Presentation on theme: "VerificationOf Pilot mode Block level Forecast (Monsoon 2013) &ForecastFor All Districts and Blocks (Monsoon 2014) Ashok Kumar."— Presentation transcript:
VerificationOf Pilot mode Block level Forecast (Monsoon 2013) &ForecastFor All Districts and Blocks (Monsoon 2014) Ashok Kumar
Introduction 1. The block level forecast was started in pilot mode and put on IMD’s website during monsoon 2013.Initially in pilot mode only one district per state. 2. 37 districts one from each state (2 each from UP and Haryana) (342 blocks) had been selected from the efficient AAS units. 3. Rainfall forecast was provided by using 0.25x0.25 resolution MME technique using the general circulation models(GCMs).Forecast for cloud amount was provided from the 0.25x0.25 resolution GCM(T-574) model after synchronizing it with MME forecast. 4. Forecast for other weather parameters up to 3-days was provided by using the 9km WRF model. Forecast for other weather parameters for day-4 and day-5 was provided by using 0.25x0.25 resolution GCM(T-574) model. 5. The forecast had been tested for its skill. Pilot mode block level forecast is coming up to the mark except that certain procedural changes are to be made. 6. Now the block level weather forecast would be made operational for 6500 blocks out of 6648 blocks, for which the latitude, longitude and altitude are readily available during monsoon 2014.
Work Completed 1.List of 6648 blocks in India had been retrieved from Panchayati Raj Ministry’s website after discussing with them. The web site is : panchayat.gov.in. Now this information is available in the website: lgdirectory.gov.in. 2. Latitude, Longitude and altitude for 6500 blocks out of 6648 blocks in India are recorded from the standard world web sites. The standard world websites are; www.worldatlas.com/aatlas/findlatlong.htm www.Veloroutes.org/elevation www.geoplaner.com www.indiamapia.com www.geopostcodes.com www.distancesfrom.com www.collinsmaps.com 3. For rest of the blocks(around 150 only) Survey of India is approached. 4. The work of getting block level maps and latitude & longitude had already been taken up with Ministry of Agriculture and NIC. 5. A prototype procedure including altitude corrections had been developed for getting forecast for all the districts and blocks in India based upon Gaussian grid output from the T-574 Model.
6. The procedure for getting forecast for 37 districts and 342 blocks in India in pilot mode including altitude corrections based upon regular(0.25x0.25) grid output from the T-574 Model, output from 9 Km WRF model and MME forecast for rainfall is implemented on 1 st June 2013. 7. The process of verification of skill of pilot mode block level weather forecast had been completed at MCs and RMCs in IMD as well as at IMD(HQ) by using the observed data obtained from RMCs/MCs. 8. The procedure for full mode block level weather forecast system that is for all the blocks had already been developed and kept ready to implement it with in the minimum stipulated time.
Sr. No.StateBlock Cloud RMSECorrelation D1D3D5D1D3D5 2Andhra PradeshRamagundam 3GujaratAnand1.942.082.190.550.440.33 4Jammu & KashmirBishna Satwari 5UttarakhandGadarpur3.432.282.360.030.490.35 6HaryanaAmbala Hissar 8Madhya PradeshBarghat1.8220.127.116.110.350.39 9SikkimMartam Pakyong Ranka Rhenock 10KeralaVellangallur1.081.381.120.510.270.43 11LakshadweepAmini1.491.81.720.430.370.38 Minicoy1.631.361.480.40.460.47 Agati1.491.531.710.560.660.6 12AssamJorhat1.441.450.540.45 13NagalandDhansiripar5.355.30.180.03 Table 2.3 RMSE and Correlation for forecast for cloud amount Using MME and T-574 model output
Sr. No.StateBlock Parameter: Cloud CorrectUsableUnusable D1D3D5D1D3D5D1D3D5 2Andhra PradeshRamagundam82.9 68.611.48.618.104.22.1684.3 3AssamJorhat6673281968 4GujaratAnand837879101411789 5Jammu & KashmirBishna Satwari 6UttarakhandGadarpur56757012 15321314 7HaryanaAmbala Hissar 9Madhya PradeshBarghat8076 1514 510 SikkimMartam Pakyong Ranka Rhenock 11KeralaVellangallur10097 003030 12LakshadweepAmini8885799918363 Minicoy91 696303 Agathy91 826915303 13NagalandDhansiripar88348988 Table 2.4 Correct, Usable and Unusable for forecast for cloud amount Using MME and T-574 model output
Conclusion (i)The rainfall forecasts are generated through Muti-Model Ensemble(MME) system by using the output from JMA, UKMO,IMD and NCEP models.MME forecast essentially gives the bias free forecast as the different NWP model forecasts are regressed against the observed data. Hence the skill for YES/NO rainfall forecast is very good (Table 2.1) even for the difficult orography regions like high terrain areas and oceanic islands. (ii)The cloud amount forecasts are generated by using the T-574 model output after synchronizing it with the MME YES/NO rainfall forecast. The skill for cloud amount forecast is also coming as very good (Table 2.3 and Table 2.4) even for the difficult orography regions. (iii)The forecast for the rainfall amount is not good uniformly for all the areas, as the rainfall amounts may get averaged out due to the use of regression techniques. Hence the procedure for rainfall amount will be changed and T-574 model forecast for rainfall amounts after synchronizing it with MME YES/NO rainfall forecast would be given.
