• Berg, H. 1948. Allgemeine Meteorologie. Duemmler’s Verlag, 337 pp.

  • Bevis, M., , S. Businger, , T A. Herring, , C. Rocken, , R A. Anthes, , and R H. Ware. 1992. GPS meteorology: Remote sensing of atmospheric water vapor using the global positioning system. J. Geophys. Res. 97:1578715801.

    • Search Google Scholar
    • Export Citation
  • Brockmann, E., and D. Inechen. 2005. COST Action 716: Final report. LPT Analysis Centre, 45–48. [Available online at http://www.oso.chalmers.se/~kge/cost716.html.].

  • Brockmann, E., , G. Guerova, , and M. Troller. 2002. Swiss activities in combining GPS with meteorology. Mitteilung des Bundesamtes für Kartographie and Geodäsie, Vol. 23, J. A. Torres and H. Hornik, Eds., EUREF, 95–99.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., , B. Navasceus, , G. Ruffini, , P. Elosegui, , A. Rius, , and J. Vila. 2000. The use of GPS to validate NWP systems: The HIRLAM model. J. Atmos. Oceanic Technol. 17:773787.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , J. Wang, , R H. Ware, , and T. Van Hove. 2002. Diurnal variation in water vapor over North America and its implications for sampling errors in radiosonde humidity. J. Geophys. Res. 107.4090, doi:10.1029/2001JD000642.

    • Search Google Scholar
    • Export Citation
  • Doms, G., and U. Schaettler. 2002. Consortium for small-scale modelling (COSMO). Newsletter Consortium for small-scale modeling, 2, G. Doms and U. Schaettler, Eds., 220 pp. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

  • Doms, G., , U. Schaettler, , and A. Montani. 2004. Model system overview: Data assimilation. COSMO Newsletter, No. 4, Deutscher Wetterdienst, Offenbach, Germany, 16–21. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

  • Elgered, G. 2001. An overview of COST Action 716: Exploitation of ground-based GPS for climate and numerical weather prediction applications. Phys. Chem. Earth A26:(6–8). 399404.

    • Search Google Scholar
    • Export Citation
  • Elgered, G., , H-P. Plag, , H. van der Marel, , S. Barlag, , and J. Nash. 2005. COST 716: Exploitation of ground-based GPS for climate and numerical weather prediction applications. European Community Final Rep., 234 pp. [Available online at http://www.oso.chalmers.se/~kge/cost716.html.].

  • Emardson, T R., , G. Elgered, , and J. Johanson. 1998. Three months of continuous monitoring of atmospheric water vapor with a network of global positioning system receivers. J. Geophys. Res. 103:18071820.

    • Search Google Scholar
    • Export Citation
  • Guerova, G., and M. Tomassini. 2003. Monitoring IWV from GPS and limited-area forecast model. University of Bern Institute of Applied Physics Research Rep. 2003-15, 9 pp. [Available online at http://www.iapmw.unibe.ch/publications/pdffiles/729.pdf.].

  • Guerova, G., , E. Brockmann, , J. Quiby, , F. Schubiger, , and C. Matzler. 2003. Validation of NWP mesoscale models with Swiss GPS network AGNES. J. Appl. Meteor. 42:141150.

    • Search Google Scholar
    • Export Citation
  • Haase, J. Coauthors 2001. The contribution of MAGIC project to the COST 716 objectives of assessing the operational potential of ground-based GPS meteorology on an international scale. Phys. Chem. Earth A26:433437.

    • Search Google Scholar
    • Export Citation
  • Haase, J., , M. Ge, , H. Vedel, , and E. Calais. 2003. Accuracy and variability of GPS tropospheric delay measurements of water vapor in the western Mediterranean. J. Appl. Meteor. 42:15471568.

    • Search Google Scholar
    • Export Citation
  • Hugentobler, U., , S. Schaer, , and P. Fridez. 2001. Bernese GPS Software version 4.2. Astronomical Institute, University of Bern, 515 pp. [Available from Astronomical Institute, University of Bern, Sidlerstrasse Bern 3012, Switzerland.].

  • Ingold, T. 2000. Monitoring atmospheric parameters from ground-based remote sensing networks in Switzerland. Ph.D. thesis, University of Bern, 238 pp. [Available online at http://www.iapmw.unibe.ch/publications/pdffiles/152.pdf.].

  • Jeannet, P., , G. Levrat, , and D. Bresch. 2003. Evaluation of the VIZ/Sippican resistive hygristors used with the SRS 400 radiosonde operated at Payerne, Switzerland. MeteoSwiss Rep., 23 pp. [Available from MeteoSwiss, Aerological Station, Les Invuardes, Payerne 1530, Switzerland.].

  • Köpken, C. 2001. Validation of integrated water vapor from numerical models using ground-based GPS, SSM/I, and water vapor radiometer measurements. J. Appl. Meteor. 40:11051117.

    • Search Google Scholar
    • Export Citation
  • Morland, J. 2002. TROWARA—Tropospheric water vapour radiometer. Radiometer review and new calibration model. University of Bern Institute of Applied Physics Research Rep. 2002-15, 33 pp. [Available online at http://www.iapmw.unibe.ch/publications/pdffiles/253.pdf.].

  • Ohtani, R., and I. Naito. 2000. Comparisons of GPS-derived precipitable water vapors with radiosonde observations in Japan. J. Geophys. Res. 105:2691726929.

    • Search Google Scholar
    • Export Citation
  • Peter, R., and N. Kämpfer. 1992. Radiometric determination of water vapor and liquid water and its validation with other techniques. J. Geophys. Res. 97:1817318183.

    • Search Google Scholar
    • Export Citation
  • Richner, H. 1999. Grundlagen aerologischer Messungen speziell mittels der Schweizer Sonde SRS 400. Veroeffentlichungen der SMA-MeteoSchweiz 61.140 pp.

