidity with B (values) and standard error (SE) of 0.995 and zero corresponding to July and 0.981 and zero corresponding to August showed the highest correlation with vegetation variable with moisture dependence parameter, which this correlation In August, with increasing the coefficient of determination of R2 to 0.997 and a significant decrease of NDVI to the value of 0.415 in July, it has increased much more. Application of Moran local index with values ​​less than 0.05 for p-value and positive values ​​for z and near positive number 1 for Moran index showed the cluster distribution of moisture variable.

The results of TOTRAM and SVR methods showed the dependence of soil moisture status on conditions and cluster moisture distribution. According to the correlation coefficients of geographical regression, there is a greater correlation between temperature and humidity variables, especially in August, due to the decrease in vegetation density. The results of the SVR algorithm maps showed that in areas with the presence of vegetation, especially dense vegetation, we see an increase and with increasing temperature, we see a decrease in humidity. Also, the coordination of moisture patterns of the SVR algorithm and precipitation showed a direct relationship between moisture and precipitation. Considering that the SVR method uses parameters such as radar scattering intensity and land cover classification, as well as the use of Sentinel 1 radar images by this algorithm, more accurate results can be expected from this algorithm.

Prediction of daily evaporation is a valuable and determinant tool in sustainable agriculture and hydrological issues, especially in the design and management of water resources systems. Therefore, in this study, the ability of artificial intelligence models of multi-layer perceptron (MLP), support vector regression (SVR), and the hybrid model of support vector regression-firefly optimization algorithm (SVR-FFA), to predict daily evaporation at Takab Station during the period 2002-2020 based on four statistical criteria have been assessed In all three models, the best scenario was the model whose input included the parameters of average temperature, minimum temperature, maximum temperature, average relative humidity, minimum relative humidity, maximum relative humidity, wind speed, and sunny hours. Among the input parameters, the sunny hours was one of the effective components on the evaporation prediction, which reduced the errors in all models. The results showed that the sixth scenario of the MLP model provided the best performance with the least error (2.18) compared to other models. It was also concluded that the sixth scenario of the SVR-FFA model had a lower error (2.20) than the other models. Among the SVR model scenarios, the sixth scenario showed the lowest error (2.27) compared to other SVR combinations. The results of this study showed that the sixth scenario of the MLP model had the best performance and the hybrid firefly algorithm improved the performance of support vector regression in estimating daily evaporation.

The effect of precipitation changes on water resources, agricultural production reveals the need for accurate methods for precipitation forecasting. In this research, one of the soft computing methods was developed in order to forecast precipitation with the data reduction approach. Input data of model was mean air temperature, dew point temperature, mean sea level pressure, mean station pressure, mean relative humidity and mean wind speed at Tabriz, Ahar and Jolfa stations. The method used in this study includes Epsilon and Nu support vector regression. In all studied stations, the use of Nu support vector regression compared to Epsilon reduced the error so that UII values ​​with Nu support vector regression in Tabriz, Ahar and Jolfa stations were decreased 19.19, 5.88 and 15.78%, respectively. The results indicate the limitation of using the data reduction approach for data with KMO factor lower than 0.5, which included Tabriz and Ahar stations. Principal component analysis in both types of support vector regression increased the performance of the model so that in Jolfa station by using principal component analysis d values ​​of Epsilon and Nu support vector regression increased by 16.6 and 17.5%. Execution of Verimax rotation in preprocessing of input data to regression was stronger than principal component analysis. In this regard, RRMSE and RMSE values ​​in Jolfa station using Epsilon support vector regression were decreased 6.66 and 6.45%. Therefore, principal component analysis is a suitable tool to improve the performance of soft computing methods by regarding the relevant constraints.

Cation exchange capacity (CEC) is one of the most important soil chemical properties that plays an important role in soil fertility. However, standard laboratory methods for measuring CEC are difficult, costly, and time-consuming. The aim of this study was to use new methods such as 1) Pedotransfer Functions (PTF) based on the basic soil properties using Multiple Linear Regression (MLR), 2) soil spectroscopy (Vis–NIR, 400 – 2500 nm) using Partial Least Squares Regression (PLSR), and Support Vector Regression (SVR), for estimating soil CEC. Also, from the regression coefficient analysis, key wavelengths were introduced. For this purpose, CEC was measured using the sodium acetate method for 72 soil samples collected, and spectral reflection of soil samples was determined using spectroscopy. These methods are made from a calibration set (70% of data) and evaluated with a validation set (30% of data). The results showed that Vis - NIR method performed better than PTF. In this study, wavelength ranges around 566, 854, 1354, 1418, 1906, 2071, 2203, 2319, and 2341 nm were investigated as the key wavelengths for estimation of CEC. Furthermore, the results of prediction models showed that SVR has a better performance than PLSR. This study proved that Vis-NIR is a promising method for soil CEC estimation.

