Investigating Soil Texture Affected by Land Use and Elevation in the East of Mazandaran Province

Extended AbstractIntroduction and Objective:Some soil properties are spatiotemporally variable. The proper handling of soil as a source of food, improving factor for climate changes and the environment, as well as for the production of fossil fuels, is definitely essential. The spatial and temporal variations of different parameters affect the formation of soil and its characteristics. The high expenses of soil sampling operation and the inaccessibility of some parts, make the use of indirect methods for the prediction of soil properties including texture unavoidable. It is essential to use and establish methods, which minimize the time and expenses of sampling and analysis for soil texture mapping. Geostatistics can suitably investigate and predict soil properties such as texture in a large scale, by reducing time and expenses. Investigating the spatial variability of soil texture under different land uses is helpful for the determination and stimulation of soil ecosystem properties subjected to the different climate and environmental variation. The spatial variability of soil texture is a function of physical parameters and land uses, and accordingly, using suitable techniques for investigating the spatial variability of soil texture is essential. For this purpose, the IPMs and ANNs consisting of Simple-Kriging, Co-Kiriging, IDW and MLPs were used to study the spatial variability of textural components of the eastern Mazandaran. Material and Methods: In the present research, the study of the spatial variability of soil texture affected by different land uses and elevation at inaccessible places, was done by kriging, co-kriging, inverse distance weights (IDW) and also artificial neural network (ANN). Compound soil samples (249) were randomly collected from the 0-15 depth. The precision of each method was determined using mean error (ME) and residual mean square error (RMSE). The data of each parameter were log converted and plotted using the method of normal box-cox and variograms. The comparison of means affected by different land uses including farming, orchard, forest, rangeland and uncultivated as well the elevation was conducted by Statistixs 9. The study area consisted of different elevations including -20-0, 0-100, 100-500, and > 500 m. The variograms were plotted by spherical, Gaussian, exponential and linear models, and the amounts of semiovariograms were calculated in the GS+ software, and the soil map was plotted by Arc GIS software. Results: The normalization of data, which indicates the distribution of soil texture, was done by the Kolmogorov-Smirnov test. The coefficient of variation of silt particles, which is affected by soil, fertilization and drainage activities was less than 50% indicating the rate of variation is not high in the study area. Accordingly, the exponential model was selected as the best fitted one for the prediction of soil texture affected by the experimental treatments. Although the amounts of clay were not significantly affected by different land uses, the amounts of sand and silt were significantly different as the amount of sand was the highest by rangeland, forest, farming, uncultivated and orchard, respectively, and the highest amount of silt was related to the orchard, farming, forest, uncultivated and rangeland. Similar research has also indicated that the change of land use from forest to farming, increases sand and decreases silt, which is due to soil erosion. In addition to different land uses, elevation (-20 - > 500 m) also significantly affected soil texture. The highest rate of sand, was related to the -20-0 m elevation. The highest and the least rate of clay was resulted by the 100-500 and > 500 m elevations, respectively, which is due to the leaching and erosion of clay particles from the higher elevation, resulting in the higher rate of sand in higher elevations, and higher rate of clay in the less elevated areas. Conclusions Among the most important objectives of the present research was to investigate the possibility of predicting soil texture with an acceptable accuracy by kriging, co-kriging, inverse distance weighting and artificial neural network. According to the results, ANNs was the most suitable method for predicting soil texture affected by different land uses and elevation. It is suggested to: 1) examine the tested methods in other places, 2) other artificial methods be also tested for predicting soil texture, 3) if possible, a regular sampling be done in such types of research areas, and 4) the prepared mapping be used as a base for investigating the spatial and temporal variability of soil texture.