    • Search Google Scholar
    • Export Citation
  • Schubiger, F. 2003. Verification of surface weather parameters at MeteoSwiss. COSMO Newsletter, No. 3, Deutscher Wetterdienst, 73–75. [Available online at http://cscs.cosmo-model.ch/Publications/Newsletters.].

  • Tomassini, M., , G. Gendt, , G. Dick, , M. Ramatschi, , and C. Schraff. 2002. Monitoring of integrated water vapour from ground-based GPS observations and their assimilation in a limited area model. Phys. Chem. Earth 27:341346.

    • Search Google Scholar
    • Export Citation
  • Turner, D., , B. Lesht, , S. Clouggh, , J. Liljegen, , H. Revercomb, , and D. Tobin. 2003. Dry bias and variability in Vaisala RS80-H Radiosondes: The ARM experience. J. Atmos. Oceanic Technol. 20:117132.

    • Search Google Scholar
    • Export Citation
  • Yang, X., , B H. Sass, , G. Elgered, , J M. Johansson, , and T R. Emardson. 1999. A comparison of precipitable water vapor estimates by an NWP simulation and GPS observations. J. Appl. Meteor. 38:941956.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Map of AGNES. The GPS sites used in the verification are marked with white bullets. The radiosonde in Payerne is presented with a black bullet, and the gray bullet is the location of the microwave radiometer in Bern. Note that the GPS sites in Payerne and Bern are collocated with the radiosonde and the MWR, respectively.

  • View in gallery

    Validation of GPS IWV with the Payerne radiosonde station: (a) Jan 2001–Jun 2003 IWV from GPS (solid line) available every 1 h and radiosonde (asterisks) available every 12 h (0000 and 1200 UTC), (b) IWV difference (GPS minus radiosonde) at 0000 UTC, and (c) IWV difference at 1200 UTC.

  • View in gallery

    Validation of GPS IWV with the MWR data for Bern in the period of Mar–Jul 2001: (a) IWV from GPS (solid line) and MWR (squares), (b) IWV difference (GPS minus MWR) at 0000 UTC, and (c) IWV difference at 1200 UTC.

  • View in gallery

    Same as Fig. 3, but for Jan–Jun 2003.

  • View in gallery

    Monthly IWV bias (black bars) and std dev (gray bars) in the period from Jan 2001 to Jun 2003 averaged over 14 Swiss GPS sites for the (a) aLMo analysis (0000–2300 UTC) and (b) aLMo forecast (0 to +23 h). Note that the monthly mean IWV from GPS (white bars) should be multiplied by 10.

  • View in gallery

    IWV diurnal cycle averaged over 14 GPS sites (aLMo points) in the period (a) May–Oct 2001 from the aLMo forecast (dashed line with circles) and from GPS (solid line) and (b) May–Oct 2002 from the aLMo analysis (dashed line with diamonds) and forecast (dashed line with circles) and from GPS (solid line). The model exhibits a constant dry bias in 2002 of approximately −2.5 kg m−2.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 25 25 2
PDF Downloads 19 19 2

An Integrated Assessment of Measured and Modeled Integrated Water Vapor in Switzerland for the Period 2001–03

View More View Less
  • a Institute of Applied Physics, University of Bern, Bern, Switzerland
  • b Swiss Federal Office of Topography, Wabern, Switzerland
  • c Federal Office of Meteorology and Climatology, Zurich, Switzerland
  • d Institute of Applied Physics, University of Bern, Bern, Switzerland
© Get Permissions
Full access

Abstract

In this paper an integrated assessment of the vertically integrated water vapor (IWV) measured by radiosonde, microwave radiometer (MWR), and GPS and modeled by the limited-area mesoscale model of MeteoSwiss is presented. The different IWV measurement techniques are evaluated through intercomparisons of GPS to radiosonde in Payerne, Switzerland, and to the MWR operated at the Institute of Applied Physics at the University of Bern in Switzerland. The validation of the IWV field of the nonhydrostatic mesoscale Alpine Model (aLMo) of MeteoSwiss is performed against 14 GPS sites from the Automated GPS Network of Switzerland (AGNES) in the period of 2001–03. The model forecast and the nudging analysis are evaluated, with special attention paid to the diurnal cycle. The results from the GPS–radiosonde intercomparison are in agreement, but with a bimodal distribution of the day-to-night basis. At 0000 UTC, the bias is negative (−0.4 kg m−2); at 1200 UTC, it is positive (0.9 kg m−2) and the variability increases. The intercomparison of GPS to MWR shows better agreement (0.4 kg m−2), with a small increase of the daytime bias with 0.3 kg m−2. The intercomparison of MWR to the radiosonde gives a bimodal distribution of the bias, with an increase in the standard deviation at the daytime measurement. The relative bias is negative (−3%) at 0000 UTC and is positive (3%) at 1200 UTC. Based on this cross correlation, it can be concluded that the bimodal distribution is a result of radiosonde humidity measurements. Possible reasons are the solar-heating correction or sensor errors. The monthly bias and standard deviation of aLMo exhibit a strong seasonal dependence with a pronounced dry bias during the warm months of May–October 2002. The diurnal IWV cycle in 2001 shows good model performance between 0000 and 0900 UTC but IWV underestimation by up to 1.5 kg m−2 for the rest of the day. In 2002 the diurnal cycle shows a systematic dry bias in both the analysis and forecast that is more pronounced in the analysis. This substantial underestimation of IWV was found to correlate with overestimation of aLMo precipitation, especially light precipitation up to 0.1 mm (6 h)−1 in 2002. There is strong evidence that this underestimation can be related to the dry radiosonde bias in midday summer observations. The aLMo dry bias is about 1.0–1.5 kg m−2 greater in the nudging analysis as compared with the forecast initialized at 0000 UTC.

