Abstract

Introduction Due to deep, complex and everlasting interaction between precipitation and climatic elements-factors, there are changes and varieties in both time and space dimensions of precipitation. So that climate experts and related scientists take their attentions to this phenomenon. An approach to do this kind of investigations is to describe spatial variations based on spatial statistics.   The major spatial non-stationary of Iran precipitation is due to variation in situation, elevation and topography characters (slope and its direction) in this country. Circumstances of every one of these characters could determine the precipitation spatial patterns. Accordingly understanding spatial distribution  of precipitation and its mechanism are important aspect in climatological researches.  One of the common statistical models in which it is possible to  determine the relation between variables as well as reconstruct, estimating and forecasting data is multivariate regression model. These sorts of models are useful for time series analyses as well as spatial modeling. One of the regression models that could be used in spatial analyses is called Geographically Weighted Regression (GWR). In current study it will be attempted to introduce this approach and using General Regression (GR) to justify spatial variation of precipitation in Iran based on 1436 stations in Iran.   Research Methodology In this research Esfezary data base have been used. This daily data based contain 15998 days and 7187 pixels (15*15 KM) of precipitation over Iran.  Accordingly the data matrix is created in 15998* 7187 and S-mode dimension. This matrix data base is estimated by using 1436 stations and Kriging method. To achieve independent variables, digital elevation map by 15*15 KM resolution has been created. So that, spatial (including longitude and latitude) and topographic (including slope magnitude and aspect) characters have been derived. Accordingly a data base has been created that contain spatial characters, topographic features and precipitation amounts.         Spatial Modeling of Annual Precipitation in Iran       The proper regression model on precipitation has been chosen based on spatial and topographical characters. Multivariate General Regression (MGR) for these m independent variables is defined as follow:     Where Ri  is precipitation in a given pixel that depends on” m “climatic factors. Geographical Weighted Regression (GWR) allows local rather than global parameters to be estimated and the above model is rewritten as:   In geographically weighted regression, the parameter estimates are made using an approach in which the contribution of a sample to the analysis is weighted based on its spatial proximity to the specific location under consideration. Thus the weighting of an observation is no longer constant in the calibration but varies with different locations. Data from observations close to the location under consideration are weighted more than data from observations far away. In this paper spatial distribution of precipitation had been modeled using General Regression and Geographical Weighted Regression. Finally the most effective variables on precipitation have been clustered using Euclidean distance method and Ward clustering method.   Discussion and Results Annual mean of Iran precipitation is about 256 mm. Spatial coefficient of variation of Iran precipitation is about 79%. Generally distribution of precipitation isohyets over Iran follows topographic features. The General Regression Model for precipitation is as follow:       This model can justify about 48% of spatial distribution of precipitation. Using GWR model tends to different coefficients of spatial variables. Accordingly three regions have been denoted: The first region in which precipitation is affected by elevation. This region is about 43% out of the all area of the country that located in northwest, inner parts and southeast of Iran. Precipitation of second region is affected by hillside of mountains in west of country that covers 18% of Iran. The precipitation of third region that is about 39% of the country is determined by slope more than other factors. This region located in small parts of northwest, west and southeast of Iran.     Conclusion In order to justifying spatial changes of precipitation over Iran, 1436 stations and GWR technique have been applied.  Based on GWR model, elevation in northwest and inner parts of Iran, direction of slop in Zagros mountain chain and the slop in northeast and Caspian coast are the spatial factors that more controlling precipitation   Keywords: Cluster analyses, Spatial autocorrelation, Geographically weighted regression (GWR), Spatial Modeling. Refrences