Genetic Algorithm Application for Soil pHSpatial Distribution Estimation with Geostatistics and Gene Expression ProgrammingCombination

Awareness about spatial distribution of soil properties plays an important role in the system of crop growth and yield in the agricultural field. In this regard, genetic algorithm was used to increase the accuracy in spatial distribution of soil pHwhich combines the estimates of linear regression, gene expression programming (GEP) and geostatistics (kriging interpolation) with data related to some parts of East Azerbaijan province (Bonab,Maragheh and Ajabshir). The linear model of kriging interpolation had the minimum error. Gene expression programming and linear regression had the lowest and highest error for soil pH spatial distribution estimation, respectively, for example the mean square error, root mean square error, relative root mean square errordecreasing from linear regression to geostatistics was 47.67%,27.58%, 26.47% and scatter index and adopted mean absolute percentage error decreasing from linear regression to GEP was 23.8% and 37.03%, respectively. The use of genetic algorithm in combination method reduced the error of spatial distribution compared to the tree types of models, for example mean square error, root mean square error, relative root mean square errordecreasing from GEP to combination method was 23.33%, 11.76%,10%, respectively. The estimates of soil pH with combination method are in the alkaline range, which is consistent with the obtained data.  The minimum and maximum value of the absolute difference between measured and estimated data were at point in Bonab (0.09) and Ajabshir(0.25), respectively. The mean absolute percentage error of combination method was in acceptable range and this shows the efficiency of combination method for soil pH spatial distribution estimation.