• 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., , G. Guerova, , and M. Troller. 2002. Swiss activities in combining GPS with meteorology. Mitteilungen des Bundesamtes für Kartographie und Geodäsie, J. A. Torres and H. Hornik, Eds., EUREF Publ. 10, Vol. 23, Frankfurt, Germany 95–99.

    • Search Google Scholar
    • Export Citation
  • Doms, G., and U. Schaettler. Eds.,. 1999. The nonhydrostatic limited-area model LM (Local Model) of DWD. Part I: Scientific documentation. Deutscher Wetterdienst Rep. LM F90 1.35, 172 pp. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

    • Search Google Scholar
    • Export Citation
  • Doms, G., and U. Schaettler. Eds.,. 2001. COSMO Newsletter. No. 1, 114 pp. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

    • Search Google Scholar
    • Export Citation
  • 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
  • Flores, A., , G. Ruffini, , and A. Ruis. 2000. 4D tropospheric tomography using GPS slant wet delays. Ann. Geophys. 18:223234.

  • Gassmann, A. 2001. Filtering of LM-orography. COSMO Newsletter, No. 1, G. Doms and U. Schaettler, Eds., 71–78. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

    • Search Google Scholar
    • Export Citation
  • Güldner, J., and D. Spänkuch. 1999. Results of year-round remotely sensed integrated water vapor by ground-based microwave radiometry. J. Appl. Meteor. 38:981988.

    • Search Google Scholar
    • Export Citation
  • Ingold, T., , B. Schmid, , C. Matzler, , P. Demoulin, , and N. Kampfer. 2000. Modeled and empirical approaches for retrieving columnar water vapor from solar transmittance measurements in the 0.72, 0.82, and 0.94 μm absorption band. J. Geophys. Res. 105:2432724343.

    • Search Google Scholar
    • Export Citation
  • Kuo, Y-H., , X. Zou, , and Y. R. Guo. 1996. Variational assimilation of precipitable water using a nonhydrostatic mesoscale adjoint model. Part I: Moisture retrieval and sensitivity experiments. Mon. Wea. Rev. 124:122147.

    • Search Google Scholar
    • Export Citation
  • Liou, Y-A., , Y-T. Teng, , T. van Hove, , and J. C. Liljegren. 2001. Comparison of precipitable water observations in the near Tropics by GPS, microwave radiometer, and radiosondes. J. Appl. Meteor. 40:515.

    • Search Google Scholar
    • Export Citation
  • MacDonald, A. E., and Y. Xie. 2000. On the use of slant observations from GPS to diagnose three dimensional water vapor using 3DVAR. Preprints, Fourth Symp. on Integrated Observing Systems, Long Beach, CA, Amer. Meteor. Soc. 6273.

    • Search Google Scholar
    • Export Citation
  • MacDonald, A. E., , Y. Xie, , and R. Ware. 2002. Diagnosis of three-dimensional water vapor using a GPS network. Mon. Wea. Rev. 130:386397.

    • Search Google Scholar
    • Export Citation
  • Majewski, D. 1985. Balanced initial and boundary values for limited area model. Contrib. Atmos. Phys. 58:147159.

  • Niell, A. E. 1996. Global mapping functions for the atmosphere delay at radio wavelengths. J. Geophys. Res. 101:32273246.

  • 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
  • Richner, H. 1999. Grundlagen aerologischer Messungen speziell mittels der Schweizer Sonde SRS 400. SMA-MeteoSchweiz Publ. 61, 140 pp. [Available from SMA-MeteoSchweiz, Kraehbuehlstrasse 58, CH-8044 Zurich, Switzerland.].

    • Search Google Scholar
    • Export Citation
  • Rocken, C. R., , R. Ware, , T. van Hove, , F. Solheim, , C. Alber, , J. Johnson, , M. Bevis, , and S. Businger. 1993. Sensing atmospheric water vapor with the global positioning system. Geophys. Res. Lett. 20:26312634.

    • Search Google Scholar
    • Export Citation
  • Rocken, C. R., , T. van Hove, , J. Johnson, , F. Solheim, , R. Ware, , M. Bevis, , S. Chiswell, , and S. Businger. 1995. GPS/STORM—GPS sensing of atmospheric water vapor for meteorology. J. Atmos. Oceanic Technol. 12:468478.

    • Search Google Scholar
    • Export Citation
  • Rothacher, M., and L. Mervart. 1996. Bernese GPS software version 4.0. Astronomical Institute, University of Bern, 418 pp.

  • Schneider, D., , E. Brockmann, , U. Marti, , A. Schlatter, , and U. Wild. 2000. Introduction of a precise Swiss Positioning Service “swipos” and progress in the Swiss national height network (LHN95). Mitteilungen des Bundesamtes für Kartographie und Geodäsie, J. A. Torres and H. Hornik, Eds., EUREF Publ. 9, Frankfurt, Germany, 315–322.

    • Search Google Scholar
    • Export Citation
  • Schraff, C. 1997. Mesoscale data assimilation and prediction of low stratus in the Alpine region. Meteor. Atmos. Phys. 64:2150.

  • Stauffer, D. R., and N. L. Seaman. 1994. Multiscale four-dimensional data assimilation. J. Appl. Meteor. 33:416434.