* Current affiliation: Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

Corresponding author address: Dr. Guergana Guerova, Laboratoire de Modélisation de la Chimie Atmosphérique, Ecole Polytechnique Fédérale de Lausanne (EPFL), EPFL-ENAC-LMCA station 6, CH-1015 Lausanne, Switzerland. guergana.guerova@epfl.ch

Abstract

In this paper an integrated assessment of the vertically integrated water vapor (IWV) measured by radiosonde, microwave radiometer (MWR), and GPS and modeled by the limited-area mesoscale model of MeteoSwiss is presented. The different IWV measurement techniques are evaluated through intercomparisons of GPS to radiosonde in Payerne, Switzerland, and to the MWR operated at the Institute of Applied Physics at the University of Bern in Switzerland. The validation of the IWV field of the nonhydrostatic mesoscale Alpine Model (aLMo) of MeteoSwiss is performed against 14 GPS sites from the Automated GPS Network of Switzerland (AGNES) in the period of 2001–03. The model forecast and the nudging analysis are evaluated, with special attention paid to the diurnal cycle. The results from the GPS–radiosonde intercomparison are in agreement, but with a bimodal distribution of the day-to-night basis. At 0000 UTC, the bias is negative (−0.4 kg m−2); at 1200 UTC, it is positive (0.9 kg m−2) and the variability increases. The intercomparison of GPS to MWR shows better agreement (0.4 kg m−2), with a small increase of the daytime bias with 0.3 kg m−2. The intercomparison of MWR to the radiosonde gives a bimodal distribution of the bias, with an increase in the standard deviation at the daytime measurement. The relative bias is negative (−3%) at 0000 UTC and is positive (3%) at 1200 UTC. Based on this cross correlation, it can be concluded that the bimodal distribution is a result of radiosonde humidity measurements. Possible reasons are the solar-heating correction or sensor errors. The monthly bias and standard deviation of aLMo exhibit a strong seasonal dependence with a pronounced dry bias during the warm months of May–October 2002. The diurnal IWV cycle in 2001 shows good model performance between 0000 and 0900 UTC but IWV underestimation by up to 1.5 kg m−2 for the rest of the day. In 2002 the diurnal cycle shows a systematic dry bias in both the analysis and forecast that is more pronounced in the analysis. This substantial underestimation of IWV was found to correlate with overestimation of aLMo precipitation, especially light precipitation up to 0.1 mm (6 h)−1 in 2002. There is strong evidence that this underestimation can be related to the dry radiosonde bias in midday summer observations. The aLMo dry bias is about 1.0–1.5 kg m−2 greater in the nudging analysis as compared with the forecast initialized at 0000 UTC.

* Current affiliation: Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

Corresponding author address: Dr. Guergana Guerova, Laboratoire de Modélisation de la Chimie Atmosphérique, Ecole Polytechnique Fédérale de Lausanne (EPFL), EPFL-ENAC-LMCA station 6, CH-1015 Lausanne, Switzerland. guergana.guerova@epfl.ch

Introduction

Water vapor plays a key role in the hydrological cycle and atmospheric radiation. It actively participates in the processes of precipitation formation, energy transfer, and atmospheric stability. In addition, water vapor has a short lifetime in the atmosphere of about 1 week and has significant spatial and temporal variations, which affect both climate and numerical weather prediction (NWP) and makes its observation a challenging task. Observations are traditionally made with balloonborne soundings. The radiosonde network provides worldwide coverage, but its poor temporal resolution—typically two soundings a day—is not sufficient to study water vapor variations at different temporal and spatial scales. Another available instrument, which provides vertically integrated water vapor measurements with very high temporal resolution, is the microwave radiometer (MWR). The microwave technique, however, is not applicable in rain events and because of its significant cost is mainly used in research. From the existing satellite observations, the extraction of water vapor is still a challenge over the earth’s vegetated surfaces. This is, however, not the case for the ground-based global positioning system (GPS). The advantage of GPS is that it operates in all weather conditions, provides high-temporal-resolution data (every hour or less), and has dense spatial coverage—more than 250 receivers are operating in Europe and provide water vapor estimates for use in NWP (Elgered et al. 2005). It is inexpensive, it is easy to operate, and its potential is expected to increase with the new European Galileo project (the European satellite navigation system) in operational service from 2008. The importance of GPS as a monitoring tool has grown as a consequence of increased network density, gains in data processing, and, last but not least, development of high-resolution mesoscale models for operational NWP.

In Europe two projects, running in parallel, funded by the European Union, have been dedicated to the use of GPS in mesoscale meteorology. The first of them—the Meteorological Applications of GPS Integrated Column Water Vapor Measurements in the Western Mediterranean (MAGIC) project (Haase et al. 2001)—studied the variability of the GPS-derived water vapor in the western Mediterranean Sea area. The second—European Cooperation in the Field of Scientific and Technological Research (COST) Action 716 (Elgered 2001), initiated in 1999—has the objective to explore the application of GPS data in operational NWP in Europe. A critical task in this project was demonstration of GPS data delivery in operational NWP mode. Within Action 716 the task of Working Group Three was to validate the data and to perform assimilation experiments (Elgered et al. 2005). The Swiss contribution to COST 716 (Brockmann et al. 2002) is a collaboration between the Swiss Federal Office of Topography (Swisstopo), Federal Office of Meteorology and Climatology (MeteoSwiss), and the Institute of Applied Physics at the University of Bern in Switzerland. The first model validation was performed using the measurements from the Automated GPS Network of Switzerland (AGNES) processed by Swisstopo. The operational NWP system of MeteoSwiss (the hydrostatic Swiss Model) and the Swiss “Alpine Model (aLMo)” version of the nonhydrostatic Local Model (LM) have been compared with the GPS measurements from six sites for the period from November 2000 to March 2001 (Guerova et al. 2003). Further work on monitoring the integrated water vapor (IWV) field of NWP models was completed by Yang et al. (1999), Köpken (2001), Cucurull et al. (2000), Haase et al. (2003), and Tomassini et al. (2002). Haase et al. (2003) performed an NWP model [High Resolution Limited-Area Model (HIRLAM)] validation for a period of more than 2 years. The results show a seasonal dependence of the model bias and standard deviation. Tomassini et al. (2002) reported a dry bias in the LM in Germany and a dry bias in the diurnal cycle in the daytime hours. They attribute them to the shortcoming of the model or/and the poor time resolution of the observations available for model analysis. It is important to note here that in the work of Ohtani and Naito (2000) and Haase et al. (2003) a difference of GPS–radiosonde statistics is reported on a day-to-night basis. Haase et al. (2003) report on bimodal distribution of the GPS–radiosonde bias, with a higher bias in the daytime launches. They attribute this to the measurement biases in the radiosonde. Ohtani and Naito (2000) report a systematically larger GPS–radiosonde bias for 1200 than for 0000 UTC. The GPS–radiosonde mean bias increases from 1.7 kg m−2 at 0000 UTC to 3.5 kg m−2 at 1200 UTC. The standard deviation of the mean bias does not vary much with the launch time. They point out that the systematic differences of mean bias at 0000 and 1200 UTC are due to neither GPS analysis software nor temporal resolution of GPS IWV estimates.

