Soil Physics, Erosion and Conservation
Heidar Ghafari; Hadi Ameri khah
Abstract
Introduction: The processes of soil erosion and sediment transport along rivers are the main causes of some socio-economic and environmental problems, such as a reduction in water quality, storage capacity of dams, destruction of aquatic habitats, failure of hydroelectric power plants, and soil degradation. ...
Read More
Introduction: The processes of soil erosion and sediment transport along rivers are the main causes of some socio-economic and environmental problems, such as a reduction in water quality, storage capacity of dams, destruction of aquatic habitats, failure of hydroelectric power plants, and soil degradation. Therefore, understanding the sedimentation status of watersheds is crucial for the effective management of soil and water resources. However, due to the lack of technical and human resources, continuous recording of sediment data is not possible in most sediment measuring stations, and sediment data are recorded only for a few days. In such a situation, a model that can estimate the amount of sediment load using auxiliary variables such as stream discharge and rainfall becomes crucial. Today, it is believed that techniques based on artificial intelligence have a much greater ability to uncover hidden relationships between variables than classical methods and are thus very useful and effective in modeling natural processes.Materials and methods: In this study, various machine learning techniques, including Artificial Neural Network (ANN), Adaptive Fuzzy-Neural Inference System (ANFIS), and Random Forest (RF), were used for sediment load modeling and sediment forecast for days without measurements. To achieve the research objectives, long-term meteorological and hydrometric data ranging from 2000 to 2020 were collected from related organizations and pre-processed before entering the model. The input variables for the models included 24-hour rainfall, flow rate, normalized difference vegetation index, maximum and minimum temperature, and daily suspended sediment as the dependent variable. Prior to modeling, the entire dataset was divided into two parts, training and testing, in a 70:30 ratio. Relationship modeling was performed using the training data, and model validation was conducted using the test dataset. The efficiency of the models was evaluated using two indicators, the coefficient of explanation (R2) and the root mean square error (RMSE). Additionally, morphometric parameters such as form factor (FF), drainage density (DF) coefficient, and relief ratio (RR) were utilized in modeling.Results and discussion: The hydrological analysis of the basin revealed that the highest annual amount of rainfall and erosivity index were recorded at the Sheyvand station in the east of the basin, while the lowest values were observed at the Ramhormoz station. The highest average monthly flow rate of 5.8 cubic meters per second was obtained at the Manjeniq station in April, and at the Mashin station, the highest average monthly flow rate of 8.8 cubic meters per second was recorded in December and January. Morphometrically, the studied basin belonged to the class of elongated basins, sloping basins in terms of relief, and the medium class in terms of drainage density. Analysis of the time series of NDVI index showed that the highest vegetation cover occurred in March, while the lowest values were recorded in September and October. The annual trend of the vegetation index indicated an overall improvement in vegetation cover in the region from 2000 to 2020, with the NDVI value increasing from 0.15 to 0.22.Among the different machine learning techniques studied, the Artificial Neural Network (ANN) model had the highest coefficient of explanation (R2=0.87) and the lowest RMSE for both sediment measuring stations in the region, making it the best model. The optimal inputs for the neural network model at Mashin station were daily average flow adjusted by the basin shape factor, daily rainfall, last day's rainfall, daily minimum temperature and daily maximum temperature. For the Manjeniq station, the optimal inputs were daily average flow, daily rainfall, last day's rainfall, cumulative rainfall for the past two days, and cumulative rainfall for the past three days. The NDVI index was removed from the model due to its low significance. The Random Forest (RF) model ranked second, and the Adaptive Fuzzy-Neural Inference System (ANFIS) model ranked third, with weak performance, especially for the Mashin station, where out-of-range errors occurred.Temporal analysis of sediment values showed that the highest sediment production occurred in December and January for Mashin station and in April for Manjeniq station. The highest production of sediment occurred in 2006 and 2002, and the trend of changes from 2011 to 2018 showed a decline, attributed to consecutive droughts and lack of rainfall. The annual average sediment production calculated using the values estimated with the neural network model was 88017 tons, equivalent to 1 ton per hectare per year. Conclusion: Overall, this research demonstrated that machine learning methods, especially the neural network model, are highly effective for modeling and predicting sediment on a daily scale. These methods can compensate for the lack of sediment measuring facilities and equipment in most existing hydrometric stations in the country and eliminate the need for continuous sediment data and other water quality parameters.