  • Ware, R. H. Coauthors,. 2000. SuomiNet: A real-time national GPS network for atmospheric research and education. Bull. Amer. Meteor. Soc. 81:677694.

    • Search Google Scholar
    • Export Citation
  • Weygandt, S. S., and N. L. Seaman. 1994. Quantification of predictive skill for mesoscale and synoptic-scale meteorological features as a function of horizontal grid resolution. Mon. Wea. Rev. 122:5771.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Map of Switzerland showing a selection of stations (white dots) from AGNES. In this study the data from Payerne (PAYE), Bern (EXWI), Jungfraujoch (JUJO), Andermatt (ANDE), Locarno (LOMO), and Davos (DAVO) were used

  • View in gallery

    (a) ZTD and ZHD vs time for the period 10–30 Nov 2000 at PAYE. (b) IWV (mm) at Payerne as derived from Eq. (3)

  • View in gallery

    Radiosonde and GPS-derived IWV in Jan 2001: (a) Radiosonde (squares), GPS observation (thick line), and LM forecast of 0000 UTC +24 to +48 h (thin line) with circle indicating forecast hour +24 at PAYE. (b) The same as (a) but for EXWI with radiosonde data from Payerne

  • View in gallery

    Histograms of the bias (radiosonde − GPS) and the GPS IWV for period Nov–Mar at PAYE: (a) Nov 2000, (b) Dec 2000, (c) Jan 2001, and (d) Mar 2001

  • View in gallery

    IWV from GPS and LM vs time in days of Dec 2000: (a) IWV from GPS (thick line), radiosonde (squares), and LM forecast of 0000 UTC +6 to +30 h (thin line) with circle indicating forecast hour +6 for PAYE (∼500 m MSL), (b) the same as (a) but for EXWI (∼550 m MSL), (c) the same as (a) but for DAVO (∼1600 m MSL), and (d) the same as (a) but for ANDE (∼2300 m MSL)

  • View in gallery

    Monthly mean IWV from GPS (black bar) and LM (gray bar) for the period Nov 2000–Mar 2001

  • View in gallery

    Comparison between GPS (thick line) and SM forecast of 0000 UTC +6 to +30 h (thin line) for summer period 10–20 August 2000: (a) IWV as measured and modeled at EXWI (∼550 m MSL), (b) IWV as measured and modeled at DAVO (∼1600 m MSL), (c) IWV as measured and modeled at ANDE (∼2300 m MSL)

  • View in gallery

    GPS IWV for Payerne, Andermatt, and Davos: (a) IWV at Payerne (PAYE: dotted solid line) and Andermatt (ANDE: circles) between 26 and 30 Nov 2000 and (b) IWV at Payerne (PAYE: dotted solid line) and Davos (DAVO: squares)

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 141 141 11
PDF Downloads 17 17 7

Validation of NWP Mesoscale Models with Swiss GPS Network AGNES

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

The importance of water vapor for the hydrological cycle of the earth and the atmosphere is well known but difficult to study and sample. In this respect, the vertically integrated water vapor (IWV) derived from global positioning system (GPS) delay is a potential source of valuable weather information. Because of the relatively dense station distribution, both the temporal and horizontal variability of water vapor are well captured by the GPS. This makes GPS data well suited for numerical weather prediction (NWP) models. In this paper the automated GPS network of Switzerland (AGNES) is used for calculation of IWV and for comparison with radiosonde data and two NWP mesoscale models from MeteoSwiss named Local Model (LM) and Swiss Model (SM). Reasonably good agreement between GPS and radiosonde data is reported. It has been identified that in some particular weather situations with low stratus clouds and temperature inversion, the radiosonde significantly overestimates the water vapor amount. The LM and SM verification with GPS data indicates good agreement during the winter period (November 2000–March 2001) and high variability and bias in the summer period (August 2000). The monthly mean IWV values from GPS and LM show a systematic deviation over the Swiss plateau region and a very good agreement for the high-altitude alpine station, Andermatt. The capability of GPS in monitoring the atmospheric phenomena has been demonstrated. Unrealistic IWV at Jungfraujoch (∼3600 m MSL) caused by GPS antenna problems is reported.

Corresponding author address: Dr. Guergana Guerova, Institute of Applied Physics, University of Bern, Sidlerstrasse 5, Bern CH-3012, Switzerland. guergana.guerova@mw.iap.unibe.ch

Abstract

The importance of water vapor for the hydrological cycle of the earth and the atmosphere is well known but difficult to study and sample. In this respect, the vertically integrated water vapor (IWV) derived from global positioning system (GPS) delay is a potential source of valuable weather information. Because of the relatively dense station distribution, both the temporal and horizontal variability of water vapor are well captured by the GPS. This makes GPS data well suited for numerical weather prediction (NWP) models. In this paper the automated GPS network of Switzerland (AGNES) is used for calculation of IWV and for comparison with radiosonde data and two NWP mesoscale models from MeteoSwiss named Local Model (LM) and Swiss Model (SM). Reasonably good agreement between GPS and radiosonde data is reported. It has been identified that in some particular weather situations with low stratus clouds and temperature inversion, the radiosonde significantly overestimates the water vapor amount. The LM and SM verification with GPS data indicates good agreement during the winter period (November 2000–March 2001) and high variability and bias in the summer period (August 2000). The monthly mean IWV values from GPS and LM show a systematic deviation over the Swiss plateau region and a very good agreement for the high-altitude alpine station, Andermatt. The capability of GPS in monitoring the atmospheric phenomena has been demonstrated. Unrealistic IWV at Jungfraujoch (∼3600 m MSL) caused by GPS antenna problems is reported.