Sr. No.StateBlock Forecast Days Day1Day3Day5 PDFRPCPDFRPCPDFRPC 1RajasthanJaipur0.620.2073.20.720.2970.50.620.3067.9 2GujaratAnand0.890.4560.00.790.3865.20.760.3764.3 3Madhya PraedeshSeoni0.930.2773.20.890.2967.90.840.3361.6 4HaryanaAmbala0.650.2569.60.680.2271.40.690.3562.5 5Andhra PradeshRamagundam0.540.3062.50.400.3222.214.171.12447.3 6KarnatakaDharwad0.250.0433.00.140.072126.96.36.1997.9 7AssamJorhat0.590.3656.30.460.3057.10.610.2763.0 8West BengalBankura1.00.3667.90.970.3961.60.940.3862.5 9NagalandDhansiripar0.530.4745.30.48 45.30.460.5340.6 10Jammu&KashmirBishanh0.820.4660.00.740.4759.00.800.4462.5 11UttarakhandGadarpur0.310.3742.00.590.2658.30.580.3055.6 12Andaman&NicobarMayabunder0.650.1364.30.700.1467.00.740.1172.3 Table 3.1 PoD, FAR and PC for YES/NO Forecast for Rainfall using T-574 model output
Sr. No.StateBlock Forecast Days Day1Day3Day5 CRUSUNCRUSUNCRUSUN 1RajasthanJaipur58.512.229.348.16.345.6188.8.131.52 2GujaratAnand29.911.958.241.111.047.940.36.952.8 3Madhya PraedeshSeoni15.815.968.311.813.275.010.111.678.3 4HaryanaAmbala51.315.433.350.012.537.540.08.651.4 5Andhra PradeshRamagundam52.95.741.4184.108.40.2069.21.918.9 6KarnatakaDharwad32.45.462.246.20.053.840.05.055.0 7AssamJorhat38.17.954.054.710.034.442.211.346.5 8West BengalBankura15.819.764.58.710.181.211.4 77.2 9NagalandDhansiripar31.212.556.345.814.639.637.220.941.8 10Jammu&KashmirBishanh41.813.444.846.910.642.544.311.444.3 11UttarakhandGadarpur51.18.940.031.811.157.130.08.361.7 12Andaman&NicobarMayabunder20.813.965.320.08.072.018.512.369.2 Table 3.2 Correct, Usable and Unusable forecast for rainfall Using T-574 Model output
Sr. No.StateBlock Forecast Days Day1Day3Day5 RMSECORRRMSECORRRMSE COR R 1RajasthanJaipur3.460.583.100.592.850.55 2GujaratAnand1.700.691.750.601.750.57 3Madhya PraedeshSeoni1.740.731.920.632.000.51 4HaryanaAmbala3.180.583.310.553.310.58 5Andhra PradeshRamagundam1.650.691.660.571.860.45 6KarnatakaDharwad0.790.470.840.500.850.52 7AssamJorhat1.620.071.500.041.370.11 8West BengalBankura3.240.043.39-0.103.70-0.15 9NagalandDhansiripar1.470.231.540.151.290.22 10Jammu&KashmirBishanh4.650.585.010.494.710.47 11UttarakhandGadarpur3.480.433.180.493.240.39 12Andaman&NicobarMayabunder5.330.295.360.175.380.07 Table 3.3 RMSE and Correlation for forecast for minimum temperature Using T-574 Model output
Sr. No.StateBlock Forecast Days Day1Day3Day5 CRUSUNCRUSUNCRUSUN 1RajasthanJaipur28.617.953.538.822.342.939.323.237.5 2GujaratAnand54.526.818.759.820.519.756.323.220.5 3Madhya PraedeshSeoni50.934.814.348.230.421.444.632.223.2 4HaryanaAmbala28.617.054.427.721.450.930.423.246.4 5Andhra PradeshRamagundam60.725.014.351.827.720.550.027.722.3 6KarnatakaDharwad83.015.21.879.518.71.8220.127.116.11 7AssamJorhat42.933.923.242.9 14.254.533.911.6 8West BengalBankura15.211.618.104.22.16822.214.171.1245.9 9NagalandDhansiripar50.033.017.056.624.518.957.529.313.2 10Jammu&KashmirBishanh21.410.767.912.510.776.812.513.474.1 11UttarakhandGadarpur25.921.352.827.8126.96.36.1995.946.3 12Andaman&NicobarMayabunder0.03.6188.8.131.524.60.03.696.4 Table 3.4 Correct, Usable and Unusable for forecast for minimum temperature Using T-574 Model output
