Comparing NEXRAD and Gauge Rainfall Data

1. Comparing NEXRAD and Gauge Rainfall Data Nate Johnson CE 394K.2 Final Project April 26, 2005

2. Motivation • USGS performs hydrologic and water quality modeling using rainfall data from gauges. • Sometimes there is greater than 20 km between gauges • NEXRAD provides rainfall estimates on 4 km2 grid • Could more acccurate hydrologic simulations could be performed with “accurate” distributed rainfall information? 10 km 6 km 9 km >20 km How does NEXRAD precipitation data compare to gauges used for hydrologic modeling?

3. Outline • Project involved 2 major tasks: • Transferring data from NEXRAD files to something more accessible • Comparing NEXRAD rainfall estimates to rain gauge measurements • Cumulative Difference • Storm Bias

4. Background – File Structure Space – Time – Variable dimensions Gridded File – One Time, all space TimeSeries File – One Location, all time • To create a TimeSeries file (for a single location) from a collection of Gridded files (for a single time) must access thousands of gridded files! 24 hours * 365 days = 8,760 files/year

5. Background – Rainfall Estimates • Compare NEXRAD to gauge based on… • Point Comparisons (Grid Cell immediately above) • Within cell spatial variability? • Spatial Averages • Interpolate between gauges? This project only considers point comparisons… Stay tuned for Jessica Watt’s presentation containing spatial comparisons

6. Methods – Extract NEXRAD Data • Extracted 2 months of hourly NEXRAD data for about 1,800 cells near San Antonio. • August and September of 2001 were very rainy • ~100,000 non-zero timeseries records

7. Methods – Compare NEXRAD/Gauge First method – Cumulative difference over two months • For each hour, calculate difference between Radar and Gauge and add to pervious total • Removes biases from timing discrepancies

8. Results – Cumulative Differences August – September 2001 > 11 inches of rain!

9. Results – Cumulative Differences S.A. Airport Gauge – 11.9 inches total NEXRAD – Rain Gauge

10. Results – Cumulative Differences Government Canyon Gauge – 10.3 inches total NEXRAD – Rain Gauge

11. Results – Cumulative Differences Bulverde Rain Gauge – 10.8 inches Total NEXRAD – Rain Gauge

12. Methods – Compare NEXRAD/Gauge Second method – Storm-to-Storm Comparison • Based on an analysis by NCAR and NOAA in 1979 • Considers each event independently and looks for systematic biases • Uses Ratio ‘G/R’ of Gauge to Radar for total storm-event precipitation

13. G/R < 1 G/R > 1 G/R < 1 G/R > 1 G/R > 1 Methods – Compare NEXRAD/Gauge Average G/R Coefficient of Variation Average Difference Average Difference (Storm Bias Removed)

14. Results – Storm Bias Comparison Original Difference Improvement? Not enough gauges? Too much scatter?

15. Conclusions / Future Work • Some significant differences between NEXRAD and Gauge measurements • May be too much scatter in G/R to remove “Storm Bias” from NEXRAD • Still need to consider effects of spatial averaging • Investigate the effects on hydrologic modeling results

18. Background – Rainfall Estimates • Hydrologic modeling traditionally relies upon estimates for “Average Drainage – Area precipitation • Some question yet as to whether or not it is appropriate to use NEXRAD for hydrologic modeling • Studies have shown some disturbing differences between NEXRAD and gauge measurements

19. Arc Hydro GDB Methods – Extract NEXRAD Data • First – Get NEXRAD data into Arc Hydro Geodatabase • Second – Get Arc Hydro into Timeseries format

20. Background – File Structure • NEXRAD data are stored in a gridded file structure • Hydrologic models require data in a TimeSeries Structure Gridded File – One Time TimeSeries File – One Location