Quantitative Precipitation Estimation using Satellite and Ground Measurements

Transcription

1 Quantitative Precipitation Estimation using Satellite and Ground Measurements Kuolin Hsu Center for Hydrometeorology and Remote Sensing University of California Irvine XVII Congress of the Spanish Association of Remote Sensing Murcia, 3-7 October, 2017

2 CHRS Remote Sensing Precipitation CHRS Multi-Satellite Precipitation Analysis Reconstruction of Historical Precipitation Estimation Integration Satellite & Gauge Measurements Adjust Bias of Satellite Estimation using Gauge Observations Merge Satellite and Gauge Observations CHRS PERSIANN Websites and Apps. Conclusions

3 X. Gao S. Sorooshian K. Hsu A. AghaKouchak D. Hohnbaum Y. Tao T. Yang D. Braithwaite N. Phu N. Karbalaee A. Akbari M. Rahnamay M. Faridzad N. Hayatbini B. Pan H. Tran R. Alharbi M. Ombadi J. Lee.

4 Center for Hydrometeorology and Remote Sensing Develop state-of-the-art systems to estimate rainfall from satellite observations at global scale and high spatial and temporal resolutions Improve the performance and reliability of hydrologic, flood, and water supply forecasting models, particularly those used by the National Weather Service and other operational agencies. Improve California s water supply management through: Decision Making Support Reservoir System Operation Improved decision optimization Utilizing Information Technology to provide world-wide access to real-time global precipitation products: U.S.-China Water-Energy Project: Sustainable Hydropower Climate Impacts on W-E system Reservoir Facility Design and Operation China USA

5 Climate Variability and Change & Extreme Hydrologic Events Floods & droughts are more frequent due to climate variability and change Long term high resolution precipitation data are needed for hydroclimate studies

6 Remote Sensing Precipitation Develop state-of-the-art systems to estimate rainfall from satellite observations at global scale and high spatial and temporal resolutions Information Technology to provide world-wide access to real-time global precipitation products: Goal: High spatial and temporal resolution of precipitation measurements at global scale for hydrological applications: Short-term operational applications Flood forecasting Data assimilation in numerical weather models Long-term climate extreme event analysis Hydro-climate studies Validation GCM models

7 Precipitation Observation

8 Coverage of the WSR-88D and gauge networks 12 3 km AGL

9 Satellite Data for Precipitation estimation Geostationary IR Cloud top data minute temporal resolution Passive Microwave (SSM/I) Some characterisation of rainfall ~2 overpasses per day per spacecraft, moving to 3-hour return time (GPM) TRMM precipitation RADAR 3D imaging of rainfall 1-2 days between overpasses ( S-35 N-35 )

10 Typical Microwave Coverage in 3 Hr TMI white SSM/I light grey AMSR-E medium grey AMSU-B dark grey

11 Global Precipitation Measurement (GPM) The GPM spacecraft will collect information that unifies data from an international network of existing and future satellites to map global rainfall and snowfall every three hours. Tanegashima Space Center, Japan Friday, Feb. 28, 2014

12 GPM Animation Courtesy: NASA s ESE

13 Integrated Multi-satellitE Retrievals for GPM (IMERG)

14 Integrated Multi-satellitE Retrievals for GPM (IMERG) Integrated Multi-satellitE Retrievals for GPM IMERG NASA TMPA: intersatellite calibration, gauge adjustment NOAA CMORPH: Lagrangian time interpolation U.C. Irvine: PERSIANN-CCS: neural-net microwave calibrated GEO-IR Address different user needs in 3 runs early (~4 hr after observation) late (~12 hr after observation) final (with gauge, ~2 months after observation)

15 Interpolation of 3-hour Precipitation Rain started between 3-hr period Missed the peak Rain ended between 3-hr period - Short-life event T T+3hr T+6hr t-hr t+3hr

16 Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Cloud Classification System (PERSIANN-CCS) T b =220K T b =235K T b =253K t=t 0 t=t 1 t=t 2 t=t k T 220K Patch Feature Extraction V 220K Patch Classification c 1 Rainfall Estimation T 235K c 2 T 253K, t=t k V 235K Image Segmentation T 253K Feaature vector V 253K c k ( V ) [ patch coldness, patch geometry, patch texture ] 80 R (mm/h) T b (K) 300 Hong et al., 2004

17 irain:

18 Continue Development Improving Precipitation Estimation over Mid-High Latitudes Improving Precipitation Estimation from Warm Cloud Adding Multi-Spectral Information Adding Lightning Detection

19 CDF: CCS Est. Before/After PMW Adjustment (January ) 5 x 5 degree CDF Curve CCS Est. CCS Est. (PMW Adj.) PMW and PERSIANN-CCS CDF for each 5X5 degree box January-winter time Karbalaee et al., 2017

