9. Significance of the research - The primary reason of implementing meteorological satellites is avoid the coverage and time gap of conventional ground-based rainfall data for a number of applications, above all hydrology and weather forecasting. - Similar to any observational data, investigating their accuracy and limitations is crucial. This is done by verifying the satellite estimates against independent data from rain gauges and radars (Levizzani, et al. 2007). - No study has compared the potential of METEOSAT 8 and METEOSAT 9 MPE products with ground based radar data and/or gauge in estimating rainfall in high spatial and temporal resolution.

12. Source data and study area (2) EUMETSAT MPE products and blending algorithm (mm/h) (mm/h) Geostationary satellite (MET8 and MET9) High temporal-spatial resolution Cloud infrared images Polar-SSM/I MW- hourly rainfall measurement Blending Algorithm Multi-sensor Precipitation Estimate (MPE)

14. Visual comparison: Met8-Radar-Met9 1 st event ( 20090514, 19:00 UTC to 20090517, 23:45 UTC)_UTM zone 32_WGS 84,3×3 km grid size Instantaneous 1 Hour accumulated 3 Hours accumulated mm/h mm/h mm/h mm/h mm/h mm/h mm/3h mm/3h mm/3h

15. Visual comparison : Met8-Radar-Met9 mm/h mm/h 2 nd event (20090525, 0000 UTC to 20090527, 23:45 UTC)_ UTM zone 32_WGS 84,3×3 km Instantaneous 1 Hour accumulated 3 Hours accumulated mm/h mm/h mm/3h mm/3h mm/3h mm/h mm/h

16. Research question and method (1) What are the differences in spatial distribution of EUMETSAT MPE products from METEOSAT 8 (5 min temporal resolution) and METEOSAT 9 (15 min temporal resolution) in comparison with reference data? Categorical comparison

17. Research question and method (2) What are the differences in spatial distribution of EUMETSAT MPE products from METEOSAT 8 (5 min temporal resolution) and METEOSAT 9 (15 min temporal resolution) in comparison with reference data?

18. Categorical statistical results obtained Instantaneous One hour accumulated Three hours accumulated Statistical Score Radar _METEOSAT 9 Radar _METEOSAT 8(RSS) Event1 Event2 Event1 Event2 POD 0.46 0.28 0.41 0.33 FAR 0.17 0.30 0.20 0.34 CSI 0.42 0.25 0.37 0.28 Accuracy 0.74 0.72 0.71 0.72 Bias 0.55 0.40 0.52 0.51 ETS 0.27 0.15 0.22 0.17 Statistical Score Radar _METEOSAT 9 Radar _METEOSAT 8(RSS) Event1 Event2 Event1 Event2 POD 0.53 0.31 0.51 0.34 FAR 0.13 0.27 0.16 0.31 CSI 0.49 0.28 0.46 0.30 Accuracy 0.73 0.69 0.71 0.68 Bias 0.61 0.43 0.60 0.50 ETS 0.30 0.15 0.26 0.15 Statistical Score Radar _METEOSAT 9 Radar _METEOSAT 8(RSS) Event1 Event2 Event1 Event2 POD 0.57 0.35 0.59 0.36 FAR 0.07 0.15 0.10 0.17 CSI 0.54 0.33 0.55 0.34 Accuracy 0.70 0.58 0.69 0.58 Bias 0.61 0.41 0.65 0.44 ETS 0.28 0.13 0.26 0.13

19. Research question and method (1) What are the differences in estimated values by EUMETSAT MPE products from METEOSAT 8 (5 min temporal resolution) and METEOSAT 9 (15 min temporal resolution) in comparison with reference data? Continuous comparison

20. Research question and method (2) What are the differences in estimated values by EUMETSAT MPE products from METEOSAT 8 (5 min temporal resolution) and METEOSAT 9 (15 min temporal resolution) in comparison with reference data?

21. Results of continuous comparison Instantaneous 1 Hour accumulated 3 Hours accumulated First event Second event

22. Instantaneous 1 Hour accumulated 3 Hours accumulated First event Second event Results of continuous comparison

23. Instantaneous 1 Hour accumulated 3 Hours accumulated First event Second event Results of continuous comparison

24. Instantaneous 1 Hour accumulated 3 Hours accumulated First event Second event Results of continuous comparison

27. Thank you

28. 2*2 contingency table Ground based RADAR Yes No RSS_METEOSAT8 YES Hits False alarms Estimated Yes NO Misses Correct negatives Estimated No Observed Yes Observed No N=Total

29. Scatter graph for MPE products, instantaneous comparison, second event The diagonal dotted line in the graph shows the ideal 1:1 relationship between reference values and estimated values by METEOSAT 8 and METEOSAT 9

30. Regressions for mean rainfall values in MPE products, α=0.05 Instantaneous One hour accumulated Three hours accumulated Radar _METEOSAT 9 Radar _METEOSAT 8(RSS) Event1 Event2 Event1 Event2 Observations (n) 288 308 288 308 Slope coefficient, p-value 7.35, 3.56E-122 1.84, 3.2E-64 7.96, 1.51E-127 1.21, 1.42E-58 Intercept coefficient, p-value -0.11, 0.047 -0.03, 0.27 -0.16, .004 0.01, 0.50 r² 0.85 0.61 0.87 0.57 Significance F 3.57E-122 3.2E-64 1.51E-127 1.42E-58 Radar _METEOSAT 9 Radar _METEOSAT 8(RSS) Event1 Event2 Event1 Event2 Observations (n) 72 77 72 77 Slope coefficient, p-value 7.60, 6.13E-35 1.84, 9.72E-17 8.23, 2.59E-36 1.20, 3.89E-15 Intercept coefficient, p-value -0.13, 0.18 -0.03, 0.58 -0.18, .06 0.01, 0.71 r² 0.89 0.60 0.90 0.56 Significance F 6.13E-35 9.72E-17 2.59E-36 3.89E-15 Radar _METEOSAT 9 Radar _METEOSAT 8(RSS) Event1 Event2 Event1 Event2 Observations (n) 24 25 24 25 Slope coefficient, p-value 7.82, 1.79E-14 1.98, 6.61E-07 8.43, 2.61E-15 1.26, 7.46E-06 Intercept coefficient, p-value -0.50, 0.19 -0.13, 0.59 -0.65, 0.09 0.02, 0.92 r² 0.93 0.67 0.94 0.59 Significance F 1.79E-14 6.61E-07 2.61E-15 7.46E-06

31. Cloud types Source :Dr. Maathuis presentation on METEOSAT-MSG

32. Illustration of the IR signal from different cloud types Source :Dr. Maathuis presentation on METEOSAT-MSG

33. METEOSAT8 METEOSAT9 0º 70 º N 12 min 15º N 5min 70 º S 52º N 140/12=11.65º/min 11min to the Netherlands from start point 3.5min to the Netherlands from start point Time stamp on images are based on the end of scan time In case of RSS, 1.5min difference between real scan time and time stamp on image In case of MET9, 4min difference between real scan time and time stamp on image