Corresponding author address: Dr. Guergana Guerova, Institute of Applied Physics, University of Bern, Sidlerstrasse 5, Bern CH-3012, Switzerland. guergana.guerova@mw.iap.unibe.ch

Introduction

The importance of water vapor for the overall state of the atmosphere in both short and long timescales has been acknowledged for many years, and a number of instrumental techniques have been developed to study it. While the most widely used method worldwide, radiosonde, is superior in providing vertically resolved water vapor profiles it is, on the other hand, relatively expensive to maintain and consequently limited in its number of sites. Microwave radiometry is another approach to the problem, providing information for vertically integrated water vapor (IWV) and cloud water. In the last two decades it has been substantially developed and is extensively used for validation studies and field campaigns (Liou et al. 2001). However, in addition to the difficulty in data processing, its main limitations remain poor spatial coverage and poor performance in rain conditions (Güldner and Spänkuch 1999). The prime objective of the global positioning system (GPS), developed three decades ago, was exact position determination (i.e., location on the earth's surface). Later, the use of GPS was extended to a number of scientific applications (Ware et al. 2000), one of which is derivation of IWV. It has to be acknowledged, however, that a significant effort was required to achieve accuracy in determination of water vapor from the GPS signal, and the possibility of producing a profile from the GPS observation is still a challenging task (Flores et al. 2000; MacDonald and Xie 2000; MacDonald et al. 2002).

On the other hand, the development of very fine resolution mesoscale models for numerical weather prediction (NWP) became possible with the improvements of both numeric and parameterization schemes and the increased power of supercomputers. A nonhydrostatic concept has been introduced in the last decade in order to achieve, among other things, the refinement of the model mesh and to provide more realistic parameterization of atmospheric waves and convection (Doms and Schaettler 1999). When mesoscale models are used to resolve boundary layer structure and dynamics, the need for reliable water vapor data is even more apparent (Schraff 1997; Weygandt and Seaman 1994; Stauffer and Seaman 1994). In addition, for realistic representation of cumulus cloud formation and decay, good parameterization schemes are required, along with as accurate as possible estimates of water vapor profiles. This is the main motivation for the implementation of GPS-derived IWV in the mesoscale models (Kuo et al. 1996).

The objective of this study is to validate the water vapor field of two operational mesoscale models against GPS-observed IWV and to draw conclusions about future assimilation of IWV in the Local Model (LM). In section 2 a description of IWV extraction from GPS, radiosonde, LM, and Swiss Model (SM) is given. The comparison between GPS and radiosonde IWV is discussed in section 3 and LM validation with GPS for the period November 2000–March 2001 is presented. For summer 2000, SM output is compared with IWV from GPS. A summary and conclusions can be found in section 4.

Description of the data used

Global positioning system data

Swiss GPS network AGNES

The Swiss Federal Office of Topography has deployed and operates the automated GPS network of Switzerland—AGNES. AGNES is mainly used for navigation and surveying purposes (postprocessing and real time) in a homogeneous reference frame (called LV95). It consists of 12 permanent receivers with a sampling rate of 1 s and contained 28 stations at the end of 2001 (Brockmann et al. 2002). The data from all stations are processed using the Bernese 4.2 software package (Rothacher and Mervart 1996; Schneider et al. 2000) with a delay of one week. For estimation of hourly zenith total delay (ZTD), GPS data above a 10° elevation are used. The Niell mapping function (Niell 1996) is used to map the slant delays to the zenith. In this study the data from six AGNES sites are used. Their locations are represented by white dots in Fig. 1. They are, from west to east, Payerne (PAYE), Bern (EXWI), Jungfraujoch (JUJO), Andermatt (ANDE), Locarno (LOMO), and Davos (DAVO).

Derivation of IWV from GPS measurements

The IWV is retrieved from the ZTD following the concept described in Bevis et al. (1992) and Emardson et al. (1998). First, the zenith hydrostatic delay (ZHD) (m) is calculated through its dependence on surface pressure ps (hPa) and a factor f(θ, h), describing the height (h) and latitude (θ) variation of the gravitational acceleration, as follows:
i1520-0450-42-1-141-e1
Second, the IWV (mm) is obtained using
i1520-0450-42-1-141-e3
where k3 = (3.776 ± 0.004) × 105 (K hPa−1), k2 = (17 ± 10) (K hPa−1), Rυ = 461.51 (J kg−1 K−1) is the specific gas constant for water vapor, and Tm is the weighted mean atmospheric temperature (K). The surface pressure (ps) and temperature (used for derivation of Tm) are obtained from the collocated meteorological stations from the Swiss surface station network (ANETZ).

