A Comparison of the First-Guess Dependence of Precipitable Water Estimates from Three Techniques Using GOES Data

Abstract

This paper evaluates and intercompares three existing algorithms for calculating precipitable water (PW) using infrared radiances from the GOES-7 VISSR (Visible and Infrared Spin Scan Radiometer) Atmospheric Sounder (VAS). The study exclusively uses simulated, rather than observed, VAS radiances in all retrievals. The National Environmental Satellite, Data, and Information Service simultaneous physical algorithm utilizes data from all 12 VAS channels and produces a vertical profile of temperature and dewpoint from which PW can be calculated. The Chesters technique and Jedlovec?s physical split-window technique retrieve PW from radiances in the two split window channels without first computing a dewpoint profile. All three algorithms also can be used with GOES-8 and GOES-9 data. These algorithms have not been intercompared previously. Each is applied on case days having wide variations in temperature and moisture. The algorithms are supplied with first-guess information of varying accuracy to assess their sensitivity to the guess data. The performance of the techniques relative to one another is described, including important similarities and differences among them. Results show that all three algorithms perform well within most temperature and moisture regimes. Each retrieves PW that is generally an improvement upon the first guess and is more accurate than PW predicted by surface temperature alone. However, each algorithm is somewhat dependent upon the first guess. Warm-biased first-guess surface temperatures are generally associated with moist-biased PW retrievals, while cold-biased first-guess surface temperatures are generally associated with dry-biased retrievals. The first-guess surface temperature errors reflect the presence, in either the first-guess or observed temperature profiles, of low-level inversions that cause the PW retrieval errors. Retrievals made where the observed contribution of low-level moisture to total column PW is small are usually moist biased, while those where the low-level contribution is large are usually dry biased. Both of these relationships exist irrespective of the sign of the first-guess PW error.