1. TEMPLATE DESIGN © 2008 www.PosterPresentations.com Assessing the Potential of the AIRS Retrieved Surface Temperature for 6-Hour Average Temperature Forecast in River Forecast Centers Feng Ding 1,2, Michael Theobald 1,2, Bruce Vollmer 1, Andrey K. Savtchenko 1,2, Thomas J. Hearty 1,3, Asghar E. Esfandiari 1,2 1 NASA, Goddard Space Flight Center, Code 610.2, Greenbelt, MD 20771, 2 ADNET Corp., 3 Wyle IS LLC. INTRODUCTION SUMMARY CONTACT INFORMATION. APPROACH Producing timely and accurate water forecast and information is the mission of National Weather Service River Forecast Centers (NWS RFCs) of National Oceanic and Atmospheric Administration (NOAA). The river forecast system in RFCs requires basin average surface temperature in the fixed 6-hour period 000- 0600, 0600-1200, 1200-1800, and 1200-0000 UTC. The current logic of RFC temperature forecast relies on ingest of point values of daytime maximum and nighttime minimum temperature. Meanwhile, the basin mean temperature for the 6-hour period is estimated from a weighted average of daytime maximum and nighttime minimum temperature. The Atmospheric Infrared Sounder (AIRS) is the first high spectral resolution infrared sounder on board the Aqua satellite which was launched in May 2002 and follows a Sun-synchronous polar orbit. It is aimed to produce high resolution atmospheric profile and surface atmospheric parameters. As Aqua crosses the equator at about 1330 and 0130 local time, the AIRS retrieved surface air temperature may represent daytime maximum and nighttime minimum value. Comparing to point observation from surface weather stations which are often sparse over the less-populated area and are unevenly distributed, satellite may obtain better area averaged observation. This study assesses the potential of using AIRS retrieved surface air temperature to forecast 6-hour average temperature for NWS RFCs. Current RFC weighted regression method Manually select 32 representative ground stations T ave : 6-h basin average surface temperature T max and T min : basin average daily maximum and minimum temperature 00-06, 06-12, and 12-18Z: T ave = W min *T min + W max *T max W min + W max = 1 18-00Z: T ave = W min *T min + W max *T max + W min2 *T min2 ; W min + W max + W min2 = 1 T min2 : minimum temperature the following day Weights are period, season, and RFC dependent Replace T max and T min with AIRS retrieved surface air temperature at ascending and descending mode T asc and T des. REGION and SEASON California Nevada River Forecast Center ― 32.5 0 N – 42 0 N, 124.5 0 W – 114 0 W ― Coarse distribution and poor coverage of ground stations in mountainous area Spring Season ― March, April, and May ― Flood caused by snow melting DATASET AIRS retrieved surface air temperature version 5 level 3 standard product archived in Goddard Earth Sciences and Data Information Services Center (GES-DISC) North American Land Data Assimilation System (NLDAS) data 2m above ground air temperature (hourly, 1/8 degree, phase 2) generate 6-h basin average as ground truth (T ave ). POLYNOMIAL FIT RESULT For 00-06Z and 12-18Z, AIRS retrieved surface air temperature can be used for 6-h temperature forecast in CNRFC, and has less error than current regression method. For 06-12Z and 18-00Z, the regression results are not as good as CNRFC because AIRS nighttime/daytime temperature is warmer/colder than daily minimum/maximum temperature. Polynomial fit can get better results, but needs more investigation to find relationship between AIRS nighttime/daytime temperature and daily minimum/maximum temperature. Using AIRS and NLDAS data for regression is more objective than using manually selected ground station data in RFCs, and easier to process. Feng Ding: feng.ding@nasa.gov AGU Fall Meeting San Francisco, CA Dec 3 – 7, 2012 H21H-1297 PRELIMINARY RESULT AIRS Data Screening All: the data will be used regardless of number of valid grid points 53%: the data will be used if at least 53% (40 out of 75) grid points are valid 67%: the data will be used if at least 67% (50 out of 75) grid points are valid 80%: the data will be used if at least 80% (60 out of 75) grid points are valid Evaluation Statistics Comparing T ave forecast from regression with T ave from NLDAS Bias: arithmetic mean error RMSE: Root Mean Square Error Regression weights CNRFC 1200-1800 1800-0000 0000-0600 0600-1200 W min W max W min W max W min2 W max W min W max W min 0.73 0.27 0.07 0.84 0.09 0.35 0.65 0.23 0.77 AIRS W des W asc W des W asc W des2 W asc W des W asc W des All 0.84 0.16 0.01 0.68 0.31 0.34 0.66 0.01 0.99 53% 0.85 0.15 0.01 0.68 0.31 0.34 0.66 0.01 0.99 67% 0.85 0.15 0.01 0.68 0.31 0.34 0.66 0.01 0.99 80% 0.84 0.16 0.01 0.67 0.32 0.33 0.66 0.01 0.99 Statistic Numbers ( 0 K) CNRFC 1200-1800 1800-0000 0000-0600 0600-1200 Bias RMSE Bias RMSE Bias RMSE Bias RMSE 0.09 1.43 0.07 1.12 0.22 1.79 0 1.39 AIRS All -0.04 1.40 -0.35 2.43 0.05 1.54 0.44 2.23 53% -0.04 1.40 -0.22 2.26 0.10 1.49 0.32 2.19 67% -0.04 1.42 -0.20 2.20 0.13 1.42 0.31 2.10 80% 0.01 1.31 -0.20 2.15 0.08 1.31 0.37 1.97 Larger bias and error at 18-00Z and 06-12Z 3-day (April 14-16, 2011) daily cycle of basin averaged hourly temperature (generated by GES-DISC Giovanni) Try polynomial fit 00-06, 06-12, and 12-18Z: T ave = a*T des + b*T asc + c 18-00Z: T ave = a*T des + b*T asc + a 2 *T des2 + c RMSE Comparison Bias Comparison Ten years of data: 2003 – 2012 Sample image generated by GES-DISC Giovanni 1.Basin averaged daily minimum/maximum temperature often occurs at 12Z/00Z. 2.AIRS passes CNRFC at about nighttime/daytime 09Z/22Z. 3.AIRS nighttime/daytime temperature is warmer/colder than daily minimum/maximum temperature, which causes poor regression and larger bias and error at 18-00Z and 06-12Z. Polynomial fit remarkably reduces error at 18-00Z and 06-12Z, but is still somewhat inferior than CNRFC regression.