Introduction
New meteorological numerical methods are developed into general circulation climate models since 1970, therefore. And different climatic models have been developed. Regional Climate Model (RegCM3) is used for climate studies and dynamic downscaling of atmosphere general circulation models output. Until now, many studies have been done using this model. Sensitivity of RegCM Model to different cumulus scheme for winter precipitation 1997 and 2000 are demonstrated by Babaeian and et al (2007). Result showed That Grell scheme with approximately 20% errors have good ability for winter precipitation simulation of Iran. Onol and et al (2009) have been investigated RegCM skill over Mediterranean region countries by using Kuo scheme. They found that annual mean simulated temperature is greater than CRU data. In this paper according to the role of autumn precipitation in water resource providing of the Khorasan area and for calibration and validation of RegCM3 as a research and seasonal prediction tool, a ten years Climate of Khorasan has been modeled for autumns 19991-2000.

Materials and methods
In this article, RegCM3 model is used for simulation of autumn precipitation and temperature over Khorasan. RegCM was originally developed by Giorgi et al (1993a, 1993b) and then improved and new generation released by Giorgi and Mearns (1999) and Pal et al. (2000). Dynamical core of the RegCM3 is fundamentally equivalent to the hydrostatic version of the NCAR/Pennsylvania State University mesoscale model MM5 (Grell et al., 1994). Surface processes are represented in the model using the Biosphere-Atmosphere Transfer Scheme, BATS. The non-local vertical diffusion scheme of Holtslag et al. (1990) is used to calculate the boundary layer physics. In addition, the physical parametrization is mostly based on the comprehensive radiative transfer package of the NCAR Community Climate Model, CCM3 (Kiehl et al., 1996). The mass flux cumulus cloud scheme of Grell (1993) is used to represent the convective precipitation with two possible closures: Arakawa and Schubert (1974) and Fristch and Chappell (1980). The Configuration of RegCM3 Model Optimize for Simulation of Precipitation and Temperature at Autumn Seasonal (September, October & November) in 1991-2000
The model domain covers whole of Iran and centered at 35°N, 57°E. Model is ran with a horizontal resolution of 15 * 15 km. The domain is roughly stretched over the latitude 55 to 61.5?N and longitude 30 to 40?E. RegCM may use initial and lateral boundary conditions from global analysis dataset, the output of a GCM or the output of a previous RegCM simulation. In our experiments these driving datasets are compiled from the Abdus Salam International Centre for Theoretical Physics. Modeled precipitations have been compared by both observed and CMAP data and modeled temperatures have been compared by observed and NCEP reanalysis data.

Result and discussion
Autumn precipitation in 1991-2000 simulated by RegCM3 and their seasonal bias is calculated for Grell-AS, Grell-FC, Kuo and Emanuel scheme relative to CMAP and observation data from equation 1:
Bias = (1)

Table 1 show Comparison between simulated and observed precipitation bias. According to table, bias maximum are found -209.9 mm in Grell-AS scheme at 1993.

Table1. Comparison between simulated and observed precipitation bias over khorasan province (mm)
Kuo Grell-FC Emanuel Grell-AS OBS year
-39.1 -39.5 -39.1 -148.7 149.2 1991
-37.1 -38.5 -39.5 -64.1 65.3 1992
20.2 16.5 19.2 -209.9 260.8 1993
38.5 28.5 28.2 27.4 505.6 1994
-12.4 -17.3 -24.2 -19.7 224.4 1995
-12.3 -12.5 -16.2 -16.7 233.9 1996
18 -3.2 2.2 -8.2 752.5 1997
-32.9 -34.4 -33.2 -35.6 68.9 1998
-1.4 8.3 5.5 6.8 421.5 1999
-13.6 -13.2 -13.8 -16.4 595.1 2000
-7.1 -10.4 -11.1 -51.8 327.7 mean

Comparison between simulated and observed temperature bias show that Grell-FC and Emanuel are better than cumulus scheme (see table 2).

Table2. Comparison between simulated and observed temperature bias over khorasan province (°C)
Kuo Grell-FC Emanuel Grell-AS OBS year
-5.4 -5.4 -5.02 -4.6 16.2 1991
-5.0 -4.7 -5.3 -4.7 16.1 1992
0.1 0.2 0.5 0.1 9.2 1993
-0.6 -0.3 0.2 -0.3 13.4 1994
-0.8 -0.5 -0.2 -0.6 12.1 1995
-1.0 -0.7 -0.6 -0.7 9.1 1996
-2.2 0.4 0.6 0.4 9.7 1997
-0.7 -0.6 -0.4 -0.6 12.9 1998
-2.0 2.3 -0.1 -0.6 9.8 1999
-0.4 -2.0 -0.6 -0.7 8.9 2000
-1.8 -1.1 -1.1 -1.2 11.7 mean


Conclusion
Comparing to the observed precipitation, minimum mean monthly and seasonal bias are found in Kuo parameterization scheme by -0.9 and -7.1 mm, respectively and minimum mean absolute errors are found in Grell-FC scheme by 19.3 and 21.3 mm. relating to the CMAP precipitation data minimum seasonal bias are found to be in Emanuel scheme by 9.3 mm.
Comparison the modeled temperature shows that the minimum monthly and seasonal biases are found in Grell-FC scheme by -1.1 °C. Minimum error was -3.4°C in Grell-FC scheme by comparing the modeled data to NCEP reanalysis data.
Overally, it can be conduced that Grell-AS has no suitable skill for modeling monthly to seasonal precipitation and temperature of the Khorasan province. The results of this paper can be used in operational seasonal predictions and research purposes as well.