Presentation is loading. Please wait.

Presentation is loading. Please wait.

MULTI-SOURCES PRECIPITATION ESTIMATION K. Tesfagiorgis, S. E. Mahani, R. Khanbilvardi (NOAA-CREST, CCNY, CUNY, NY-10031) David Kitzmiller (NOAA-NWS Collaborator)

Similar presentations


Presentation on theme: "MULTI-SOURCES PRECIPITATION ESTIMATION K. Tesfagiorgis, S. E. Mahani, R. Khanbilvardi (NOAA-CREST, CCNY, CUNY, NY-10031) David Kitzmiller (NOAA-NWS Collaborator)"— Presentation transcript:

1 MULTI-SOURCES PRECIPITATION ESTIMATION K. Tesfagiorgis, S. E. Mahani, R. Khanbilvardi (NOAA-CREST, CCNY, CUNY, NY-10031) David Kitzmiller (NOAA-NWS Collaborator) (NOAA NWS/HL, Silver Spring, MD-20910) 1 NOAA-NESDIS CoRP 7 th Annual Symposium Fort Collins, CO., August, 2010

2 Outline 2  Objectives  Data sets  Comparison of the different precipitation estimation algorithms  The different bias correction techniques  Spatial corrections  Results for study cases  Conclusion  Future work

3 Objectives 3  To improve Satellite Precipitation Estimation (SPE) by selecting appropriate bias correction technique.  To develop a Multi-Sources Rainfall Estimation algorithm to help optimal rainfall estimations.  Be capable of extending radar like outputs inside radar gap regions using satellite and the surrounding radar rainfall estimations.

4 Data Sets 4  Hourly 4kmx4km resolution for the Oklahoma region bounded by 94.50-100 o W Longitude 34.50-37.0 o N Latitude  Satellite Rainfall Estimations selected from the following NESDIS models  AE (Auto-Estimator)  GMSRA (GOES Multispectral Rainfall Algorithm)  HE (Hydro-Estimator)  SCaMPR-(Self Calibrating Multivariate Precipitation Retrieval)  Blend-(IR/Microwave Blended Algorithm)  Radar Rainfall Estimation  Radar Stage IV (ST-IV)  Rain-gauge Measurements

5 8677 cases, 8677x62x137=73,702,438 pixels considered Comparison of the Different Rainfall Estimations 5 Satellite Rainfall Estimations Radar Rainfall Estimation YesNo YesHitsFalse Alarms NoMissesCorrect negatives Observed yesObserved no Satellite Rainfall Estimation Bias ScoreFalse Alarm Ratio GMSRA2.710.63 HE1.730.46 SCaMPR2.410.68 Auto-Estimator2.080.53

6 Bias Corrections 6 Field Bias Correction Generally it helps for: – Intensity correction – Frequency correction Methods of bias corrections: – Ratios of Mean, Median, Maximum – Mean of Ratio of the corresponding rainy pixels in both Satellite and Radar Rainfall Estimations – Bias ratio field using Inverse Distance method

7 Bias Corrections: continued… 7 We need to calculate the Multiplicative factors (F) for Bias corrections RR-Radar Rainfalls and SPE-Satellite Rainfall Estimates Method 1 of Bias correction The ratio of Max and Mean gave a better output. However, ratio of max is not stable and reliable.

8 Bias Corrections: continued… 8 How about Spatial Errors that might have already existed? Before working on the Bias Corrections, it is important to make spatial corrections between the satellite and the radar rainfall estimations. Spatial Correction using the Method of Least Squares (Brogan 1985): – Apply the method of Hill Climbing to cluster rainy pixels; because the clustered corresponding rainy pixels are easier to pick up – Pick corresponding points (Rainy Pixels) – Write Least Square equations and apply the method of least squares on these points as shown in equations shown below.

9 Spatial Correction CoefficientsInterpretations Shift in longitude Scale in longitude Shear in the longitude CoefficientsInterpretations Shift in latitude Scale in latitude Shear in the latitude 9 Linear form of the equations with N=3 R-Radar S-Satellite

10 Spatial Correction 10 Corresponding Pixel Locations

11 Method 2 of Bias Correction Bias Corrections: continued… 11 - Calculate the bias ratio between ST IV and HE - Calulate the bias field using Inverse distance weight technique - Multiply HE by the mean field bias Method 2 provides a more radar like output both spatially and intesity wise.

12 Bias Correction: continued… 12 The performance of bias field method for a winter case

13 Bias Correction: continued… 13 Ensembles of rainfall for a pixel around the center of the study area Ensemble generation of bias fields Instead of 1, 100 realizations

14 Bias Correction continued…. 14 UncorrectedCorrected Pixels used in the development of the algorithm are not part of the CORR analysis Case-2006071022 (YYYYMMDDHH) Case-2006122917 (YYYYMMDDHH)

15 Conclusion 15  Hydro-Estimator has a better detection capability than the others, so that it is chosen for further studies that will include radar estimations and rain-gauge measurements.  There are cases where the alignment algorithm faces difficulties. When rainfall is very cluttered in radar and continuous in satellite estimations.  In these cases it is difficult to pick up corresponding rainy pixels.  However we can still apply the Bias field generation Algorithm without doing the alignment.  Ensemble generation helps to account other errors (Eg. physical, paralax).  Generation of bias fields can potentially be used to correct satellite estimations in radar gap regions.

16 Ongoing Works  Check the performance of the model in other geographical locations.  Implement a technique that will give a multi-sources rainfall algorithm by merging the radar and the satellite estimations.  Produce gridded rain-gauge measuremts using Bayesian Kriging and/or inverse distance method.  Merge the gridded rain-gauge with the combined radar- satellite rainfall estimation. 16

17 Acknowledgements  This study was partially supported and monitored by the National Oceanic and Atmospheric Administration (NOAA) under grant number NA06OAR4810162. The statements contained within this presentation are not the opinions of the funding agency or the U.S. government, but reflect the author’s opinions.  I would like to thank Robert Kuligowski (Ph.D.) for providing all the necessary data. 17


Download ppt "MULTI-SOURCES PRECIPITATION ESTIMATION K. Tesfagiorgis, S. E. Mahani, R. Khanbilvardi (NOAA-CREST, CCNY, CUNY, NY-10031) David Kitzmiller (NOAA-NWS Collaborator)"

Similar presentations


Ads by Google