1. Introduction
Atmospheric water vapor significantly influences many processes of the earth’s weather and climate. Water vapor is one of the main variables controlling the greenhouse effect and it plays a crucial role in the global energy cycle. Accurate knowledge of the water vapor distribution in the atmosphere and its change with time is indispensable for the description and understanding of global climate processes. In contrast to other greenhouse gases such as carbon dioxide or methane, water vapor has a much higher temporal and spatial variability. Given the current state of the global upper-atmospheric observing system consisting primarily of radiosondes and satellites, water vapor is generally observed insufficiently compared to other meteorological parameters.
Traditionally, humidity was mainly measured by radiosondes. Today, the radiosonde network consists of more than 900 stations worldwide. However, the radiosonde observations are distributed very inhomogenously. Most of the stations are situated in industrial countries. There are nearly no radiosonde observations over the oceans and in the polar regions. Additionally, the interpretation of the radiosonde data is complicated because of the differences in the instruments used and the applied analysis strategies of the individual countries (Garand et al. 1992). The measurements of the relative humidity from distinct sensors can differ by up to 20%. The systematic error in the relative humidity mainly depends on the calibrations of the humidity sensor performed prior to the start of the radiosonde (Soden and Lanzante 1996). Globally, nearly 40 different types of radiosondes are in use.
Today, numerous remote sensing methods exist that are able to observe water vapor globally. However, each of these methods has its limitations. Since 1979, passive radiometers have been working in the infrared band, but their observations are of low accuracy and limited to clear sky conditions (Chaboureau et al. 1998; Divakarla et al. 2006). Humidity measurements from downward-looking microwave radiometers are independent of the cloud coverage and provide high accuracy over the oceans in nonprecipitating atmospheres (Sun 1993). However, over land and ice surfaces the accuracy of the water vapor estimates from these microwave radiometers is degraded because of assumptions that have to be made about surface emissivity (Miao et al. 2001; Wang and Manning 2003; Karbou et al. 2005; Deeter 2007).
Radio occultations, between low earth-orbiting satellites and satellites of the Global Positioning System (GPS), are independent of both weather and surface conditions. The sounding of upper-tropospheric humidity by GPS radio occultations is of high accuracy but the accuracy decreases in the lower troposphere (e.g., Marquardt et al. 2001; Wickert et al. 2002; Beyerle et al. 2002, 2004; Hajj et al. 2004; Liou et al. 2005; Larsen et al. 2005). However, the lowest 5-km layer of the troposphere includes about 95% of the total atmospheric water vapor. Ground-based GPS measurements have proven to estimate the content of integrated water vapor in an air column very accurately (e.g., Bevis et al. 1992, 1994; Rocken et al. 1995; Tregoning et al. 1998; Basili et al. 2001; Niell et al. 2001). Water vapor estimates from ground-based GPS observations do generally not provide information on the vertical distribution of the water vapor, but they are of high temporal resolution from a few minutes up to 2 h and they are independent of all weather conditions (Yuan et al. 1993; Rocken et al. 1997; Yunck et al. 2000).
Concerning water vapor data, numerical weather prediction (NWP) models were originally based on radiosonde observations. Today satellite observations from infrared and microwave radiometers are also operationally assimilated into the NWP models (Kanamitsu et al. 1991; Kistler et al. 2001; Andersson et al. 2007). Studies on weather and climate change in the last 20 to 40 yr are increasingly based on analyses of NWP models. For example, investigations of the general atmospheric circulation and the feedback mechanism between temperature and humidity or the estimation of energy and humidity transports between the atmosphere and the Earth surface were carried out on the basis of NWP model data (Allan et al. 2002; Bengtsson et al. 2004; Philipona et al. 2005).
The use of NWP models for climate studies requires an exact knowledge of the accuracy of the analyses results provided by these models. In particular, the parameters of heat and humidity transports are only to a small extent based on direct observations. In regions where they are needed most, humidity observations with sufficient accuracy and precision are not available. Hence, these parameters are less certain than other analysis output parameters such as temperature or air pressure for which many more observations are available (Kalnay et al. 1996; Bengtsson et al. 2004). Therefore, water vapor information generated by NWP models particularly needs to be validated to allow a reliable interpretation of these results.
One major advantage of the validation conducted in this study compared to previous NWP validations is that the GPS-derived water vapor represents an independent data source. Validations using, for example, the New Global Water Vapor (NVAP) dataset from Randel et al. (1996) are not independent because the NVAP dataset mainly consists of the same radiosonde and satellite radiometer observations that were already assimilated in the NWP model (Trenberth and Guillemot 1998). Possible systematic errors caused by the assimilated observations cannot be discovered from comparisons with the NVAP dataset. Former studies using GPS observations for validating the water vapor in NWP models were carried out with an objective of improving the short- and medium-range weather forecast in different countries. These investigations are mainly based on regional weather models and are restricted to some days or months. Additionally, most of the investigations are concentrated on Europe, North America, and Japan (e.g., Cucurull et al. 2000; Köpken 2001; Johnsen and Rockel 2001; Vedel et al. 2001; Gutman and Benjamin 2001; Okamura and Kimura 2003; Bock et al. 2005; Guerova et al. 2006). Smith et al. (2007) investigated the impact of data assimilation from GPS-derived water vapor on numerical weather forecasts over the United States using 6 yr of data. Validation of the water vapor in global weather models performed with GPS data were conducted for the operational analysis of the European Centre for Medium-Range Weather Forecasts (ECMWF) and the Canadian Meteorological Center (Hagemann et al. 2003; Deblonde et al. 2005). Both studies are based on data of 4 to 7 months only. They mainly concentrate on investigating the bias between the model and the GPS-derived water vapor and the scatter of the water vapor differences. The following analysis covers the complete time period from the beginning of 1994 up to the end of 2004. The main focus is the validation of the seasonal signals and interannual variations in the water vapor of the NWP model from the National Centers for Environmental Prediction (NCEP). The variability of the water vapor values from the model output will also be intercompared and validated with observations.
2. Estimation of precipitable water from GPS
The atmospheric constituents (including the water vapor) along the signal’s path from the GPS satellite to the receiver cause a propagation delay of the GPS signal. The delay is estimated as an additional parameter—the so-called zenith total delay (ZTD)—within the GPS data processing. This study is based on tropospheric parameters estimated in a common GPS reprocessing project of the Technical Universities in Munich and Dresden (Steigenberger et al. 2006). The reprocessing of a global GPS network covers the period from the beginning of 1994 up to the end of 2004 and was carried out with a modified version of the Bernese GPS software, version 5.0 (Dach et al. 2007).