Negar Hafezi; Mohammad Javad SheikhDavoodi; Houshang Bahrami; Seyed Enayatallah Alavi
Abstract
Introduction Sugarcane is a tropical, perennial grass that forms lateral shoots at the base to produce multiple stems. It is the main source of sugar production and one of the most important sources of energy production in the world. Today, the use of artificial intelligence and data mining findings ...
Read More
Introduction Sugarcane is a tropical, perennial grass that forms lateral shoots at the base to produce multiple stems. It is the main source of sugar production and one of the most important sources of energy production in the world. Today, the use of artificial intelligence and data mining findings to help predict product production is considered. Determining the relationship between inputs and outputs of production process using artificial intelligence (AI) has drawn more attention rather than mathematical models to find the relationships between input and output variables by training, and producing results without any prior assumptions. The adaptive neuro-fuzzy inference system (ANFIS), as a form of AI, is a combination of artificial neural network (ANN) and fuzzy systems that uses the learning capability of the ANN to derive the fuzzy if-then rules with appropriate membership functions worked out from the training pairs, which in turn leads to the inference.Particle swarm optimization (PSO) is an algorithm modeled on swarm intelligence, in a search space, or model it finds a solution to an optimization problem and predict social behavior in the presence of objectives. The PSO is a population-based stochastic computer algorithm, modeled on swarm intelligence. Swarm intelligence is based on social psychological principles and it provides insights into social behavior, also helps to many engineering applications. Feature selection is becoming very important in predictive analytics. Indeed, many data sets contain a large number of features, so we have to select the most useful ones. One of the most advanced methods to do that is the genetic algorithm (GA). Genetic algorithms can select the best subset of variables for predictive model. The purpose of this research is to evaluate the applicability of one artificial intelligence technique including adaptive neuro-fuzzy inference system and also combining this technique with particle swarm optimization to increase the accuracy and speed of training of the neuro-fuzzy system in prediction of yield and recoverable sugar percentage (R.S%) of sugarcane. Materials and Methods In this paper, one main pattern of adaptive neuro-fuzzy inference system (ANFIS) and one synthetic model of adaptive neuro-fuzzy inference system with particle swarm optimization (PSO) were used to predict the studied properties by MATLAB version 2017. Initial data for this study were collected from Debal-Khozaie Agro-industry Company in Khouzestan province, Iran. The actual data for the seven periods of sugarcane harvest from 2010 to 2017 were used for modeling. The studied parameters included a set of agronomic factors, soil characteristics, irrigation and climate in the study area. The test data sets were used for comparison of selected ANFIS and ANFIS with PSO, as well as for the observation values. This comparison was performed by using three statistical indices: Determination Coefficient (R2), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). Results and DiscussionFrom all of the studied parameters, eleven parameters were selected as the effective features by the binary genetic algorithm (BGA). In feature selection, the function to optimize is the generalization performance of a predictive model. More specifically, in this method, purpose was to minimize the error of the model on an independent data set not used to create the model. The data were randomly divided into two groups: training and testing. Each pattern was modeled separately and then the results were compared. The results showed that the combination of adaptive neuro-fuzzy inference system with particle swarm optimization algorithm (ANFIS-PSO) had better performance in predicting cane yield and recoverable sugar percentage. In ANFIS-PSO model the root mean square error, mean absolute percentage error and coefficient of determination values were found 0.0181, 0.0217, 0.9237 and 0.0086, 0.0138, 0.9847 respectively for two variables of cane yield and recoverable sugar percentage. In relation to the predicted cane yield by the neuro-fuzzy network with particle swarm algorithm, it can be concluded that among the effective factors, with increasing plant age and use of resistant varieties, the amount of yield was decreased and increased, respectively. Conclusion The hybrid pattern of adaptive neuro-fuzzy inference system with the particle swarm optimization has been directed against the mere neuro-fuzzy system to a more accurate and stronger solution. Indeed, it can be concluded that ANFIS model with the PSO has the ability for precise estimation of sugarcane yield and recoverable sugar percentage.