Sr. No.StateBlock Forecast Days Day1Day3Day5 RMSECORRRMSECORRRMSE COR R 1RajasthanJaipur3.300.753.410.753.220.68 2GujaratAnand1.680.862.150.772.690.57 3Madhya PraedeshSeoni2.450.882.600.833.420.63 4HaryanaAmbala4.620.715.890.525.050.54 5Andhra PradeshRamagundam2.050.862.350.743.130.47 6KarnatakaDharwad2.210.782.210.741.940.69 7AssamJorhat3.810.233.570.193.620.32 8West BengalBankura2.570.543.030.473.430.38 9NagalandDhansiripar4.590.234.660.255.020.30 10Jammu&KashmirBishanh4.760.655.970.585.220.63 11UttarakhandGadarpur5.960.06.19-0.045.440.01 12Andaman&NicobarMayabunder2.74-0.032.980.03.0-0.01 Table 3.5 RMSE and Correlation for forecast for maximum temperature Using T-574 Model output
Sr. No.StateBlock Forecast Days Day1Day3Day5 CRUSUNCRUSUNCRUSUN 1RajasthanJaipur28.617.054.424.1 51.825.931.242.9 2GujaratAnand47.332.120.636.629.533.925.028.646.4 3Madhya PraedeshSeoni39.325.934.8 28.636.631.318.750.0 4HaryanaAmbala21.415.263.414.320.565.219.618.861.6 5Andhra PradeshRamagundam45.528.625.940.220.539.331.320.548.2 6KarnatakaDharwad37.527.734.832.233.034.841.136.622.3 7AssamJorhat22.325.052.725.920.553.623.218.858.0 8West BengalBankura27.725.047.324.118.757.221.4 57.2 9NagalandDhansiripar10.411.378.310.412.3184.108.40.2069.6 10Jammu&KashmirBishanh13.420.566.18.912.578.68.017.974.1 11UttarakhandGadarpur14.816.768.512.017.670.413.917.668.6 12Andaman&NicobarMayabunder21.420.558.120.5 59.020.517.062.5 Table 3.6 Correct, Usable and Unusable for forecast for maximum temperature Using T-574 Model output
Conclusion (i)The forecast for rainfall amounts using T-574 model output is coming better than the MME forecast, as the unusable forecast is not so high and skill is coming good for some of the stations(Table3.1 and 3.2). Hence the rainfall forecast would be provided by using the T-574 model output after synchronizing it with the MME YES/NO rainfall forecast. (ii) The forecast for maximum/minimum Temperatures using the T-574 model output is giving the good skill as for the RMSE and correlation is concerned(Table 3.3 and 3.5). Correct, Usable and unusable forecast is also coming good for many stations(Table 3.4 and 3.6). Although the high terrain and oceanic islands are showing the problem. Hence day-4 and day-5 forecast for maximum/minimum temperatures and also for the other weather parameters(discussed in next section 5)would be provided by using the T-574 model output and day-1 to day-3 forecast would be provided by using the WRF model output as it is marginally better(discussed in next section) than the T-574 model output.