The most important modifications in the reprocessing compared to the processing strategies of the analysis centers of the International Global Navigation Satellite System (GNSS) Service (IGS) are (i) the application of the isobaric hydrostatic mapping function (IMF) based on data of the NWP model of the ECMWF (Niell 2000; Vey et al. 2006) and (ii) the consideration of higher-order ionospheric effects (Fritsche et al. 2005). Some further processing-related aspects are that absolute antenna phase center variations for GPS receiver and satellites were used (Schmid et al. 2005), the cutoff elevation angle was set to 3°, and an elevation-dependent weighting was applied to the observations. The parameters for the zenith wet delay were estimated using continuous piecewise-linear interpolation with a 2-h parameter spacing. The ZTD parameter results from the sum of the estimated zenith wet delay and the a priori zenith hydrostatic delay. Tropospheric gradients in the north–south and east–west directions were estimated in 24-h intervals. For further details on the strategy of the GPS reprocessing, see Steigenberger et al. (2006).
- PGPS air pressure at GPS antenna height (hPa),
- PS air pressure at the height of the pressure sensor (hPa),
- ΔH = HGPS − HS height difference (m),
- g gravity acceleration,
- Rd = 287.053 J K−1 kg−1 gas constant of dry air,
- T mean temperature of the layer between the GPS antenna and the meteorological sensor (K).
- ρw = 1025 kg m−3 density of liquid water,
- Rν = 461.51 J K−1 kg−1 specific gas constant of water vapor,
- k′2 and k3 atmospheric refraction constants, k′2 = 22.1 ± 2.2 K hpa−1, k3 = 373900 ± 1200 K2 hpa−1.
3. PW results
Precipitable water time series with a daily resolution were estimated for 141 stations. However, many PW time series cover the period from 1994 to 2004 only partly. For further analysis 62 stations covering a period of at least 5 yr with data gaps smaller than 3 months were selected as a compromise between the time series length and the number of available stations (Fig. 1).
a. Accuracy
The accuracy of the PW results strongly depends on the accuracy of the ZTD values. Here only the main error sources concerning the estimation of ZTD will be discussed. One important error source is the modeling of the location of the phase center of the satellite and receiver antennas relative to their geometrical reference point. These models consist of a constant term (phase center offsets, or PCOs) and elevation- and azimuth-dependent parts (phase center variations, or PCVs). Relative antenna phase center models can be computed using short baselines and a reference antenna (Mader 1999). Unfortunately, the models are relative to a reference antenna (a Dorne Margoin T model) that has its own elevation dependence. Absolute PCVs for receiver antennas estimated by robot calibrations are independent of a reference antenna (Menge et al. 1998). Additionally, they were used to estimate a consistent set of PCVs for satellite antennas (Schmid and Rothacher 2003). The change from relative to absolute PCVs causes 4–5-mm smaller zenith delays, reducing the bias between the ZTD estimates from GPS and Very Long Baseline Interferometry (VLBI) significantly (Schmid and Rothacher 2003; Zhu et al. 2003; Steigenberger et al. 2007). Absolute PCVs from field and from robot calibrations agree at the level of few millimeters (Görres et al. 2006). Radomes installed on top of the antennas delay the GPS signal and modify its directional dependence on the azimuth and the elevation angle of the incoming signal. Radomes cause systematic errors in the estimated ZTD parameters that can reach some millimeters depending on the shape of the radome (Johansson 1998; Schupler 2001). The effect of radomes was not considered in this study. Hence, the installation or change of radomes can cause significant jumps in the ZTD time series.
A further important error source in the estimation of the ZTD parameters is uncertainties in the mapping functions that model the elevation dependence of the tropospheric delay. The widely used Niell mapping function (NMF) is based on radiosonde data of the United States (Niell 1996). By assuming longitudinal homogeneity and symmetry between the Southern and the Northern Hemisphere, the NMF was extended to the entire globe. Seasonal variations are approximated by a sine function. A small error in the hydrostatic mapping function of 0.1% at an elevation angle of 5°—as used in the NMF—causes an error of 3 mm in the ZTD estimates (Niell 1996). Uncertainties in the mapping function can cause systematic as well as seasonal errors in the GPS-derived ZTD time series (Vey et al. 2006).
In the last years the modeling of the elevation-dependence of the tropospheric parameters could be improved by mapping functions based on data of numerical weather prediction models (Niell 2001; Boehm et al. 2006). The application of mapping functions based on NWP model data—such as the isobaric mapping function used in this study—reduces the elevation dependence of the estimated ZTD parameters by 20% compared to a solution based on the NMF (Vey et al. 2006). Additionally, the bias between the ZTD parameters from GPS and VLBI decreases when the IMF is applied instead of the NMF (Tesmer et al. 2006). Mapping functions based on NWP data are closer to reality than the NMF. Therefore, the errors in the ZTD estimation due to uncertainties in the IMF can be stated to be smaller than 3 mm (Vey et al. 2006). Site-specific effects like multipath signal masking or snow accumulation on top of a GPS antenna or radome can significantly influence the parameter estimation and cause errors in the ZTD estimates of up to 2 mm (Jaldehag et al. 1996; Park et al. 2004). Also, higher-order ionospheric effects have a visible impact on the GPS parameter estimation (Kedar et al. 2003). If higher-order ionospheric effects are neglected, errors in the ZTD values of up to 1 mm can be expected (Fritsche et al. 2005).
A comparison of the ZTD parameters from six GPS–GPS-collocated stations shows a standard deviation of 2.5 to 5.4 mm (Steigenberger et al. 2007). Compared to the IGS combined solution, the precision of the tropospheric delay estimates in our reprocessing could be improved by 20%. The improvement is related to the application of enhanced models in the reprocessing (Steigenberger et al. 2007). Additionally, a validation of the troposphere parameters from the GPS and a VLBI reprocessing using the same analysis strategy for both techniques was carried out at 27 collocated GPS–VLBI stations by Steigenberger et al. (2007). For most of the stations the bias in the troposphere parameters estimated from both techniques ranges between −5 and 9 mm and the rms of the differences ranges between 4 and 10 mm. This translates into a systematic error of 0.7 to 1.7 mm in the PW estimates.
b. Identification of consistent time series
The homogeneity of the GPS-derived PW time series is a prerequisite for the validation of the seasonal and interannual changes in the modeled PW. The combined tropospheric parameter time series from the IGS solution, which are provided from 1997 until 2006, show inconsistencies of several millimeters due to changes in the reference frame used and the models and processing strategies applied (Steigenberger et al. 2007). Such processing-related inhomogeneities are not present in our reprocessed series because a consistent processing strategy was applied for the whole period of time.
However, changes in the hardware or in the data recording of the GPS stations can significantly affect the GPS parameter estimation. Gradinarsky et al. (2002) compared PW time series from GPS and radiosonde observations and found that the installation of a radome on top of a GPS antenna can cause significant jumps of 0.4 to 1.3 mm in the GPS-derived PW time series. Figure 2a shows a jump of 0.3 mm in the PW difference time series from the two collocated GPS stations TROM and TRO1 in Tromsø, Norway. This jump is caused by an antenna change of the station TRO1 on 22 December 1998. In the PW time series of station TRO1 this jump is not visible because it is covered by a strong seasonal signal of several centimeters (Fig. 2b).