L. Rasoli; K. Nabiollahi; R. Taghizadeh Mehrjardi
Abstract
Introduction Rapid population growth in developing countries implies that more food will be required to meet the demands of this population. Wheat as one of the most important grain crops in the world is a great source of food for human which is planted under a wide range of environments and its production ...
Read More
Introduction Rapid population growth in developing countries implies that more food will be required to meet the demands of this population. Wheat as one of the most important grain crops in the world is a great source of food for human which is planted under a wide range of environments and its production influences on local food security. The production of wheat per unit area in Iran is low compared to developed countries in the world. One of the main causes for this low yield is that the suitable land for planting has not been recognized. Therefore, to overcome this problem, land suitability assessment is needed, which can help to increase crop yield. The first step in agricultural land use planning is land-suitability assessment which is often conducted to determine which type of land use is suitable for a particular location Digital mapping approach have been applied to link between soil observations and auxiliary variables to understand spatial and temporal variation in soil class and other soil properties. Little attempt has been made for using Digital mapping approach to digitally map land suitability classes Therefore, this paper applied land suitability assessment framework and digital soil mapping approach to map land suitability for rain-fed wheat in Kurdistan province. Materials and Methods The study area is located in Kurdistan Province, western Iran. It surrounds the city of Ghorveh and covers a region of 6500 ha. The climate is semi-arid whose features can be performed using a cold and rainy winter and a moderate and dry summer. The mean yearly rainfall is 369.8 mm and over 90% of the rain falls between November and March. The mean temperature (10.8℃) is relatively cool. Soil moisture and temperature regimes are Xeric and Mesic, respectively. The physiography units include piedmont, fan, hills, and mountain and slope varies from gentle to very steep. At first land unit component map was prepared by Mahler physiography method, then, 17 representative profiles in each land unit component were dug and described. 105 auger samples also were taken at three depths (0-20, 20-50 and 50-100 cm). Soil texture, acidity, organic carbon, CaCO3, gypsum, ESP, electrical conductivity and gravel were measured in all soil samples. Topography and climate data were also recorded. Numeric ratings of soil, topography and climate parameters based on land requirements of wheat were determined and land suitability index using parametric method were calculated. Then land suitability classes of wheat were determined. A set of auxiliary variables (i.e. land unit component, terrain attributes and remotely sensed data) to predict land suitability classes of rain-fed wheat. In order to generate land suitability class map, artificial neural network were applied to make relation between auxiliary variables and land suitability classes. Results and Discussion The results showed that the area has about 36.61% N2 class, 40.32% N1 class and 22.53% S3 class. The validation results of the model based on the statistical indices including root mean square error, mean error, and determination coefficient (6.56, 4.81, 0. 68, respectively), indicates that the artificial neural network model has suitable accuracy. Auxiliary data including MrVBF index, LS factor, MRRTF index, slope, Land unit component, VDCN and band 2 were the most important for prediction of wheat land suitability index in digital method. The major limitation of the study area to plant rain-fed wheat were rainfall in the flowering stage, sever slope, shallow soil depth, high pH and gravel. Therefore, to increase production and sustainable agricultural system it is suggested land improvement operations such as terracing, decreasing pH, supplementary irrigation and gathering gravel. The highest values of rain-fed land suitability index were observed in the units physiographic of river plain and plateau, while the lowest value were observed in the units physiography of mountain and hill which had high slope, shallow soil and high gravel. These results were confirmed by one-way ANOVA and Duncan tests. Conclusion Based on the results of statistics indices artificial neural network had suitable accuracy for predicting land suitability index of wheat. In general, the study area, because of limitation of sever slope, shallow soil, high pH, and gravel, has low land suitability index for rain-fed wheat. Hence, to improve land suitability of the study area and increasing its production, suitable land improvement operations is required.
