Water Vapor Tomography with Low Cost GPS Receivers

Transcription

1 Water Vapor Tomography with Low Cost GPS Receivers C. Rocken, J. Braun, C. Meertens, R. Ware, S. Sokolovskiy, T. VanHove GPS Research Group University Corporation For Atmospheric Research P.O. Box 3000, Boulder CO Abstract We propose to apply low cost single frequency GPS receivers, deployed in a dense array, to tomographic estimation of small scale three-dimensional (3-D) atmospheric water vapor fields. The proposed proof-of-concept experiment will be conducted during the Atmospheric Radiation Measurement (ARM) program s Intensive Observation Period (IOP) of July 1999 near the Lamont Cloud and Radiation Testbed (CART). The system will utilize GPS receivers deployed at 1 to 3 km spacing in a 10 x 10 km or larger area. Carrier phase data from this array will be analyzed to determine line-of-sight tropospheric delays caused by atmospheric water vapor. These line-ofsight delays shall be inverted, using tomographic techniques, and optionally first guess fields based on other atmospheric data available at the ARM CART site, to estimate 3-D atmospheric water vapor fields at 30 min or smaller time intervals. An alternative approach will be to assimilate the slant measurements into a high resolution numerical weather model like MM5. The purpose of the proposed effort is to develop and demonstrate a new atmospheric sensing technique to measure small scale atmospheric water vapor fields, which are important to provide the initial and boundary conditions for the Single Column Modeling (SCM) efforts conducted under the ARM program. Background Improved knowledge of the water vapor field is needed for a variety of atmospheric research applications and for improved weather forecasting. Within the ARM program an all weather, continuously operating system to monitor water vapor fields is needed to provide the initial conditions for the SCM efforts, aimed at improving the treatment of atmospheric radiation in Global Climate Models. Recently developed methods for sensing precipitable water vapor (PWV) using the Global Positioning System (GPS) promise to provide needed water vapor information for weather and climate models [Bevis et al., 1992; Rocken et al., 1993; Rocken et al., 1995a; Businger et al., 1996; Duan et al., 1996]. For example, the GPS Research Group at UCAR/UNAVCO is currently analyzing GPS data from the NOAA Forecast Systems Laboratory (FSL) GPS network [Gutman et al., 1994] in near real time, as shown in Figure 1. Scientists at FSL and at the National Center for Atmospheric Research (NCAR) are working on assimilating these observations into weather models to determine their effect on short term forecasting.

2 Figure 1: PWV estimates of the NOAA/FSL GPS network. These results are taken from the near real time analysis of the GPS Research Group at UCAR. These two figures show a significant gradient of water vapor through the center portion of the network, where the proposed tomographic network will be located. PWV measurements are derived by analyzing delays on the GPS signal caused by the neutral atmosphere. Traditionally this zenith delay is estimated as a parameter that maps as the cosecant of the satellite elevation angle [Niell, 1996]. The parameter is defined as: zenith delay = where dz has units of length in the zenith direction, and 10 6 N( z) dz antenna N = 77.6 P e T T 2 approximates the refractivity at a point in the atmosphere with air pressure P (mb), water vapor pressure e (mb), and temperature T (K). The first or dry term in N contributes up to 240 cm to zenith delay and the second or wet term contributes up to 40 cm. If P is known and hydrostatic equilibrium is assumed, the first term in the refractivity equation can be removed from the GPS solution. Zenith wet delay can then be estimated as a least squares fit to GPS residuals that varies as the cosecant of the elevation angle, typically over intervals of 15 min to several hr. Zenith wet delay and PWV are related by: PWV = Π zenith wet delay where Π is a dimensionless conversion factor approximately equal to 0.15 [Hogg et al., 1981].

3 Surface temperature measurements can be used to estimate Π with an error less than 2% [Bevis et al., 1992; 1994]. GPS sensed PWV is modeled using a fit of all observed rays seen by a receiver. These rays are spaced throughout an inverted cone in the atmosphere whose tip is the GPS antenna (see figure 2). For inhomogeneous atmospheres, this model is inadequate. Rocken et al. [1991a] and Davis et al. [1993] found more than 20% azimuthal variation in radiometer measurements of integrated water vapor at 30 o elevation. Observations that include azimuthal variations in atmospheric water vapor should be more useful than column PWV measurements. to GPS satellites to GPS satellites GPS Receiver Figure 2: PWV estimates from GPS observations are typically estimated by sampling throughout a conical section of the atmosphere above the GPS receiver (left). Typically, between 5 and 12 satellites are in view at a time. The green lines in the plot on the right shows the satellite tracks of GPS satellites over a station in Colorado for a 24 hour period. The blue circles indicate the position of all satellites in view at a instant in time. For a 30 minute time period, GPS satellites will move approximately 15 degrees across the sky. The technique of measuring these azimuthal variations using GPS was shown to be possible by Ware et al., [1997]. Examples of the slant path water vapor (SWV) measurements are shown in figure 3. The SWV is estimated as the delay in the direction of the satellite signal: slant wet delay = 10 6 satellite N wet ( s) ds antenna where N wet is the second term in Equation and slant wet delay and ds have units of length along the ray path. Slant wet delay is related to SWV by: SWV = Π slant wet delay