To detect inhomogeneities in the PW time series, the differences of the PW estimates from neighboring GPS stations were systematically investigated. In addition to radome and antenna changes, changes in the number of observations due to modified elevation cutoff angles or signal masking were also found to cause inhomogeneities. Only one-third of the PW time series can be assumed to be homogeneous. For the PW time series with jumps, an offset correction was applied. In the following time series analysis, only homogeneous time series or time series with an offset correction of better than 0.3 mm were included. PW time series with a slowly and irregularly changing number of observations were excluded from the subsequent analysis. Further details on the analysis of the homogeneity of the PW time series can be found in Vey et al. (2009).
c. Time series analysis
d. Cluster analysis
To conclude how well NCEP represents the water vapor in different areas, the stations were clustered regionally. Gaffen et al. (1992) defined five humidity regimes from 56 globally distributed radiosonde stations. Using these humidity regimes and geographical information, the a priori clusters were formed. In a second step a variance analysis of the seasonal precipitable water signal of a station and the mean seasonal precipitable water signal of the cluster were conducted, and all stations with a variance of more than 2 mm were removed from the cluster. Typical seasonal signals in the PW of different regions are shown in Fig. 4. In the Northern Hemisphere the seasonal signals in different areas exhibit very similar behavior with a clear minimum at the end of January and a maximum at the end of July following the annual temperature cycle. The seasonal signal in the PW of the stations in North Australia exhibits a maximum at the end of February and a minimum in mid-August, being influenced by the monsoon and seasonal temperature variations. The seasonal variations in the PW over Antarctica with a range of 3 to 4 mm are very small, which is related to the very low amount of water vapor in this area. In climate-sensitive regions such as in the tropics and in Asia, the generation of clusters turned out not to be reasonable because of a low observation density and high moisture variability (Figs. 4c,e). In these cases the results will be given separately for each station. An overview of the created clusters can be found in Fig. 1.
4. Validation of NCEP
In a common project, two analysis centers, NCEP and the National Center for Atmospheric Research (NCAR), reanalyzed global atmospheric observations for the period from 1948 to 2002 (Kalnay et al. 1996; Kistler et al. 2001). The reanalysis is based on the operational weather model implemented in 1995 at NCEP (Kanamitsu et al. 1991). The model has a temporal resolution of 6 h and a horizontal resolution of 2.5° × 2.5° and consists of 28 vertical layers. To prevent inhomogeneities, the model and the assimilation scheme were kept constant for the complete reanalysis period. However, changes in the type and number of assimilated observations can still influence the reanalysis results. For example, the assimilation of satellite observations since 1979 has a significant impact on the reanalysis results. A repetition of the reanalysis was carried out by the Department of Energy (DOE) in relation to the Atmospheric Model Intercomparison Project (AMIP-II) for the years from 1979 up to present (Kanamitsu et al. 2002). The NCEP/DOE AMIP-II Reanalysis (R-2) is based on an improved model physics and more consistent observations. The validation was carried out using the data from this NCEP Reanalysis (hereafter referred to simply as NCEP; data are available online at http://www.cdc.noaa.gov/data/gridded/data.ncep.reanalysis2.html).
While NCEP provides mean values for grid cells with a ground surface of more than 10 000 km2, the GPS measurement is only representative for a region with a diameter of about 100 km. The PW values of the four closest grid cells were interpolated linearly to the position of the GPS station. Different reference heights between the GPS station and the model topography were considered by a vertical interpolation assuming a linear temperature lapse rate of the standard atmosphere. The GPS-derived PW data have a daily resolution. Because we are interested in seasonal and interannual signals, the 6-h values from NCEP were averaged to daily values. Additionally, only those data points from the continuous NCEP time series were extracted for which corresponding GPS observations were available.
a. Correlation
The PW values from GPS and from NCEP are highly correlated as shown in Figs. 5a,d for the stations Metsahovi, Finland, and Wuhan, China. The correlation coefficient for most of the stations in the Northern Hemisphere is close to 1 (Fig. 6a). Lower correlation coefficients were found for the Southern Hemisphere, especially for the Antarctic Peninsula and South America. This can be attributed to the very low number of observations being assimilated in the model in these regions. The strong influence of the observations on the analysis results can be seen in Australia. A higher observation density over Australia yields a much better agreement between the PW values from NCEP and GPS in this area compared to the rest of the Southern Hemisphere.
b. Bias
The differences in the long-term mean of the PW estimates from GPS and NCEP are for 50% of the stations in the range of ±1.5 mm and for 80% of the stations in the range of ±3 mm (Fig. 6b). Larger biases are mainly related to large height differences between the NCEP topography and the GPS antenna height. In the case of the station Lhasa, Tibet, situated at an elevation of 3658 m, the reference height of the model is 4563 m. The large height difference causes a significant error in the height reduction of the PW data due to the assumption made about the temperature lapse rate. Therefore, all stations with an elevation difference of more than 400 m were excluded from the validation.1 Additionally, for stations at high elevation the pressure reduction from mean sea level to the GPS antenna height is significantly influenced by errors in the temperature lapse rate. The hypsometric equation used is based on the temperature lapse rate of 6.5 K km−1 of the standard atmosphere. Depending on the location and the season, the temperature lapse rate can vary between 3 and 8 K km−1. For a station at an elevation of 900 m such as Zimmerwald, Switzerland, an error of 3 K in the mean atmospheric temperature can cause a significant error in the reduced pressure of 1.2 hPa (see Appendix A). Stations with an elevation of more than 900 m were also excluded from the calculation of the cluster mean values.2
The bias between the PW values from GPS and NCEP of the remaining stations is, except for very few stations, in the range of ±3 mm (Table 3). The GPS-derived PW values have an accuracy of 1.1 to 1.9 mm (section 3a). Therefore, reasons for larger biases between the PW estimates from GPS and NCEP are related to the NCEP analysis. For large parts of North America NCEP shows, compared to the GPS PW, 2 to 3 mm smaller values over the last decade. This confirms the results of local case studies that found a wet bias of the NCEP analysis over North America (Gutman and Benjamin 2001). Comparisons with water vapor values from GPS radio occultations also reveal more humid results for the NCEP analysis (Kursinski and Hajj 2001). The humidity in the analysis strongly depends on the assimilated observations (Andersson et al. 2007). Over land surfaces, mainly observations from radiosondes and satellite radiometers working in the infrared band are assimilated.
Over Europe values from the NCEP analysis were found to be drier than the GPS results. The systematic differences in the biases between North America and Europe can be related to differences in the radiosonde observations between both regions. The use of different radiosonde types and the application of different algorithms for the calibration of the radiosondes and the conversion of the measured relative humidity to the assimilated dewpoint temperature can cause significant differences between the radiosonde observations from different countries (Soden and Lanzante 1996; Miloshevich et al. 2006). Also, different limits applied for the recording of the humidity at low temperature or at high humidity can cause systematic differences in the radiosonde records (Garand et al. 1992).
