Studying the process of space-time ground water level by support vector machine and Kriging Method in GIS (case study: silakhor plain)

Since ground water and dwindling water resources is important for the operation of research and modeling is important. Assessment and prediction of groundwater level to help predict groundwater resources. The use of artificial intelligence methods based on the theory of data mining is used to predict the water table fluctuation. The support vector machine in artificial intelligence methods and the methods of geostatistical Kriging method has considerable precision in order to predict the time and location of the water table is level. In this study, the combination of support vector machine and Kriging model as a new way to predict when and where water table fluctuation in the plain area Silakhor is used. In the first phase, modeling when using support vector machine model data 11 piezometric wells in the region were carried out using support vector machine and secondly to predict the location of monthly data output of the first stage as an input earth model Statistics were used.
Data of 11 observation wells in the Silakhor plain data collected in the course of the past ten years in both normal and abnormal SVM model were used as input. Using the software Matlab function algorithm support vector machine was configured as an input at each stage of a well is this model. In addition, the maximum error in the calculation of the wells with 0.2 Keyvareh is due to lack of observation data in interval or lack of access to the region is to read water level. Absolute estimation interpolation and location estimate is a major feature. This means that estimates the quantity of sampling points is equal to the measured value and variance estimate is zero. In fact this model to estimate the amount of variance minimizes unknowns. Thus, the curves are drawn based routing and thus go beyond the boundary drawn. This feature makes the model the spatial distribution of data which are dependent on terrain, Kriging in calculations of high accuracy. Locate underground water level role in reducing the cost of drilling a well in the region. This way you can reach the height of water to be achieved in the region or even decline or rise of water table revealed
The ability of Support Vector Machine, and of course, a relatively new as a useful tool in water resource management to predict fluctuations in groundwater levels were evaluated in Silakhor plain. According to the accuracy of this method in predicting groundwater level can be comprehensive and appropriate program management discussion groundwater resources to be expected. In this study, using geostatistical methods in the area of water resources in plain Silakhor predicted with high accuracy are discussed. The most important issue in the analysis of spatial-temporal data to determine the dependence structure of the data. Whatever the choice of models and model are more accurate, the prediction will be more accurate. In terms of predicting when it should be noted that the time for the distance away from the last viewing time predicted prediction accuracy will be reduced. According to the study area in the last ten years has been well observed that in the best case scenarios predicted by the model, the highest coefficient of determination for wells Chughadun (DC = 0/96), sugar (DC = 0 / 94) and Valian (DC = 0/93) was calculated, which represents a hybrid model to predict groundwater level is the wells could be decided.