Improvement of Systematic Biases of Climate Forecast System (CFS) Model through Revised Convection-Microphysics and Superparameterization

Abstract:

It is well known that Coupled General Circulation Model (CGCM) shows limited skill in capturing the Tropical Intraseasonal Oscillations (TISO). As the NCEP Climate Forecast System (CFS) is adopted for operational monsoon forecast of India, improving its bias will directly benefit the operational forecast and finally the society. Keeping this in background, we felt, improving the cloud and convection parameterization of CFS (version2) is the need of the hour. A first attempt is made to improve the Indian Summer Monsoon (ISM) rainfall from diurnal through daily to seasonal scale. Experiments with Simplified Arakawa Schubert (SAS) and a revised SAS (RSAS) schemes are carried out to make 15 years climate run (free run). It is clearly seen that the use of RSAS is able to improve some of the biases of CFSv2 with SAS. Improvement is seen in the annual seasonal cycle, onset and withdrawal but most importantly the rainfall probability distribution function (PDF). The PDF of diurnal rainfall has significantly improved with respect to even a high resolution CFSv2 T382. The improvement of diurnal cycle of total rainfall is found to be contributed by the improvement of convective rainfall. However, the cold tropospheric temperature bias, low cloud fractions need further improvement. As the RSAS could only improve the convective rainfall but not the resolved scale process, the existing Zhao and Carr microphysics scheme is replaced by WSM6 scheme with six class of hydrometeors. The incorporation of WSM6 along with RSAS appears to make a significant improvement in the systematic biases of CFSv2 in the intraseasonal scale. It is able to capture the cloud processes much realistically and show a significant improvement in simulating the tropical waves and TISO. As a part of improving the cloud processes in CFSv2, we have attempted Superparameterization technique and have developed a SP-CFS (superparameterized (SP)). The CFSv2 used for SP is at T62 resolution. SP-CFS simulates an improved precipitation distribution over the globe; better temperature structures, variance for the synoptic and low-frequency TISO and reduction of the dry precipitation bias. These developments give a clear message that improvement in representation of cloud and convection processes improve the systematic biases of CFSv2 and improve model fidelity.