Sr. No.StateBlock Forecast Days Day1Day2Day3 PDFRPCPDFRPCPDFRPC 1RajasthanJaipur0.980.4560.20.960.4757.60.910.4656.8 2GujaratAnand1.000.4951.70.980.4753.41.000.4458.5 3Madhya PraedeshSeoni0.990.3071.21.000.3367.80.990.3564.4 4HaryanaAmbala1.000.4162.70.950.4062.70.940.4161.0 5Andhra PradeshRamagundam1.000.4752.51.000.4655.10.970.4556.8 6KarnatakaDharwad1.000.1288.11.000.1189.01.000.1189.0 7AssamJorhat0.970.3765.30.940.3566.90.930.3665.2 8West BengalBankura0.960.4060.21.000.3766.10.990.3961.9 9NagalandDhansiripar1.000.4456.31.000.4456.21.000.4456.2 10Jammu&KashmirBishanh0.960.4662.70.870.4760.20.860.4859.3 11UttarakhandGadarpur0.850.2968.40.880.3463.20.860.3264.9 12Andaman&NicobarMayabunder1.000.1980.51.000.1980.51.000.1980.5 Table 4.1 PoD, FAR and PC for YES/NO Forecast for Rainfall using WRF model output
Sr. No.StateBlock Forecast Days Day1Day2Day3 CRUSUNCRUSUNCRUSUN 1RajasthanJaipur25.411.363.322.17.370.622.417.959.7 2GujaratAnand4.914.880.34.819.076.210.218.871.0 3Madhya PraedeshSeoni10.711.977.46.317.576.22.622.475.0 4HaryanaAmbala24.313.562.223.014.962.120.816.762.5 5Andhra PradeshRamagundam1.611.387.17.716.975.413.420.965.7 6KarnatakaDharwad1.911.586.64.824.870.44.825.769.5 7AssamJorhat13.0 74.019.013.967.119.57.071.5 8West BengalBankura9.95.684.5220.127.116.11.912.380.8 9NagalandDhansiripar0.012.718.104.22.1689.34.817.577.8 10Jammu&KashmirBishanh32.412.255.438.014.147.938.611.450.0 11UttarakhandGadarpur24.412.862.812.59.777.822.214.171.124 12Andaman&NicobarMayabunder0.010.589.51.013.785.30.013.786.3 Table 4.2 Correct, Usable and Unusable forecast for rainfall Using WRF Model output
Sr. No.StateBlock Forecast Days Day1Day2Day3 RMSECORRRMSECORRRMSE COR R 1RajasthanJaipur2.280.572.150.582.480.54 2GujaratAnand1.300.571.260.561.320.48 3Madhya PraedeshSeoni1.470.621.450.611.500.54 4HaryanaAmbala2.150.602.260.522.370.51 5Andhra PradeshRamagundam1.340.671.460.541.540.48 6KarnatakaDharwad1.060.571.020.711.110.57 7AssamJorhat1.430.131.470.081.390.16 8West BengalBankura2.44-0.172.68-0.222.81-0.31 9NagalandDhansiripar1.430.291.450.271.490.32 10Jammu&KashmirBishanh3.300.503.340.463.270.44 11UttarakhandGadarpur2.700.332.690.382.900.46 12Andaman&NicobarMayabunder4.210.024.230.054.210.04 Table 4.3 RMSE and Correlation for forecast for minimum temperature Using WRF Model output
Sr. No.StateBlock Forecast Days Day1Day2Day3 CRUSUNCRUSUNCRUSUN 1RajasthanJaipur41.533.125.442.425.432.246.623.729.7 2GujaratAnand67.821.211.066.922.910.266.120.313.6 3Madhya PraedeshSeoni57.626.316.162.722.914.463.622.014.4 4HaryanaAmbala126.96.36.1992.422.035.640.721.238.1 5Andhra PradeshRamagundam59.328.811.955.130.514.444.942.412.7 6KarnatakaDharwad67.826.35.9188.8.131.520.323.76.0 7AssamJorhat52.535.611.953.434.711.960.228.011.8 8West BengalBankura17.834.144.111.035.653.413.629.656.8 9NagalandDhansiripar46.939.314.352.733.014.361.619.618.8 10Jammu&KashmirBishanh35.616.947.524.626.349.130.522.946.6 11UttarakhandGadarpur35.125.439.536.825.437.730.728.141.2 12Andaman&NicobarMayabunder16.126.357.616.928.854.321.224.654.2 Table 4.4 Correct, Usable and Unusable for forecast for minimum temperature Using WRF Model output
