Modeled and empirical approaches for retrieving columnar water vapor from solar transmittance measurements in the 0.72, 0.82, and 0.94 mu m absorption bands

A Sun photometer (18 channels between 300 and 1024 nm) has been used for me asuring the columnar content of atmospheric water vapor (CWV) by solar tran smittance measurements in absorption bands with channels centered at 719, 8 17, and 946 nm. The observable is the band-weighted transmittance function defined by the spectral absorption of water vapor and the spectral features of solar irradiance and system response. The transmittance function is app roximated by a three-parameter model. Its parameters are determined from MO DTRAN and LBLRTM simulations or empirical approaches using CWV data of a du al-channel microwave radiometer (MWR) or a Fourier transform spectrometer ( FTS). Data acquired over a 2-year period during 1996-1998 at two different sites in Switzerland, Bern (560 m above sea level (asl)) and Jungfraujoch ( 3580 m asl) were compared to MWR, radiosonde (RS), and FTS retrievals. At t he low-altitude station with an average CWV amount of 15 mm the LBLRTM appr oach (based on recently corrected line intensities) leads to negligible bia ses at 719 and 946 nm if compared to an average of MWR, RS, and GPS retriev als. However, at 817 nm an overestimate of 2.7 to 4.3 mm (18-29%) remains. At the high-altitude station with an average CWV amount of 1.4 mm the LBLRT M approaches overestimate the CWV by 1.0, 1.4, and 0.1 mm (58, 76, and 3%) at 719, 817, and 946 nm, compared to the FTS instrument. At the low-altitud e station, CWV estimates, based on empirical approaches, agree with the MWR within 0.4 mm (2.5% of the mean); at the high-altitude site with a factor of 10 less water vapor the agreement of the SPM with the FTS is 0.0 to 0.2 mm (1 to 9% of the mean CWV there). Sensitivity analyses show that for the conditions met at the two stations with CWV ranging from 0.2 to 30 mm, the retrieval errors are smallest if the 946 nm channel is used.