The difference (ZTD − ZHD) is often referred to as zenith wet delay (Bevis et al. 1992). The conversion constant K is on the order of 0.15. The uncertainty in derivation of ZHD [specified in Eq. (1)] is on the order of 0.2%. However, the uncertainty introduced in IWV varies from ±0.25 to ±0.35 mm for IWV amounts of 10 mm and larger than 10 mm, correspondingly. The uncertainty in estimation of the parameter K [specified in Eq. (4)] is reported to be 2% (Ohtani and Naito 2000). The overall accuracy of IWV estimation is reported to be better than 1.5 mm (Rocken et al. 1993, 1995).

One demonstration of the GPS concept can be found in Fig. 2 where ZTD, ZHD, and IWV are plotted for the GPS station PAYE. From the first plot (Fig. 2a), it can be clearly seen that the ZHD variation for the period of 20 days is between 2.15 and 2.20 m (2150–2200 mm). The ZTD, however, varies in the range of 2.20–2.35 m (2200–2350 mm). It is obvious that the high time fluctuation in the ZTD signal has less to do with the hydrostatic component of the delay (ZHD), but is essentially contributed by the wet delay. Thus, being in the range of 0.05–0.15 m, the wet delay induced by the water vapor content of the lower troposphere is a prime source of high temporal variability in the GPS signal. Indeed, the dashed and dash–dotted lines give evidence for the good correspondence between the low and high values of IWV (Fig. 2b) and ZTD (Fig. 2a).

Numerical weather prediction models

Local model

The increase in the capability and speed of supercomputers has provided an opportunity for the implementation of a new generation of high-resolution nonhydrostatic models for mesoscale numerical weather prediction. At the Federal Office of Meteorology and Climatology (MeteoSwiss) a nonhydrostatic model, called the Local Model, has been running operationally twice every day (at 0000 and 1200 UTC) since 1 April 2001, with a preoperational test period from September 2000 to March 2001. The model horizontal grid of 7 km × 7 km covers the region of central and southwestern Europe and part of the Atlantic Ocean. The model has 45 vertical levels from the surface up to 20 hPa in a generalized terrain-following coordinate system (Doms and Schaettler 1999, 2001). The vertical resolution in the lowest 2 km of the atmosphere is about 100 m. Since November 2000, filtered orography was introduced into the model to solve for unrealistic precipitation forecasts in mountain regions (Gassmann 2001). The LM analyses are obtained via interpolation of driving model analyses—the German Global Model (GME). The model prognostic variables are temperature, perturbation pressure, horizontal and vertical wind velocity, water vapor, cloud water, and air density. The LM has been developed in close cooperation with Deutscher Wetterdienst (DWD) within the framework of the Consortium of Small-Scale Modelling (COSMO; Doms and Schaettler 2001), which integrates the national weather services of Germany, Switzerland, Italy, and Greece.

Swiss model

The Swiss Model is a hydrostatic limited-area mesoscale model. It has a horizontal resolution of 14 km and a grid mesh of 145 × 145 points. The domain covers western and central Europe (from Ireland, Denmark, and Poland in the north, to Spain and southern Italy in the south). The vertical domain is divided into 31 layers. It has been used for operational weather forecasts by MeteoSwiss since 1994 and it has been developed in collaboration with DWD. The SM analysis is obtained via interpolation of the GME analysis using the method of Majewski (1985). The SM data were used for GPS validation in summer 2000.

Derivation of IWV from LM and SM

The IWV is calculated from the model specific humidity fields so that it can be compared with the GPS observations. For the comparison, the model grid points have been selected based on the smallest height difference to the GPS antenna. The spatial search radius is about 10 and 20 km, correspondingly, for LM and SM. In addition, the starting integration level is taken to be as close as possible to the height of the GPS site; that is, the model IWV is calculated from the level with smallest height difference. The integration is performed using
i1520-0450-42-1-141-e5
where ρwv is water vapor density, h is height in meters, and ρw is density of liquid water, and IWV is in millimeters. For intercomparison with GPS the LM forecast initialized at 0000 UTC is used. The forecast hours from +6 to +30 h are compared with corresponding GPS measurements. In only one case the LM forecast period from +24 to +48 h (next-day forecast) is used.

Radiosonde data

There is one radiosonde sounding site in Switzerland located at Payerne (Swiss plateau region). A balloon sounding (sonde type SRS 400, MeteoLabor, Switzerland) is performed twice a day (0000 and 1200 UTC) measuring temperature, pressure, humidity, and wind profiles. The IWV amount is calculated using Eq. (5).

A fast response VIZ 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).

Validation study

Comparison of IWV derived from GPS and radiosonde

In order to validate the GPS IWV, a comparison with the collocated radiosonde in Payerne is performed. The monthly mean difference between radiosonde and GPS over the period of November 2000–March 2001 is in the range of 0.27–0.64 mm, showing a slight dry bias of GPS. This dry GPS bias, however, is below the estimated GPS error range of ±1 mm. From this comparison it was identified that during some particular weather situations in winter (low stratus clouds and temperature inversion), when very low amounts of water vapor were measured by GPS, the radiosonde measurements had a significant positive bias. Such a case is presented in Figs. 3a and 3b. It can be clearly seen that between 14 January and 16 January the radiosonde (squares) and GPS IWV (thick line) have a significant offset for the GPS antenna in PAYE as well as for the GPS site EXWI in Bern (the distance between Payerne and Bern is about 40 km). The same situation was observed on 23 December 2000 (Fig. 5a).