Over the oceans infrared and microwave radiometer observations are assimilated into the model. The Special Sensor Microwave Imager (SSM/I) reveals a systematically higher humidity compared to radiosonde measurements (Andersson et al. 2007). The positive bias in the NCEP PW for many stations in the coastal areas can be explained by the assimilation of SSM/I data. Inconsistencies between the analysis and local radiosonde data can occur if the assimilated local observations are not representative for the region. The high biases at the stations Noumea (NOUM), New Caledonia, and Townsville (TOW2), North Australia, might be caused by such inconsistencies as their neighboring stations show a much smaller bias. Other examples are the stations Hoefn and Reykjavik, Iceland, where the steep topography at the coast has a remarkable impact on the water vapor behavior at smaller scales. Such local effects in the frontier between ocean and land are insufficiently represented by NCEP because of the poor spatial resolution of 2.5° × 2.5° of the global model.
c. Seasonal signal
The time series of the difference of the PW estimates from GPS and NCEP are represented exemplarily for the stations Metsahovi (METS), Finland, and Wuhan (WUHN), China, in Figs. 5c,g. At Metsahovi the standard deviation of the differences is very small (a value of 1.2 mm). In the case of Wuhan the PW difference time series shows a clear seasonal signal, causing a much higher standard deviation of the differences compared to Metsahovi. At Wuhan the spectrum of the PW estimates shows a significantly smaller amplitude of the annual and semiannual cycle in the PW from NCEP than in the PW from GPS (Fig. 5h).
As can be seen in Fig. 7, the PW differences between GPS and NCEP increase with a higher amount of water vapor. In the case of Wuhan, the difference in the water vapor is anticorrelated to the absolute PW values; hence, the regression line has a negative slope. For the station Magadan (MAG0), Siberia, a positive correlation and regression slope indicates a larger seasonal signal in the NCEP PW time series compared to the seasonal signal in the GPS PW time series.
The PW differences in the seasonal signals from GPS and NCEP represented in Fig. 8 show a similar behavior for the stations in every specific common cluster. In Europe the differences in PW are independent of the season (Fig. 8a). The standard deviation of the differences is smaller than 1.5 mm and the relative error of the seasonal signal (2%) is very low. The theoretically estimated precision of the GPS-derived PW values is consistent with the average differences. The results also testify to a high precision of the modeled PW over Europe. The water vapor values from GPS and NCEP also agree very well for the stations in the central and southeast regions of the United States as well as in South Australia. The relative errors in the seasonal signals of the water vapor in these regions are better than 5% (Table 3).
Large relative errors in the seasonal signals of about 25% occur in the tropics and in Antarctica. Because of the extremely low water vapor content in Antarctica, the absolute differences in the seasonal signal between GPS and NCEP in this area are smaller than 1 mm whereas these differences reach 3 to 4 mm in the tropics (Figs. 8g,h). For most of the regions the differences in the seasonal signals of the PW from NCEP and GPS are larger than the errors in the GPS-derived PW. Hence, the observed differences can be attributed to deficiencies in the water vapor modeling of NCEP. Further evidence for this interpretation is the strong dependence of the differences on the total water vapor content.
For most of the stations NCEP underestimates the amplitude of the seasonal signal in the precipitable water. This underestimation is pronounced in the tropics. For example, at the stations Cocos Island (COCO) in the Indian Ocean and BAHR in the Persian Gulf, the amplitude in the seasonal signal of PW is captured by NCEP with only 75% and 60%, respectively (Table 3). In the tropics, the definition of the atmospheric structure in NCEP is inferior compared to other regions (Trenberth and Guillemot 1998). Additionally, the number of radiosonde observations in the tropics is very limited and high cloud coverage is responsible for the low quality of the observations from satellite infrared radiometers. The low number of accurate observations available for the assimilation in the tropics causes the analysis results of the humidity to be of much lower quality than in other regions.
At the station Bermuda (BRMU) in the North American Basin the water vapor is underestimated by 4 mm in the summer months, corresponding to a relative error in the seasonal signal of 25% (Fig. 8b). In this case, the discrepancies are caused by small-scale effects related to the location of the station Bermuda, which is situated on an island. The stations on the west coast of North America show errors in the seasonal signal of the PW of 10%–20%. The topography of the Rocky Mountains complicates the correct reproduction of the water vapor by the model in this area. An overestimation of the amplitude of the seasonal signal in PW by NCEP was found for the stations in Siberia and northern Canada (Figs. 7 and 8c). The explanation for the differences at these stations is dry biases of radiosonde observations, which occur at low temperatures (Soden and Lanzante 1996; Soden et al. 2004). The assimilation of these observations causes a dry bias in the NCEP analysis data in the winter months. This conclusion corresponds to the results of a validation of the PW in ECMWF from Hagemann et al. (2003). They detected a dry bias in the ECMWF analysis in the high northern latitudes for the winter months. NCEP and ECMWF are mostly assimilating the same observations. The amplitude of the seasonal signal in the PW from NCEP was found to be overestimated by up to 7% in the high northern latitudes (Table 3).
d. Anomalies
PW anomalies were validated at the GPS stations containing time series with a length of at least 7 yr. As shown for the station KOSG in Fig. 9a, as an example, the interannual variations in the PW are well represented by NCEP for stations in Europe. In this region, the differences between the water vapor anomalies from GPS and NCEP are very small, having a standard deviation of 0.3 mm (Table 3). The anomalies from both methods also agree very well for the stations in Iceland, North America, and Antarctica. They confirm the high accuracy of the PW anomalies from GPS and show that NCEP correctly represents the interannual variations in the water vapor of the regions mentioned above.
However, the stations in southeastern North America show differences in the PW anomalies of up to 1 mm (Table 3). In some cases, such as for the station COCO in the Indian Ocean in Fig. 9c, the anomalies from GPS and NCEP can differ by almost 3 mm. In most cases the water vapor anomalies are underestimated by the model. An exception represents the PW anomaly at the station BAHR, which for 1998 is overestimated by the model (Fig. 9d). That year is characterized by a strong influence of El Niño (McPhaden 1999).
e. Variability
The standard deviation of the PW values of one month was defined as monthly variability of the water vapor. The example of the station at Pency (GOPE), Czech Republic, in Fig. 10a shows a mean variability in the PW of 4 mm over the last decade. Additionally, the monthly variability is marked by significant seasonal signals. The variability in the PW at GOPE varies by ±1 mm during one year, showing a higher variability in the summer months. For the PW variability the same time series analysis was applied as for the PW time series itself (section 3c).
The agreement of the PW variabilities between GPS and NCEP strongly depends on the region. NCEP very well represents the PW variability over Europe. As exemplarily shown for GOPE in Fig. 10c, the bias and the standard deviation of the differences in the variabilities from GPS and NCEP are both smaller than 0.1 mm. In Europe the error in the seasonal signal of the PW variability is smaller than 5%. Over large parts of North America and Iceland NCEP also well represents the seasonal signal in the PW variability (Table 4).
