Expected future developments of ground based GNSS meteorology • Improvements of ”existing” system, e.g.: – – – – widespread assimilation. better geographical coverage higher homogeneity of GNSS delay data active quality control, withhold or flagging of questionable data. • Use of IWV maps in now-casting. • Introduction of real-time (in contrast to near real-time) processing. – Clearly useful in now-casting. – Unclear if it is beneficial for NWP. – Already done today, and data are available at E-GVAP server, but only uploaded once per hour. • Extraction and use of extra information from the GNSS data processing, e.g.: – GNSS slant total delays – GNSS ZTD gradients • Derivation of 3D humidity maps from GNSS delay data in combination with other meteorological data via tomography Observing geometry On slants • The ZTD is an average property. Information about the local variability of the atmosphere around a given GNSS site is lost. • The slant total delays (STD), or the residuals from the estimation of the ZTD, contain this information. • But besides ZTD, many other properties are estimated (via least square fits to models) in the GNSS processing, on different timescales. • How much atmospheric information is in the slants and how much noise? On slants, 2 • Software for extraction of slants from the GNSS data processing made at Techn. Univ. Delft, including correction for multipath effects, and at GFZ. • Data format in E-GVAP useful for slants, but no current derivation and upload of slants in regi of E-GVAP. • Software exists for assimilation of STDs in HIRLAM – In a small assimilation experiment the impact was found to be more or less like the impact of ZTDs. – Clear that slants are noisy. • Florian Zus work on slants (report at expert team workshop tomorrow). Based on MM5?! • Area for reseach, no near future operational use expected. On gradients • Estimation of East-West and North-South gradients of ZTD provides additional information to ZTD, but reduces somewhat the noise of STDs. • Inclusion of extra GNSS satellites in processing (GLONASS, future Galileo) will clearly improve gradient estimates, less clear is improvement of STDs. • Some work has been done on the comparison of GNSS and NWP ZTD gradients (e.g., Walpersdorf with ALADIN model). • No (known) work on assimilation of gradients. • In reseach, more work should be devoted to estimation of ZTD gradients from GNSS data, and validation and assimilation of those in NWP models. Could become useful on a relatively short timescale. GPS Tomography Based largely on original material from D. Leuenberger, MeteoSwiss and E. Troller, ETH Zürich • Tomography is a method by which a large number of line of sight integral observations in different directions and or different locations are used to derive 3D images of a distribution. • Tomography is widely used in certain types of medical scanning techniques, as an elegant, low impact, non toxic way of deriving 3D images of human bodies. • For about a decade research has been done on ground based GNSS tomography, in which line of sight delay, or IWV (integrated water vapour), observations from a dense network are combined in a tomographic reconstruction of the 3D delay or water varpour field. • The 3D water vapour field can be assimilated into almost all types of NWP models, including those based using a „nudging“ assimilation system (such as MeteoSwiss and DWD). • The 3D water vapour field can be used for now-casting (consider for example potential in combining with wind profiler and orther radar wind information). • For 3 and 4DVar assimilation systems use of the original observations is expected to be beneficial – always use observations with as little preprocessing as possible. Example from a project joint project of MeteoSwiss, ETH Zürich and swisstopo, in which humidity profiles are derived at a number of locations – ETH • GPS tomography – MeteoSwiss • Validation and assimilation of humidity profiles – swisstopo • Data provider height [m] Swiss Tomography system GPS Station Radiosonde Wet refractivity [ppm] M. Troller, ETH Zürich GPS tomography • • • • • • • 18 hourly profiles up to 6000m a.s.l, ∆x=50km System of linear equations on N in each voxel Largely under-determined Surface humidity constraints Spatial and temporal neighbour constraints No use of radiosonde/model data Least squares solver Validation of humidity profiles • Comparison of GPS humidity profiles with – radiosonde of Payerne – model analyses and forecasts • More than one year of data • Comparison of wet refractivity N – Radiosonde: – NWP COSMO model: Yearly humidity distribution GPS 2006 2006 2007 Radiosonde 2007 Yearly humidity distribution GPS 2006 2007 BIAS Jul/Aug 2006 STD bold: GPS thin: COSMO bold: GPS thin: COSMO black: 00UTC red: 12UTC black: 00UTC red: 12UTC Jul/Aug 2006 (with model BIAS STD constraints) bold: GPS thin: COSMO bold: GPS thin: COSMO black: 00UTC red: 12UTC black: 00UTC red: 12UTC Dashed: + model constraint Dashed: + model constraint BIAS Nov/Dec 2006 STD bold: GPS thin: COSMO bold: GPS thin: COSMO black: 00UTC red: 12UTC black: 00UTC red: 12UTC Nov/Dec 2006 (with model BIAS STD constraints) bold: GPS thin: COSMO bold: GPS thin: COSMO black: 00UTC red: 12UTC black: 00UTC red: 12UTC Dashed: + model constraint Dashed: + model constraint Statistics of July/August 2006 BIAS STD Statistics with Bias Correction BIAS STD Summary • Winter: – Large dry bias: -4ppm or -40% (max. at 2000m a.s.l.) – Std dev larger than model forecasts • Summer: – Wet bias (15ppm or 30%) (max. at 1000m a.s.l.) – Dry bias (-10ppm or -30%) (max. at 2000m a.s.l.) – Std dev comparable to model forecasts • Problem at ~2000m a.s.l. • Model first guess does not seem to solve the problem • Bias hard to handle • -> Currently: quality not sufficient for assimilation
© Copyright 2026 Paperzz