The LM forecast (thin line) is also plotted in Figs. 3a and 3b. The model forecast from +24 to +48 h is chosen in order to minimize the observation (e.g., assimilated radiosonde data) influence. It can be observed that the model predicts IWV amounts that are close to the GPS-derived one and, as well, below 5 mm. The histograms in Fig. 4 show the bias (radiosonde − GPS) and the GPS-derived IWV for four months in the period from November 2000 to March 2001 (the data from February are not presented because the number of comparisons is small). From the histograms in Figs. 4b and 4c, the periods with very low IWV at the end of December and mid-January can be identified. It is to be noticed that, for these particular cases, the bias can reach up to 5 mm. It can also be seen that the bias is mostly positive and is particularly high for IWV amounts below or close to 10 mm. For example, from 17 cases in December when the difference is in the range of ±1 to ±2 mm, 12 cases have a positive sign (excluding the already discussed case with very low IWV at the end of December).

Independent measurements of IWV at Bern by a sunphotometer (Ingold et al. 2000) were available on 15 January at 1300 UTC. The IWV from the sunphotometer is 5.23 mm and the amount from the collocated GPS site EXWI is 5.0 mm, that is, the difference is about 0.3 mm. At approximately the same time (1200 UTC) in Payerne, the GPS-derived IWV is 4.2 mm while the radiosonde measures 6.66 mm, that is, 2.4 mm more.

Verification of LM water vapor field with GPS observations

Since November 2000, the verification of the Local Model has begun. This work is a first step in investigating the potential use of ZTD measurements in numerical weather prediction in Switzerland. Comparisons of hourly IWV from LM and GPS are performed. Results for four GPS sites and corresponding model grid points are presented in Fig. 5. The site selection is based on availability of collocated surface meteorological measurements from ANETZ.

The period of 5–31 December 2000 is plotted in Fig. 5. Additionally the corresponding bias (LM − GPS) and standard deviation (std dev) values are given. With only one exception (PAYE) the bias is well below 1 mm. However, some days with substantial differences between LM IWV (thin line) and GPS IWV (thick line) are observed. One case is the +6 to +30 h model forecast on 13 December (+6 h is marked by an arrow on Fig. 5a). For Payerne (Fig. 5a), a substantial offset is visible toward the end of the forecasting period, while for Andermatt (Fig. 5d) and Davos (Fig. 5c) the IWV variation is poorly predicted over the entire forecast period. A careful investigation of the 1200 UTC surface pressure map from 13 December indicates a cold front passing through Switzerland. Indeed, the IWV minimum is associated with the cold-air advection. The LM forecast for Andermatt and Davos correctly predicts the minimum IWV values but with time offsets of 9 and 12 h, respectively. Thus, it is to be concluded that the forecast failure is due to inaccurate modeling of the atmospheric processes after frontal zone passage.

The second half of the month shows reasonably good agreement except for the forecast on 23 December where the very first forecast hours (+6 to +12 h) are particularly poorly predicted for the Payerne grid point. The reason is once again humidity overestimation from the radiosonde (squares in Fig. 5a). This demonstrates the model sensitivity to the radiosonde data. It must be emphasized that in general, good agreement between the model and GPS was obtained during the winter period under study.

Monthly mean water vapor amounts from the LM and GPS

Monthly mean IWV estimates from LM and GPS are presented in Fig. 6 using 1-h time-resolution data. The sites are as follows: Payerne (PAYE) and Bern (EXWI) located on the Swiss plateau, Andermatt (ANDE), Davos (DAVO), and Jungfraujoch (JUJO) located in the Swiss Alps, and Locarno (LOMO) in southern Switzerland. In Table 1, the heights above sea level (MSL) of GPS antenna and model topography are listed.

The results over the monitored period from November 2000 to March 2001, shown in Fig. 6, reveal that GPS (black bar) and LM (gray bar) have good correspondence. It can be clearly seen that the lowest IWV measured at PAYE and EXWI is in the middle of the period, namely, in January 2001. However, it should be noted that for Payerne and Bern the amounts from LM are always higher than those from GPS. Keeping in mind that the height differences between the model grid point and the GPS antenna are 30 and 79 m (see Table 1), respectively, it is logical that the forecast IWV amount should be lower than that measured by the GPS. This leads to the conclusion that LM overestimates the water vapor amount in the Swiss plateau region. A further comparison between LM, radiosonde, and GPS shows a positive model bias of 0.8 mm relative to the radiosonde and 1.1 mm relative to the GPS for November 2000 at Payerne. This trend remains valid for the entire period of investigation. From the third set of bars in Fig. 6 it is seen that the GPS receiver in DAVO measures slightly more water vapor in all cases when compared with the values obtained from LM (gray bars). The excess amount reported by the GPS varies from 3.7% to 7.7% and correlates well with the fact that the height difference is a little more than 200 m. For the station at ANDE the bars show very good agreement between GPS and LM. In addition it is to be noted that the values are below 6 mm because of the high altitude of the site (2300 m). On the other hand this good agreement is an indication of correct representation of WV amounts in the model layers above 2300 m.

