1. Application of low- resolution ETA model data to provide guidance to high impact weather in complex terrain Juha Kilpinen Finnish Meteorological Institute (FMI), Finland Juan Bazo, Gerardo Jacome & Luis Metzger Servicio National de Meteorologia e Hidrologia (SENAMHI), Peru

2. Co-operation between FMI and SENAMHI Institutional development Among the items: 1.Forecast verification (training, application development: an entry level verification system) 2.Post-processing of NWP output (training, application development: testing of Kalman filtering for NWP data ) 4.10.20152Kilpinen-Bazo-Jacome-Metzger

3. 4.10.2015Kilpinen-Bazo-Jacome-Metzger3 EUMETCAL training tools used

4. User-interface for the verifcation system at SENAMHI 4.10.2015Kilpinen-Bazo-Jacome-Metzger4

5. Introduction Peruvian ETA-model data is available for guidance at Peruvian Hydro-meteorological service (SENAMHI). The application of rather low resolution (22 km grid) data is not straight forward in complex terrain in terms of topography, climatic zones and sharp land-sea gradient. The applicability of the data is evaluated and if some problems are encountered a solution will be searched. So far data has been partly verified and tests with Kalman filter have started. The test data includes maximum and minimum temperature. Another focus is the heavy precipitation in Machu Picchu area resulting flooding and danger to life and property. 4.10.20155Kilpinen-Bazo-Jacome-Metzger

6. ETA-model Resolution 22 km 18 vertical levels Output 6 h interval up to 72 h Kain Frish cumulus parameterization the GFS global model data is used for the lateral boundaries No data assimilation is made Test data periods: temperature 2009-2010 Precipitation 2010 Observations (21 stations) Tumbes Piura Huánuco Pucallpa Lima Andahuaylas Cuzco Tacna Chiclayo Chimbote Cajamarca Tarapoto Yurimaguas Iquitos Tingo Maria Trujillo Ayacucho Puerto Maldonado Arequipa Juliaca Pisco 4.10.2015Kilpinen-Bazo-Jacome-Metzger6 Data

7. Methods – post-processing/Kalman filter Only for temperature forecasts: Two state parameters T Kalman = B 1 + B 2 * T ETA Optimized estimation of measurement noise R and system noise Q 4.10.2015Kilpinen-Bazo-Jacome-Metzger7

8. Methods - verification For temperature forecasts: ME (Mean Error) RMSE (Root Mean Squared Error) HR (Hit Rate) (not shown) Also other scores For precipitation forecasts (categories): B (Bias) PC (Percent correct) POD (Probability of Detection) FAR (False Alarm Rate) KSS (Kuipers Skill Score) TS (Threat Score) ETS (Equility Threat Score) … 4.10.2015Kilpinen-Bazo-Jacome-Metzger8

9. Stations in focus: Cusco 3399 m Pucallpa 154 m Andahuaylas 2866 m Huanuco 1859 m Lima 13 m Tacna 452 m Machu Picchu 4.10.2015Kilpinen-Bazo-Jacome-Metzger9

10. TMAXMEME_KALRMSERMSE_KAL Day 1 -7.09-0.337.612.99 Day 2 -7.07-0.347.683.00 Day 3 -7.09-0.377.843.06 TMINMEME_KALRMSERMSE_KAL Day 1 -2.88-0.164.002.35 Day 2 -2.12-0.133.502.36 Day 3 -2.00-0.123.482.38 Temperature verification Cusco (near Machu Picchu) 3399 m

11. TMAXMEME_KALRMSERMSE_KAL Day 1 -3.00 -0,104,012,36 Day 2 -2.99 -0,104,292,52 Day 3 -3.18 -0,094,582,52 TMINMEME_KALRMSERMSE_KAL Day 1 1,580,133,202,68 Day 2 2,680,163,902,68 Day 3 2,820,184,082,70 Pucallpa 154m

12. TMAXMEME_KALRMSERMSE_KAL Day 1 0,56-0,0153,423,12 Day 2 0,760,0053,453,13 Day 3 0,790,0253,503,09 TMINMEME_KALRMSERMSE_KAL Day 1 3,430,154,593,31 Day 2 4,920,265,763,42 Day 3 5,120,275,923,38 Andahuaylas 2866 m

13. TMAXMEME_KALRMSERMSE_KAL Day 1 -8,99-0,589,242,38 Day 2 -8,84-0,479,112,44 Day 3 -8,52-0,538,902,50 TMINMEME_KALRMSERMSE_KAL Day 1 -2,79-0,183,472,12 Day 2 -1,50-0,102,622,01 Day 3 -1,20-0,142,692,06 Huanuco 1859 m

14. TMAXMEME_KALRMSERMSE_KAL Day 1 -1,25-0,132,261,48 Day 2 -1,36-0,102,241,46 Day 3 -1,48-0,152,401,54 TMINMEME_KALRMSERMSE_KAL Day 1 -0,44-0,020,970,76 Day 2 -0,020,020,800,74 Day 3 -0,050,020,800,74 Lima 13 m

15. TMAXMEME_KALRMSERMSE_KAL Day 1 0,960,092,961,38 Day 2 0,920,043,461,40 Day 3 0,620,013,441,54 TMINMEME_KALRMSERMSE_KAL Day 1 -1,02-0,042,181,38 Day 2 0,740,101,971,33 Day 3 0,630,102,001,34 Tacna 452 m

16. Machu Picchu area verification of ETA model 4.10.2015Kilpinen-Bazo-Jacome-Metzger16 Precipitation forecasts

17. Machu Picchu 2010 4.10.2015Kilpinen-Bazo-Jacome-Metzger17

18. Categorical results: Machu Picchu 2010 Day 1Day 2 R>3mm yesno yesno yes10437yes10143 no47158no41152 R>15mm yesno yesno yes1319yes1939 no40247no31248 R>3mmR>15mmR>3mmR>15mm B0,930,6 1,011,16 PC0,760,830,750,79 POD0,690,250,710,38 FAR0,260,590,30,67 KSS0,50,180,490,24 TS0,550,180,540,21 ETS0,230,11 0,220,12 4.10.201518Kilpinen-Bazo-Jacome-Metzger

19. Conclusions The preliminary results indicate that ETA model has problems with temperature forecasts in most regions; in tropical rain forest, at mountains and near coastline affected by cold sea current. The application of Kalman filter was used to minimize systematic errors from the ETA-model and rather encouraging results was received. For precipitation forecasts ETA model is not able to forecast higher precipitation amounts correctly A higher resolution (e.g. WRF) NWP model would be a way to inprove the forecast quality in complex terrain 4.10.2015Kilpinen-Bazo-Jacome-Metzger19