جستجوی مقالات مرتبط با کلیدواژه « پرسپترون چند لایه » در نشریات گروه « آب و خاک »

This study was conducted to evaluate the relationship between leaf iron and some easily-available soil properties in citrus orchards in the southern region of Kerman province by artificial neural network modeling and stepwise regression. For this purpose, 40 orchards were selected from the study area and the physical and chemical properties of soil and iron in the plant leaves were measured. Using artificial neural network in different models with different data from soil properties as input and leaf iron as output, the ability of these models to predict leaf iron concentration was evaluated. The results showed artificial neural network with variables of organic carbon, pH, clay, phosphorus, TNV and electrical conductivity with explanation coefficient of 0.86 and 0.81 and root mean square error (RMSE) of 14.60 and 20.13 mg.kg-1 for data Training and testing were the best models in estimating leaf iron. Comparison of regression and neural network models in the test data showed that the neural network had a higher accuracy with an explanation coefficient of 0.81 than stepwise regression with an explanation coefficient of 0.2. The amount of RMSE in the neural network also improved and increased from 27.72 mg.kg-1 in the stepwise regression model to 20.13 mg.kg-1 in the neural network. Artificial neural networks have been able to predict the iron in plant leaves based on the easily-available properties of the soil, so that by choosing organic carbon as the input of the first model to the best model by selecting organic carbon, pH, clay, phosphorus, TNV and electrical conductivity, model accuracy increased.

Short-term prediction of minimum temperature is important in mitigation chilling and frost injury in agriculture. In current study, the frequency of early autumn and late spring frosts in Sanandaj synoptic station, Iran were worked out. Then, using six variables, i.e. dry and wet bulb temperature, relative humidity, wind speed, wind direction, and cloud cover as the inputs, a multilayer perceptron artificial neural network model (MPL/ANN) based on the Levenberg-Marquardt training algorithm of MATLAB software package was applied for prediction of the minimum temperature for the next 3, 6, 9 and 12 hours ahead. The selected Transfer Functions were hardlims, logsig, polsin, radbas, satlins, softmax, tansig, and tribas. The statistical measures of MAD, MSD, RMSD, and R were used for comparisons. The results showed that in case of late spring frost, the poslin, logsig, tansig, and satlin functions in April with a correlation coefficient greater than 0.8 and error values of 1.17, 1.61, 1.88 and 2.00 (◦C) for the different times steps, respectively are the best options. Similarly, in October, the radbas, poslin, poslin, and tribas functions with a correlation more than 0.7 and error values of 1.60, 1.96, 1.99, and 1.36, were found to be the most suitable ones for prediction of the minimum temperature at 21:30, 00:30, 03:30 and 06:30 local time. Also, among the selected functions, the poslin with the highest frequency has the best performance in predicting nocturnal frosts in Sanandaj. The results confirmed the good performance of the ANN approach in short-term prediction of minimum temperature and frost occurrence in study region.

The growing demand for agricultural products and the reduction of access to conventional water resources have highlighted the need for proper management of water resources in agricultural sector. On the other hand, measuring and acquiring field data problems make it necessary to employ efficient models that can accurately predict yield and sensitivity analysis of yield over various parameters. Therefore, in this research, the effects of different parameters on the yield of forage maize in the country have been investigated and the effectiveness of the artificial neural network models to predict forage maize yield has been evaluated. The parameters of irrigation water salinity, soil salinity, plant growth period, water consumption and number of irrigation of 104 farms in 8 provinces that were collected during field studies were used to develop models in the artificial neural network. Also, by estimating the water requirements of selected farms, the expected yield was estimated using the optimal selected model and compared with the measured yield in the farms. The results showed that the artificial neural network model with two hidden layers and the final structure of 1-15-19-5 can accurately estimate forage maize yield (R2=0.85). The results of the sensitivity analysis of the selected optimal model showed that the parameters of irrigation events and plant growth period were the most effective and least effective parameters on the forage maize yield, respectively. Also, the results of the modeling showed that there is a gap between the measured yield in the farms and the expected yield, and if the full irrigation requirement is applied, the forage maize yield in the studied areas can be increased by an average of 12.5 tons per hectare.

Selecting a model that simulate the runoff with high accuracy and less error, can be helpful in favorable management of water resources plans and increasing the performance of these plans. Also, increasing the accuracy of runoff simulation in the basins with no meteorological data, is of great significance in efficient management of water resources in these basins.

Estimating total sediment load in rivers is an important and practical issue for water resources planning and management. The sediment concentration can be calculated by both direct and indirect measurements, but direct methods are usually costly and time-consuming. Further, total sediment load can be determined by several sediment load transport models. These equations, however, are applicable in certain circumstances, and in most cases the outcomes do not agree with each other and with measured data. The objective of this study was to propose a method based on artificial neural networks (ANN) to predict total sediment load concentration. Consequently, two ANNs including multilayer perceptrone (MLP) and radial basis function (RBF) with 200 data were used for the modeling purposes. For training and testing the ANN models, 75 and 25 percent of data were used, respectively. The input variables were designated to be average flow velocity, average depth, water surface slope, canal width and median particle diameter of sediment, while the models output was total sediment load concentration. The input variables were included to the models step wisely and the results were evaluated to find out the most suitable ANN models. The predicted values were then compared with five known sediment load transport equations. The conducted statistical analyses indicated that ANNs models in particular MLP can provide better prediction for total sediment load with correlation coefficient of 0.96. It was further concluded that to enhance the accuracy of ANN model, training of the network should be accomplished using both hydrological and sediment data. The Ackers and White equation was very overestimating the total sediment load, while all other equations were underestimating. Based on the results obtained in this study, the ANN-based models provide better concurrence with the observed data, particularly MLP network which can reasonably well predict the peak point of total sediment.

AbstractInfiltration process is one of the most important components of the hydrological cycle. On the other hand, the direct measurement of infiltration process is laborious, time consuming and expensive. In this study, the possibility of predicting cumulative infiltration in specific time intervals, using readily available soil data and Artificial Neural Networks (ANNs) was investigated. For this purpose, 210 double ring infiltration data were collected from different regions of Iran. Basic soil properties of the two upper pedogenic layers including initial water content, bulk density, particle-size distributions, organic carbon, gravel content (>2mm size), CaCO3 percent and soil water contents at field capacity and permanent wilting point were determined on each soil sample. The feedforward multilayer perceptron was used for predicting the cumulative infiltration at times 5, 10, 15, 20, 30, 45, 60, 90, 120, 150, 180, 210, 240, 270 minutes after the start of the infiltration test and the time of basic infiltration rate. The developed ANNs were categorized into two groups; type 1 and type 2 ANNs. For developing type 1 ANNs, the basic soil properties of the first upper soil horizon were used as inputs, hierarchically. In developing the type 2 ANNs, the available soil properties of the two upper soil horizons were used as inputs, using principal component analysis technique. Results of Reliability test for developed ANNs indicated that type 1 ANNs with a RMSE of 1.136 to 9.312 cm had the best performance in estimating the cumulative infiltration. Also, type 1 ANNs with the mean RMSD of 6.307 cm had the best performance in estimating the cumulative infiltration curve.