Sr. No.StateBlock Forecast Days Day1Day2Day3 RMSECORRRMSECORRRMSE COR R 1RajasthanJaipur2.880.763.410.753.610.75 2GujaratAnand1.590.852.170.782.270.76 3Madhya PraedeshSeoni2.410.813.040.732.900.72 4HaryanaAmbala2.940.713.870.644.050.63 5Andhra PradeshRamagundam1.760.862.890.603.300.45 6KarnatakaDharwad1.340.711.370.711.400.67 7AssamJorhat184.108.40.2060.312.160.16 8West BengalBankura2.110.512.340.472.440.44 9NagalandDhansiripar1.850.341.780.301.760.34 10Jammu&KashmirBishanh4.320.685.430.625.790.59 11UttarakhandGadarpur4.08-0.024.74-0.024.78-0.03 12Andaman&NicobarMayabunder220.127.116.11-0.042.24-0.10 Table 4.5 RMSE and Correlation for forecast for maximum temperature Using WRF Model output
Sr. No.StateBlock Forecast Days Day1Day2 CRUSUNCRUSUNCRUSUN 1RajasthanJaipur29.726.344.024.622.952.528.826.344.9 2GujaratAnand46.633.120.343.223.733.134.830.534.7 3Madhya PraedeshSeoni33.934.731.423.735.640.723.029.642.4 4HaryanaAmbala33.028.039.023.017.818.104.22.1688.3 5Andhra PradeshRamagundam55.125.419.533.125.441.522.214.171.124 6KarnatakaDharwad61.927.111.060.227.112.757.627.115.3 7AssamJorhat42.426.633.046.629.723.741.522.935.6 8West BengalBankura35.632.2 39.025.435.639.824.635.6 9NagalandDhansiripar49.127.723.257.125.017.94126.96.36.199 10Jammu&KashmirBishanh26.317.855.916.918.664.518.616.964.5 11UttarakhandGadarpur27.218.454.423.720.2188.8.131.522.6 12Andaman&NicobarMayabunder32.239.828.037.333.129.634.837.327.9 Table 4.6 Correct, Usable and Unusable for forecast for maximum temperature Using WRF Model output
Conclusion (i)Although the probability of detection is coming very good for YES/NO rainfall forecast obtained from WRF model, but the false alarm rate is also higher(Table 4.1). It had been observed that WRF model is giving rainfall on most of the days and its percentage of unusable rainfall amounts is uniformly very high for all the areas(Table 4.2). Hence this forecast could not be used for rainfall forecast. (ii)The skill of maximum/minimum temperatures forecast by using the WRF model output is coming very good as far as the RMSE and Correlation is concerned(Table 4.3 and 4.5). Even for the correct, usable and unusable skill of the forecast the WRF model is showing a very good skill and it is marginally better than T-574 model forecast(Table 4.4 and 4.6), although difficult orography regions are showing the problems. Hence WRF model output would be used for maximum/minimum temperatures and also for the other weather parameters(discussed in next section 5) for giving the forecast for Day-1 to Day-3.
(i)For other weather parameters(other than rainfall and maximum/minimum temperatures) the forecast for Day-1 to Day-3 was provided by using the WRF model output and forecast for Day-4 and Day-5 was provided by using the T-574 model output.
