Presentation is loading. Please wait.

Presentation is loading. Please wait.

Center for Hydrometeorology and Remote Sensing, University of California, Irvine Basin Scale Precipitation Data Merging Using Markov Chain Monte Carlo.

Similar presentations


Presentation on theme: "Center for Hydrometeorology and Remote Sensing, University of California, Irvine Basin Scale Precipitation Data Merging Using Markov Chain Monte Carlo."— Presentation transcript:

1 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Basin Scale Precipitation Data Merging Using Markov Chain Monte Carlo Method K. Hsu, F. Boushaki, S. Sorooshian, and X. Gao Center for Hydrometeorology and Remote Sensing University of California Irvine The 3rd Workshop of the International Precipitation Working Group, 23-27 October, 2006

2 Center for Hydrometeorology and Remote Sensing, University of California, Irvine  PERSIANN Rainfall  Precipitation Data Merging  Grid-Based Precipitation Data Merging  Basin Scale Precipitation Data Merging  Case Study  Summary Outline

3 Center for Hydrometeorology and Remote Sensing, University of California, Irvine PERSIANN System “Estimation” Global IR MW-RR (TRMM, NOAA, DMSP Satellites) Merged Products - Hourly rainfall - 6 hourly rainfall - Daily rainfall - Monthly rainfall ANN Error Detection Quality Control Merging Satellite Data Ground Observations Products High Temporal-Spatial Res. Cloud Infrared Images Feedback Hourly Rain Estimate Sampling MW-PR Hourly Rain Rates Hourly Global Precipitation Estimates Gauges Coverage GPCC & CPC Gauge Analysis Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) Center for Hydrometeorology and Remote Sensing, University of California, Irvine

4 PERSIANN-CCS (Cloud Classification System)

5 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Global PERSIANN: http://hydis8.eng.uci.edu/hydis-unesco/ US PERSIANN-CCS: http://hydis8.eng.uci/CCS 0.25 o x0.25 o Hourly 0.04 o x0.04 o Hourly PERSIANN Precipitation Products

6 Center for Hydrometeorology and Remote Sensing, University of California, Irvine A SHORT MOVIE OF PERSIANN PRODUCTS (PERSIANN: Precipitation estimation from Remote Sensing Information using Artificial Neural Network) PERSIANN (0.25°  0.25°) 07/25-27/2006 PERSIANN CCS (0.04°  0.04°) 07/24-27/2006 High resolution precipitation data are needed for hydrologic applications in SW. Severe storms propagate from mountains to low-elevated areas. Acknowledgement. This research is partially funded by NSF/SAHRA and NASA/GPM programs

7 Center for Hydrometeorology and Remote Sensing, University of California, Irvine RESEARCH TO SUPPORT MODELING EFFORTS Flash Flood Monitoring (7/27-28/2006) Poor radar coverage over mountainous southwest can result in missing flood warning for the areas radar network does not cover (Maddox et al., 2003). The demo shows our on-going study to check how the missing portions of a severe storm can be retrieved by the concurrent PERSIANN storm images and also reduce false warning. Strong convections start over mountains where radar coverage is poor. PERSIANN monitors the lifetimes of storm systems and provides information for early warning. Radar beams (3-km above ground level) are blocked by mountains in southwest United States. Differences between PERSIANN and radar images exist. Red: PERSIANN Rain vs. Radar No Rain Blue: PERSIANN No Rain vs. Radar Rain

8 Center for Hydrometeorology and Remote Sensing, University of California, Irvine 6-Hour Accumulated Rainfall: Hurricane Ivan hydis8.eng.uci.edu/CCS

9 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Precipitation Measurement is one of the KEY hydrologic Challenges

10 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Hydrologic Models QBQB QRQR t q R IAIA  i t API Model Sacramento Model Mike SHE Model, DHI VIC Model

11 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Streamflow Simulation vs. Precipitation Uncertainty:

12 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Streamflow Simulation vs. Precipitation Uncertainty:

13 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Streamflow Simulation vs. Precipitation Uncertainty:

14 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Radar Gauge Surface Temperature Soil Moisture Vegetation LABZ Multiple Sources for Rainfall Estimation Geosynchronous Satellites VIS, IR, Sounding Low Orbiting Satellites VIS, IR, MV, and Radar

