Calculation of zenith delays from meteorological data comparison of NWP model, radiosonde and GPS delays

We discuss aspects of how to derive precise estimates of zenith total delay s (ZTD's) of the neutral atmosphere (its non-ionized constituents) from met eorological data in order to validate ZTD's derived from ground based GPS ( Global Positioning System) observations. The goal is that a later stage GPS ZTD's may conversely be used in numerical weather prediction (NWP) model v alidation and forecasting. We determine ZTD by numerical integration over m odeled or measured profiles of the atmosphere. Doing so, we find it is nece ssary to discriminate between geopotential and geometric heights. A term is added for the delay arising above the known atmospheric profile. Results o btained for the dry delay by numerical integration and by the Saastamoinen compare well, offsets being of sub millimeter scale only. The offsets we at tribute to the time variations of the atmospheric temperature and humidity profile not being accounted for in the Saastamoinen formula. It is found th at conversion of the dewpoint temperatures appearing in radiosonde (RS) rep orts to relative humidities may constitute a problem, and give an estimate of the maximum error associated with that. The GPS sites are in general not co-located with the RS sites nor with the grid-points of the model fields. Before deriving and comparing ZTD's it is necessary to correct for such po sitional offsets. We find that the correction for vertical offsets between GPS and RS sites or model orography is ambiguous, which can introduce error s. Using our ZTD calculation algorithms we compare ZTD's based on data from th e first one and half year of the MAGIC project. The results are: (ZTD(GPS) - ZTD(RS)) = 6.0 +/- 11.7 mm, (ZTD(GPS) - ZTD(model)) = 3.2 +/- 17.1 mm, (ZTD(RS) - ZTD(model)) = -1.6 +/- 14.7 mm, (ZHD(RS) - ZHD(model)) = 1.2 +/- 3.7 mm. (C) 2001 Elsevier Science Ltd. All rights reserved.