مدل سازي تغييرات مكاني عناصر اقليمي مطالعه موردي : بارش سالانه استان اصفهان

Spatial Change Modeling of Climatic Data A case study: Annual Precipitation of Esfahan Province

Due to the phenomenon of climatic interaction, multivariate statistical techniques have gained more attention in climatic modeling. One of the applied fields of multivariate techniques is spatial change in climatic elements modeling. Proper application of these techniques makes it possible to model and simulate these elements with more certainty. In this way, correlation between independent variable may cause wrong results. Then it is necessary to pay more attention to this correlation that is called multicolinearity. For example, if we look for interaction of climatic factors such as longitude, latitude and altitude with precipitation, we must remove the correlation between these factors and then calculate the correlation of every one of them with precipitation. This can be obtained by Ridge Regression method and partial correlation function. In this paper, the author defines multivariate regression and its application in analysis of precipitation in Esfahan province spatial change. To define the relationship between precipitation and the other factors (longitude, latitude and altitude) 32 maps of precipitation for Esfahan province based on Kriging interpolation prepared for 1969-2000. Based on algebraic map technique the 32-year map of mean precipitation was created. At last the average map changed to digital values for 2845 points representative of all pixel in the average map of Esfahan precipitation. Then the statistical methods were used for these 2845 points of longitude, latitude, altitude and precipitation. The independent variables (longitude, latitude and altitude) have multicolinearity. Then there are unreality coefficients in spatial function of precipitation because of multicolinearity in longitude (A), latitude (0) and altitude (h). The regression function obtained as follow: R = 1765.578 - 27.789¢ - 6.254Landa+ 0.003921h Therefore, Ridge regression method was used to describe precipitation spatial variation in Esfahan province. The regression coefficients were adjusted based on Variation Inflation Factor (VIF) by 0.01 interval. The VIF measured variance of coefficients. We can minimize it until it becomes unbiased .The result by this method is as follow: R =1738.95-19.91Landa -8.76¢+0.03h Based on the above model longitude (Landa), latitude (¢) and altitude (h) affect precipitation by magnitude coefficients respectively. But because of more elevation variability in Esfahan province (900 - more than 4000 meter), the effect of this factor is more significant. The regression and determination coefficient for the model are 0.67934 and 46.15 percent respectively.