4 SWV DD (mm) :00 11:00 12:00 13:00 14:00 Day 141: Satellites 29 & 18: 1.2 mm rms SWV DD (mm) :00 18:30 19:00 19:30 20:00 20:30 Day 141: Satellites 9 & 4: 0.9 mm rms SWV DD (mm) :00 20:00 21:00 22:00 23:00 Day 143: Satellites 24 & 5: 1.6 mm rms Figure 3: Examples of GPS (jagged) and radiometer (smooth) sensed SWV double differences and their agreement (rms). Observations similar to the ones shown here would be used for tomographic inversion of water vapor fields. Overview of Proposed System Development Typically, the GPS systems used for atmospheric estimation cost approximately $15K. For a network of 20 stations, the cost of the equipment would be around $300K. To reduce this cost, we will utilize L1 GPS systems that are being developed under a NASA grant for volcano monitoring. We anticipate the cost of each of these receivers to be approximately $3K each. We believe that these systems will have sufficient accuracy to acquire the SWV observations needed for this project. The main tasks in the implementation of this project include the following: Development and testing of low cost receivers Development and testing of low cost, low multipath antennas Development and testing of GPS analysis software to obtain SWV measurements Development and testing of tomographic techniques Field tests and installation of the new hardware Operation of the tomographic network Evaluation of the recovered 3-D water vapor fields. GPS Hardware The system that we will be deploying in the field will be an independent system that requires no external power supply or data communications network. We plan on using solar power and batteries to power the systems. Radio modems equipped with Time Delay Multiple Access (TDMA)

5 technology will transmit he data to a central processing site in real time. We believe that this will allow for greater flexibility in deploying and operating the systems near the Lamont CART facility. Figure 4 shows diagrams and pictures of a prototype of the proposed GPS system. L1 antenna and ground plane GPS + MODEM A B C solar panel radio antenna guy wire 1.5 meters anchor 10 cm Battery base point monument Site 2 slave TDMA Site 1 slave-repeater TDMA Data Processing Computer Master Site 3 slave TDMA Site 25 slave TDMA Figure 4: The cartoon in the top left depicts the schematic diagram of the GPS system. The picture on the top right shows a prototype of the system. These systems will be designed to be independent of power and communications infrastructure. They will have solar and battery power to last up to four days without sun, and will transmit data back to a central site using the Time Delay Multiple Access (TDMA) technology in the Freewave radio modems. The picture in the bottom left shows a close up view of the Canadian Marconi Company (CMC) L1 GPS receiver and the Freewave TDMA radio modem. A schematic of a proposed network is shown in the bottom right corner.

6 The L1 carrier phase data from this L1 Canadian Marconi Company (CMC) system is comparable to the data obtained from the more expensive dual frequency receivers used to obtain the PWV and SWV examples shown in figures 1 and 3. Figure 5 shows double difference residuals for a 10 Km baseline in Boulder, CO. The baseline was measured with Trimble 4000 SSI receivers and the L1 CMC board. We believe that the L1 data from the CMC receivers is as good as the Trimble receiver and will be sufficient to observe estimate SWV measurements. Trimble 4000 SSI Canadian Marconi Difference Between Trimble and CMC Figure 5: Double difference residuals for a 10 Km baseline with a pair of Trimble 4000 SSI receivers (top) and a pair of CMC L1 receives (middle). The double difference pair is SV2 - SV21. The bottom trace is the difference between the top two double difference pairs. This indicates that the noise level of the CMC receiver is on the same level as the dual frequency Trimble receiver. In addition to having a high quality GPS receiver, the antenna will also need to be a low noise antenna with stable phase center characteristics. Specifically, this antenna will be required to have high multipath suppression over all elevation and azimuth angles. Multipath will be the dominant error sources in this system. Currently, we are planning to use an L1 patch antenna manufactured by MicroPulse. This antenna will be equipped with an ground plane or choke ring attachment for multipath suppression. Figures 6 shows the MicroPulse antenna and signal to noise observations taken with the antenna. Figure 7 shows the MicroPulse antenna with an additional Frisbee attachment for multipath suppression. The 30% reduction in signal to noise ration noise indicates that this configuration has better multipath characteristics than just the antenna.

