1. THE USE OF DUAL-POLARIMETRIC RADAR DATA TO IMPROVE RAINFALL ESTIMATION ACROSS THE TENNESSEE RIVER VALLEY W.A. Petersen NASA – Marshall Space Flight Center, Huntsville, AL P. N. Gatlin, L. D. Carey University of Alabama in Huntsville – Earth Systems Science Center, Huntsville, AL S. R. Jacks Tennessee Valley Authority, Knoxville, TN

2. Motivation  Reduction of costs associated with maintenance of large rain gauge network  Provide a custom-tailored rainfall product specific to the end-user’s needs  Independent validation of ARMOR rain rate algorithms  Ground-validation for TRMM satellite measurements

3. Tennessee River Watershed AL MS TN GA KY SC NC  112 sub-basins  1840 km 2  189 rain gauges maintained by TVA   11 sub-basins within 100 km of the ARMOR dual-pol. radar

4. Advanced Radar for Meteorological & Operational Research  Location: Huntsville International Airport, Huntsville, AL (Altitude 206m)  C-band dual-polarimetric Doppler radar  Simultaneous transmit and receive of H, V  Variables: Z, V, W, ZDR, Φ DP, ρ hv  Operations: 24-hrs a day / 7 days Rain volumetric scans at least every 5-min (tilts: 0.7°,1.5 °,2.0 °) Also used in research mode (e.g., RHIs, full volumes, vertically pointing scans)  Routine calibration: Receiver calibrations Solar scans Self-consistency amongst variables Comparisons with TRMM and rain gauges

5. ARMOR Rainfall Estimation Processing System (AREPS) Grid rain rates (1 km 2 spacing) T1-line ARMOR NSSTC End-user Summation of rain rates Compute point and areal N-hr rainfall estimates Raw Iris Files

6. ICE PRESENT? NO YES K DP  0.3 and Z H  35? R = R(K DP ) YE S NONO Z H  BAD? YE S R = R(Z H RAIN ) R=BAD NONO K DP  0.3 , Z H  35.0 dBZ Z DR  0.5 dB? YE S R > 50 mm/hr, dBZ > 50,or Z, ZDR corr. too large ? ZH > 30 dBZ, Z DR  0.5 dB? R = R(Z H,Z DR ) R = R(Z H ) ARMOR RAIN RATE ALGORITHM (1) R(K DP,Z DR ) (2) R(K DP ) (3) R(Z H,Z DR ) R = R(Z H ) GOOD DATA? YES NO R=BAD KDP ≥ 0.5? KDP< 0.5? YE S R = R(K DP ) YE S R =R(K DP,Z DR ) YE S R =R(Z H,Z DR ) no NONO YE S NONO 1-hr Accumulation 6-hr (N-hr) Accumulation

7. AREPS Coverage  100 km from ARMOR  11 sub-basins  42 rain gauges

8. AREPS Distributed Rainfall Products  Rainfall products created every 5-min: 1-hr and 6-hr basin/sub-basin rainfall statistics (mean, max, min, etc) 1-hr and 6-hr basin/sub-basin rainfall statistics (mean, max, min, etc) Rainfall at critical locations (e.g., dams) Rainfall at critical locations (e.g., dams) rainfall accumulation images (1-hr, 6-hr) rainfall accumulation images (1-hr, 6-hr)  Text files transmitted every hour to TVA Contain previous hour’s rainfall Contain previous hour’s rainfall used as input by inflow model input used as input by inflow model input 6-hour accumulation statistics 6-hr Basin Mosaic 1-hr rainfall (also create 6-hr rainfall)

9. Verification: Point Comparisons ARMOR vs. TVA rain gauges (October 2007 – June 2008)   Original bias and error targets achieved (+/-20%, +/-25% respectively)   Constant monitoring of calibration maintains precision and accuracy of product Before Calibration Correction Bias = -10% (-0.99 mm) Error = 12% Bias = -17% (-1.80 mm) Error = 18% After Correction Radar Rainfall Estimate Improved

10. Verification: Sub-basins ARMOR vs. rain gauge-derived areal mean (January 2008 – July 2008)  Radar rainfall estimates averaged over each sub-basin  rain-gauge network used by TVA to compute Theissen polygon values to represent each sub-basin  Radar underestimates sub- basin rainfall by only 8%  Random error = 20% Largely attributed to Theissen polygons (i.e, density of rain gauge network with respect to sub-basin boundaries) Gauge derived accum. (mm) Radar derived accum. (mm)

11. Gauge-Estimated Basin Means vs. Radar BASIN GAUGE (in) ARMOR (in) Decatur-Wheeler0.790.25 Guntersville-Decatur0.430.46 Upper Bear Creek0.000.06 Town Creek 0.000.12 Why are their gauge-radar differences? Case 1 (no gauge rain when there is rain) Rain narrowly missed gauge, but radar captured Case 2 (isolated gauge “deluge”) Single gauge located in heavy rain maximum- single point translated to entire basin- results in overestimate of basin mean Case 3 (Gauge and radar match) More gauges, broader rain distribution Result: Distributed Radar Rainfall Measurement Benefits TVA Water management impacts? How might the application of distributed rainfall measurements be extended? 6-Hour Rain Accumulation (in): 12 – 6 PM, 7/9/2008 1 1 2 3

12. What’s next?  Employ NCAR hydrometeor identification algorithm to remove clutter and improve precipitation calculations  Correct for partial beam blockage  Use ARMOR to polarimetrically “tune” nearby NEXRAD until upgraded  Examine radar dominated rainfall estimates in a distributed model vs gauge only estimates