The objective of this study is 1) to evaluate the IWV derived from different techniques, including radiosonde, MWR, and GPS, and 2) to validate the water vapor field of the operational NWP system of MeteoSwiss. In section 2, a description of the datasets is given. In section 3, the intercomparison of GPS and radiosonde (Payerne) and GPS and microwave radiometer (Bern) is described. The verification of aLMo with GPS data from AGNES for the period of January 2001–June 2003 is presented in section 4. A summary and conclusions can be found in section 5.

Datasets

GPS data

AGNES, the ground-based GPS network of Switzerland, is operated by Swisstopo. It is primarily used for navigation and surveying purposes in postprocessing and real-time mode. The network was completed in 2001 and has 30 permanent receivers at a spacing of 50 km (Brockmann and Inechen 2005). For estimation of hourly zenith total delay (ZTD) the Bernese 4.2 software package (Hugentobler et al. 2001) with Niell mapping function and 10° minimum elevation angle is used. In this study, the data from 20 AGNES sites are compared with the aLMo. The site locations are represented by white bullets in Fig. 1.

GPS IWV data are retrieved from the ZTD following the standard extraction procedure described in Bevis et al. (1992) and Emardson et al. (1998). The meteorological data, surface temperature ts and pressure ps, are obtained from the surface observation network of MeteoSwiss. The data are interpolated to the height h of the GPS antenna by using the method of Berg (1948):
i1520-0450-44-7-1033-e1
i1520-0450-44-7-1033-e2

There is no spatial interpolation applied because 11 out of 14 meteorological stations are within a radius of 3 km (last column in Table 1).

NWP data

At MeteoSwiss the nonhydrostatic Consortium of Small Scale Modeling (COSMO; Doms and Schaettler 2002) model has been in operational use for NWP since 1 April 2001, after a preoperational phase from September 2000 to March 2001. The Swiss implementation of LM (aLMo) has a horizontal grid of 7 km × 7 km (1/16°) and 45 vertical levels from the surface up to 20 hPa in a generalized terrain-following coordinate system. The model prognostic variables are temperature, perturbation pressure, horizontal and vertical wind velocity, water vapor, and cloud water. The aLMo lateral boundary conditions are obtained by interpolation of driving-model forecast from the German Global Model (GME).

The aLMo analyses are produced with a nudging data-assimilation scheme. Only conventional observations are assimilated (Doms et al. 2004), including synoptic (synop), ship, and buoy (surface pressure, 2-m relative humidity, 2-m temperature and 10-m wind for stations below 100 m MSL); temperature (temp) and pilot (wind, temperature, and specific humidity profiles); and air reports (AIREP) and aircraft meteorological data reporting (AMDAR) (wind and temperature). The time window of the assimilated vertical profiles extends 3 h before the observation time to 1 h after. The time window of the surface observations (synop) and single upper-air observations extends 1.5 h before the observation time and 0.5 h after. The half-width horizontal influence radius is 110 km close to the ground. The aLMo IWV is computed as in Eq. (3), defined in section 2c. The gridpoint selection is based on smallest height difference to the GPS antenna. The first 14 sites in Table 1 are selected for validation because the height difference is less than 55 m. The remaining six sites, with significant height difference greater than 55 m, are given for completeness.

Radiosonde data—Payerne

There is one radiosonde sounding (RS) station in Switzerland located at Payerne (Swiss Plateau region; black bullet in Fig. 1). A balloon sounding (sonde type SRS 400, MeteoLabor, Switzerland) is performed 2 times per day (0000 and 1200 UTC), measuring temperature, pressure, humidity, and wind profiles.

A fast-response VIZ, Inc., ACCU-LOK carbon hygristor is used to measure relative humidity. During the radiosonde preflight procedure, the lock-in humidity resistance is introduced in the data acquisition software and the sensor. The operating range extends from 0% to 100% relative humidity (RH) and from −60° to +40°C, with an accuracy of 2% RH (rms; Richner 1999).

The result from a recent investigation at MeteoSwiss (Jeannet et al. 2003) of the radiosonde humidity hygristor accuracy is of relevance to this work and is summarized here. Until mid-March of 2001 the sensors used in Payerne were produced by VIZ in the United States. VIZ was bought by Sippican, Inc., and the production was transferred to a factory in Mexico. The hygristors delivered by Sippican encountered serious problems during the preflight checks in mid-2001. The problems were not limited to preflight conditions; they also occurred in flight conditions. One of the findings was that the sensors that give lower humidity when compared with the reference during the preflight test tend toward unrealistic low humidity in atmospheric layers characterized by low humidity. On the basis of three different analyses, Jeannet et al. (2003) concluded that the quality and reproducibility of the hygristors delivered by Sippican were reduced in comparison with hygristors manufactured by VIZ. The problems in the sensors, used in the period from March 2001 to early 2002, were reported to the company, and the new hygristors used since then seem to have better reproducibility based on the preliminary results from dual soundings.

The radiosonde IWV amount is calculated using
i1520-0450-44-7-1033-e3
where ρ is water vapor density, h is height in meters between the ground level h0 and the top level htop, and IWV is in kilograms per square meter.

