Using Independent NCDC Gauges to Analyze Precipitation Values from the OneRain Corporation Algorithm and the National Weather Service Procedure Steven M. Martinaitis iti Henry E. Fuelberg John L. Sullivan Chandra S. Pathak American Meteorological Society New Orleans, LA
Objective Water management and regulatory decisions in Florida are made by the Department of Environmental Protection (FDEP) and five Water Management Districts (WMDs) However, FDEP and the WMDs use different multi-sensor datasets for hydrologic modeling Can the two datasets be used interchangeably? We compare and contrast the two datasets against an independent set of gauges
Datasets Florida State University Historical Dataset Employs the National Weather Service (NWS) Multi- sensor Precipitation Estimator (MPE) algorithm Uses quality controlled (QC) hourly gauge data provided by the National Climatic Data Center (NCDC) and the WMDs FSU conducts a second QC of hourly gauges Uses hourly digital it precipitation it ti arrays (HDPAs) Final product is hourly on the ~ 4 4 km Hydrologic Rainfall Analysis Project (HRAP) grid Hereafter denoted FSU/NWS MPE
Datasets OneRain Corporation Algorithm Algorithm is proprietary--little is known about its characteristics Provides a near-real time and an end-of-month product (only evaluate the end-of-month product) Final end-of-month product is at 15 min intervals on a Cartesian 2 2 km grid Hereafter denoted d OneRain
Example Monthly Sums FSU/NWS MPE OneRain June 2005 June 2005
Independent NCDC Gauges Evaluation over SFWMD Thirteen NCDC daily co-op gauges from 2004-2005 Each gauge records over different 24 h intervals Not used in either multisensor algorithms
Methodology Place FSU/NWS MPE and OneRain data on same spatial and temporal resolution Sum OneRain 15 min data into hourly intervals Transfer the OneRain data from its 2 2 km grid to the HRAP ~ 4 4 km grid using area weighted averaging in ArcGIS Sum each multi-sensor product over the 24 h period of each NCDC gauge
Results Statistics for Combined Two Year Period of Gauges vs. Algorithms Time Mean Difference (mm) Standard Deviation of R 2 Period Difference (mm) FSU/NWS MPE OneRain FSU/NWS MPE OneRain FSU/NWS MPE OneRain ALL 0.80 0.26 10.38 11.12 0.568 0.532 WARM 0.98 0.12 11.87 12.51 0.500 0.472 COOL 0.42 0.57 6.06 7.03 0.791 0.735 FSU/NWS MPE has larger mean differences, or bias, (positive values means multi-sensor is underestimating) but smaller standard deviations of those differences FSU/NWS MPE has larger R 2 values
Results: Monthly Box-Whisker Plots Multi-sensor vs. NCDC FSU/NWS MPE OneRain Top (bottom) of boxes represent 75 th (25 th ) percentile Top (bottom) whiskers represent 90 th (10 th ) percentile
Results: Monthly R 2 Values FSU/NWS MPE has greater R 2 for 10 of 12 months (exceptions are August and December) Overall R 2 characterize seasonal rainfall dynamics
Results: Time Series es Accumulation for 2005 (2 of 13 Gauges) Fort Lauderdale d Ot Ortana Lock k2 There are no coherent spatial patterns in accumulation differences
Results: Winter Case Study 31 January to 1 February 2004 FSU/NWS MPE OneRain Both multi-sensor schemes handle event very well compared to gauges (slope near 1.00 and R 2 0.90) Little difference between FSU/NWS MPE and OneRain
Hurricane Wilma Case Study Made landfall as a category 3 hurricane near Cape Romano, FL at 1030 UTC on 24 October 2005 Over a three day period, 15 of 39 daily gauge observations either were missing or reported zero precipitation during the heavy rain event Remains 24 daily ygauge g observations over a three day period; only 10 of 24 gauges reported daily precipitation 10 mm Consider under-estimation from wind-driven di rain Compare hourly multi-sensor data only
Results: Hurricane Wilma Case Study Pi Prior to Landfall After Landfall 0400 UTC 24 October 1200 UTC 24 October Pixel-to-pixel comparison demonstrate that OneRain had greater values prior to landfall while FSU/NWS MPE had greater values post-landfall, especially with the larger precipitation amounts
Summary Both multi-sensor products performed better during the cool season than the warm season OneRain produced smaller mean differences while FSU/NWS MPE had smaller standard deviations of those differences FSU/NWS MPE had overall slightly greater R 2 values Both products under-estimated large precipitation amounts but FSU/NWS MPE appeared to handle the amounts better
Summary Since the majority of differences are usually not major, the degree to which data from the two procedures can be used interchangeably depends on criteria specified by the user However, this is only a comparison of the precipitation input for hydrological models Need to determine which h algorithm provides most accurate streamflow data for water management decisions
Ongoing Research Currently inserting FSU/NWS MPE, OneRain, and rain gauge data at their native resolutions into the MIKE SHE hydrologic model Results compared against observed streamflow data for the Big Cypress Basin, which includes the Florida Everglades Determine the interchangeability of the two algorithms from a hydrological standpoint
We thank the following organizations for their support of this project South Florida Water Management District Florida Department of Environmental Protection NOAA / National Weather Service D.J. Seo (Office of Hydrologic Development) Jay Briedenbach (WFO Boise) Judy Bradberry (Southeast t River Forecast Center) Joel Lanier (WFO Tallahassee) United States Geological Survey University of Florida