However, in the Southern Hemisphere, large differences occur between the variabilities from GPS and NCEP. The example of station Karratha (KARR) in North Australia in Figs. 10b,d shows a 1-mm smaller mean variability for the PW from NCEP than for that from GPS. Additionally, the difference in the variability from both methods reveals a seasonal signal. Significant differences in the seasonal signal of the variabilities with amplitudes of up to 1 mm were detected for all stations in North Australia (Fig. 11c). The magnitude of the differences in the PW variability strongly depends on the total water vapor content. This dependency indicates a weakness of the NCEP analysis. The accuracy of the PW variability from GPS is mainly independent of the amount of water vapor. From the results found over Europe the accuracy of the PW variability can be stated to be in the range of 0.1 mm. The large differences in the variabilities from GPS and NCEP over North Australia of more than 1 mm are interpreted as deficiencies in the NCEP analysis. For most of the stations NCEP underestimates the mean variability as well as the seasonal signal in the PW variability (Table 4). For the Antarctic stations represented in Fig. 11d the differences between the variabilities from GPS and NCEP are small. However, because of the low water vapor content in Antarctica the relative error in the seasonal signal of the PW variability is 35%.
5. Conclusions
Comparison of NCEP water vapor with independent GPS data over a time period of up to one decade shows that the differences between modeled and observed PW values in some locations are time dependent. Hence, the differences in the PW values from GPS and NCEP need to be analyzed at seasonal and interannual time scales. Biases estimated from previous studies covered only few months (Hagemann et al. 2003; Deblonde et al. 2005) and are mostly not valid for longer periods of time.
Over Europe and large parts of North America the seasonal cycle and the interannual variations in the PW from GPS and NCEP agree very well. The monthly variability in the PW is also very well represented by NCEP. In these regions NCEP depicts a highly accurate database for studies of long-term changes in the atmospheric water vapor. The results of the validation reveal a submillimeter accuracy of the GPS-derived PW anomalies. Hence, GPS observations are a valuable data source for climate studies, supposing the GPS processing strategy is kept unchanged and modifications in the tracking hardware or in the number of the recorded GPS observations are considered in the analysis.
In the Southern Hemisphere large differences in the seasonal signals and in the anomalies of the PW as well as in the PW variability were found between the data from GPS and from NCEP. In the tropics and in Antarctica the seasonal signal in the PW is strongly underestimated by NCEP by up to 40% and 25%, respectively. In these regions the number of observations being assimilated in the model is very limited and they are often of lower accuracy. Hence, the global observing system needs to be extended in the Southern Hemisphere and differences between PW values measured by satellites over the oceans and GPS need to be sorted out and resolved. Additionally, GPS measurements provide a data source of highly accurate PW estimates that could be assimilated into the model.
Climate change studies based on water vapor data from NCEP should consider the high uncertainties in the analysis when interpreting these model outputs, especially in the tropics. The validation carried out here exemplarily for NCEP can be extended to other numerical weather prediction or climate models.
We thank NCEP for providing model data of the precipitable water on its website. We gratefully acknowledge the constructive comments of the reviewers. The investigations presented here were partly supported by the German Research Foundation (DFG). Sincere thanks go to Johannes Böhm for making the data of the height of the 200-hPa surface and the mean atmospheric temperature of the ECMWF available to us. One of the authors (S. Vey) was supported by the Studienstiftung des deutschen Volkes, which is gratefully acknowledged.
REFERENCES
Allan, R. P., , A. Slingo, , and V. Ramaswamy, 2002: Analysis of moisture variability in the European Centre for Medium-Range Weather Forecasts 15-year reanalysis over the tropical oceans. J. Geophys. Res., 107 , 4230. doi:10.1029/2001JD001132.
Andersson, E., and Coauthors, 2007: Analysis and forecast impact of the main humidity observing systems. Quart. J. Roy. Meteor. Soc., 133 , 1473–1485.
Basili, P., , S. Bonafoni, , R. Ferrara, , P. Ciotti, , E. Fionda, , and R. Arnbrosini, 2001: Atmospheric water vapor retrieval by means of both a GPS network and a microwave radiometer during an experimental campaign in Cagliari, Italy, in 1999. IEEE Trans. Geosci. Remote Sens., 39 , 2436–2443.
Bengtsson, L., and Coauthors, 2003: The use of GPS measurements for water vapor determination. Bull. Amer. Meteor. Soc., 84 , 1249–1258.
Bengtsson, L., , K. I. Hodges, , and S. Hagemann, 2004: Sensitivity of large-scale atmospheric analyses to humidity observations and its impact on the global water cycle and tropical and extratropical weather systems in ERA40. Tellus, 56A , 202–217.
Bevis, M., , S. Businger, , T. A. Herring, , C. Rocken, , R. A. Anthes, , and R. H. Ware, 1992: GPS meteorology: Remote sensing of the atmospheric water vapor using the Global Positioning System. J. Geophys. Res., 97 , (D14). 15787–15801.
Bevis, M., , S. Businger, , S. Chiswell, , T. A. Herring, , R. A. Anthes, , C. Rocken, , and R. H. Ware, 1994: GPS meteorology: Mapping zenith wet delays onto precipitable water. J. Appl. Meteor., 33 , 379–386.
Beyerle, G., , K. Hocke, , J. Wickert, , T. Schmidt, , C. Marquardt, , and C. Reigber, 2002: GPS radio occultations with CHAMP: A radio holographic analysis of GPS signal propagation in the troposphere and surface reflections. J. Geophys. Res., 107 , 4802. doi:10.1029/2001JD001402.
Beyerle, G., , J. Wickert, , T. Schmidt, , and C. Reigber, 2004: Atmospheric sounding by global navigation satellite system radio occultation: An analysis of the negative refractivity bias using CHAMP observations. J. Geophys. Res., 109 , D01106. doi:10.1029/2003JD003922.
Bock, O., , C. Keil, , E. Richard, , C. Flamant, , and M. N. Bouin, 2005: Validation of precipitable water from ECMWF model analyses with GPS and radiosonde data during the MAP SOP. Quart. J. Roy. Meteor. Soc., 131 , 3013–3036.
Boehm, J., , B. Werl, , and H. Schuh, 2006: Troposphere mapping functions for GPS and very long baseline interferometry from European Centre for Medium-Range Weather Forecasts operational analysis data. J. Geophys. Res., 111 , B02406. doi:10.1029/2005JB003629.
Chaboureau, J-P., , A. Chédin, , and N. A. Scott, 1998: Remote sensing of the vertical distribution of atmospheric water vapor from the TOVS observations. Method and validation. J. Geophys. Res., 103 , 8743–8752.
Cucurull, L., , B. Navascues, , G. Ruffini, , P. Elósegui, , A. Rius, , and J. Vilà, 2000: The use of GPS to validate NWP systems: The HIRLAM model. J. Atmos. Oceanic Technol., 17 , 773–787.