Microwave radiometer data—Bern

The MWR operated at the University of Bern’s Institute of Applied Physics (Peter and Kämpfer 1992) consists of two total power radiometers at 21.3- and 31.5-GHz frequency. The instrument has been operating automatically since late 1994. The view direction was southeast in 2001 with a sampling rate of two measurements per minute. After an instrument revision in the spring/summer of 2002, the view direction was northeast and the sampling rate was increased to 30 measurements per minute. For every sky measurement, hot-load (312 K) and cold-load (35 K) observations are also made. An initial internal calibration is carried out using these measurements and a radiometer model (Morland 2002). This calibration is adjusted using an external calibration calculated from sky observations using the tipping-curve method.

Radiometric retrieval of IWV was carried out using the linear algorithm described in Ingold (2000). The microwave radiometer could possibly have a diurnal bias as a result of solar heating during the daytime, but this effect has not been noticed and should be compensated by the temperature measurements used in the radiometer model. In this study, 1-h-averaged IWV is used for the periods of March–July 2001 and January–June 2003.

Results from evaluation of IWV from radiosonde, MWR, and GPS

Intercomparison of GPS with radiosonde—Payerne

The radiosonde IWV data are compared with the GPS data from the collocated site at Payerne (PAYE). The results are plotted in Fig. 2 for the period from January 2001 to June 2003 and show overall good agreement, with a small positive bias of 0.3 kg m−2. However, to investigate the dependence of statistics on day/night observation, as reported by Ohtani and Naito (2000) and Haase et al. (2003), the bias is computed and plotted separately in Fig. 2b for 0000 UTC and in Fig. 2c for 1200 UTC launches. In both plots a clear seasonal dependence is seen, being substantial in the summer to early autumn period. The standard deviation (std dev) increases from 1.8 kg m−2 at 0000 UTC to 2.1 kg m−2 at 1200 UTC. Note that at 0000 UTC the IWV from GPS presents an overall negative bias of −0.4 kg m−2, whereas for the 1200 UTC sounding the bias is positive at 0.9 kg m−2. The main contributions to the positive bias at 1200 UTC are from 2002 and 2003. The 1200 UTC biases are 1.5 and 0.9 kg m−2 for those years; that is, there is a consistent wet GPS bias or dry radiosonde bias in the midday observations for 2002–03. The relative bias at 1200 UTC in 2003 is 3.5%. At 0000 UTC of the same period, the agreement between the IWV from GPS and radiosonde is much better: 0.3 and −0.6 kg m−2. The relative bias at 0000 UTC is negative: −2.3%.

The statistics from 2001 are commented upon separately because of reported problems in the preflight and in-flight humidity measurements. In 2001, both the 0000 and 1200 UTC observations have considerable variations, which can be seen in Figs. 2b and 2c. The std dev increases from 2.0 kg m−2 at 0000 UTC to 2.4 kg m−2 at 1200 UTC. The difference in the bias remains at the 1.2 kg m−2 level, but this time it is negative (−1.1 kg m−2) at midnight, whereas the midday soundings show good agreement. Note that the reported hygristor problems in 2001 are most likely the reason for the differences in the day-to-night statistics in 2001 and the following 2002–03 period.

Note that radiosonde observations at 1200 UTC are corrected for the impact of solar heating (P. Jeannet 2004, personal communication). This correction has possible influences on the retrieval of the humidity, but maybe it is not the only reason. In an investigation using a Vaisala radiosonde, Turner et al. (2003) reported problems in humidity measurements even after introduction of a radiative heating correction.

Intercomparison of GPS with microwave radiometer—Bern

The microwave radiometer IWV is compared with the GPS data from the collocated site in Bern (EXWI). The microwave data are used first to assess the accuracy of the GPS-derived IWV and second to evaluate the day-to-night bias differences already seen in the radiosonde intercomparison (section 3a).

The results presented in Fig. 3 (from 2001) and Fig. 4 (from 2003) show overall agreement between GPS and MWR IWV at a level below 0.5 kg m−2. In 2001 the agreement between GPS and MWR is very good; in 2003 the GPS shows a positive bias of 0.4 kg m−2 in comparison with MWR. The std dev is 1.1 and 1.4 kg m−2 in 2001 and 2003, respectively. For the day-to-night differences in the statistics, similar results are obtained in 2001. In this period the bias is very small, in the range of 0.1 kg m−2, and it could not be concluded that any significant differences from day to night are observed. The January–June period of 2003 is different, however. There the midday bias is almost 0.6 versus 0.3 kg m−2 at midnight. There is a slight increase in the midday std dev to 1.7 versus 1.4 kg m−2 at midnight. This result suggests that IWV from the GPS station EXWI in Bern exhibits a small overestimation in the midday observation. This day-to-night variation in the range of 0.3 kg m−2 is well within the accuracy limit of the two techniques.

To identify better the reason for the differences in the day-to-night GPS–RS statistics, a cross correlation with the MWR is offered; that is, the MWR in Bern is compared with the radiosonde in Payerne for the period of March–July 2003. Note that Payerne and Bern are located in the Swiss Plateau at a distance of 50 km apart (Fig. 1). During the period considered, mean measured GPS IWV is 26.2 kg m−2 in Payerne and 25.9 kg m−2 in Bern. The MWR IWV compares surprisingly well to that of the radiosonde for this period. There is no bias, and the std dev is 2.2 kg m−2. Further, the MWR–RS statistics are computed separately at 0000 and 1200 UTC. At 0000 UTC the MWR underestimates IWV: the bias is −0.8 kg m−2 with a std dev of 1.8 kg m−2. The relative bias is a little smaller than −3%. At 1200 UTC the MWR overestimates the IWV in comparison with the radiosonde: the bias is 0.8 kg m−2, and the variability is increased to 2.2 kg m−2. The relative bias has the same magnitude of 3% but is opposite in sign.