7 Figure 6: The MicroPulse GPS antenna shown in the picture on the left was used to collect the observations in the plot on the right which shows signal to noise ratios observations (blue) as a function of satellite elevation angle. The rms (1.62dB) of the polynomial fit (red line) indicates a high level of multipath error in the observations. Figure 7: The MicroPulse antenna equipped with a frisbee ground plane is shown in the picture on the left. The plot on the right shows signal to noise observations (blue) as a function of satellite elevation angle. The rms fit of the observations (1.17dB) to the polynomial fit (red line) is improved by almost 30% as compared to the MicroPulse antenna without ground plane.

8 GPS Processing Software The objective of the GPS data analysis is to remove all effects from the measured L1 carrier phase data except the delay from the integrated atmospheric water vapor between the antenna and GPS satellite. This delay can vary between 0 and 40 cm in the zenith direction, and can reach several meters for observations below a 10 o elevation angle. Overall analysis will be carried out in the following steps: GPS data collection and translation to RINEX (Receiver Independent Exchange Format) Forming of differences to remove clock errors Removal of geometric effects, using known receiver coordinates and satellite positions. Removal of dry hydrostatic delay based on pressure and temperature data from the ARM/ CART site and mapping functions. For the L1 network, which is small, first order ionospheric effects will difference out. We will correct second order effects using an ionospheric model. Generation of SWV residual files including station to satellite azimuth and elevation angles. These files will be used as input to the tomographic inversion. An automatic L1 carrier phase processing system will be implemented on the data reception PC using the Bernese 4.0 processing software [Beutler et. al., 1996; Rothacher et. al., 1996]. An ionospheric model such as the one shown in figure 8 will be used to correct for errors in the L1 observations. The ionosphere models will be derived from the surrounding dual frequency network already installed throughout the South Great Plans (SGP) region. Figure 8: The two plots shown above are ionosphere models derived from dual frequency GPS observations. The left plot is from the UCAR GPS Research Group s real time analysis. The right plot is from the Jet Propulsion Laboratory (JPL). Models like the ones shown would be used to remove ionosphere effects on the L1 carrier phase observations.

9 Tomographic Inversion We propose to develop software for tomographic inversion of the SWV values. To determine optimal strategies for dealing with these data we will perform computational simulations of the tomographic reconstruction of the 3-D water vapor fields with data from a 2-D network of groundbased receivers. The goal of these simulations is to test our software and to estimate the capability of such a network to reconstruct refractivity field inhomogeneities of different scales. In addition we need to determine an optimal discretization and interpolation scheme of the refractivity field to be used for the processing of observational data. Observational Operator: We will calculate simulated observables (i.e. excess phases) s i, j along all given rays {j} for all receivers {i}) and thus construct the model of the observational operator. For each ray we will integrate refractivity (ray bending can be ignored above 5 to 10 o elevation) along the ray from the receiver to the top of the boundary layer by interpolating it onto a 1-D grid, and then by applying either sliding polynomial interpolation or spline interpolation along the ray. Once we have obtained the model of the observational operator H i, j and the observational vector y i, we may solve a least squares problem for the vector of state: 2 H i, j x j y i = min The quality of different elements of the solved-for vector of state x may be different depending on the chosen grid, interpolation scheme, configuration of rays, and the structure of inhomogeneities of the refractivity field. Clearly, space domains containing many elements of the vector of state, but not transsected by a sufficient number of rays may provide artifacts. So the spatial density of the chosen grid should roughly correspond to the density of rays. In particular along the horizon the increment of the grid should be consistent with the distance between the receivers. The potential vertical resolution depends more strongly on the structure of refractivity field. For a horizontally homogeneous field for example, all receivers provide the same observables and the inverse problem is reduced to that of one receiver, which is known to have poor vertical resolution. On the other hand, isotropic inhomogeneities (having approximately equal vertical and horizontal dimensions) are expected to be resolved quite well. As an alternative to the tomographic solution we will also attempt to assimilate the measured slant observations into high resolution numerical weather models in collaboration with NCAR/MMM meteorologists. Vertical Resolution: To improve vertical resolution, observations providing profiling data (like lidars, profiling radiometers, radiosondes, high-resolution model fields, observations from orbiting GPS receivers, etc.) will be included into the analysis.