Dach, R., , U. Hugentobler, , P. Fridez, , and M. Meindl, Eds. 2007: Bernese GPS Software version 5.0. Astronomical Institute, University of Bern, 612 pp. [Available online at http://www.bernese.unibe.ch/docs/DOCU50.pdf].
Deblonde, G., , S. Macpherson, , Y. Mireault, , and P. Héroux, 2005: Evaluation of GPS precipitable water over Canada and the IGS network. J. Appl. Meteor., 44 , 153–166.
Deeter, M. N., 2007: A new satellite retrieval method for precipitable water vapor over land and ocean. Geophys. Res. Lett., 34 , L02815. doi:10.1029/2006GL028019.
Divakarla, M. G., , C. D. Barnet, , M. D. Goldberg, , L. M. McMillin, , E. Maddy, , W. Wolf, , L. Zhou, , and X. Liu, 2006: Validation of Atmospheric Infrared Sounder temperature and water vapor retrievals with matched radiosonde measurements and forecasts. J. Geophys. Res., 111 , D09S15. doi:10.1029/2005JD006116.
Fritsche, M., , R. Dietrich, , C. Knöfel, , A. Rulke, , S. Vey, , M. Rothacher, , and P. Steigenberger, 2005: Impact of higher-order ionospheric terms on GPS estimates. Geophys. Res. Lett., 32 , L23311. doi:10.1029/2005GL024342.
Gaffen, D. J., , A. Robock, , and W. P. Elliott, 1992: Annual cycles of tropospheric water vapor. J. Geophys. Res., 97 , (D16). 18185–18193.
Garand, L., , C. Grassotti, , J. Hallé, , and G. L. Klein, 1992: On differences in radiosonde humidity-reporting practices and their implications for numerical weather prediction and remote sensing. Bull. Amer. Meteor. Soc., 73 , 1417–1423.
Gendt, G., , G. Dick, , C. Reigber, , M. Tomassini, , Y. Liu, , and M. Ramatschi, 2004: Near real time GPS water vapor monitoring for numerical weather prediction in Germany. J. Meteor. Soc. Japan, 82 , 361–380.
Görres, B., , J. Campbell, , M. Becker, , and M. Siemes, 2006: Absolute calibration of GPS antennas: Laboratory results and comparison with field and robot techniques. GPS Solutions, 10 , 136–145.
Gradinarsky, L. P., , J. M. Johansson, , H. R. Bouma, , H. G. Scherneck, , and G. Elgered, 2002: Climate monitoring using GPS. Phys. Chem. Earth, 27 , 335–340.
Guerova, G., , J-M. Bettems, , E. Brockmann, , and C. Matzler, 2006: Assimilation of COST 716 near-real time GPS data in the nonhydrostatic limited area model used at MeteoSwiss. Meteor. Atmos. Phys., 91 , 149–164. doi:10.1007/s00703-005-0110-6.
Gutman, S. I., , and S. G. Benjamin, 2001: The role of ground-based GPS meteorological observations in numerical weather prediction. GPS Solutions, 4 , 16–24.
Hagemann, S., , L. Bengtsson, , and G. Gendt, 2003: On the determination of atmospheric water vapor from GPS measurements. J. Geophys. Res., 108 , 4678. doi:10.1029/2002JD003235.
Hajj, G. A., and Coauthors, 2004: CHAMP and SAC-C atmospheric occultation results and intercomparisons. J. Geophys. Res., 109 , D06109. doi:10.1029/2003JD003909.
Jaldehag, R. T. K., , J. M. Johansson, , J. L. Davis, , and P. Elósegui, 1996: Geodesy using the Swedish Permanent GPS Network: Effects of snow accumulation on estimates of site positions. Geophys. Res. Lett., 23 , 1601–1604.
Janes, H. W., , R. B. Langley, , and S. P. Newby, 1991: Analysis of tropospheric delay prediction models: Comparisons with ray-tracing and implications for GPS relative positioning. Bull. Geod., 65 , 151–161.
Johansson, J. M., 1998: GPS antenna and site effects. Advances in Positioning and Reference Frames, F. Brunner, Ed., Springer, 229–235.
Johnsen, K. P., , and B. Rockel, 2001: Validation of the NWP model HRM with ground-based GPS data. Phys. Chem. Earth, 26 , 463–466.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Kanamitsu, M., and Coauthors, 1991: Recent changes implemented into the Global Forecast System at NMC. Wea. Forecasting, 6 , 425–435.
Kanamitsu, M., , W. Ebisuzaki, , J. Woollen, , S-K. Yang, , J. J. Hnilo, , M. Fiorino, , and G. L. Potter, 2002: NCEP–DOE AMIP-II Reanalysis (R2). Bull. Amer. Meteor. Soc., 83 , 1631–1643.
Karbou, F., , F. Aires, , C. Prigent, , and L. Eymard, 2005: Potential of Advanced Microwave Sounding Unit-A (AMSU-A) and AMSU-B measurements for atmospheric temperature and humidity profiling over land. J. Geophys. Res., 110 , D07109. doi:10.1029/2004JD005318.
Kedar, S., , G. A. Hajj, , B. D. Wilson, , and M. B. Heflin, 2003: The effect of the second order GPS ionospheric correction on receiver positions. Geophys. Res. Lett., 30 , 1829. doi:10.1029/2003GL017639.
Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Reanalysis: Monthly means CD-ROM and documentation. Bull. Amer. Meteor. Soc., 82 , 247–267.
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 , 1105–1117.
Kumar, B., , R. S. C. Dutta, , and H. Shah, 1992: On the design of FIR digital differentiators which are maximally linear at the frequency π/p, p ∈ {positive integers}. IEEE Trans. Signal. Process., 40 , 2334–2338. doi:10.1109/78.157235.
Kursinski, E. R., , and G. A. Hajj, 2001: A comparison of water vapor derived from GPS occultations and global weather analyses. J. Geophys. Res., 106 , (D1). 1113–1138.
Larsen, G., , K. Lauritsen, , F. Rubek, , and M. Sørensen, 2005: Processing of CHAMP radio occultation data using GRAS SAF software. Earth Observation with CHAMP: Results from Three Years in Orbit, C. Reigber et al., Eds., Springer Verlag, 543–548.
Li, Z., , J-P. Muller, , and P. Cross, 2003: Comparison of precipitable water vapor derived from radiosonde, GPS, and Moderate-Resolution Imaging Spectroradiometer measurements. J. Geophys. Res., 108 , 4651. doi:10.1029/2003JD003372.
Liou, Y. A., , C. Y. Huang, , and Y. T. Teng, 2000: Precipitable water observed by ground-based GPS receivers and microwave radiometry. Earth Planets Space, 52 , 445–450.
Liou, Y. A., , A. G. Pavelyev, , J. Wickert, , T. Schmidt, , and A. A. Pavelyev, 2005: Analysis of atmospheric and ionospheric structures using the GPS/MET and CHAMP radio occultation database: A methodological review. GPS Solutions, 9 , 122–143. doi:10.1007/s10291-005-0141-y.