In this research, monthly temperature, precipitation and discharge values of Jamishan Dam in 1988-2017 have been considered as the base period. Simulation of resources and uses of Dam-d-Gomishan basin in WEAP model with existing cropping patterns in the region was investigated. To evaluate the effect of climate change on precipitation and temperature parameters in this region, the outputs of the RCP8.5 scenario of the HADGEM2_ES models were used from the cmip5 fifth report series and the output of these models was downscaled for the desired area. In this study, using the method of change factor downscaling the climatic model and monthly temperature and precipitation parameters of Jamishan Dam for the period 2021-2050 were generated. In order to investigate the runoff of the area due to climate change, SVM, GEP and IHACRES models were investigated and compared. The results of the climatic model show an average temperature increase of 1.5 °C and an increase in precipitation of 5%. In general, results of discharge forecasting in all three models indicate a decrease in runoff that the highest runoff reduction was related to SVM with 21.6% and the lowest runoff reduction was related to IHACRES with 4%. In this study, IHACRES and GEP models have more favorable accuracy than SVR method. Simulation of different cropping patterns in WEAP shows that the highest supply is equal to 74.9% in GEP model and the lowest supply is equal to 70% in SVR model which is related to cropping pattern 1 and cropping pattern 3, respectively.

Accurate and correct prediction of surface water flow plays an important role in the principled planning and proper management of water resources. To achieve this, various prediction models using mathematical relationships based on hydrological information can be used. In this study, monthly discharge of Polechehr hydrometric station for a 48-year has been used (Sep. 2018-October 1971). Two main scenarios with and without pre-processing (standardization), two time series or non-time series approaches were considered. Also, two cases with and without feature selection have been considered by a random forest algorithm. In all cases, 80% and 20% of the data are intended for model training and testing, respectively. The entire coding process is done in the Python programming platform. Genetic algorithm was used to optimize the parameters of the support vector regression method. The results showed that standardization, non-time series approach, reducing the dimensionality of the model input to select and also using genetic algorithm to optimize the parameters of the support vector regression model have the greatest effect on prediction accuracy, respectively. So that the highest coefficient of explanation for training data is 0.85 and for testing is equal to 0.6.If standardization is not applying on the data, adopting a time series approach and feature selection will lead to better results in predicting the SVR model, and also the use of genetic algorithm optimizer compared to the simple model will have a significant effect on improving results.

Piano-Key weirs are a new type of overflow that are designed to increase the drainage capacity of dams and canals. If the keys forming this overflow model are placed on an arc of a circle, it is called curved piano-key weir. In this research, the performance of three models of Intelligent Support Vector Regression (SVR), Support Vector Regression- Firefly (SVR-FA) and Support Vector Regression- Grasshopper (SVR-GOA) to predict curved piano-key weir flow rate were evaluated. Determination Coefficient (R2), Mean Squared Error (MAE), Root Mean Squared Error (RMSE), and Scattering Index (SI) are four statistical indicators that are used to determine the accuracy of intelligent models. The result of these evaluation criteria during the test period is that the SVR-GOA model with values ​​of 0.99275, 0.01220, 0.00026 and 0.00046 compared to the SVR-FA model with values ​​of  0.95666, 0.03844, 0.00200 and 0.00342 and SVR with values ​​of 0.94249, 0.04013, 0.06027 and 0.00410 for R2, MAE, RMSE and SI indicators, are more accurate in predicting curved piano-key weir flow rate

Correct estimation of the river sediment volume is important for many water resources projects.In this study, the empirical equations of sediment transport, support vector regression (SVR), and developed k-nearest neighbor regression (KNN) were used in order to estimate suspended sediment load in the Sistan River. In this regard, in addition to the maximum temperature, minimum temperature, and dischargeand Suspended Loud In the period 1374 to 1390, it was paired on 1682, discharge, the classified discharge was used as effecting variable to suspended sediment load modeling. For each of the input combinations in support vector regression and development k-nearest neighbor regression the best structure of the regression model determined based on the performance criteria. The Result showed that the Toffaleti method is the best method between empirical equations with R2 equal to 0.705 and RMSE equal to 66558 Ton/day. Also, the support vector regression model with discharge, minimum temperature, maximum temperature and classified discharge as the set of input data is the best model in estimation of suspended sediment load with R2 equal to 0.96 and RMSE equal to 2809.3 Ton/day. The results indicate that the regression method estimate suspended sediment load much better than empirical equations in the Sistan River.