The results for JUJO (3600 m MSL) show strong discrepancy between LM and GPS. The reason is unrealistic IWV amounts from GPS (including negative estimates). The problem was identified to be the unmodeled antenna phase center of this special antenna, which is covered by a heated dome to protect the antenna against the impact of snow. After detection of the problem the antenna was calibrated, and from now on, more realistic values of GPS IWV at Jungfraujoch are expected.

The last set of bars in the histogram (Fig. 6) is for LOMO, the southernmost GPS site in Switzerland. The IWV minima for LOMO are not measured in January 2001 (as they are for Payerne) but instead in February 2001. This is probably due to the fact that the atmospheric circulation in this region is influenced by the Mediterranean circulation patterns. Unfortunately the model orography is 400 m above the antenna height, so any conclusions about the model WV field are not feasible. However, it should be pointed out that the 400-m difference adds from 6.5% to 16.4% to the overall amount reported by LM.

In order to solve the problem of height difference between GPS and LM grid points (e.g., Davos and Locarno), an additional experiment was performed. First, the IWV amounts for a layer thickness of 200 and 400 m are calculated, assuming constant relative humidity (observed at ANETZ station height). Second, the GPS IWV values are extrapolated to the corresponding LM height by subtracting IWV amounts due to the height difference. As a result, a substantial decrease of the GPS value is obtained with a negative offset as compared with the LM (opposite to the one plotted on Fig. 6). It is to be concluded that such a simple approach does not provide representative validation results. The reason is mainly the complex spatial and vertical variability of water vapor in the lower troposphere. Therefore, the assimilation of GPS data from stations with substantial height differences could produce questionable results and should be either avoided or treated with special care. This problem is particularly important for Switzerland where the model orography poorly matches steep slopes and narrow valleys in the Alps.

Investigation of summer water vapor with the Swiss Model

Because the LM forecast was not available in the summer of 2000, the hydrostatic Swiss Model was used for comparison with GPS. The summer periods are of particular interest for countries like Switzerland because they are characterized with high IWV amounts and strong cumulus convection and precipitation. One further reason is that the SM has the same parameterization schemes as LM, the main differences being the doubled mesh distance and noncompressibility of the flow in SM. For the summer period of 2000, the ZTD data derived by GPS are available with a 2-h time resolution. No measurements were collected at the Payerne GPS station because the GPS receiver was used in a field campaign. In Fig. 7 a comparison between SM (thin line) and GPS (thick line) IWV is plotted for 10 days in mid-August 2000 for Bern (Fig. 7a), Davos (Fig. 7b), and Andermatt (Fig. 7c). The day-to-day variation of IWV is poorly modeled and differences of 5 mm and more occur. The bias and std dev values also indicate increased variability. For the GPS station at Davos, the bias in the summer period is twice that in the winter period. For completeness the radiosonde data are plotted in Fig. 7a, and it is to be noted that they are in relatively good agreement with the GPS data.

The SM validation over the winter period of 2000/01 showed similar results to the one already discussed in section 3b (LM verification). Therefore, it may be concluded that it is more difficult to accurately predict the IWV amount in summer.

Tracking regional and temporal variability of IWV

The potential of GPS-retrieved water vapor as a valuable source of information for mesoscale numerical modeling lies in its ability to track short-term local variations in the water vapor field. This is illustrated in Fig. 8, where the IWV amounts are plotted for the period of 26–30 November 2000 with 1-h time resolution. It is observed that at midday on 27 November, a gradual increase in IWV is initiated in Payerne (dotted solid line on both Figs. 8a and 8b). The maximum amount is reached on 0000 UTC of 28 November. As compared with Payerne, a short delay of 1 h is observed for the IWV maxima at Andermatt. The maxima at Davos (squares on Fig. 8b) has a 3-h delay relative to Payerne. A careful investigation of the weather situation on this day confirms that this time delay correlates well with a warm atmospheric front passing from west to east. As seen from Fig. 1, Payerne is located in the west part of Switzerland and the distances to Andermatt and Davos are 120 and 220 km, respectively. The sensitivity of the GPS IWV to the atmospheric phenomena on short temporal and spatial scales can be observed over the entire period presented. The daily cycle of IWV is a topic of further interest, related to the IWV dynamics in the boundary layer, that could be studied with GPS; however, this has not yet been done.