Mader, G. L., 1999: GPS antenna calibration at the National Geodetic Survey. GPS Solutions, 3 , 50–58. doi:10.1007/PL00012780.
Marquardt, C., , K. Labitzke, , C. Reigber, , T. Schmidt, , and J. Wickert, 2001: An assessment of the quality of GPS/MET radio limb soundings during February 1997. Phys. Chem. Earth, 26 , 125–130.
McPhaden, M. J., 1999: Genesis and evolution of the 1997–98 El Niño. Science, 283 , 950–954.
Menge, F., , G. Seeber, , C. Völksen, , G. Wübbena, , and M. Schmitz, 1998: Results of absolute field calibration of GPS antenna PCV. Proc. 11th Int. Tech. Meeting of the Satellite Division of the Institute of Navigation, ION, Nashville, TN, 31–38. [Available online at http://www.ife.uni-hannover.de/aoa-kalibration/ion98.pdf].
Miao, J., , K. Kunzi, , G. Heygster, , T. A. Lachlan-Cope, , and J. Turner, 2001: Atmospheric water vapor over Antarctica derived from Special Sensor Microwave/Temperature 2 data. J. Geophys. Res., 106 , 10187–10203.
Miloshevich, M. L., , H. Vömel, , D. N. Whiteman, , B. M. Lesht, , F. J. Schmidlin, , and F. Russo, 2006: Absolute accuracy of water vapor measurements from six operational radiosonde types launched during AWEX-G and implications for AIRS validation. J. Geophys. Res., 111 , D09S10. doi:10.1029/2005JD006083.
Niell, A. E., 1996: Global mapping functions for the atmosphere delay at radio wavelengths. J. Geophys. Res., 101 , (B2). 3227–3246.
Niell, A. E., 2000: Improved atmospheric mapping functions for VLBI and GPS. Earth Planets Space, 52 , 699–702.
Niell, A. E., 2001: Preliminary evaluation of atmospheric mapping functions based on numerical weather models. Phys. Chem. Earth, 26 , 475–480. doi:10.1016/S1464-1895(01)00087-4.
Niell, A. E., , A. J. Coster, , F. S. Solheim, , V. B. Mendes, , P. C. Toor, , R. B. Langley, , and C. A. Upham, 2001: Comparison of measurements of atmospheric wet delay by radiosonde, water vapor radiometer, GPS, and VLBI. J. Atmos. Oceanic Technol., 18 , 830–850.
Okamura, O., , and F. Kimura, 2003: Behavior of GPS-derived precipitable water vapor in the mountain lee after the passage of a cold front. Geophys. Res. Lett., 30 , 1746. doi:10.1029/2003GL017572.
Park, K-D., , R. S. Nerem, , M. S. Schenewerk, , and J. L. Davis, 2004: Site-specific multipath characteristics of global IGS and CORS GPS sites. J. Geod., 77 , 799–803. doi:10.1007/s00190-003-0359-9.
Philipona, R., , B. Dürr, , A. Ohmura, , and C. Ruckstuhl, 2005: Anthropogenic greenhouse forcing and strong water vapor feedback increase temperature in Europe. Geophys. Res. Lett., 32 , L19809. doi:10.1029/2005GL023624.
Randel, D. L., , T. H. Vonder Haar, , M. A. Ringerud, , G. L. Stephens, , T. J. Greenwald, , and C. L. Combs, 1996: A new global water vapor dataset. Bull. Amer. Meteor. Soc., 77 , 1233–1246.
Rocken, C., , 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 , 468–478.
Rocken, C., , T. Van Hove, , and R. Ware, 1997: Near real-time GPS sensing of atmospheric water vapor. Geophys. Res. Lett., 24 , 3221–3224.
Saastamoinen, J., 1972: Atmospheric corrections for the troposphere and stratosphere in radio ranging of satellites. The Use of Artificial Satellites for Geodesy, Geophys. Monogr., Vol. 15, Amer. Geophys. Union, 247–251.
Schlittgen, R., , and B. H. J. Streitberg, 1999: Zeitreihenanalyse. R. Oldenbourg Verlag, 571 pp.
Schmid, R., , and M. Rothacher, 2003: Estimation of elevation-dependent satellite antenna phase center variations of GPS satellites. J. Geod., 77 , 440–446. doi:10.1007/s00190-003-0339-0.
Schmid, R., , M. Rothacher, , D. Thaller, , and P. Steigenberger, 2005: Absolute phase center corrections of satellite and receiver antennas: Impact on global GPS solutions and estimation of azimuthal phase center variations of the satellite antenna. GPS Solutions, 9 , 283–293. doi:10.1007/s10291-005-0134-x.
Schüler, T., 2001: On ground-based GPS tropospheric delay estimation. Ph.D. thesis, Universität der Bundeswehr München, Studiengang Geodäsie und Geoinformation, 364 pp.
Schupler, B. R., 2001: The response of GPS antennas—How design, environment, and frequency affect what you see. Phys. Chem. Earth, 26A , 605–611.
Smith, T. L., , S. G. Benjamin, , S. I. Gutman, , and S. Sahm, 2007: Short-range forecast impact from assimilation of GPS-IPW observations into the Rapid Update Cycle. Mon. Wea. Rev., 135 , 2914–2930.
Soden, B. J., , and J. R. Lanzante, 1996: An assessment of satellite and radiosonde climatologies of upper-tropospheric water vapor. J. Climate, 9 , 1235–1250.
Soden, B. J., , D. D. Turner, , B. M. Lesht, , and L. M. Miloshevich, 2004: An analysis of satellite, radiosonde, and lidar observations of upper tropospheric water vapor from the Atmospheric Radiation Measurement Program. J. Geophys. Res., 109 , D04105. doi:10.1029/2003JD003828.
Steigenberger, P., , M. Rothacher, , R. Dietrich, , M. Fritsche, , A. Rülke, , and S. Vey, 2006: Reprocessing of a global GPS network. J. Geophys. Res., 111 , B05402. doi:10.1029/2005JB003747.
Steigenberger, P., , V. Tesmer, , M. Krügel, , D. Thaller, , R. Schmid, , S. Vey, , and M. Rothacher, 2007: Comparisons of homogeneously reprocessed GPS and VLBI long time-series of troposphere zenith delays and gradients. J. Geod., 81 , 503–514. doi:10.1007/s00190-006-0124-y.
Stier, W., 2001: Methoden der Zeitreihenanalyse. Springer-Verlag, 400 pp.
Stull, R. B., 1995: Meteorology for Scientists and Engineers—A Technical Companion Book. West Publishing, 528 pp.
Sun, J., 1993: Effects of vertical distribution of water vapor and temperature on total column water vapor retrieval error. J. Geophys. Res., 98 , (C4). 7069–7079.
Tesmer, V., , J. Boehm, , R. Heinkelmann, , and H. Schuh, 2006: Effect of different tropospheric mapping functions on the TRF, CRF and position time-series estimated from VLB. J. Geod., 81 , 409–421. doi:10.1007/s00190-006-0126-9.
