The model of estimating   the daily relative humidity from the relative humidity of the local standard time Iran and the parameters of daily precipitation, minimum, maximum and average daily temperature in desert and semi-desert is the subject of this article. Clustering the149 stations by 9 available phenomena (temperature, precipitation, evaporation, relative humidity, range of annual temperature, altitude climate stations and three Indices Demartan, Ivanov and Torrent White) were done.  60 stations Clustered in the desert and semi-arid region by partitioning around mediods. Systematic sampling design for selecting a sample was used. Different patterns of linear and nonlinear regression fitted on the sample data and the best models were determined in monthly and annual scales. Calibration models presented in this paper and compared with conventional models of old patterns showed that the accuracy of new patterns is better.  Our criterion is Mean square errors