Design and development of groundwater level monitoring network using geostatistical and statistical methods in Arak plain

Nowadays, groundwater level estimator models have important roles in development and management of water supplies for agriculture. The analysis of the data obtained from quality and quantity monitoring of groundwater resources is related to the environmental factors and spatial continuous variation is an important common characteristic. For a quantitative description of the distribution patterns of the environmental variables, in addition to the determination of the amount of factor, geographical location of the observations data must be considered. Since the accuracy of the estimator models depends on the quantity of the measured input data, an optimization process is necessary to provide an adequate number of data points like the monitoring wells. Therefore, it is necessary to design and develop a monitoring network in the management programs of groundwater resources. The results of various studies have shown that making use of a geostatistic technique, especially the Kriging method, the groundwater resources could be predicted and monitored with a good performance. Also, the distribution conditions of the observation data and its adequacy can be studied with this method. However, monitoring investigation in Iranian watersheds has been carried out extensively. The subject is very important because in those country plains that are fertile and have the potential to grow various crops depending on the climate, the groundwater level monitoring networks must be well managed. Arak unchained aquifer with an area of 1946 square kilometers has 1458 deep wells and 1554 semi-deep wells. The purpose of this research was to prepare a methodology to investigate on the adequacy of groundwater level monitoring wells in the Arak Plain using the Kriging geostatistical method and the evaluate its advantages with respect to statistical methods. To analyse the data, 45 wells were selected from the Arak Plain. First, Kolmogorov-Smirnov test was carried out for data normality. Then, the best semivariogram of the Kriging method was determined based on R2 and RSS. For accuracy evaluation of the monitoring network model, statistical analysis m,measures including the interpolation error, cross validation, and time variation. According to results of R2 and RSS, an experimental variogram in the kriging method was best model. Also, nugget/sill is 0.22 that indicated spatial correlation was high. Based on the interpolation error, in borders of the Arak plain, uncertainty was high due to the lack of observation data. It indicates that the monitoring network needs further development to make it denser in this regions. Based on cross validation, in wells 6, 29, 35 and 36, the difference between observation and estimation data is high, showing that uncertainty in those regions has increased. Therefore, observation data in those regions are important and more wells are necessary. Because Meyghan wetland is located in the center plain, management programs should be considered in monitoring network. It is possible to eliminate 5 observation wells from the middle of the Arak plain, while 5 other wells are needed in the east and northeast of the region. Also, time variation analysis showed that in 10 monitoring wells, short time scale measurement is not needed and it suffices to carry out seasonal scale estimates for groundwater level monitoring.