Kamran Azizi; Kamal Nabiollahi; Masoud Davari
Abstract
Introduction Soil salinity and alkalization are recognized worldwide as a major threat to agriculture, particularly in arid and semi-arid regions. To manage these soils a lot of data are needed and laboratory measurement is costly and time-consuming. Therefore, indirect methods that are cheap, fast and ...
Read More
Introduction Soil salinity and alkalization are recognized worldwide as a major threat to agriculture, particularly in arid and semi-arid regions. To manage these soils a lot of data are needed and laboratory measurement is costly and time-consuming. Therefore, indirect methods that are cheap, fast and easy to access are one of the research priorities. One of these methods is visible near infrared diffuse reflectance spectroscopy. Visible and near infrared diffuse reflectance spectroscopy is a time and cost-effective approach that has been successfully used for characterizing soil properties. Materials and Methods The study area is located in Kurdistan Province, about 20 km northeast of Ghorveh city, west of Iran, and covers 260 km2. Average annual precipitation and temperature are 369.8mm and 10.8 °C, respectively. Soil moisture and temperature regimes are Xeric and Mesic, respectively. In the study area, 100 soil samples were collected (0–30 cm depth). The main land use types consist of cropland and rangeland. The soil samples were air-dried at room temperature and then, passed through a 2mm sieve. EC, pH, SAR, OC, CaCO3 and ΔMWD were measured. Sodium Adsorption Ratio (SAR) was calculated using results from the saturated paste extracts of sodium, calcium, and magnesium. The stability aggregate was measured using the difference between distributions of particle size in dry and wet sieve methods. Spectral analysis of soil samples was done using a spectrophotometric instrument with a wavelength of 350 to 2500 nm and recorded using RS3 software. After recording the spectra, different preprocessing methods were evaluated. Two models of multiple linear regression and artificial neural network were used to predict soil properties using spectral data. Results and Discussion The soil salinity of the study area ranged between low and high. The highest amount of salinity was observed in the center, south and southwest of the study area and the least amount of salinity was observed in northwest, southeast, northeast and north. The maximum amounts of acidity and sodium adsorption ratio showed that the central part of the study area has saline and sodium soils. The results showed that the best method for preprocessing of spectral data is the 1st Derivative + Savitzky-Golay filter + Mean center + SNV. The Pearson correlation coefficient between the soil properties and the spectral reflection values for each wavelength in the range of 2450-400 nm showed that there is a relatively high correlation between the measured characteristics and the spectral values of the soil. The results showed that the correlation coefficient can be positive or negative. The maximum positive correlation coefficients for electrical conductivity, soil acidity, sodium adsorption, organic carbon, calcium carbonate and aggregate stability at the wavelengths 1229, 2397, 2399, 1298, 2090, 2014, and two spectra 2257 and 660 were 0.45**, 0.43**, 0.46**, 0.61**, 0.53** and 0.40**, respectively. The maximum negative correlation coefficients for electrical conductivity, soil acidity, sodium adsorption ratio, organic carbon, calcium carbonate and aggregate stability at the wavelengths 630, 2289, 630, 1904, 1379 and 2107 were -0.47**, -0.42**, -0.44**, -0.46**, -0.55** and -0.44**, respectively. Based on the determination coefficient statistic, artificial neural network model (0.88, 0.25, 0.59, 0.68, 0.52 and 0.48 to electrical conductivity, PH, SAR, calcium carbonate and aggregate stability, respectively) had better results compared to the multiple linear regression model (0.45, 0.13, 0.23, 0.66, 0.48 and 0.28 to electrical conductivity, PH, SAR, calcium carbonate and aggregate stability, respectively). Conclusion In this study, visible near infrared diffuse reflectance spectroscopy was evaluated to estimate some properties of salt-affected soils. After recording the spectral data, the continuity curve and pre-processing of spectral data were performed. The results showed that the best method for pre-processing of spectral data is the first derivative + Savitzky filter and Glair + Mid filter + Normal standard variable. Multiple linear regression and artificial neural network models were used to estimate some properties of salt-affected soils (EC, pH, SAR, OC, CaCO3 and ΔMWD) using spectral data. Based on the statistics of mean error, root mean squared error, and correlation coefficient, the artificial neural network model had better results in estimateing the properties of salt-affected soils compared to the multiple linear regression model. Therefore, based on these findings it is suggested that soil spectral data be used as an indirect method to the estimate soil properties.