20 PERSIANN-CCS Hourly Estimates: Jan 4, 2012 CCS Estimation CCS (PMW Adjusted) Estimation

21 PERSIANN-CCS Hourly Estimates: July 5, 2012 CCS Estimation CCS (PMW Adjusted) Estimation

22 Improve Warm Rain Estimation (PERSIANN-CCS) PERSIANN-CCS, :00 UTC, Tb = 253 k mm/hr Radar(Q2), :00 UTC mm/hr PERSIANN-CCS, :00 UTC, Tb = 260 k mm/hr Infrared from GEOs, :00 UTC K PERSIANN-CCS, :00 UTC, Tb = 280 k mm/hr # of rainfall captured : TP(t = 253 K) = 7143 pixels # of rainfall captured : TP(t = 280 K) = pixels # of rainfall captured specifically from warm clouds: TP( t > 253 K ) = 1151 pixels

23 Include Other Channels d) Ch5 ETS=25 POD=74 FAR=45 ETS=29 POD=77 FAR=42 ETS=27 POD=78 FAR=44 f) f) Ch3+Ch5 g) Ch4+Ch5 ETS=36 POD=76 FAR=35 ETS=30 POD=80 FAR=42 ETS=30 POD=72 FAR=39 ETS=35 POD=79 FAR=37 i) Ch3+Ch4+Ch5 ETS=37 POD=78 FAR=35 ETS=37 POD=80 FAR=36 ETS=48 POD=75 FAR=22 ETS=49 POD=79 FAR=24 Hit Under Estimation Over Estimation Ch 1 : 0.6 µm Ch2 : 3.9 µm Ch3 : 6.5 µm Ch4:10.7µm Ch5 : 13.3µm

24 Multi-Satellite & Conceptual Cloud Model for Rainfall Estimation GOES GMS METEOSAT (30-minute Coverage) Passive Microwave Samples GPM, NOAA, DMSP Satellites Dynamic Cloud Tracking Conceptual Cloud Model Q f g e a C R c R s PMW Rainfall R( t) C( t) f ( t) aq( t) dq( t) C( t) g( t) (1 a e) Q( t) dt LMODEL+KF Rainfall Lagrangian Model + Kalman Filter (LMODEL+KF) A Multi-satellite Cloud Motion Tracking and Rainfall Estimation system High resolution cloud advection tracking Synchronisation of PMW precipitation to IR clouds Conceptual model of precipitation development Kalman filter for state adjustment based on PMW rainfall observations mm/hr Bellerby et al., J. Hydrometeorology, 2009

25 Tropical Storm Harvey - August 26, 2017 K K mm/hr

26 PERSIANN-CDR Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks -Climate Data Record Reconstruction of more than 30 years of daily precipitation data (1983 ~2016)

27 PERSIANN-CDR PERSIANN estimation at 0.25 o every 3-hr from GridSat B1 IRWIN Monthly accumulation and bias adjusted using GPCP monthly estimation at 2.5 o Bias adjustment of short-term 3-hr estimation GridSat-B1 IRWIN High Temporal-Spatial Res. Cloud Infrared Images Artificial Neural Network PERSIANN 3-Hourly Rainfall (0.25 o x0.25 o ) PERSIANN Monthly Rainfall (2.5 o x2.5 o ) Adjusted PERSIANN 3-Hourly Rainfall (0.25 o x0.25 o ) GPCP Bias Adjustment GPCP Monthly Precipitation (2.5 o x2.5 o ) Spatiotemporal Accumulation Ashouri et al, BAMS, 2015

28 PERSIANN -CDR Daily Precipitation Data Data Period: 1983~2016 Coverage: 60 o S ~ 60 o N Spatial Resolution: 0.25 o x0.25 o Ashouri, Hsu et al., BAMS, 2015.

29 PERSIANN-CDR daily rainfall During Katrina 2005 Center for Hydrometeorology and Remote Sensing (CHRS)

30 Sierra-Nevada Mountain Region Area: 63,100 square kilometers (24,370 sq mi) Length: 400 mile, Width: 64 mile. Source: Google Earth Center for Hydrometeorology and Remote Sensing (CHRS)

31 Sierra-Nevada Mountain (California and Nevada) Center for Hydrometeorology and Remote Sensing (CHRS)

32 Testing of PERSIANN-CDR: Number of Rainy days >= 10 mm/day Center for Hydrometeorology and Remote Sensing (CHRS)

33 Evaluation Indices ID Definition Unit RR95p The 95th percentile of annual precipitation on wet days (precipitation 1mm) mm/day R10mmTOT Annual total precipitation when daily precipitation 10mm mm R10mm Annual count of days when precipitation 10mm Days Extreme precipitation indices used in the analysis

34 Results: Entire China EA PERSIANN-CDR Pixel correlation Scatterplot of mean RR95p R10mmTOT R10mm Miao et al., JHM 2015

35 PERSIANN-CDR: Potential Usefulness for Model testing Center for Hydrometeorology Center for Hydrometeorology and Remote & Sensing Remote Sensing, (CHRS) University of California, Irvine Ashouri et al., BAMS 2015

