The GTS data on-line decoding is implemented on VAX 8810/Dec Alpha systems running on VMS/Degital UNIX operating system to categorise various types of observations and store them in Report Data Base(RDB) format using ECMWF Fortran Decoders. All other components of the GDAFS related with complex quality control, data analysis/assimilation and forecast model are implemented on CRAY XMP/216 system which runs on UNICOS 6.0 operating system in Fortran. The end-to-end operational suite is also implemented on Dec Alpha Cluster to generate operational products in stand-by mode. All forecast fields are archived in CRAY Binary/GRIB format. Meteorological Application Graphics and Integrated Colour System(MAGICS) / Grid Analysis and Display System(GRADS) softwares of the ECMWF/University of Maryland is implemented on front-end systems for generating output graphical plots.

The dissemination of forecast fields valid for different geographical locations of the country is carried out by using Very Small Aperture Terminal(V-SAT) network. V-SATs operate in STAR configuration using TDM/TDMA technique with inbound data rate of 16Kbps and outbound data rate of 64Kbps. The network is linked with the Transponder Number - 9 of the INSAT-2D satellite. Presently 67 V-SATs are installed at different places all over India.

At the moment, only cloud tracked winds from geostationary satellites(like INSAT), temperature/moisture profiles from polar orbital satellites of NOAA, USA are the most widely employed satellite derived products in NWP. Global communication satellites transmit these data sets by collecting from land based communication networks spread over various countries and transmit the global observational data in real-time to the global weather centers. This service is part of global telecommunication system(GTS) of World Meteorological Organisation(WMO). The quantum of data received through GTS at this Centre is limited to the bandwidth of the communication line between the RTH, Tokyo and RTH,Delhi.

The results of the quality control checks as mentioned above are combined to produce a single final quality flag for each sample of observation indicating whether or not the observation is to be used by the data assimilation procedure subsequently.

At the NCMRWF, all the data after decoding are checked for their quality before being written to the Report Data Base(RDB) files. Earlier the Quality control procedures were embedded in the decoder programmes but in the new version they have been developed as a separate module and called after each report is decoded. Besides the quality control checks which were applied in the earlier version, i.e. HYDROSTATIC CHECK of the TEMP data has been included in the new version. Residuals are computed for all the elements by converting the guess from spectral space to grid space and interpolating the same at the observation locations as earlier. All the quality control programmes were modified to adapt to the changes made in the data preprocessing procedures.

The core of the operational suite, consisting quality control, assimilation and forecast, is executed on the CRAY-XMP/216 Computer or the DEC ALPHA Cluster after the prescribed cut-off time. Most of the post-processing and graphics related jobs are implemented on front-end computers. Real-time forecast generation up to 5-days is carried out daily based on 00UTC initial conditions and a Regional Spectral Model(RSM) at 0.5°lat./long. Resolution is run as well by taking appropriate lateral boundary fields from the global model forecasts. Extended range forecasts on monthly scale are also generated once in a month. Recently, an 4-Member Ensemble Prediction System(EPS) based on breeding method is tested.

The analysis scheme of the data assimilation system employed at the NCMRWF is the Spectral Statistical Interpolation(SSI) Scheme. While running the SSI for the last two years, it was felt that the mass and momentum fields were not properly balanced and their were large ageostrophic components specially over the tropics. One of the reasons for the same could be the weak balance introduced in the analysis scheme between mass and wind fields and hence a new set of analysis variables were chosen in the new version to introduce a better balance between mass and wind fields.

The analysis variables chosen being the Vorticity, Mixing ratio and unbalanced part of Divergence, Temperature and log of Surface Pressure instead of the full fields as used in the old scheme. The balanced component of these fields are computed from Vorticity using the first six EOF's and based on statistical considerations. As a result of this, new set of statistics had to be computed for forecast error covariances and for computing the empirical constants used in calculating the balanced part of the above variables.

The second level of balance is achieved by including an explicit fitting of the full divergence tendency to the guess divergence tendency which is used as additional observations for the analysis. Observational errors in the case of Satellite data which were considered to be correlated only in the horizontal but not in the vertical are now correlated both in horizontal as well as vertical.

The concept of superobing has been removed and the residuals are used at the observation location itself instead of at the Guassian grid as used earlier. The concept of superobing is still operative in the case of Satellite data. To give a 4 Dimensional touch to the analysis scheme the previous 6-hrly analysis is used and the residuals were corrected for the time difference between the observation time and the analysis time by taking the trend from the previous analysis and current guess values. Surface reports, significant level data from TEMP/PILOT reports and Precipitable water data from TOVS are the additional data which are being used in the new analysis scheme.

7.2     Medium Range Forecast Model

Brief Description of Global Spectral Model

The model uses a simple land-surface scheme includes:

1. exchange coefficients computations based on Monin Obukov similarity theory

2. Penman Monteith method of evapotranspiration over land which includes vegetation effects

3. prognostic surface temperature equation

4. 3 layer of surface and soil temperature prediction based

5. interactive bucket hydrology

6. evaporation by bulk method over ocean and

7. Charnock's roughness length computation of ocean.

21 Regional Spectral Model

The model output generated by the GDAFS include the following important fields besides many other parameters. The following parameters are produced involving the post-processing at 12 standard pressure levels viz. 1000, 850, 700, 500, 400, 300, 250, 200, 150, 100, 70 and 50hPa, for synoptic assessment of the forecasts at 24hrly interval.

