| Implementation of geostatistical tools into Natural Resource Management studies such as soil science, entomology, ecology or forestry marked the end of 1980s and the beginning of 1990s. During this time the spatial analysis in natural resource oriented research found significant support in geostatistical methods. The objective of this study is to develop and test digital elevation models based on a kriging interpolation algorithm to predict the elevation values in any unsampled location. This methodology, however, can be applied to many other related problems in natural resource management such as air or water pollution assessment, soil properties, spatial distribution of mineral resources or insect outbreaks. Vegetation mapping and many other projects would also benefit highly from modeling procedures developed and tested in this research.; The Geographic Information System (GIS) was designed for manipulation and analysis of spatial data. Hence the linkage between geostatistical methods and GIS was mutually useful. It increased the efficiency of spatial analysis and accelerated many intermediate steps in the model building process. Display, graphing and built-in GIS functionality appeared to be useful at different stages of this research for checking and testing of intermediate steps. For example, the generate function produced coverages from predicted values, contouring capabilities were useful for anisotropy modeling, thiessen function, Grid functions and number of other additional functions allowed efficient manipulation of spatial data. Statistical analysis in GIS, however, is limited and the research often required implementation of additional systems such as SAS, GSLIB, FORTRAN 77, Gauss, Delta graph and Excel, packages to develop models for spatial data with graphical output.; The variogram analysis played significant role before the kriging interpolation procedure took place. The accuracy of predicted results from the kriging was highly depended on precise identification of variogram parameters. The quality of data and the character of earth's surface was another significant factor with high impact on the accuracy of the predicted results. The kriging variance and the kriging estimates responded sensitively to the relief differences in all testing sites. Particularly, the abrupt changes in elevations along mountainous rims and the jagged mountainous areas significantly increased the mean error of prediction and the error variance of predicted values.; The use of stochastic methods in the natural recourse management better corresponds to the character and behavior of the earth science phenomena. Hence, kriging, as a stochastic method, was selected to generate four digital elevation models. The results of testing corroborated the hypothesis that the elevation models (EMs) can be generated with a good level of accuracy with probabilities methods and also confirmed the assumption that the accuracy of testing models decreases with increasing relief diversity. |