Sr. No.StateBlock Minimum Relative Humidity RMSECorrelation D1D3D5D1D3D5 2Andhra PradeshRamagundam 3GujaratAnand15.5520.5524.50.660.530.23 4Jammu & KashmirBishna4.670.78 Satwari5.060.77 5UttarakhandGadarpur5.375.345.640.350.480.28 6HaryanaAmbala Hissar 8Madhya PradeshBarghat24.222.5184.108.40.206 9SikkimMartam Pakyong Ranka Rhenock 10KeralaVellangallur3.422.742.7220.127.116.11 11LakshadweepAmini2.32.093.420.090.20.16 Minicoy2.312.832.510.230.120.18 Agati2.082.132.49-0.02-0.11-0.07 12AssamJorhat2.743.110.460.34 13NagalandDhansiripar3.113.050.410.32 Table 5.1 RMSE and Correlation for forecast for minimum relative humidity Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Sr. No.StateBlock Parameter: Minimum Relative Humidity CorrectUsableUnusable D1D3D5D1D3D5D1D3D5 2Andhra PradeshRamagundam51.445.754.331.437.1 17.1 8.6 3AssamJorhat4230 28 42 4GujaratAnand523126.230.33225173748 5Jammu & KashmirBishna142660 Satwari42671 6UttarakhandGadarpur276202415786979 7HaryanaAmbala Hissar 9Madhya PradeshBarghat22372642262936 45 10SikkimMartam Pakyong Ranka Rhenock 11KeralaVellangallur6271681826322130 12LakshadweepAmini9497416356003 Minicoy85 151215030 Agathy94 856615000 13NagalandDhansiripar273128274542 Table 5.2 Correct, Usable and Unusable for forecast for minimum relative humidity Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Sr. No.StateBlock Maximum Relative Humidity RMSECorrelation D1D3D5D1D3D5 2Andhra PradeshRamagundam 3GujaratAnand4.735.3512.830.680.530.55 4Jammu & KashmirBishna3.770.63 Satwari3.430.72 5UttarakhandGadarpur18.104.22.168.350.40.25 6HaryanaAmbala Hissar 8Madhya PradeshBarghat22.214.171.124.70.60.4 9SikkimMartam Pakyong Ranka Rhenock 10KeralaVellangallur2.512.592.99-0.15-0.25-0.2 11LakshadweepAmini1.791.743.550.310.430.19 Minicoy2.212.263.030.180.340.13 Agati2.32.342.88-0.1-0.05-0.12 12AssamJorhat3.263.230.020.13 13NagalandDhansiripar2.913.13-0.10.002 Table 5.3 RMSE and Correlation for forecast for maximum relative humidity Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Sr. No.StateBlock Parameter: Maximum Relative Humidity CorrectUsableUnusable D1D3D5D1D3D5D1D3D5 2Andhra PradeshRamagundam40.0 37.140.031.425.720.028.637.1 3AssamJorhat272220.52752.551 4GujaratAnand9695484543.4008.2 5Jammu & KashmirBishna423523 Satwari602218 6UttarakhandGadarpur6459532131 151016 7HaryanaAmbala Hissar 9Madhya PradeshBarghat817381162415444 10SikkimMartam Pakyong Ranka Rhenock 11KeralaVellangallur797453212647000 12LakshadweepAmini100 240076000 Minicoy9197629338000 Agathy88916512935000 13NagalandDhansiripar402322353842 Table 5.4 Correct, Usable and Unusable for forecast for maximum relative humidity Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Sr. No.StateBlock Parameter: Wind Speed RMSER D1D3D5D1D3D5 1Andhra PradeshRamagundam 2AssamJorhat 3GujaratAnand7.438.3513.820.340.190.34 4Jammu & KashmirBishna100.36 Satwari10.720.32 5UttarakhandGadarpur9.3611.075.260.150.380.42 6HaryanaAmbala Hissar 7West BengalBankura 8Madhya PradeshBarghat 13.5217.069.85-0.04 -0.01 9SikkimMartam Pakyong6.456.676.44-0.180.180.03 Ranka Rhenock 10KeralaVellangallur7.057.8614.940.33-0.070.10 11LakshadweepAmini11.9815.060.570.50 Minicoy14.310.31 Agathy9.88.440.180.38 12NagalandDhansiripar Table 5.5 RMSE and Correlation for forecast for wind speed Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Sr. No.StateBlock Parameter: Wind Speed CorrectUsableUnusable D1D3D5D1D3D5D1D3 D5D5 1 Andaman and Nicobar Mayabandar Rangat 2Andhra PradeshRamagundam 3AssamJorhat34.4323.7728.6931.9736.8944.26 4GujaratAnand57.3847.5414.7540.985043.441.642.46 41.8 5 Jammu & Kashmir Bishna38.5252.469.02 Satwari22.1368.039.84 6UttarakhandGadarpur61.025085.8929.6632.214.419.3217.80 7HaryanaAmbala Hissar 8West BengalBankura 9Madhya PradeshBarghat31182835276834554 