Conclusions

Atmospheric water vapor has well-pronounced temporal and spatial variability, which is far from being satisfactorily determined. Humidity data are particularly important for a country like Switzerland because it is a “crossroad” for different weather circulation patterns. Thus, monitoring IWV in the atmosphere using the automated GPS network of Switzerland, AGNES, provides a unique opportunity to gain insight into the local variation of water vapor. It also proves to be a good tool in verifying the NWP models' performance. In this paper, one goal is to validate the GPS-derived IWV with the IWV from the collocated radiosonde station in Payerne. The agreement of monthly mean difference is in the range of 0.27–0.64 mm over the period from November 2000 to March 2001. In some cases of low stratus cloud and temperature inversion the radiosonde was found to overestimate the water vapor amount. One possible reason is the slow recovery time of the humidity sensor after cloud passage. The validation of the NWP mesoscale models was performed for GPS sites at Payerne, Bern, Davos, and Andermatt. Comparison of LM-forecasted IWV with GPS data in December shows very good agreement with some exceptions where substantial differences are observed indicating partial failure of model forecasts. The LM monthly mean IWV amounts indicate systematic overestimation for the Swiss plateau region and agreement for the high-altitude station at Andermatt (∼2300 m MSL). The IWV variation during the summer period is relatively poorly matched by the SM forecast. The bias and standard deviation are significant in summer. The GPS data sensitivity to high-temporal-resolution atmospheric phenomena, that is, atmospheric fronts, has been demonstrated. The passage of a warm front from west to east has been traced. It introduces time shifts of the IWV maxima, with respect to Payerne, of 1 and 3 h for Andermatt and Davos, respectively. The IWV data from the alpine station Jungfraujoch (∼3600 m MSL) give unrealistic, that is, some negative, values due to the unmodeled effects of this particular antenna. Based on the study of the impact of height difference it is concluded that care should be taken when assimilating GPS data from stations with model-to-station height differences of more than 200 m: for instance, Davos and Locarno.

As demonstrated in this paper, the GPS-retrieved IWV is a useful tool for monitoring atmospheric humidity. Although humidity information provided by GPS is not in the form of a profile, its temporal and horizontal coverage is better than that provided by the meteorological sounding network. If treated properly, the GPS information could fill the existing gap in model data input. This is the main motivation to continue the work for more efficient use of GPS data in LM. It is planned to start assimilation via nudging of GPS ZTD. The assimilation work will be performed in close cooperation with the DWD, Swiss Federal Office of Topography, and MeteoSwiss and will be part of the Swiss contribution to COST Action 716.

Acknowledgments

This work is supported by the Swiss Federal Office of Education and Science under Grant C97.0027. It is in the frame of the Swiss contribution to COST Action 716 and is considered a first step in preparation for assimilation of GPS measurements in LM.

REFERENCES

  • 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., , G. Guerova, , and M. Troller. 2002. Swiss activities in combining GPS with meteorology. Mitteilungen des Bundesamtes für Kartographie und Geodäsie, J. A. Torres and H. Hornik, Eds., EUREF Publ. 10, Vol. 23, Frankfurt, Germany 95–99.

    • Search Google Scholar
    • Export Citation
  • Doms, G., and U. Schaettler. Eds.,. 1999. The nonhydrostatic limited-area model LM (Local Model) of DWD. Part I: Scientific documentation. Deutscher Wetterdienst Rep. LM F90 1.35, 172 pp. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

    • Search Google Scholar
    • Export Citation
  • Doms, G., and U. Schaettler. Eds.,. 2001. COSMO Newsletter. No. 1, 114 pp. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

    • Search Google Scholar
    • Export Citation
  • 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
  • Flores, A., , G. Ruffini, , and A. Ruis. 2000. 4D tropospheric tomography using GPS slant wet delays. Ann. Geophys. 18:223234.

  • Gassmann, A. 2001. Filtering of LM-orography. COSMO Newsletter, No. 1, G. Doms and U. Schaettler, Eds., 71–78. [Available from Deutscher Wetterdienst, P.O. Box 100465, 63004 Offenbach, Germany.].

    • Search Google Scholar
    • Export Citation
  • Güldner, J., and D. Spänkuch. 1999. Results of year-round remotely sensed integrated water vapor by ground-based microwave radiometry. J. Appl. Meteor. 38:981988.

    • Search Google Scholar
    • Export Citation
  • Ingold, T., , B. Schmid, , C. Matzler, , P. Demoulin, , and N. Kampfer. 2000. Modeled and empirical approaches for retrieving columnar water vapor from solar transmittance measurements in the 0.72, 0.82, and 0.94 μm absorption band. J. Geophys. Res. 105:2432724343.

    • Search Google Scholar
    • Export Citation
  • Kuo, Y-H., , X. Zou, , and Y. R. Guo. 1996. Variational assimilation of precipitable water using a nonhydrostatic mesoscale adjoint model. Part I: Moisture retrieval and sensitivity experiments. Mon. Wea. Rev. 124:122147.

    • Search Google Scholar
    • Export Citation
  • Liou, Y-A., , Y-T. Teng, , T. van Hove, , and J. C. Liljegren. 2001. Comparison of precipitable water observations in the near Tropics by GPS, microwave radiometer, and radiosondes. J. Appl. Meteor. 40:515.

    • Search Google Scholar
    • Export Citation
  • MacDonald, A. E., and Y. Xie. 2000. On the use of slant observations from GPS to diagnose three dimensional water vapor using 3DVAR. Preprints, Fourth Symp. on Integrated Observing Systems, Long Beach, CA, Amer. Meteor. Soc. 6273.

    • Search Google Scholar
    • Export Citation
  • MacDonald, A. E., , Y. Xie, , and R. Ware. 2002. Diagnosis of three-dimensional water vapor using a GPS network. Mon. Wea. Rev. 130:386397.

    • Search Google Scholar
    • Export Citation
  • Majewski, D. 1985. Balanced initial and boundary values for limited area model. Contrib. Atmos. Phys. 58:147159.

