Developing a new grid-based optimization model for groundwater depth and comparing with some traditional statistical methods: An application to the Lower Seyhan Basin

The main objective of this research was to optimize the number of groundwater (GW) observation wells in the monitoring network for the sake of saving time and money in data collection process. To this end, a new grid-based optimization model was developed by transforming the research area into matrices with the help of gridding. The objective function of the optimization procedure was to minimize the penalty point [mean (M)+ standard deviation (STD)] which was determined as the loss of information in GW depth measurements. This study was carried out in the Lower Seyhan Basin, covering an area of 9495 ha. GW depth measurements in 107 locations (NW_107), which were done in 2011, 2012 and 2013 years, was utilized in the study. GW Depth measurements were done in four seasons a year, i.e., in rainy season (in Jan/Feb), before irrigation season (in Mar/Apr), in the peak irrigation season and at the end of irrigation season. GW depth data of existing drainage observation network (NW_107) in the basin and depths of different network combinations {NCW=(n¦k)} of 107 drainage wells were mapped, in turn, by using Inverse Distance Weighting (IDW) interpolation technique. The derived maps were dubbed as "basin matrices". The penalty points (PP=M+STD) of different network combinations established earlier were calculated by referencing Ms and STDs of 12-season observations of NW_107. In this regard, 23 different NCWs which were resulted in minimum information loss were considered and optimized accordingly. Statistical test criteria of RMSE and percent prediction error (PE%), and additional analysis of GW depth-area hypsometric and frequency curves, and classical statistical tests suggested that the GW observation network of 67 wells (NCW_67) was the optimal one with the minimum loss information compared to the existing network of 107 observation wells. Overall results helped us to recommend that NCW_67 network with the prediction error (PE%) of 3.0% might be, hereinafter, adopted in monitoring process instead of existing network of 107 wells in Basin. Key Words: Groundwater depth, Observation well network, Inverse distance weighting, Optimization, Lower Seyhan Basin