Jump to DART Documentation Main Index
The COSMIC project provides data from a series of satellites. There are two forms of the data that are used by DART: GPS Radio Occultation data and Electron Density. The programs in this directory extract the data from the distribution files and put them into DART observation sequence (obs_seq) file format.
The COSMIC satellites measure the phase delay caused by deviation of the straight-line path of the GPS satellite signal as it passes through the Earth's atmosphere when the GPS and COSMIC satellites rise and set relative to each other. This deviation results from changes in the angle of refraction of light as it passes through regions of varying density of atmosphere. These changes are a result of variations in the temperature, pressure, and moisture content. Vertical profiles of temperature and moisture can be derived as the signal passes through more and more atmosphere until it is obscured by the earth's horizon. There are thousands of observations each day distributed around the globe, including in areas which previously were poorly observed. These data are converted with the convert_cosmic_gps_cdf.f90 program and create DART observations of GPSRO_REFRACTIVITY.
The COSMIC satellites also provide ionospheric profiles of electron density. The accuracy is generally about 10-4 ∼ 10-5 cm-3. These data are converted with the convert_cosmic_ionosphere.f90 program and create DART observations tagged as COSMIC_ELECTRON_DENSITY.
Data from the COSMIC Program are available by signing up on the data access web page. We prefer delivery in netCDF file format.
The files we use as input to these conversion programs are
the Level 2 data, Atmospheric Profiles (filenames include
the string 'atmPrf').
Currently each vertical profile is stored in a separate file, and there are between 1000-3000 profiles/day, so converting a day's worth of observations involves downloading many individual files. There are plans in place by the data providers to bundle these profiles together in a tar file to make it simpler to download the raw data.
The scripts in the shell_scripts directory now by default try to download profiles from any of the available satellites that return GPS RO data to the CDAAC web site. See the do_convert.csh script for a list of what satellites are checked by default.
The files we have used as input to these conversion programs are from
the COSMIC 2013 Mission and have a data type of 'ionPrf'.
The file naming convention and file format are described by COSMIC here and there can be more than 1000 profiles/day, so converting a day's worth of observations involves downloading many individual files. COSMIC has instructions on ways to download the data at
The data are distributed in
file format. DART requires all observations to be in a proprietary
format often called DART "obs_seq" format.
The files in this directory (a combination
of C shell scripts and a Fortran source executable)
do this data conversion.
The shell_scripts directory contains several example scripts, including one which downloads the raw data files a day at a time (download_script.csh), and one which executes the conversion program (convert_script.csh). These scripts make 6 hour files by default, but have options for other times. Each profile is stored in a separate netcdf file and there are usually between 1000-3000 files/day, so the download process can be lengthy. You probably want to download as a separate preprocess step and do not use the script options to automatically delete the input files. Keep the files around until you are sure you are satisified with the output files and then delete them by hand.
The conversion executable convert_cosmic_gps_cdf, reads the namelist &convert_cosmic_gps_nml from the file input.nml.
The namelist lets you select from one of two different forward operators. The 'local' forward operator computes the expected observation value at a single point: the requested height at the tangent point of the ray between satellites. The 'non-local' operator computes values along the ray-path and does an integration to get the expected value. The length of the integration segments and height at which to end the integration are given in the namelist. In some experiments the difference between the two types of operators was negligible. This choice is made at the time of the conversion, and the type of operator is stored in the observation, so at runtime the corresponding forward operator will be used to compute the expected observation value.
The namelist also lets you specify at what heights you want observations to be extracted. The raw data is very dense in the vertical; using all values would not results in a set of independent observations. The current source code no longer does an intermediate interpolation; the original profiles appear to be smooth enough that this is not needed. The requested vertical output heights are interpolated directly from the full profile.
Each profile is interpolated to a set of desired levels that are specified at run time.
During the conversion process, each profile is checked for negative values of
electron density above the minimum desired level. If negative values are found,
the entire profile is discarded. If an observation sequence file already exists,
the converter will simply add the new observations to it. Multiple profiles may
be converted in a single execution, so it is easy to consolidate all the profiles
for a single day into a single observation sequence file, for example.
convert_cosmic_ionosphere reads the namelist &convert_cosmic_ionosphere_nml from the file input.nml.
The original observation times are preserved in the conversion process. If it is desired to subset the observation sequence file such that observations too far away from desired assimilation times are rejected, a separate post-processing step using the obs_sequence_tool is required. A script will be necessary to take a start date, an end date, an assimilation time step, and a desired time 'window' - and strip out the unwanted observations from a series of observation sequence files.