  • Niell, A. E. 1996. Global mapping functions for the atmosphere delay at radio wavelengths. J. Geophys. Res. 101:32273246.

  • 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
  • Richner, H. 1999. Grundlagen aerologischer Messungen speziell mittels der Schweizer Sonde SRS 400. SMA-MeteoSchweiz Publ. 61, 140 pp. [Available from SMA-MeteoSchweiz, Kraehbuehlstrasse 58, CH-8044 Zurich, Switzerland.].

    • Search Google Scholar
    • Export Citation
  • Rocken, C. R., , R. Ware, , T. van Hove, , F. Solheim, , C. Alber, , J. Johnson, , M. Bevis, , and S. Businger. 1993. Sensing atmospheric water vapor with the global positioning system. Geophys. Res. Lett. 20:26312634.

    • Search Google Scholar
    • Export Citation
  • Rocken, C. R., , T. van Hove, , J. Johnson, , F. Solheim, , R. Ware, , M. Bevis, , S. Chiswell, , and S. Businger. 1995. GPS/STORM—GPS sensing of atmospheric water vapor for meteorology. J. Atmos. Oceanic Technol. 12:468478.

    • Search Google Scholar
    • Export Citation
  • Rothacher, M., and L. Mervart. 1996. Bernese GPS software version 4.0. Astronomical Institute, University of Bern, 418 pp.

  • Schneider, D., , E. Brockmann, , U. Marti, , A. Schlatter, , and U. Wild. 2000. Introduction of a precise Swiss Positioning Service “swipos” and progress in the Swiss national height network (LHN95). Mitteilungen des Bundesamtes für Kartographie und Geodäsie, J. A. Torres and H. Hornik, Eds., EUREF Publ. 9, Frankfurt, Germany, 315–322.

    • Search Google Scholar
    • Export Citation
  • Schraff, C. 1997. Mesoscale data assimilation and prediction of low stratus in the Alpine region. Meteor. Atmos. Phys. 64:2150.

  • Stauffer, D. R., and N. L. Seaman. 1994. Multiscale four-dimensional data assimilation. J. Appl. Meteor. 33:416434.

  • Ware, R. H. Coauthors,. 2000. SuomiNet: A real-time national GPS network for atmospheric research and education. Bull. Amer. Meteor. Soc. 81:677694.

    • Search Google Scholar
    • Export Citation
  • Weygandt, S. S., and N. L. Seaman. 1994. Quantification of predictive skill for mesoscale and synoptic-scale meteorological features as a function of horizontal grid resolution. Mon. Wea. Rev. 122:5771.

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

Map of Switzerland showing a selection of stations (white dots) from AGNES. In this study the data from Payerne (PAYE), Bern (EXWI), Jungfraujoch (JUJO), Andermatt (ANDE), Locarno (LOMO), and Davos (DAVO) were used

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Fig. 2.
Fig. 2.

(a) ZTD and ZHD vs time for the period 10–30 Nov 2000 at PAYE. (b) IWV (mm) at Payerne as derived from Eq. (3)

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Fig. 3.
Fig. 3.

Radiosonde and GPS-derived IWV in Jan 2001: (a) Radiosonde (squares), GPS observation (thick line), and LM forecast of 0000 UTC +24 to +48 h (thin line) with circle indicating forecast hour +24 at PAYE. (b) The same as (a) but for EXWI with radiosonde data from Payerne

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Fig. 4.
Fig. 4.

Histograms of the bias (radiosonde − GPS) and the GPS IWV for period Nov–Mar at PAYE: (a) Nov 2000, (b) Dec 2000, (c) Jan 2001, and (d) Mar 2001

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Fig. 5.
Fig. 5.

IWV from GPS and LM vs time in days of Dec 2000: (a) IWV from GPS (thick line), radiosonde (squares), and LM forecast of 0000 UTC +6 to +30 h (thin line) with circle indicating forecast hour +6 for PAYE (∼500 m MSL), (b) the same as (a) but for EXWI (∼550 m MSL), (c) the same as (a) but for DAVO (∼1600 m MSL), and (d) the same as (a) but for ANDE (∼2300 m MSL)

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Fig. 6.
Fig. 6.

Monthly mean IWV from GPS (black bar) and LM (gray bar) for the period Nov 2000–Mar 2001

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Fig. 7.
Fig. 7.

Comparison between GPS (thick line) and SM forecast of 0000 UTC +6 to +30 h (thin line) for summer period 10–20 August 2000: (a) IWV as measured and modeled at EXWI (∼550 m MSL), (b) IWV as measured and modeled at DAVO (∼1600 m MSL), (c) IWV as measured and modeled at ANDE (∼2300 m MSL)

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Fig. 8.
Fig. 8.

GPS IWV for Payerne, Andermatt, and Davos: (a) IWV at Payerne (PAYE: dotted solid line) and Andermatt (ANDE: circles) between 26 and 30 Nov 2000 and (b) IWV at Payerne (PAYE: dotted solid line) and Davos (DAVO: squares)

Citation: Journal of Applied Meteorology 42, 1; 10.1175/1520-0450(2003)042<0141:VONMMW>2.0.CO;2

Table 1. 

GPS, LM, GPS − LM, and SM heights

Table 1. 
Save