نوع مقاله : مقاله پژوهشی

نویسندگان

• مهتاب بادکوبه هزاوه ¹
• محسن نجارچی ¹
• محمد رضا جلالی ¹
• حسین مظاهری ²
• سعید شعبانلو ³

¹ گروه مهندسی عمران، واحد اراک، دانشگاه آزاد اسلامی ، اراک ، ایران

² گروه مهندسی شیمی، واحد اراک، دانشگاه آزاد اسلامی ، اراک ، ایران

³ گروه مهندسی آب، واحد کرمانشاه، دانشگاه آزاد اسلامی، کرمانشاه،ایران

چکیده

کلیدواژه‌ها

• ضخامت معادل آب زیرزمینی
• ماهواره GRACE
• GA-ANN
• ICA-ANN
• PSO-ANN

موضوعات

• منابع آب

عنوان مقاله [English]

Prediction of fluctuations in the equivalent thickness of groundwater using satellite information and artificial intelligence hybrid models

نویسندگان [English]

• mahtab Badkoube Hezaveh ¹
• Mohsen Najarchi ¹
• Mohammad Reza Jalali ¹
• hossein mazaheri ²
• saeid shabanlou ³

کلیدواژه‌ها [English]

• equivalent thickness of groundwater
• GRACE satellite
• GA-ANN
• ICA-ANN
• PSO-ANN

Prediction of fluctuations in the equivalent thickness of groundwater using satellite information and artificial intelligence hybrid models

EXTENDED ABSTRACT

Introduction

Fluctuations in the equivalent thickness of groundwater are one of the main components of the hydrogeological cycle and one of the required variables for many water resources exploitation models. The lack of reliable and comprehensive data is one of the most important challenges in analyzing the decline and predictions of the equivalent thickness of groundwater in water management. In recent years, the use of different satellite information has been noticed as a reliable solution. The aim of this research is to predict fluctuations in the equivalent thickness of groundwater using GRACE satellite data and modeling it using artificial intelligence hybrid models.

Methods and Materials

The study area of this research is the basin area of Lake Urmia located in the northwest of Iran. For this purpose, 180 GRACE satellite data between April 2002 and March 2017 were used. GRACE satellites point information is taken as 1º x 1º, which leads to a 360 x 180 matrix for the whole earth. The output of GRACE satellites includes 6 pixels located on the selected watershed, of which 2 points that overlapped the most with the watershed area were selected for modeling with artificial intelligence tools. One of the effective methods in this field is combining the MLP model with the optimization algorithm in the form of a hybrid model. The GA-ANN, ICA-ANN and PSO-ANN hybrid models were used for this purpose. In the structure of these models, optimal weights are obtained by optimization algorithms. The objective function in these models is to minimize the RMSE value. The generation and modification of weights in the model structure continued until the minimum error was reached, and the number of iterations of the algorithm was adjusted accordingly.

Results and Discussion

The performance evaluation of the GA-ANN, ICA-ANN and PSO-ANN hybrid artificial intelligence models showed that these models are very accurate in predicting fluctuations in the equivalent thickness of groundwater. The results showed that the output of the ICA-ANN model had the best fit with the observational data with a correlation coefficient equal to 0.915 and 0.942 in the two selected pixels 2 and 5 in the test phase, and the results of this model had the best and closest distribution of points. It was 45 degrees around the line and it is considered the most accurate model. Also, the ICA-ANN model had the lowest RMSE value so that the value of RMSE in this method in the two stages of train and test in the Urmia lake basin for study point 2 was 7.3 and 5.73 respectively and for study point 5 was 7.5 and 5.75 respectively. Considering the importance of knowing the changes in the equivalent thickness of groundwater as one of the most important parameters of the water budget, the artificial intelligence models used in this research can be recommended, especially for areas without basic statistics or in situations where it is not possible to use mathematical models. did In this case, without the need for complex relationships and equations to investigate the effect of surface and groundwater interaction and only based on satellite data, the equivalent thickness of groundwater can be predicted in the studied plain in dry and wet periods with great accuracy.

Conclusion

The possibility of predicting the equivalent thickness of groundwater for a long-term period based on a very small amount of information compared to complex models and using only satellite data is one of the most important achievements of this research. In this case, without the need for extensive information and without the need for complex maps and software, and without spending a lot of time and money for the calibration and validation of mathematical models, the equivalent thickness of groundwater based on artificial intelligence methods and using GRACE satellite data is forecasted. This is of great help to experts in the water resources sector in basins that lack statistics or aquifers that lack basic information and accurate maps, or plains that are faced with widespread statistical deficiencies. Because by using artificial intelligence models, very valuable management information regarding the prediction of the equivalent thickness of groundwater in dry and wet years can be obtained with very little time and cost.