Validation of the Alpine Model with GPS

Monthly IWV from GPS and aLMo analysis and forecast

The bias and std dev of the aLMo analysis (0000–2300 UTC) and aLMo forecast (from 0 to +23 h) are calculated on a monthly basis and are presented in Fig. 5. These statistics are averaged over 14 stations (model grid points) in the period from January 2001 to June 2003. In Fig. 5a, clear seasonal dependence of both the bias and the std dev is seen. The relative bias from the forecast and analysis is presented in Table 2 for winter [December–February (DJF)], spring [March–May (MAM)], summer [June–August (JJA)], and autumn [September–November (SON)]. In general, the bias and the relative bias are significant in the warm months from JJA and are pronounced in 2002 in both analysis and forecast. The model tends to underestimate the IWV amount in comparison with the GPS measurements; that is, the model exhibits a dry bias.

The bias in the aLMo analysis, presented in Fig. 5a, is in the range from −1.5 to −2.5 kg m−2 in JJA. When compared with the forecast (Fig. 5b), it is to be noted that the dry bias is stronger in the analysis, produced by the nudging scheme. In the period from December to March the model forecast bias rarely exceeds −0.5 kg m−2. This result is consistent with model verification with the German Network reported in Tomassini et al. (2002). In contrast to the aLMo analysis, a difference is seen in the forecast between 2001 and 2002. In 2002 the dry model bias can be as low as −1.5 kg m−2, and the std dev ranges from 1.5 to 2.8 kg m−2 in the period from May to October. In 2001 the dry bias in the model is less pronounced and varies between −0.4 and −0.9 kg m−2. The monthly mean IWV in summer 2001 and 2002 are in the same range, which does not explain this difference in the bias.

For the aLMo forecast an investigation of the site-dependent bias and std dev was made. For the PAYE site the model dry bias is pronounced in the summer of 2002. During the winter months, the model forecast successfully predicts the water vapor content and the bias is small—in the range from −0.2 to −0.5 kg m−2. The model performs very well for the Lausanne (EPFL) site, located some 60 km south from PAYE. There the typical dry bias is identified in the three summer months of 2002. For two other sites, at St. Gallen (STGA) and Zimmerwald (ZIMM), the model dry bias is present during the entire period under investigation. Note that the altitude of the sites is about 200 and 400 m above the average Swiss Plateau height (about 500 m), which could be a reason for the persistent dry bias.

For the remaining 10 sites a pronounced dry bias is observed in the summer months of 2002. A bias in the range from −1.0 to −1.8 kg m−2 is obtained for EXWI, Muttenz (FHBB), Geneva (GENE), and Lucerne (LUZE). Note that for the two alpine sites at Saanen (SAAN; 1368 m MSL) and Andermatt (ANDE; 2317 m MSL), the IWV monthly bias in June, July, and August of 2002 exceeds −2.0 kg m−2. The same is true for the Stabio (STAB) site located in southern Switzerland. The bias reported for these three sites contributes substantially to the overall bias seen in the summer of 2002. The GPS IWV derived from the Jungfraujoch (JUJO) site continuously underestimates the model IWV.

Diurnal IWV cycle

The diurnal variation of atmospheric water vapor affects atmospheric longwave radiation and atmospheric absorption of solar radiation (Dai et al. 2002). It is also related to many other processes, such as atmospheric stability, diurnal variation of moist convection and precipitation, surface wind convergence, and evapotranspiration. The diurnal water vapor cycle from GPS and aLMo has been investigated from May to October in 2001 and 2002. In 2001 (Fig. 6a), there is good agreement between the GPS and aLMo forecast in the hours between 0000 and 0900 UTC and a dry model bias in the range from 0.5 to 1.5 kg m−2 for the rest of the time. The monthly results for July, August, and September at 1200 UTC show that the dry bias in the forecast reaches 2 kg m−2. This amount is about 1 kg m−2 larger than the one reported by Tomassini et al. (2002) and Guerova and Tomassini (2003).

The diurnal IWV cycle in 2002 (Fig. 6b) clearly demonstrates that both the analysis and forecast underestimate the IWV amount measured by GPS. For the period from May to October a systematic IWV underestimation in the range of 2.0–2.8 kg m−2 is found in the forecast. The aLMo analysis shows a dry bias of up to 3 kg m−2 in the hours between 1200 and 2300 UTC, stronger than the forecast bias. From the precipitation verification in the summer of 2002, reported by Schubiger (2003), it was found that precipitation was overestimated by ∼50%–60% for 0.1 mm (6 h)−1 threshold. High precipitation amounts [larger than 10 mm (6 h)−1] were overestimated by ∼10%. This verification of the aLMo precipitation and IWV diurnal cycle in 2002 could suggest that there is a coupling: too little water vapor and too much light precipitation. In addition, the strong dry bias in 2002 can be related to dry radiosonde bias in midday summer observations, that is, to the underestimation of the IWV in the 1200 UTC radiosounding. Strong support for this relation is given by differences in the range of 1–1.5 kg m−2 between the model forecast and analysis in the hours between 1000 and 2300 UTC. Note that the Payerne radiosonde is the only sounding station located in western Switzerland (Fig. 1) and that in the assimilation a large area is influenced through the observation-spreading function implemented in the model (section 2b). In addition, the assimilated-in-the-model European radiosoundings (except Payerne) that are assimilated in aLMo are of Vaisala type and, as reported in Haase et al. (2003), have an increased bias and std dev for the 1200 UTC sounding when compared with those from 0000 UTC. This result suggests that the aLMo IWV field contains larger error in the daytime analysis than in the forecast, where the observations have a limited influence.

Summary and conclusions

IWV measurements from the radiosonde, microwave radiometer, and GPS are evaluated in the period of 2001–03 for Switzerland. Further, the IWV forecast and analysis of the Alpine Model, the operational NWP model of MeteoSwiss, have been monitored using the GPS measurements. The ground-based GPS data from 20 permanent Swiss sites, processed by the Swiss Federal Office of Topography in postprocessed mode, have been selected based on availability of surface meteorological observations.