10 Summary This system has been proposed for development and deployment for the July 1999 Intensive Observation Period field campaign near Lamont, Oklahoma. It will utilize GPS technology being developed under a NASA grant for volcano research. We propose to used Slant Water Vapor (SWV) observations to construct fields of water vapor using tomographic or data assimilation techniques. References Alber, C., Millimeter Precision GPS Surveying and GPS Sensing of Slant Path Water Vapor, Ph.D. Thesis, Univ. Colorado, Dec Beutler, G., E. Brockman, S. Frankhauser, W. Gurtner, J. Johnson, L. Mervart, M. Rothacher, S. Schaer, T. Springer, and R. Weber, Bernese GPS Software Version 4.0, Univ. Berne, September, Bevis, M., S. Businger, T. Herring, C. Rocken, R. Anthes and R. Ware, GPS Meteorology: Remote Sensing of Atmospheric Water Vapor Using the Global Positioning System, J. Geophys. Res., 97, 15,787-15,801, Bevis, M., S. Businger, S. Chiswell, T. Herring, R. Anthes, C. Rocken, and R. Ware, GPS Meteorology: Mapping Zenith Wet Delays onto Precipitable Water, J. Appl. Meteorology, 33, , Businger, S., S. Chiswell, M. Bevis, J. Duan, R. Anthes, C. Rocken, R. Ware, T. Van Hove, and F. Solheim, The Promise of GPS in Atmospheric Monitoring, Bulletin of the American Meteorological Society, 77, 5-18,1996. Campen, C., R. Cunningham, and V. Plank, Electromagnetic wave propagation in the lower atmosphere, Handbook of Geophysics, New York, Macmillan, ch. 13, 5-11, Davis, J., G. Elgered, A. Niell, and C. Kuehn, Ground-based measurement of gradients in the wet radio refractivity of air, Rad. Sci., 28, , Duan, J., M. Bevis, P. Fang, Y.Bock, S. Chiswell, S. Businger, C. Rocken, F. Solheim, T. Van Hove, R. Ware, S. McClusky, T. A. Herring, and R. W. King, GPS Meteorology: Direct Estimation of the Absolute Value of Precipitable Water, J. of Applied Met., 35, , June Gutman, S., R. Chadwick, D. Wolfe, A. Simon, T. Van Hove, and C. Rocken, Toward an Operational Water Vapor Remote Sensing System Using GPS, FSL Forum, September Hogg, D., F. Guiraud, and M. Decker, Measurement of Excess Radio Transmission Length on Earth-Space Paths, Astron. Astrophys., 95, , Kuo, Y.-H., Y.-R. Guo, and E. Westwater, Assimilation of Precipitable Water Measurements into a Mesoscale Numerical Model, Mon. Wea. Rev. 121, , Kuo, Y.-H., X. Zou, and Y.-R. Guo, Variational Assimilation of Precipitable Water Using a Nonhydostatic Mesoscale Adjoint Model Part 1: Moisture Retrieval and Sensitivity Experiments, Mon. Wea. Rev., 124, , Rocken, C., J. Johnson, R. Neilan, M. Cerezo, J. Jordan, M. Falls, L. Nelson, R. Ware, M. Hayes, The Measurement of Atmospheric Water Vapor: Radiometer Comparison and Spatial Variations, IEEE Transactions on Geoscience and Remote Sensing, 29, 3-8, Rocken, C., R. Ware, T. Van Hove, F. Solheim, C. Alber, and J. Johnson, Sensing Atmospheric Water Vapor with the Global Positioning System, Geophys. Res. Lett., 20, , Rocken, C., T. Van Hove, J. Johnson, F. Solheim, R. Ware, M. Bevis, S. Chiswell, and S. Businger, GPS/ STORM-GPS Sensing of Atmospheric Water Vapor for Meteorology, Journal of Atmospheric and Ocean Technology, 12, , 1995a. Rothacher M., L. Mervart, Bernese GPS Software Version 4.0, Astronomical Institute University of Bern, pp 418, Sep Ware, R. H., C. Alber, C. Rocken, F. Solheim, Sensing integrated water vapor along GPS ray paths, Geophys. Res. Lett., 24, , Zou, X, Y.-H. Kuo, and Y.-R. Guo, Assimilation of Atmospheric Radio Refractivity Using a Nonhydrostatic Mesoscale Model, Monthly Weather Review, 123, , 1995.