Consistency and representativeness of integrated water vapour from ground-based GPS observations and ERA-Interim reanalysis

Abstract

<p><strong>Abstract.</strong> This study examines the consistency and representativeness differences of daily IWV data from ERA-Interim reanalysis and GPS observations at 120 global sites over a 16-year period (1995–2010). Various comparison statistics are analysed as a function of geographic, topographic, and climatic features. A small (±1 kg m<sup>−2</sup>) bias is found in the reanalysis across latitudes (moist in northern and southern mid-latitudes and dry in the tropics). The standard deviation of daily IWV differences is generally below 2 kg m<sup>−2</sup> but peaks in the northern and southern storm-tracks regions. In general, the larger IWV differences are explained by increased representativeness errors, when GPS observations capture some small-scale variability that is not resolved by the reanalysis. A representativeness error statistic is proposed which measures the spatiotemporal variability in the vicinity of the GPS sites, based on reanalysis data at the four surrounding grid points. It allows to predict the standard deviation of daily IWV differences with a correlation of 0.73. In general, representativeness differences can be reduced by temporal averaging and spatial interpolation from the four surrounding grid points. A small number of outlying cases (15 sites) which don’t follow the general tendencies are further examined. It is found that their special topographic and climatic features strongly enhance the representativeness errors (e.g. steep topography and coast-lines, strong seasonal cycle in monsoon regions). Discarding these sites significantly improves the global ERA-Interim and GPS comparison results. The selection of site a priori, based on the representativeness error statistic, is able to detect 11 out of the 15 sites and improve the comparison results by 20 to 30 %.</p>

Homogenizing GPS Integrated Water Vapor Time Series: Benchmarking Break Detection Methods on Synthetic Data Sets

“…For these stations, either a phase shift between ERAI and GPS IWVs is found (e.g., IISC) or a difference in the amplitude of the seasonal cycle is observed (e.g., KIT3 and POL2). The same stations have also been identified by Bock and Parracho (2019) as outlying sites due to representativeness differences between both IWV data sets. The mean amplitude of the annual signal is 0.38 kg/m 2 , and this mean value is nearly halved for every subsequent harmonic (see Table S1).…”

“…Therefore, in this study, we have used ERA‐Interim reanalysis IWV data extracted at the site locations. The drawbacks of this approach are representativeness differences between the two IWV data sets at a number of sites (Bock & Parracho, 2019) and the possible presence of inhomogeneities in the reference ERA‐Interim data set as well (Ning et al, 2016; Schröder et al, 2016; Schröder et al, 2019). Representativeness differences might result in seasonality and trends in the difference time series, while remaining inhomogeneities in ERA‐Interim, even after 4‐D variational bias correction in the data assimilation, might be responsible for trends in the differences time series.…”

“…Although for the large majority of the sites, the IGS repro 1 and ERA‐interim IWV time series are highly correlated (see, e.g., Fig. 2 in Van Malderen et al, 2017, and Bock & Parracho, 2019), larger IWV differences are explained by increased representativeness errors, when GPS observations capture some small‐scale variability that is not resolved by the reanalysis (Bock & Parracho, 2019). For a number of outlying cases (15 sites), their special topographic and climatic features strongly enhance the representativeness errors (e.g., steep topography, coastlines, and strong seasonal cycle in monsoon regions).…”

“…The reasons for the different components included in this model are the following. Trend and seasonal signals in Δ IWV s are present due to representativeness differences between modeled and observed IWV estimates, model biases, and GPS biases (Bock & Parracho, 2019). Seasonal signals in Δ IWV s appear either due to a phase shift or a difference in the amplitude of the seasonal signal between the GPS and ERAI IWV time series.…”

“…Of those 970 events, 177 dates were confirmed by a visual inspection of the ERAI‐GPS IWV difference time series (see Figure 2a), and 50 more new dates were added, without reference in the IGS site logs. For four stations the visual detection of breaks is not reliable (see example in Figure 2b), as these sites have excessive (relative) standard deviation of IWV differences (see also Bock & Parracho, 2019). We hence end up with a list of 227 break positions (1.93 dates per station on average), which are inserted in equation to adjust the model parameters.…”