The GPS–radiosonde statistics shows differences for 0000 and 1200 UTC measurements. The observed GPS IWV has a positive bias of 0.9 kg m−2 at 1200 UTC. At 0000 UTC, the bias is negative and is reduced to −0.4 kg m−2. The GPS intercomparison with the microwave radiometer data from Bern indicates a positive bias of about 0.6 kg m−2 at 1200 UTC observations in 2003. In the same period, but for 0000 UTC, a positive bias is reported in the range of 0.3 kg m−2. The cross correlation of MWR with radiosonde confirms the dry midday bias of the radiosonde sensor.

The model monthly bias and standard deviation for 14 sites exhibit a seasonal dependence with pronounced dry bias in the summer of 2002. The diurnal IWV cycle in aLMo exhibits an underestimation of the IWV in the daytime hours of 2001 and over the entire day in 2002. A significant dry bias, reaching 2.8 kg m−2, is seen in the model forecast in 2002. The precipitation verification in 2002 gives an overestimation of about 50% of the threshold of 0.1 mm (6 h)−1. The increased IWV underestimation up to 1.5 kg m−2 between 1000 and 2300 UTC in the aLMo analysis when compared with the forecast is a sign of the influence of 1200 UTC radiosonde IWV underestimation. Note that the forecast initialized at 0000 UTC is presented.

It can be concluded that GPS-derived IWV can be efficiently used in monitoring the performance of the NWP models as well as in assimilation process. The integration of GPS in the operational verification package of MeteoSwiss was successfully completed in 2003, and since January 2004 the GPS data have been used for daily model verification.

Acknowledgments

We are grateful to Dr. U. Hugentobler (Astronomical Institute, University of Bern) for the useful comments concerning the ionospheric contribution in the GPS data processing. This work was supported by the Swiss Federal Office of Education and Science under Grant C99.0046 and by MeteoSwiss. The work also contributes to the Swiss NCCR Climate Program of the National Science Foundation.

REFERENCES

  • Berg, H. 1948. Allgemeine Meteorologie. Duemmler’s Verlag, 337 pp.

  • Bevis, M., , S. Businger, , T A. Herring, , C. Rocken, , R A. Anthes, , and R H. Ware. 1992. GPS meteorology: Remote sensing of atmospheric water vapor using the global positioning system. J. Geophys. Res. 97:1578715801.

    • Search Google Scholar
    • Export Citation
  • Brockmann, E., and D. Inechen. 2005. COST Action 716: Final report. LPT Analysis Centre, 45–48. [Available online at http://www.oso.chalmers.se/~kge/cost716.html.].

  • Brockmann, E., , G. Guerova, , and M. Troller. 2002. Swiss activities in combining GPS with meteorology. Mitteilung des Bundesamtes für Kartographie and Geodäsie, Vol. 23, J. A. Torres and H. Hornik, Eds., EUREF, 95–99.

    • Search Google Scholar
    • Export Citation
  • Cucurull, L., , B. Navasceus, , G. Ruffini, , P. Elosegui, , A. Rius, , and J. Vila. 2000. The use of GPS to validate NWP systems: The HIRLAM model. J. Atmos. Oceanic Technol. 17:773787.

    • Search Google Scholar
    • Export Citation
  • Dai, A., , J. Wang, , R H. Ware, , and T. Van Hove. 2002. Diurnal variation in water vapor over North America and its implications for sampling errors in radiosonde humidity. J. Geophys. Res. 107.4090, doi:10.1029/2001JD000642.

    • Search Google Scholar
    • Export Citation
  • Doms, G., and U. Schaettler. 2002. Consortium for small-scale modelling (COSMO). Newsletter Consortium for small-scale modeling, 2, G. Doms and U. Schaettler, Eds., 220 pp. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

  • Doms, G., , U. Schaettler, , and A. Montani. 2004. Model system overview: Data assimilation. COSMO Newsletter, No. 4, Deutscher Wetterdienst, Offenbach, Germany, 16–21. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

  • Elgered, G. 2001. An overview of COST Action 716: Exploitation of ground-based GPS for climate and numerical weather prediction applications. Phys. Chem. Earth A26:(6–8). 399404.

    • Search Google Scholar
    • Export Citation
  • Elgered, G., , H-P. Plag, , H. van der Marel, , S. Barlag, , and J. Nash. 2005. COST 716: Exploitation of ground-based GPS for climate and numerical weather prediction applications. European Community Final Rep., 234 pp. [Available online at http://www.oso.chalmers.se/~kge/cost716.html.].

  • Emardson, T R., , G. Elgered, , and J. Johanson. 1998. Three months of continuous monitoring of atmospheric water vapor with a network of global positioning system receivers. J. Geophys. Res. 103:18071820.

    • Search Google Scholar
    • Export Citation
  • Guerova, G., and M. Tomassini. 2003. Monitoring IWV from GPS and limited-area forecast model. University of Bern Institute of Applied Physics Research Rep. 2003-15, 9 pp. [Available online at http://www.iapmw.unibe.ch/publications/pdffiles/729.pdf.].

  • Guerova, G., , E. Brockmann, , J. Quiby, , F. Schubiger, , and C. Matzler. 2003. Validation of NWP mesoscale models with Swiss GPS network AGNES. J. Appl. Meteor. 42:141150.

    • Search Google Scholar
    • Export Citation
  • Haase, J. Coauthors 2001. The contribution of MAGIC project to the COST 716 objectives of assessing the operational potential of ground-based GPS meteorology on an international scale. Phys. Chem. Earth A26:433437.

    • Search Google Scholar
    • Export Citation
  • Haase, J., , M. Ge, , H. Vedel, , and E. Calais. 2003. Accuracy and variability of GPS tropospheric delay measurements of water vapor in the western Mediterranean. J. Appl. Meteor. 42:15471568.

