Prediction of Monsoon in a Seamless Framework

ABSTRACT

During last two decades, accuracy of weather prediction has improved significantly through improvement in understanding of underlying physical and dynamical processes, enhanced quality and quantity of meteorological observations, and availability of progressively increasing computing power. However, the predictability of day-to-day weather patterns over the Indian monsoon region is still limited to 4-5 days.

Under the National Monsoon Mission, steps have been undertaken to improve the forecast system with special emphasis on the short and medium range scale up to two weeks in advance by NCMRWF. The advanced seamless unified modelling (UM) system of Met Office, UK has been implemented at NCMRWF. The UM's dynamical core uses a semi-implicit semi-Lagrangian formulation to solve the non-hydrostatic, fully compressible deep-atmosphere equations of motion (Davies et al., 2005). The UM at NCMRWF has a resolution of N512L70 (approximately 25 km in the mid-latitudes and 70 vertical levels). It has a 4-D VAR assimilation system which has a huge advantage as it as it extracts more information from observations consistently, in a better way. Special efforts have been made to maximise the use of observations from the Indian satellites. The UM is being run in real time and its evaluation has been carried out with special emphasis on monsoon. The average Day 3 forecast error for winds at 850 hPa against radiosondes during monsoon season is about 3-4 % less in the UM as compared to the currently operational GFS (Global Forecast System).

In order to predict severe weather events more accurately during the pre-monsoon, monsoon and post monsoon seasons, a very high resolution (1.5 km) nested regional UM has recently been implemented at NCMRWF. Initial evaluation shows improvements in location and amount of precipitation associated with the recent severe weather events.

For extending the temporal range of weather forecasts beyond seven days, it is pertinent to utilize coupled ocean atmosphere model to capture the space-time variability of monsoon more accurately. A coarse resolution coupled model (atmosphere-ocean) using Unified Model (UM) for the atmosphere and NEMO for the ocean has been implemented at NCMRWF and is being evaluated. Quality of observations over Indian oceans is monitored and evaluated for ocean data assimilation.

As the short and medium range weather forecasts are sensitive to the initial state of atmosphere, representation of NWP model uncertainties is now considered imperative. This is particularly true over tropical region. In order to account for the uncertainties, ensemble data assimilation and prediction system based on ETKF method (Ensemble Transform Kalman Filter) have also been implemented at NCMRWF. The probabilistic forecasts thus generated have been found to be useful for planners and decision makers.