1. 1 GOES-R AWG Hydrology Algorithm Team: Rainfall Probability June 14, 2011 Presented By: Bob Kuligowski NOAA/NESDIS/STAR

2. 2 Outline  Executive Summary  Algorithm Description  ADEB and IV&V Response Summary  Requirements Specification Evolution  Validation Strategy  Validation Results  Summary

3. 3 Executive Summary  This Rainfall Probability Algorithm generates the Option 2 product of predicted probability rainfall accumulation exceeding 1 mm for the next 3 h on the ABI grid.  Version 3 (80%) was delivered in September 2010. Version 5 (100%) delivery has been delayed 12 months in response to GPO guidance.  An algorithm has been developed that produces conditional probabilities based on the outputs and intermediate products of the Rainfall Potential algorithm.  Preliminary validation of the version 3 algorithm has been performed against 20 days of SEVIRI data from 2005.  Validation against Nimrod radar indicates partial spec compliance, and modifications have brought it closer to spec.

4. 4 Algorithm Description

5. 5  This Rainfall Probability Algorithm generates the Option 2 product of predicted probability rainfall accumulation exceeding 1 mm for the next 3 h on the ABI grid.  The algorithm derives the probability of rainfall during the next 3 h from the current Rainfall Rate product and intermediate (every 15 min) nowcasts of rainfall from the Rainfall Potential algorithm.  The algorithm was calibrated using regression against the Rainfall Rate product instead of ground measurements to: »Eliminate uncertainties associated with errors in the Rainfall Rate algorithm; »Allow much more spatially widespread calibration (ground truth is generally available over Western Europe only)

6. 6 (Original) Algorithm Description Probability of Rainfall if pixel has one or more intermediate Rainfall Potential forecasts of non-zero rainfall Probability of Rainfall if pixel Rainfall Potential is zero but a neighbor pixel has a non-zero Rainfall Potential More intermediate forecasts at pixel with non-zero rain rate  higher probability Greater distance to nearest neighbor with non-zero rain rate  lower probability

7. Example Probability of Rainfall Output Probability of Rainfall from 1500- 1800 UTC 8 July 2005 derived from intermediate output of Rainfall Potential algorithm (which in turn uses SEVIRI-derived Rainfall Rate fields from 1445 and 1500 UTC as input).

8. 8 ADEB and IV&V Response Summary  ADEB did not recommend advancing the Rainfall Probability algorithm to 100% based on »Concerns about the 1 mm/h threshold—should be higher »Concerns about the “method and utility” of the algorithm  As a result, the algorithm development schedule has been delayed by 12 months.  Regarding the 1 mm/h threshold, the AT surveyed users from OSPO/SAB, NWS/HPC, NWS/OHD, and NWS/WGRFC users who indicated interested in probabilities at multiple thresholds. »However, the thresholds and additional data volume would need GPO approval.

9. 9 ADEB and IV&V Response Summary  Regarding concerns about the “method”, the algorithm is currently being re-developed based on feedback from the ADEB, testing new predictors and evaluating the skill of the resulting algorithm.  Regarding concerns about the “utility” of the algorithm, the user survey found significant interest in the product and for it to be available at launch.

10. 10 ADEB and IV&V Response Summary  The ADEB also raised general concerns about the amount of validation.  High-quality rainfall data at time scales of 3 h or less have been extremely difficult to find. Planned solutions: »CHUVA (Brazil) field campaign data »Global Precipitation Climatology Center (GPCC) gauge data (can only be used in-house—visit in summer 2012?) »EUMETSAT Hydrology SAF—reaching out to see if their data (also in-house only) could be used via a visit.

11. 11 Requirements Specification Evolution 11 Name User & Priority Geographic Coverage (G, H, C, M) Vertical Resolution Horizontal Resolution Mapping Accuracy Measurement Range Measurement Accuracy Product Refresh Rate / Coverage Time (Mode 3) Refreshment Rate / Coverage Time (Mode 4) Vendor Allocated Ground Latency Product Measurement Precision Probability of Rainfall GOES-RFDN/A2.0 km1.0 km0 to 100% 25%FD: 15 min FD: 5 min 266 sec40% Temporal Coverage Qualifiers Product Extent Qualifier Cloud Cover Conditions Qualifier Product Statistics Qualifier Day and nightQuantitative out to at least 70 degrees LZA or 60 degrees latitude, whichever is less, and qualitative beyond N/AOver rainfall cases and mesoscale-sized surrounding regions

12. 12 Validation Strategy

13.  Since the requirement is for rainfall accumulations of 3 h, radar and short-term gauges are the only available source of data for validation against spec  Ground-based radars: »Nimrod radars in UK and Western Europe—5-km grid composite Sample Nimrod 3-h accumulation

14. 14 Validation Strategy  Derive probability of rainfall from intermediate Rainfall Potential outputs (in turn derived from rainfall rates retrieved from GOES-R ABI proxy datasets) µm »Meteosat-8 SEVIRI inputs to rainfall rates: Bands 5 (6.2 µm), 6 (7.3 µm), 7 (8.7 µm), 9 (10.8 µm), and 10 (12.0 µm)  Collocate (in space and time) derived rainfall potential with ground validation rainfall accumulations »3-hour observed probability of rainfall (0 or 1) derived from Nimrod radar  Generate comparative statistics (satellite rainfall accumulation – ground truth rainfall accumulation) »Accuracy »Precision

15. 15 Validation Results

16. 16 Rainfall Potential Validation (Original algorithm) Reliability diagram for Probability of Rainfall vs.Nimrod radar data (Western Europe only) for the 5 th -9 th of April, July, and October 2005. Reliability diagram for Probability of Rainfall vs. Nimrod radar data (Western Europe only) for the 5 th -9 th of April, July, and October 2005.

17. 17 Rainfall Potential Validation (Original algorithm) CDF’s of (absolute) errors of Probability of Rainfall vs.Nimrod radar data (Western Europe only) for the 5 th -9 th of April, July, and October 2005. CDF’s of (absolute) errors of Probability of Rainfall vs. Nimrod radar data (Western Europe only) for the 5 th -9 th of April, July, and October 2005.

18. 18 Rainfall Potential Validation vs. Spec Original algorithm validation versus Nimrod radar data (covering Western Europe only) for 15 days of data: 5 th -9 th of April, July, and October 2005: F&PSEvaluation vs. Nimrod radar %AccuracyPrecisionAccuracyPrecision Probability of Rainfall25401171 F&PSEvaluation vs. Nimrod radar %AccuracyPrecisionAccuracyPrecision Probability of Rainfall25401557 Preliminary new algorithm validation versus Nimrod radar data (covering Western Europe only) for 15 days of data: 5 th -9 th of April, July, and October 2005:

19. 19 Next Steps  The Rainfall Probability calibration cannot be finalized until the Rainfall Potential algorithm is finalized.  In the meantime, additional predictors are being explored for use in the Rainfall Probability algorithm and have brought the algorithm closer to spec.

20. 20 Summary  This Rainfall Probability Algorithm generates the Option 2 product of predicted probability rainfall accumulation exceeding 1 mm for the next 3 h on the ABI grid.  Version 3 (80%) was delivered in September 2010. Version 5 (100%) delivery has been delayed 12 months in response to GPO guidance.  An algorithm has been developed that produces conditional probabilities based on the outputs and intermediate products of the Rainfall Potential algorithm.  Preliminary validation of the version 3 algorithm has been performed against 20 days of SEVIRI data from 2005.  Validation against Nimrod radar indicates partial spec compliance, and modifications have brought it closer to spec.