The Study Of Retrieving2D/3D Water Vapor Distribution Using Ground-based GNSS Meteorology

As an important part of the earth’s atmosphere, water vapor plays a key role in global climate change and many synoptic processes. It accounts for a very small proportion in the atmosphere which is about from0.1%to3%while it has the most significant temporal and spatial variations, which makes it difficult to monitor the distribution of water vapor with high precision. The traditional measuring means have many disadvantages such as low spatial or temporal resolution, finite precision, and small scope of application, etc. With the development of modern numerical weather prediction (NWP) models, lack of accurate knowledge about the highly-resolved distribution of water vapor gradually becomes a key factor restricting the precision of NWP, especially for the quantitative precipitation.Ground-Based GNSS meteorology is a very promising technique which can retrieve highly resolved and accurate water vapor fields in space and time. It has the advantages of near-real-time processing, all-weather, no requirement for instrument calibration, high precision, low cost for each station and so on. Either total water vapor content or water vapor vertical profiles can be provided respectively by GNSS precipitable water vapor (PWV) measuring and GNSS tomography. So far the time length and the observed points distribution density of the accumulated ground-based GNSS observation data have been enough for studying many climate and weather issues, so make full use of these observations will greatly exert the application range of such GNSS networks in meteorology and other fields.The main objective of this thesis is to study several important issues in ground-based GNSS meteorology. The method that obtains the essential meteorological elements for retrieving GNSS PWV, such as surface pressure and weighted average temperature, of any location in China for PWV inversion from NWP reanalysis data is presented and the result precision is analyzed using real data. An improved tomographic numerical quadrature approach is proposed which represents the inhomogeneous distribution of water vapor and the curvature of earth’s surface more accurately without significant increase in calculation amount. The effects of several factors on the tomographic accuracy are studied by simulation experiments. Near-real-time tomographic system which fuses both ground-based GNSS slant water vapor observations and surface point moisture measurements is developed and a1-year term experiment using real data is carried out in Hong Kong. The main research contents and conclusions include:1. The method that interpolates surface pressure, temperature and weighted average temperature from NWP reanalysis data to any position in China is assessed using1-year term and large real data. The comparisons between the interpolated surface pressures and the real measurements on616sites included in the Integrated Surface Database (ISD) and Crustal Movement Observation Network of China (CMONOC) revealed that77.7%of the root-mean-square-errors (RMSEs) are less than1hPa, causes errors in PWV less than0.5mm, and98.5%of RMSEs are less than2.8hPa that causes PWV errors less than1mm. The real surface pressure variations can still be well represented even under extreme weather conditions. Errors of interpolated surface temperatures are smaller than5K on95.9%of all stations, corresponding relative errors of PWV are below1.8%. Weighed averaged temperature on22CMONOC stations with radiosonde observations and other10CMONOC stations without radiosondes are compared, the results show that the impact of interpolation errors on the GNSS PWV is smaller than1%in most cases and the results are still reliable in extreme weather. However, large errors exist on those sites that have much lower elevations than the four neighbor grids in NWP data for any interpolated meteorological element, and such errors presented significant seasonal changes, but the change trends are quite different in various regions. The absolute biases and RMSEs between the PWV derived from GPS and the PWV derived from radiosonde or NCEP data are smaller than2mm and3mm, respectively.2. The PWV distribution during the period of the severe torrential rain over Beijing region on21July2012reconstructed from CMONOC GNSS network and NWP reanalysis data are compared. The results demonstrate that a wide range of distribution and transportation of total water vapor content can be well retrieved by a wide ground-based GNSS network. Analyses on some PWV time series at specific stations reveal that GPS PWV often continues to rise and the temperature falls from about3-4hours before the rainfall, so detecting PWV in real time has certain indicating significance for heavy rainfall prediction.3. The current tomographic numerical quadrature algorithms are improved and simplified for better representations of the real inhomogeneous distribution of water vapor. Simulation studies about the effects of the earth surface’s curvature show that the impact increases very rapidly with the descent of slant path’s elevating angle, so a tomographic approach that takes such factor into account is proposed in this paper, and the number of zero coefficient parameters can therefore be reduced.4. Simulation experiments reveal that the vertical distribution of ground-based GNSS network has the greatest impact on the precision of tomographic results. Higher precisions are achieved at the layers which are lower than the top sites expect for the bottom layer. The overall tomographic precision is related to observed satellite constellation. Combination of multi-GNSS observation can improve tomographic precision; however, this improvement is very limited when the network is relatively flat. Using the slant observations with rather low elevations can also improve the results’precisions significantly especially for the middle-low troposphere while the Geometric Dilution of Precision (GDOP) has not considerably change. A balance should be achieved between the vertical resolution of tomographic results and the precision at each tomographic layer Overall precision would be higher if the divisions of tomographic grids in vertical direction were uneven.5. A near-real-time tomography software is developed in the thesis based on GAMIT software, TEQC software and standard c++programming language. A1-year term real experiment has been carried out in Hong Kong area. Time variations of three-dimensional water vapor fields are modeled through the long-term analyses on the surface meteorological and radiosonde data, and then this model is applied to describe the dynamic noise of Kalman filter to solve the tomographic observation equations. The performance of the weighted-exponential vertical interpolation formulas proposed in this paper is better than the other methods. Accuracies of water vapor density inversions are significantly improved at the middle-low levels due to the fusion of surface moisture observations, as the bias and RMSE both decrease by half in the ground level with respect to balloon soundings. The RMSE at the layers lower than4km falls down17.74%. However, such improved effects weaken gradually with the rise of altitude until becoming slightly adverse above5km. Quality of individual profile is also improved as the success rate of tomographic solution is increased from44.44to63.82%while the failure rate is reduced from55.56%to36.18%, and most of the correlation coefficients and RMSEs between tomography and radiosonde individual profiles are better. Various water vapor profiles with different characteristics are successfully reconstructed, but this study also indicates the inability of our tomography approach to reconstruct the spike layers in upper levels higher than1km which are often "spread" into the neighbor layers.