Spatial and Temporal Variability Analysis of Shallow Water Table Using Deterministic and Geostatistical Methods (Case study: Dashte- Naz of Sari)

Shallow groundwater table changes are important and effective factors in studying and ground water resources management especially in agriculture and drainage issues. Practically, because of lack of the number of observation points, reading and observation, because of time and financial problems, the information availability is limited. Thus, the application of interpolation methods and estimation water level in specific points base to information adjacent points studies have special position in the groundwater resources studies. In this study, deterministic and Geostatistical methods evaluated in order to determine the most appropriate method for Spatial and temporal variability analysis of shallow water table in 1700 hectare areas of agricultural land of Dashte_naz Sari company, in 81 observation well with 500*500 meters regulation grids. The results show that in deterministic methods, Inverse Distance Weighting and radial basis function have more accurate than other methods with a root mean square error of 54 cm to 55.9, and the mean absolute error of 42.1 to 45% and the correlation coefficient of 0.21 and 0.27 for data of 15 March and 21 December. Due to the high spatial variation of water level because of localized drainage problems in this area, deterministic methods and kriging haven’t sufficient precision. For the data of 21 January and 15 April, cokriging method (with the co-variable of water level in the previous period) with 26.7 and 34 cm root mean square and mean absolute error of 17 and 24 percent and correlation coefficient 0.83 and 0.75, have more accuracy. K-Bessel Variogram model also had stronger spatial structure. In the case of high Spatial variability of shallow water table, cokriging geostatistics method with the help of a co-variable will be significant effect on increasing the precision of the spatial distribution of shallow water table and resulting 37% and 50% decrease of root mean square error and 44% and 59% decrease of the average absolute error.