    • Search Google Scholar
    • Export Citation
  • Hugentobler, U., , S. Schaer, , and P. Fridez. 2001. Bernese GPS Software version 4.2. Astronomical Institute, University of Bern, 515 pp. [Available from Astronomical Institute, University of Bern, Sidlerstrasse Bern 3012, Switzerland.].

  • Ingold, T. 2000. Monitoring atmospheric parameters from ground-based remote sensing networks in Switzerland. Ph.D. thesis, University of Bern, 238 pp. [Available online at http://www.iapmw.unibe.ch/publications/pdffiles/152.pdf.].

  • Jeannet, P., , G. Levrat, , and D. Bresch. 2003. Evaluation of the VIZ/Sippican resistive hygristors used with the SRS 400 radiosonde operated at Payerne, Switzerland. MeteoSwiss Rep., 23 pp. [Available from MeteoSwiss, Aerological Station, Les Invuardes, Payerne 1530, Switzerland.].

  • Köpken, C. 2001. Validation of integrated water vapor from numerical models using ground-based GPS, SSM/I, and water vapor radiometer measurements. J. Appl. Meteor. 40:11051117.

    • Search Google Scholar
    • Export Citation
  • Morland, J. 2002. TROWARA—Tropospheric water vapour radiometer. Radiometer review and new calibration model. University of Bern Institute of Applied Physics Research Rep. 2002-15, 33 pp. [Available online at http://www.iapmw.unibe.ch/publications/pdffiles/253.pdf.].

  • Ohtani, R., and I. Naito. 2000. Comparisons of GPS-derived precipitable water vapors with radiosonde observations in Japan. J. Geophys. Res. 105:2691726929.

    • Search Google Scholar
    • Export Citation
  • Peter, R., and N. Kämpfer. 1992. Radiometric determination of water vapor and liquid water and its validation with other techniques. J. Geophys. Res. 97:1817318183.

    • Search Google Scholar
    • Export Citation
  • Richner, H. 1999. Grundlagen aerologischer Messungen speziell mittels der Schweizer Sonde SRS 400. Veroeffentlichungen der SMA-MeteoSchweiz 61.140 pp.

    • Search Google Scholar
    • Export Citation
  • Schubiger, F. 2003. Verification of surface weather parameters at MeteoSwiss. COSMO Newsletter, No. 3, Deutscher Wetterdienst, 73–75. [Available online at http://cscs.cosmo-model.ch/Publications/Newsletters.].

  • Tomassini, M., , G. Gendt, , G. Dick, , M. Ramatschi, , and C. Schraff. 2002. Monitoring of integrated water vapour from ground-based GPS observations and their assimilation in a limited area model. Phys. Chem. Earth 27:341346.

    • Search Google Scholar
    • Export Citation
  • Turner, D., , B. Lesht, , S. Clouggh, , J. Liljegen, , H. Revercomb, , and D. Tobin. 2003. Dry bias and variability in Vaisala RS80-H Radiosondes: The ARM experience. J. Atmos. Oceanic Technol. 20:117132.

    • Search Google Scholar
    • Export Citation
  • Yang, X., , B H. Sass, , G. Elgered, , J M. Johansson, , and T R. Emardson. 1999. A comparison of precipitable water vapor estimates by an NWP simulation and GPS observations. J. Appl. Meteor. 38:941956.

    • Search Google Scholar
    • Export Citation
Fig. 1.
Fig. 1.

Map of AGNES. The GPS sites used in the verification are marked with white bullets. The radiosonde in Payerne is presented with a black bullet, and the gray bullet is the location of the microwave radiometer in Bern. Note that the GPS sites in Payerne and Bern are collocated with the radiosonde and the MWR, respectively.

Citation: Journal of Applied Meteorology 44, 7; 10.1175/JAM2255.1

Fig. 2.
Fig. 2.

Validation of GPS IWV with the Payerne radiosonde station: (a) Jan 2001–Jun 2003 IWV from GPS (solid line) available every 1 h and radiosonde (asterisks) available every 12 h (0000 and 1200 UTC), (b) IWV difference (GPS minus radiosonde) at 0000 UTC, and (c) IWV difference at 1200 UTC.

Citation: Journal of Applied Meteorology 44, 7; 10.1175/JAM2255.1

Fig. 3.
Fig. 3.

Validation of GPS IWV with the MWR data for Bern in the period of Mar–Jul 2001: (a) IWV from GPS (solid line) and MWR (squares), (b) IWV difference (GPS minus MWR) at 0000 UTC, and (c) IWV difference at 1200 UTC.

Citation: Journal of Applied Meteorology 44, 7; 10.1175/JAM2255.1

Fig. 4.
Fig. 4.

Same as Fig. 3, but for Jan–Jun 2003.

Citation: Journal of Applied Meteorology 44, 7; 10.1175/JAM2255.1

Fig. 5.
Fig. 5.

Monthly IWV bias (black bars) and std dev (gray bars) in the period from Jan 2001 to Jun 2003 averaged over 14 Swiss GPS sites for the (a) aLMo analysis (0000–2300 UTC) and (b) aLMo forecast (0 to +23 h). Note that the monthly mean IWV from GPS (white bars) should be multiplied by 10.

Citation: Journal of Applied Meteorology 44, 7; 10.1175/JAM2255.1

Fig. 6.
Fig. 6.

IWV diurnal cycle averaged over 14 GPS sites (aLMo points) in the period (a) May–Oct 2001 from the aLMo forecast (dashed line with circles) and from GPS (solid line) and (b) May–Oct 2002 from the aLMo analysis (dashed line with diamonds) and forecast (dashed line with circles) and from GPS (solid line). The model exhibits a constant dry bias in 2002 of approximately −2.5 kg m−2.

Citation: Journal of Applied Meteorology 44, 7; 10.1175/JAM2255.1

Table 1.

Station name, height above mean sea level (MSL) of the GPS station, the aLMo grid point, the GPS–aLMo difference, data availability, and distance to the meteorological sensor.

Table 1.
Table 2.

GPS–aLMo relative bias from the analysis and forecast.

Table 2.
Save