1. Wind Field(Flow Pattern)

2. Geopotential Height

3. Temperature

4. Specific Humidity

5. Vertical Velocity

In addition, Mean Sea Level Pressure(MSLP) and its 24hrly changes from the initial time distribution; Rainfall(accumulated for 24hrs) in quantitative terms estimated both from stratiform and convective type of clouds; Weekly Sub-Divisional Rainfall Distribution on every Wednesday; Location specific surface weather elements viz. Rainfall, Cloud cover, Maximum and Minimum Temperatures and Surface Wind (speed and direction) and Humidity etc.(twice weekly for issuing forecast to farmers) are also computed.

Also, several diagnostic fields(low, medium and high cloud cover; surface fluxes, radiative fluxes etc.) from different physical processes are archived.

• Agrometeorological Advisory Service(AAS)

The main objective of NCMRWF is to provide medium range weather forecast based AAS to the farmers of the country. To accomplish the task, it has been decided to open an Agro meteorological Field Unit (AMFU) one each in 127 agroclimatic zones of the country. At present, 81 AMFUs are established covering that many agroclimatic zones. Currently, the Centre issues weekly forecast (on every Tuesday) to 76 AAS units out of which 43 Units are given forecast bi-weekly (on every Tuesday and Friday). Location specific forecasts to various AMFUs is disseminated through VSATs, FAX/Telephone etc.

• Services rendered to other organizations in the Country

Besides AAS, We also cater to the needs of several other governmental and non-governmental organisations. Following is the list of some of the important users and forecast products provided to them :

• Direct Model Output(DMO) Forecast

For the purpose of preparing location specific weather forecasts for AAS, the DMO forecast is prepared using the predicted values of the required surface weather elements at the grids directly from the global model output. The values of the meteorological elements (analysis/forecast) are obtained from the nearest grid point or from surrounding 4 grid points. For certain stations where orographic influence is large, these values are obtained depending on how best the station is represented by these grid point forecasts. The model output for each of the time step (15 minutes) is accumulated to obtain forecast values over 24-hours ending at 0830 hours IST. The forecast values of the following variables is obtained:

• Total precipitation (mm)

• Mean sea level pressure (hPa)

• Average wind speed (Kmph)

• Predominant wind direction (degrees)

• Maximum temperature (⁰ C)

• Minimum temperature ( ⁰ C)

• Maximum relative humidity (%)

• Minimum relative humidity (%)

• Statistical Interpretation(SI) Forecast

Meteorological parameters like rainfall, maximum temperature and minimum temperature are highly dependent on local topographic and environmental conditions. Numerical weather prediction (NWP) models especially general circulation models(GCM) do not normally represent these local conditions adequately. Therefore, model forecast for a particular location may have certain errors. Upper air features over a location are however not so much dependent on local conditions and can be obtained from the analysis or forecast using a GCM easily. A statistical relation developed between upper air circulation around the location of interest and observed values of the surface weather element at that particular location, is more likely to take account of the effect of these local conditions. This indicates that statistical interpretation(SI) forecast so obtained will have better skill as compared to DMO forecast. In this Centre, Perfect Prog Method(PPM) is adopted in which a statistical relation is derived that relates large sample of observed surface elements (predictands) to concurrent observed surface and upper air reports i.e. analysis (predictors). In order to get a forecast for the appropriate valid time, values of the predictors obtained from NWP models is substituted in the relation developed. This approach assumes that the model forecasts are "perfect". Hence this approach is having a disadvantage that it does not account for systematic biases and errors of the model. This problem can be solved by using the unbiased model forecast, which can be obtained just on the basis of model analysis and forecast data for last one or two months. A major advantage of this method is that stable forecasting relations can be derived from a long period of record. Its forecast improves as NWP model forecast is improved.

• Final Forecast Preparation

It is well known that a great deal of experience is required in order to fully take advantage of NWP products in the operational forecast preparation. The final forecast preparation at this Centre is essentially a man-machine mix approach involving an experienced panel of forecasters. While due weightage is given to the DMO, an experienced synoptician may at times, modify the specific forecast element based on the synoptic climatology of regional specific circulation characteristics. Subsequently, the final weather forecasts issued are verified against observations collected at the respective locations. The following table shows verification scores for selected stations of India during monsoon season of 1999 in respect of forecasts produced by i) Direct Model Output (DMO) ii) Statistical Interpretation(SI) and iii) Final forecast. It is noted that the skill scores for SI forecasts are generally higher than those of DMO and comparable to that of final forecasts.

• Use of high resolution temperature profiles from NOAA satellites
  [globally 120Km and locally 80Km]

• Height Reassignment of CMVs

• Assimilation of direct satellite measured radiance in the global data assimilation system

DATA ASSIMILATION

• Assimilation of INSAT OLR estimates

• Assimilation of SSM/I or OCEANSAT measured total precipitable water content field

• Assimilation of METEOSAT-5 derived winds

• Assimilation of analysed NOAA SST field

• Assimilation of soil moisture

HORIZONTAL RESOLUTION

• Increase the horizontal resolution from existing T-80(150Km) to T-170(75Km)

• Increase the vertical resolution from existing 18 levels to 28 levels

PHYSICAL PROCESSES

• Incorporation of Non-Local Closure Scheme in Boundary Layer

• Incorporation of Simplified Arakawa Schubert Scheme of Deep Cumulus Convection

• Incorporation of improved treatment of Land-Surface Processes

Extension of MRF range to 7-days

Introduction of rainfall verification system

Introduction of Regional scale analysis system

Introduction of Meso-scale forecast