15 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Bias Correction and Downscaling of Daily Rainfall to Hourly Rainfall Time Step: Day CPC Daily Analysis PERSIANN Rainfall (non-adjusted) PERSIANN Rainfall (bias adjusted) PERSIANN Rainfall Daily Rainfall: Summer 2005 Downscaled to Hourly Rainfall Grid size: 0.25 o x0.25 o Grid size: 0.04 o x0.04 o CPC Daily Gauge Analysis Grid-Based Data Merging

16 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Basin Scale Precipitation Data Merging

17 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Gages used by NWS Hydrologic Model Sacramento Soil Moisture Accounting Model (NWS) (RFC parameters) Input time step : 6 hours Output time step : 24 hours Leaf River Near Collins Mississippi USGS # 02472000 Basin Area : 753 mi 2 PERSIANN Rainfall Estimates in Hydrologic Simulation Observed Radar/Gage Merged OBSERVED vs. SIMULATED DISCHARGE (RADAR/GAGE MERGED RAINFALL ESTIMATES) Radar/Gauge 6-hour Rainfall Observed Radar/Gage Merged TRMM/Multi Satellite OBSERVED vs. SIMULATED DISCHARGE (TRMM-MULTI SATELLITE RAINFALL ESTIMATES) PERSIANN 6-hour Rainfall

18 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Basin Scale Precipitation Data Merging  i : Hydro. Model parameters Q : Output P : Input  : Errors  I : Weighting parameters  I : Bias parameters  output Hydrologic Model (  i) Optimization Q t obs t comp Q t comp  ()  (  I, Model ) (  g,  g ) (  s,  s ) PiPi Ps Pg Hydrologic Model (SAC-SMA Model)

19 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Probability distribution to be maximized = observations = simulated flows * Hours Flow Parameter Calibration

20 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Uncertainty of Parameters Hours Uncertainty associated with parameters Total Uncertainty including structural errors Probability distribution to be maximized 95%

21 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Bayesian Model Analysis Learn model parameters from data: p(ө): Priori distribution of parameters p(D|ө): Likelihood function p(ө|D): Posterior distribution of parameters

22 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Markov Chain Monte Carlo (MCMC) Sampling Probability distribution to be maximized w.r.t Current guess

23 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Always accept New guess > 1 Markov Chain Monte Carlo (MCMC) Sampling 100% acceptance of new points having higher probability than the old point

24 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Accept if  R ~ Uniform (0,1) MCMC –Acceptance of New Points Having Lower Probability than the Old Point is Probabilistic If the  ratio is small, then the probability of acceptance is small < 1 Markov Chain Monte Carlo (MCMC) Sampling α% acceptance of new points having lower probability than the old point

25 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Rainfall Runoff Time Series Gages used by NWS Leaf River Near Collins Mississippi USGS # 02472000 Basin Area : 753 mi 2 Streamflow (CMSD) Precipitation (mm/day) Gauge PERSIANN Time: Day

26 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Runoff Forecasting from Gauge, PERSIANN, and Merged Rainfall 50 100 50 100 50 100 Gauge Rainfall Satellite: PERSIANN Rainfall Merged Rainfall Rainfall (mm/day) 500 1000 0 250 750 Streamflow (m 3 /day) Gauge PERSIANN Merged RMSE 51.82 80.78 34.91 CMSD Corr. 0.876 0.706 0.901 Bias 15.34 -17.68 -3.52 CMSD 0 100 200 300 Time (Day)

27 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Parameter Distribution Distribution of Merging Parameters(5000 samples) Weighting factor (α g ) Weighting factor (α s ) Bias parameter (β g ) Bias parameter (β s )

28 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Interaction Between Parameters Parameter: α g Parameter: α s Parameter: β g Parameter: α s Parameter: β g Parameter: β s

29 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Confidence Interval of Merged Rainfall (95%) 95% confidence interval

30 Center for Hydrometeorology and Remote Sensing, University of California, Irvine Rainfall (mm/day) 40 80 120 0 0 200 400 600 800 Streamflow (m 3 /day) 0 100 200 300 Precipitation 95% Uncertainty Bound 99% Uncertainty Bound 95% Uncertainty Bound Observed Streamflow


Download ppt "Center for Hydrometeorology and Remote Sensing, University of California, Irvine Basin Scale Precipitation Data Merging Using Markov Chain Monte Carlo."

Similar presentations


Ads by Google