10SikkimMartam Pakyong 83.61 80.33 76.2316.3919.67 23.77 0 00 Ranka Rhenock 11KeralaVellangallur7459122441533035 12LakshadweepAmini322947262144 Minicoy241244293259 Agathy5662593229321299 13NagalandDhansiripar Table 5.6 Correct, Usable and Unusable for forecast for wind speed Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Sr. No.StateBlock Parameter: Wind Direction RMSER D1D3D5D1D3D5 1Andhra PradeshRamagundam 2AssamJorhat 3GujaratAnand52.5954.7126.96.36.1990.09 4Jammu & KashmirBishna71.94 Satwari60.69 5UttarakhandGadarpur68.6268.185.090.08-0.090.00 6HaryanaAmbala Hissar 7West BengalBankura 8Madhya PradeshBarghat 57.872.465.80.940.950.96 9SikkimMartam Pakyong Ranka Rhenock 10KeralaVellangallur102.38 59.550.040.110.29 11LakshadweepAmini25.7229.826.360.290.12-0.16 Minicoy42.20.16 Agathy22.1923.625.080.680.620.48 12NagalandDhansiripar Table 5.7 RMSE and Correlation for forecast for wind direction Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Sr. No.StateBlock Parameter: Wind Direction CorrectUsableUnusable D1D3D5D1D3D5D1D3D5 1Andhra PradeshRamagundam 2AssamJorhat 3GujaratAnand69.67 76.2310.667.381.6419.6722.95 22.1 3 4Jammu & KashmirBishna8.1910.6581.14 Satwari11.4715.5772.95 5UttarakhandGadarpur43.2233.0528.816.7816.105.085050.84 66.1 0 6HaryanaAmbala Hissar 7West BengalBankura 8Madhya PradeshBarghat4933501285405944 9SikkimMartam Pakyong Ranka Rhenock 10KeralaVellangallur263235391571550 11LakshadweepAmini7471881518612 6 Minicoy534488915638416 Agathy88768231869612 NagalandDhansiripar Table 5.8 Correct, Usable and Unusable for forecast for wind direction Using WRF(Day-1 to 3) and T-574(Day-4 &5) Model output
Conclusion (i)Verification results for maximum/minimum relative humidity have shown the good skill(Table 5.1 to 5.4), although some of the difficult orography(high terrain and oceanic islands) areas have shown the problems. (ii)Verification results for wind speed have also shown the good skill(Table 5.5 to 5.6), in this case also some of the difficult orography areas have shown problems. (iii)Verification results for wind direction have not shown the good results for many stations(Table 5.7 to 5.8). This could be due to the verification procedure being followed presently. The verification procedure needs to be changed for the typical weather parameter like wind direction. Here one suggestion could be calculating the simple success rate for wind direction by putting the values into the eight quadrants and if the observed and forecasted vales lies in the same quadrant then it is success case and total number success cases divided by the total number of cases would give the success rate.
Final Conclusion (i)The rainfall forecast would be provided by using the T-574 model output after synchronizing it with the MME YES/NO rainfall forecast for all the five days(Day-1 to Day-5). (ii)The cloud amount forecast would be provided by using the T- 574 model output after synchronizing it with the MME YES/NO rainfall forecast for all the five days(Day-1 to Day-5). (iii)Weather forecast for all other weather parameters including maximum and minimum temperatures would be given by using the WRF model output for Day-1 to Day-3 and by using the T-574 model output for Day-4 to Day-5. (iv)The difficult orography regions that is high terrain areas and oceanic islands have shown the problems in the skill of the forecast. This is a typical problem for which the NWP models needs to be improved based upon the results obtained from the special projects like Himalayan Meteorology and Coupled Ocean Atmospheric model. The alternative way could be trying the -Neural Network Technique for MME forecast for rainfall. -Kalman filter for Maximum/minimum temperatures. -Model Output Statistics(MOS) guidance for different weather parameters.