36 Future Modeling Scenarios IPCC AR5 RCP2.6 Time period: CNRM-CM5 (France GCM) CSIRO-MK (Australian GCM) GISS-E2-R (U.S. GCM) HadGEM2-ES (U.K. GCM) Dr. Chiyuan Miao - BNU

37 Future Modeling Scenarios IPCC AR5 CNRM-CM5 (France GCM) CSIRO-MK (Australian GCM) GISS-E2-R (U.S. GCM) RCP8.5 Time period: HadGEM2-ES (U.K. GCM) Dr. Chiyuan Miao - BNU

38 CMIP5 Models historical simulation ( ): CHINA bcc_csm1_1_m (Chinese GCM) CCSM4 (NCAR, USA GCM) HadGEM2-ES (U.K GCM) MIROC5 (Japan GCM) MPI-ESM-MR (Germany GCM) Observation (CRU Dataset) Observation (PERSIANN-CDR)

39 Integration of Satellite & Ground Measurements Bias Adjustment of Satellite Estimation using Gauge Measurements

40 Bias Adjustment of Satellite Estimation Using Gauge Observation Bias adjustment of PERSIANN-CCS Estimation Using Local Gauge Measurements: A Case Study over Chile Gauge Observation 456 gauges are selected Data period: (5 years) Observations (see figures on the right) PERSIANN-CCS Match to the same time period for evaluation Evaluation zones 14 zones are divided by every 2 latitudes over the adjusted region (shown as the right figure) Yang et al., JGR Atmosphere 2016 & 2017

41 F(r): CDF Quantile Mapping of CCS to Gauge Estimation Estimate cumulative distribution of satellite and gauge rainfall estimation at 0.04 o x0.04 o under 1 o x1 o coverage CCS gird 1 o x1 o 0.04 o x0.04 o Gauge Observation F sat (r): CDF of Satellite Estimation F G (r): CDF of Gauge Measurement rainfall (mm/day) R sat : Satellite Estimates R sat-adj : Adjusted Satellite Estimations

42 Step III: Calculate CDF curves of CCS and gauge Estimates 5-year of daily gauge and CCS estimations are collected Estimate the corresponding CDF curve for each 1 o x1 o box for each season (or month) If gauge is not available for a 1 o x1 o box, extend the box size to 3 o x3 o or larger to collect gauge & satellite samples to estimate CDF for that 1 o x1 o box Gauge 456 gauges 98(1 x 1 ) boxes

43 Bias Correction of Monthly Rainfall (5-year average) July October

44 Results: Monthly Time Series (Zonal Average: #1, #7, #13)

45 Results: Cumulative Time Series (Zonal Average: #1, #7, #13)

46 Merge Real-time Satellite and Gauge Daily Estimates Merging daily gauge and satellite estimation Merging gauge and bias adjusted CCS estimation based on the uncertainty (RMSE) of estimation CCS gird 1 o x1 o 0.04 o x0.04 o Gauge Observation

47 Monthly Average Rainfall ( ) October

48 Results: Monthly Time Series (Zonal Average: #1, #7, #12)

49 Daily Time Series (Zonal Average: #1, #7, #12)

50 PERSIANN Websites and Apps - CHRS irain - CHRS RainSphere - CHRS Data Portal

51 PERSIANN Extensions: Weather-Related CHRS irain An Integrated System for Global Real-time Precipitation Observation

52 CHRS s Real-time Global Extreme Precipitation Events Observation System irain.eng.uci.edu

53 CHRS irain capturing Tropical Storm Harvey - August 2017 irain.eng.uci.edu

54 CHRS irain App

55 PERSIANN Extensions: Climate-Related CHRS RainSphere An Integrated System for Global Satellite Precipitation Data and Information

56 CHRS RainSphere Rainsphere.eng.uci.edu Search Location Map Layers Rain Data Type Rain Layers Rain Comparison Rain Statistics and Trends Legend

57 CHRS RainSphere

58 CHRS RainSphere

59 CHRS RainSphere

60 CHRS RainSphere Rainsphere.eng.uci.edu Countries Watersheds Political Divisions Major Basins Rainfall Trends Across Spatial Scales

61 CHRS Data Portal

62 Who Uses CHRS Products irain.eng.uci.edu

63 Conclusions Remote sensing precipitation is a way to extend rainfall estimation to regions where ground observations are limited. Operation requirement of precipitation data: - High quality near real-time estimation is needed for hydrologic applications Precipitation Climate Data: PERSIANN-CDR: A more than 30 year daily precipitation data is introduced. PERSIANN-CDR is a very useful source for hydro-climate studies. Integration of Satellite & Ground Measurements Effective use of ground observations to reduce bias of satellite-based estimation CHRS PERSIANN Websites and Data Service PERSIANN related Apps and Data Service are available through CHRS webpage.

64 Gracias por su invitación