Behrooz pourmohamadali; M.H. Salehi; S.J. Hosseinifard; H. Shirani; I. Esfandiarpour Borujeni
Abstract
Introduction In recent decades, due to a significant increase of pistachio cultivation and uncontrolled exploitation of groundwater resources as well as reducing rainfall precipitation, groundwater level has dropped and the quantity and the quality of water has also been reduced. Therefore, agricultural ...
Read More
Introduction In recent decades, due to a significant increase of pistachio cultivation and uncontrolled exploitation of groundwater resources as well as reducing rainfall precipitation, groundwater level has dropped and the quantity and the quality of water has also been reduced. Therefore, agricultural producers, researchers and policy makers need to pay more attention to appropriate land management as an important strategy to achieve greater yield per unit area and optimal use of soil and water resources. Crop yield prediction regarding its temporal and spatial variations has an important role in developing proper management programs. However, few studies have been carried out in relation to pistachio yield prediction using an acceptable range of features on regional scale. In the present study, pistachio yield modeling was performed by multivariate linear regression and artificial neural networkbased on soil, water and management features. Materials and Methods 129 orchard plots in different areas of Rafsanjan and Anar were identified and selected. The study area is located between 54° 56′ and 56° 41′ E, 29° 54′ and 31° 13′ N. Soil sampling, was performed from the areas under pistachio canopy and three soil depths of 0 to 40, 40 to 80 and 80 to 120 cm in each plot, fully expanded sub-terminal leaflets were randomly collected from non-fruiting branches, during the late July through August. Irrigation water of all orchards was also sampled. Moreover, for each orchard, a questionnaire was prepared to collect management and yield data. Soil quality indicators including particle size distribution, pH in saturated soil paste, electrical conductivity of saturated extract, soluble sodium, soluble calcium, soluble magnesium, available phosphorus and available potassium were determined for soil samples. The concentrations of phosphorus, potassium, iron, zinc, copper, manganese, calcium and magnesium in leaf samples and electrical conductivity in water samples, were also calculated. Finally, a dependent variable (pistachio yield) and 50 independent variables including soil, water and plant characteristics were used for modeling. For this purpose, stepwise multiple linear regression and artificial neural network technique were applied. Then, the study area was divided into 4 parts with the highest pistachio orchards densities and regression models were run for each part, separately. The ability of models to yield prediction was evaluated using the root mean square error (RMSE), relative root mean square error (% RMSE), adjusted coefficient of determination (adj - R2) and Durbin - Watson statistic (D – W). Results and Discussion The average of yield in the study area is about 1,700 kilograms per hectare. Results indicated that multiple linear regression could explain only 26 percent of the pistachio yield variation, but its accuracy increased when data became more homogeneous via dividing the study area into four parts. The model adjusted-R2 for Noogh, Anar, eastern suburbs and western suburbs orchards rose to about 92.4, 81.5, 95 and 53.6 percent, respectively. In all regression models except the model of western suburbs, at least one of the characteristics associated with irrigation water was significant. Artificial neural network with 9 neurons in a hidden layer, Tangent - sigmoid activation function and Levenberg - Marquardt training function, has a 98.3 percent accuracy in predicting pistachio yield in the study area (% RMSE = 13.8). Conclusion Multivariate linear regression model did not accurately predict the pistachio yield for the whole of study area whereas increasing data homogeneity and decreasing sources of variations, reduced complexity of relationships between features which resulted in increasing of the efficiency of linear regression to modeling these relationships. These models were highly sensitive to irrigation water features. Therefore, special attention should be paid to modern irrigation techniques and proper management approaches in order to enhance water efficiency. Overall, artificial neural network had greater accuracy compared to multivariate linear regression for pistachio yield modeling. This indicates the existence of non-linear and complex relationships between pistachio yield and the factors affecting yield and also the necessity of using modern and robust data mining tools for crop yield estimating. It seems that artificial intelligence techniques can be used as an efficient tool for developing proper management programs.