Future Plans i) The block level weather forecast would be made operational for 6500 blocks out of 6648 blocks, for which the latitude, longitude and altitude are readily available. The procedure had already been developed and most likely it would be tried to implement the block level weather forecast for almost all the blocks from 1 st June 2014. ii) Following procedural changes in the weather forecast would be carried out in future; -Including the block level maps for putting forecast on web. -Putting the value addition table for districts showing the different type of forecasts. -Ikonical representation of forecast for rainfall and max./min. temperature. -Putting the Meteograms for all the districts and blocks on web. iii) Kalman Filter technique would be applied for getting the bias free Maximum/Minimum temperature forecast after the regular grid (0.25 degree resolution) temperature data would become available. iv) Neural Network technique would be applied for improving the MME technique used for rainfall forecast. v) The procedure for the block level weather forecast would be developed and implemented again by using the output form the high resolution(10-12Km) GCM, when it would become operational.
vi) Verification at each grid point for the Indian region would be carried out for rainfall and maximum/minimum temperature, once the regular grid data would become available. vii) These procedures for bias free block level location specific weather forecast would be tried for other weather parameters as well. viii) After the success of block level weather forecast, location specific weather forecast would be attempted for over 2 lakh village panchayat using the WRF model output at 9 Km resolution for three days. ix) Developing these procedures into an expert system for getting the forecast and the meteogram for any place with the specific latitude, longitude and altitude. x) Automation of forecast bulletin and agro-advisory for block and village panchayat level on mobile. xi) Redevelopment of the above mentioned procedures and regeneration of the forecasts required due to the up-gradation of the GCM to the higher resolution.
District and Block Level Forecast And Its Verification For Monsoon 2014
District And Block Level Forecast -Rainfall forecast would be provided after synchronizing the MME forecast with the T-574 model forecast. -Cloud amount forecast would also be provided based upon T-574 model after synchronizing it with MME rainfall forecast. -Altitude correction would be applied for maximum and minimum temperature. -Day-1 and Day-2 forecast for other weather parameters would be provided from WRF model.(considering that one day would be lost in running the forecast procedures) -Day-3,Day-4 and Day-5 forecast for other weather parameters would be provided from T-574 Model. -All these forecasts would be generated at IMD’s HPC systems and comprehensive forecast tables containing the above mentioned comprehensive forecast for 655 districts and 6500 blocks would be generated every day and put on ftp-server and IMD’s web browser.
Forecast files from ftp-server For accessing the data one has to type on the web browser as follows; ftp://188.8.131.52 user: anonymous password: anonymous Then go to “pub” directory and then “DIST_BLOCK_LEVEL_FCST” directory which contains four directories dfcst, dfct, bfcst, bfct and each such directory contains the 35 states and states directories contains the forecast files with name as f_name00zyymmdd. To have formatted forecast files, which would be utilized for forecast verification and other studies, one has to see the following directories; /pub/DIST_BLOCK_LEVEL_FCST/dfcst/f_name00zyymmdd :-for districts /pub/DIST_BLOCK_LEVEL_FCST/bfcst/f_name00zyymmdd :-for blocks (There is “1-how-to-read” file in each of dfcst and bfcst directories as well) To have files containing the forecast tables, which would be utilized for generation of value added forecast and issuing the forecast to AAS units, one has to see the following directories; /pub/DIST_BLOCK_LEVEL_FCST/dfct/f_name00zyymmdd :- for districts /pub/DIST_BLOCK_LEVEL_FCST/bfct/f_name00zyymmdd :-for blocks ***Remark:- For getting forecast file on any particular day, the file with date stamp of previous day is to be looked into.***
Forecast files from web browser After opening the web site www.imd.gov.in, one has to open up the following ikons; NWP PRODUCTS SERVICES FORECAST DISTRICT AND BLOCK LEVEL And then click on the drop down window or on map. For the time being drop down window is available and very soon district maps would also be included. Note:- On any particular day if the comprehensive forecast for districts and blocks under the directory DIST_BLOCK_LEVEL_FCST, is not available due to HPC maintenance and other unforeseen circumstances, then one can look for the earlier version of the district level forecast only in the following directory on ftp-server; /pub/DIST_ENS_FCST/files
Forecast Verification and report Formatted forecast files are already being put on ftp-server which would be utilized for forecast verification at district and block level separately. The verification report would be entitled “Report on verification of operational district and block level weather forecast” containing two sections one is on District level forecast and another is on block level forecast. Please include at least 20 to 50 blocks for which observed data is available for having state wise scores. But care should be taken that only 10% blocks should be high terrain and oceanic islands and not all. As 48 hr forecast is issued as Day-1 forecast and so on, hence care should be taken that starting from 48hr forecast would be verified as Day-1, and 144 hr forecast would be verified as Day-5 forecast. These forecasts are for next 24hrs as shown against each date starting from 8.30 A.M. in the formatted files, where as in the forecast table files it for the previous 24 hours.