Tregoning, P., , R. Boers, , D. O’Brien, , and M. Hendy, 1998: Accuracy of absolute precipitable water vapor estimates from GPS observations. J. Geophys. Res., 103 , (D22). 28701–28710.
Trenberth, K. E., , and C. J. Guillemot, 1998: Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/NCAR reanalyses. Climate Dyn., 14 , 213–231.
Vedel, H., , K. S. Mogensen, , and X-Y. Huang, 2001: Calculation of zenith delays from meteorological data comparison of NWP model, radiosonde and GPS delays. Phys. Chem. Earth, 26 , 497–502.
Vey, S., , and R. Dietrich, 2008: Validation of the atmospheric water vapour content from NCEP using GPS observations over Antarctica. Geodetic and Geophysical Observations in Antarctica: An Overview in the IPY Perspective, A. Capra and R. Dietrich, Eds., Springer, 125–135.
Vey, S., , R. Dietrich, , K. P. Johnsen, , J. Miao, , and G. Heygster, 2004: Comparison of tropospheric water vapour over Antarctica derived from AMSU-B data, ground-based GPS data and the NCEP/NCAR reanalysis. J. Meteor. Soc. Japan, 82 , 259–267.
Vey, S., , R. Dietrich, , M. Fritsche, , A. Rülke, , M. Rothacher, , and P. Steigenberger, 2006: Influence of mapping function parameters on global GPS network analyses: Comparisons between NMF and IMF. Geophys. Res. Lett., 33 , L01814. doi:10.1029/2005GL024361.
Vey, S., , R. Dietrich, , M. Fritsche, , A. Rülke, , P. Steigenberger, , and M. Rothacher, 2009: On the homogeneity and interpretation of precipitable water time series derived from global GPS observations. J. Geophys. Res., 114 , D10101. doi:10.1029/2008JD010415.
Wang, J., , L. Zhang, , and A. Dai, 2005: Global estimates of water-vapor-weighted mean temperature of the atmosphere for GPS applications. J. Geophys. Res., 110 , D21101. doi:10.1029/2005JD006215.
Wang, J., , L. Zhang, , A. Dai, , T. Van Hove, , and J. Van Baelen, 2007: A near-global, 2-hourly data set of atmospheric precipitable water from ground-based GPS measurements. J. Geophys. Res., 112 , D11107. doi:10.1029/2006JD007529.
Wang, J. R., , and W. Manning, 2003: Retrievals of low integrated water vapor using MIR and SSM/T-2 measurements. IEEE Trans. Geosci. Remote Sens., 41 , 630–639.
Wickert, J., , G. Beyerle, , G. A. Hajj, , V. Schwieger, , and C. Reigber, 2002: GPS radio occultation with CHAMP: Atmospheric profiling utilizing the space-based single difference technique. Geophys. Res. Lett., 29 , 1187. doi:10.1029/2001GL013982.
Yuan, L. L., , R. A. Anthes, , R. H. Ware, , C. Rocken, , W. D. Bonner, , M. G. Bevis, , and S. Businger, 1993: Sensing climate change using the Global Positioning System. J. Geophys. Res., 98 , 14925–14937.
Yunck, T. P., , C. H. Liu, , and R. Ware, 2000: A history of GPS sounding. Terr. Atmos. Ocean. Sci., 11 , 1–20.
Zhu, S. Y., , F-H. Massmann, , Y. Yu, , and C. Reigber, 2003: Satellite antenna phase center offsets and scale errors in GPS solutions. J. Geod., 76 , 668–672. doi:10.1007/s00190-002-0294-1.
APPENDIX A
Pressure Reduction Errors
APPENDIX B
Station List
Locations of the stations used in the analysis are given in Table B1.
Range of the seasonal signals in the PW. The denotation of the clusters (Cl) is given in Table 1.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
(a) Difference PW time series of the collocated GPS stations TROM and TRO1 in Tromsø, Norway (TROM − TRO1). The jump with an offset of 0.3 mm corresponds to an antenna change at the station TRO1 on 22 Dec 1998. (b) Daily PW values at TRO1 show a dominant seasonal signal.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Time series analysis of the PW values from the station Cape Ferguson, Australia. (a) Daily values (black) and 30-day median filtered data (gray), (b) interannual component, (c) spectra of the time series (black) and the applied low pass filter (gray), and (d) seasonal signal.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Seasonal signals of the PW in different regions;
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Comparison of the daily PW estimates from NCEP and GPS. (a) GPS-derived PW time series for Metsahovi (METS), Finland; (b) correlogram; (c) difference PW time series (GPS − NCEP); and (d) amplitude spectra (black: GPS, gray: NCEP). (e)–(h) As in (a)–(d), but for Wuhan (WUHN), China.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
(a) Correlation coefficients between the PW from NCEP and GPS. (b) Bias between the PW estimates from NCEP and GPS. Stations marked with a black dot exhibit a bias of <1.5 mm.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Dependency of the differences (NCEP − GPS) on the total water vapor content at the stations Wuhan (WUHN), China, and Magadan (MAD0), Siberia: (a) DOY and (b) PW.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Regional differences (NCEP − GPS) of the seasonal signals in the PW.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Comparison of the PW anomalies from GPS (black dotted line) and NCEP (gray solid line) for the stations Kootwijk (KOSG), Netherlands; Mawson (MAW1), Antarctica; Cocos Island (COCO), Indian Ocean; and Bahrain (BAHR), Persian Gulf.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Comparison of the monthly variability in the PW from GPS (black dotted line) and NCEP (gray solid line) for the stations (a) Ondrejov (GOPE), Czech Republic, and (b) Karratha (KARR), Australia. (c),(d) Differences in the PW variability (NCEP − GPS) For the same stations.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Regional differences (NCEP − GPS) in the seasonal signal of the PW variability.
Citation: Journal of Climate 23, 7; 10.1175/2009JCLI2787.1
Main error sources affecting the estimation of GPS-derived precipitable water. For better comparability, all effects are converted to PW using a conversion factor Π of 0.17.
Validation statistics of the PW from NCEP. An overview of the cluster names and location is given in Fig. 1; N = number of stations used for the validation of the entire signal (first value) and the interannual component (second value), r = correlation coefficient, Bias = PW(NCEP) − PW(GPS), Std dev = standard deviation of the differences of the entire signal ΔPW and the interannual component ΔPWa, and σ(PWs) = relative error of the seasonal signal.
Validation statistics of the variability V in the PW from NCEP. An overview of the cluster names and location is given in Fig. 1; N = number of stations used for the validation of the entire signal (first value) and the interannual component (second value); Bias(V) = bias in the PW variability (NCEP − GPS); Std dev (ΔV) = standard deviation of the differences in the PW variability; σ(Vs) = relative error in the seasonal signal of the PW variability; Std dev (ΔVa) = standard deviation of the differences in the interannual component of the PW variability.
Table B1. Station list.
