نوع مقاله : پژوهشی
نویسندگان
1 استاد گروه منابعطبیعی دانشکده کشاورزی و منابعطبیعی، دانشگاه هرمزگان، بندرعباس، ایران
2 استادیار گروه مدیریت و کنترل بیابان، دانشکده علوم محیطی، برنامهریزی و توسعه پایدار، دانشگاه سراوان، سراوان، ایران
3 کارشناس ارشد گروه تحقیقات کاربردی شرکت سهامی آب منطقه ای استان هرمزگان، بندرعباس، ایران
4 کارشناس ارشد گروه پژوهش منابع آب زیری شرکت سهامی آب منطقه ای استان هرمزگان، بندرعباس، ایران
5 کارشناس ارشد گروه حفاظت و بهره برداری از منابع آب میناب، میناب، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Introduction and Goal
Groundwater resources as a major portion of the world's freshwater reserves, play a vital role in meeting human needs and maintaining ecosystems. On the other hand, unsustainable extraction, climate change, and pollution have posed serious challenges to these valuable resources. Declining water levels, deteriorating quality, and land subsidence are among the negative consequences of excessive groundwater extraction. Effective and sustainable management of these resources requires precise monitoring and continuous assessment of their quantitative and qualitative status. Groundwater monitoring networks, by collecting data on water levels and quality, enable the evaluation of resource conditions and the identification of potential issues. On the other hand, the optimal design of these networks is challenging due to cost and operational constraints. The objective of this research was to investigate and optimize the existing network of observation wells in the Shamil-Ashkara aquifer, located in the northeast of Hormozgan province, using data mining and artificial intelligence methods, including Analytic Hierarchy Process (AHP), Principal Component Analysis (PCA), and the Non-dominated Sorting Genetic Algorithm II (NSGA-II). In addition to examining improvements in monitoring network efficiency and reducing associated costs, this study also identified optimal locations for constructing new observation wells.
Materials and Methods
The study area, the Shamil-Ashkara plain in the northeast of Hormozgan province, covers an area of 321.93 km2 and features diverse geological formations (including the Bangestan, The Aghajari, Gachsaran, Asmari, Razak, Hormuz, Gurpi, Mishan, and Quaternary alluvium). Studied aquifer was located within the Quaternary formations, and the land use comprised poor rangeland, agriculture, and residential areas. In this research, suitable locations for constructing observation wells were initially determined using the Analytic Hierarchy Process (AHP) method. Eight criteria were considered, including the long-term average groundwater level, annual groundwater decline rate, slope of decline changes, density of exploitation wells, distance from the river, geological formation, land use, and distance from faults. The weight of each criterion was calculated using pairwise comparisons and Expert Choice software. Then, using ArcGIS software, final map prioritising suitable locations was prepared. In the next step, the current monitoring network was compared with the site selection results. Using Thiessen networks and calculating the average site selection scores in each polygon, less significant wells were identified. Homogeneity tests were performed on the less important wells to ensure the selection was correct. Then, using the PCA method, the relative importance of the monitoring wells was determined. By sequentially removing each well and calculating the correlation coefficient between the data of the remaining wells and the first principal component (PC1), the relative importance of each well was calculated. Additionally, the lack of certainty resulting from the removal of insignificant wells was assessed using the average coefficient of variation of groundwater levels. To investigate the effect of removing insignificant wells on interpolation error, the kriging method was used. Two scenarios were examined, including the use of all wells and the removal of insignificant wells, and the standard error and RMSE were calculated. At the end, using the NSGA-II algorithm, the optimal network of observation wells was determined. Two objective functions, including the number of observation wells and the minimum RMSE of groundwater level, were minimised. The IDW method was used to estimate the groundwater level in the deleted wells.
Results and Discussion
The results of AHP show that the highest priority for constructing observation wells was due to the concentration of agricultural activities, the high density of wells, and the severe decline in water levels in the southern and southeastern. In this analysis, the most important factor was the density criterion of exploitation wells with a weight of 0.327, followed by the criterion of groundwater level reduction (0.245) and land use (0.15). The inconsistency rate was calculated as 0.04, indicating an appropriate consistency of the expert judgments. Comparing the current monitoring network with the AHP results showed that five wells (W7, W9, W12, W14, and W16) had an average location score of less than 0.113, indicating they were unsuitable from a location perspective. The homogeneity tests of the data using the Pettitt method and a p-value of less than 0.05 also indicated the lack of homogeneity of the data from these wells. The results of PCA showed that the insignificant wells identified by this method (W8, W10, W18, W21 with correlation coefficients of less than 0.20) were different from the wells identified by the AHP method, and this difference was due to the differing structure and input data of the two methods. This is because in the PCA method, only water level data is used, whereas in the AHP method more criteria are considered. The analysis of the kriging results showed that the removal of insignificant wells based on AHP led to a decrease in the interpolation standard error from an average of 4.8 to 4.2 m. Additionally, the removal of wells based on PCA caused an increase in this error from 4.8 to 3.5 m. On the other hand, in both methods, the removal of less important wells led to an increase in RMSE. In the AHP method, the root mean square error decreased from 22.14 to 17.08 m (a 1.20% increase), while in the PCA method it changed from 22.14 to 41.14 m (a 3.1% increase). The NSGA-II algorithm for the monitoring network proposed an optimal configuration of 16 wells, which represented a 43% reduction in the number of wells compared to the initial network of 28 wells. Moreover, the RMSE only increased by 7.3% (from 22.14 to 26.15 m). The coefficient of variation of groundwater level in the optimal network was calculated to be 0.89, which was close to its value in the main network (0.92). This finding indicates the preservation of important hydrogeological information in the reduced network.
Conclusion and Suggestions
The AHP method for determining optimal locations for observation wells was effective and its use is also recommended in other plains. Furthermore, it is suggested to employ other variables and the output of groundwater models to enhance the efficiency of this method. Additionally, the use of PCA and NSGA-II methods will lead to the optimisation of monitoring networks.
کلیدواژهها [English]
)