Monitoring groundwater level network of Dezful-Andimeshk plain

Introduction
Preservation and proper management of water resources are one of the essential fields of study in the world. In arid and semi-arid regions like Iran, quantitative and qualitative management of underground water resources is particularly important. In most hydrological issues and groundwater resources studies, groundwater statistics and information availability are critical. To collect information without side effects, comprehensive and sufficient data collection with the help of a groundwater monitoring network is very important. In line with the sustainable management of renewable water resources, the need for a network of underground water observation (monitoring) wells to accurately measure the water level is necessary and necessary. Considering the complexities of the underground water environment and the high costs of conventional monitoring methods, inventing new technologies and using advanced methods in this matter will significantly help improve the underground water systems. One of the parameters of particular importance in monitoring groundwater quantity is the groundwater level. Therefore, this parameter should be measured or estimated as accurately as possible. In recent decades, the use of computer and calculation models to monitor the level of underground water has developed significantly. Considering the importance of underground water resources and network monitoring, to save time and money, in this research, principal component analysis and Shannon's entropy theory were used to monitor the underground water network of the Dezful-Andimeshk Plain.
 
Materials and Methods
This research used monthly groundwater level information from 77 observation wells in the Dezful-Andimeshk Plain during 2018-2019. Groundwater level information is collected twice a month. Principal component analysis and Shannon entropy methods were used for monitoring. In the current research, the number of statistical periods for each well is 24, less than the total number of observation wells. Twenty-four observation wells around it were used to monitor each well. In groundwater level monitoring, the relative importance of each well is defined by the ratio of the number of times that well is recognized as a compelling well to the number of times that well is included in the analysis of the main components. This ratio shows the importance of each well compared to other wells. Therefore, to save time and costs, less important wells can be removed in the monitoring of the underground water level. In 1948, Shannon showed that events with a high probability of occurrence show less information, and on the contrary, the lower the probability of an event, the more information it provides.  In this method, the weight of each well was obtained using Shannon's entropy theory. Any well that has a higher Shannon entropy weight contains more important and unpredictable information and should be preserved. On the contrary, a well that has a lower Shannon entropy weight can be removed from the network. Principal component analysis and Shannon's entropy method in the current research were done with the help of coding in Matlab software due to the high volume of calculations.
 
Results and Discussion
To rank the wells, the threshold limits are equal to zero, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9 and one considered. At threshold one, only wells that have a rank of one remain (wells that are recognized as effective wells in all analyses) and threshold zero includes all wells (effective and ineffective). According to the obtained results, increasing the error in the threshold zero to 0.7 is gradual, but in the thresholds 0.8, 0.9, and one, the error value increases with a high slope. So, the amount of error in the thresholds of 0.7, 0.8, 0.9, and 1 has been calculated as 12.2, 17.7, 25.3 and 34.2 respectively. Therefore, the threshold limit in the current research is considered to be 0.7. However, the number of wells effective in monitoring the underground water level is reduced from 77 to 32. Shannon's entropy weight values were also calculated for all wells. 11 wells have the highest value of Shannon's entropy weight, which shows that they contain the most information.
 
Conclusion
The general comparison of the results of the two methods showed that all 11 wells with the highest weight in the Shannon entropy method were also observed as effective wells in the principal component analysis method. By knowing the effective wells in the region, firstly, in the face of lack of time and money, it is possible to use known effective wells for monitoring secondly, by removing ineffective wells, there will be little change in the average level of underground water. It is not possible, or in other words, the tracking error does not increase significantly. Comparing the results of the two methods showed that the remaining wells in Shannon's entropy theory are among the wells identified in the principal component analysis method. Also, considering that the wells in the region were built by the Khuzestan Water and Electricity Organization considering the types of uses, removing the ineffective wells will not affect the process of using the information of the wells. It is recommended to use principal component analysis and Shannon entropy for groundwater quality monitoring in the study area. Additionally, it is suggested to monitor the quality of the underground water network in the study area using the methods used in future research.