There are multiple ways of specifying the observation error variance at run time. They are implemented in a routine named electron_density_error() and are selected by the namelist variable observation_error_method.
|'constant'||a scalar value for all observations|
|'scaled'||the electron density is multiplied by a scalar value|
|'lookup'||a lookup table is read - UNSUPPORTED for now - see Known Bugs|
|'scaled_lookup'||the lookup table value is multiplied by a scalar value and the electron density value. Also (temporarily) UNSUPPORTED.|
I-Te Lee: " ... the original idea for error of ionospheric observation is 1%. Thus, I put the code as "oerr = 0.01_r8 * obsval". Liu et. al and Yue et al investigated the Abel inversion error of COSMIC ionosphere profile, both of them figure out the large error would appear at the lower altitude and push model toward wrong direction at the lower ionosphere while assimilating these profiles. On the other hand, the Abel inversion error depends on the ionospheric electron density structure, which is a function of local time, altitude and geomagnetic latitude. To simplify the procedure to define observation error of profiles, Xinan Yue help me to estimate an error matrix and saved in the file which named 'f3coerr.nc'. ... The number in the matrix is error percentage (%), which calculated by OSSE. Here are two reference papers. In the end, the observation error consists of instrumentation error (10%) and Abel error."
- X. Yue, W.S. Schreiner, J. Lei, S.V. Sokolovskiy, C. Rocken, D.C. Hunt, and Y.-H. Kuo (2010),
Error analysis of Abel retrieved electron density profiles from radio occultation measurements.
Annales Geophysicae: Atmospheres, Hydrospheres and Space Sciences. 28 No. 1, pp 217-222, doi:10.5194/angeo-28-217-2010
- J.Y. Liu, C.Y. Lin, C.H. Lin, H.F. Tsai, S.C. Solomon, Y.Y. Sun, I.T. Lee, W.S. Schreiner, and Y.H. Kuo (2010),
Artificial plasma cave in the low-latitude ionosphere results from the radio occultation inversion of the FORMOSAT-3/COSMIC}, Journal of Geophysical Research: Space Physics. 115 No. A7, pp 2156-2202, doi:10.1029/2009JA015079
It is possible to create observation sequence files for perfect model experiments that have realistic observation sampling patterns and observation error variances that do not have any actual electron densities. The COSMIC data files are read, but the electron density information is not written. Keep in mind that some methods of specifying the observation error variance require knowledge of the observation value. If the observation value is bad or the entire profile is bad, no observation locations are created for the profile.
This namelist is read from the file input.nml. Namelists start with an ampersand '&' and terminate with a slash '/'. Character strings that contain a '/' must be enclosed in quotes to prevent them from prematurely terminating the namelist.
&convert_cosmic_gps_nml obs_levels = -1.0 use_original_kuo_error = .false. local_operator = .true. ray_ds = 5000.0 ray_htop = 15000.0 gpsro_netcdf_file = 'cosmic_gps_input.nc' gpsro_netcdf_filelist = '' gpsro_out_file = 'obs_seq.gpsro' /
|obs_levels||integer(200)||A series of heights, in kilometers, where observations from this profile should be interpolated. (Note that the other distances and heights in the namelist are specified in meters.) The values should be listed in increasing height order.|
|use_original_kuo_error||logical||If .true. use the observation error variances for a refractivity observation that come from a Kuo paper and were implied to be used for the CONUS domain. If .false. use observation error variances similar to what is used in GSI.|
|local_operator||logical||If .true. compute the observation using a method which assumes all effects occur at the tangent point. If .false. integrate along the tangent line and do ray-path reconstruction.|
|ray_ds||real(r8)||For the non-local operator only, the delta stepsize, in meters, to use for the along-path integration in each direction out from the tangent point.|
|ray_htop||real(r8)||For the non-local operator only, stop the integration when one of the endpoints of the next integration step goes above this height. Specify in meters.|
|gpsro_netcdf_file||character(len=128)||The input filename when converting a single profile. Only one of the file or filelist items can have a valid value, so to use the single filename set the list name 'gpsro_netcdf_filelist' to the empty string (' ').|
|gpsro_netcdf_filelist||character(len=128)||To convert a series of profiles in a single execution create a text file which contains each input file, in ascii, one filename per line. Set this item to the name of that file, and set 'gpsro_netcdf_file' to the empty string (' ').|
|gpsro_out_file||character(len=128)||The output file to be created. To be compatible with earlier versions of this program, if this file already exists it will be read in and the new data will be appended to that file.|
A more useful example follows:
&convert_cosmic_gps_nml gpsro_netcdf_file = '' gpsro_netcdf_filelist = 'flist' gpsro_out_file = 'obs_seq.gpsro' local_operator = .true. use_original_kuo_error = .false. ray_ds = 5000.0 ray_htop = 13000.1 obs_levels = 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.2, 2.4, 2.6, 2.8, 3.0, 3.2, 3.4, 3.6, 3.8, 4.0, 4.2, 4.4, 4.6, 4.8, 5.0, 5.2, 5.4, 5.6, 5.8, 6.0, 6.2, 6.4, 6.6, 6.8, 7.0, 7.2, 7.4, 7.6, 7.8, 8.0, 8.2, 8.4, 8.6, 8.8, 9.0, 9.2, 9.4, 9.6, 9.8, 10.0, 10.2, 10.4, 10.6, 10.8, 11.0, 11.2, 11.4, 11.6, 11.8, 12.0, 12.2, 12.4, 12.6, 12.8, 13.0, 13.2, 13.4, 13.6, 13.8, 14.0, 14.2, 14.4, 14.6, 14.8, 15.0, 15.2, 15.4, 15.6, 15.8, 16.0, 16.2, 16.4, 16.6, 16.8, 17.0, 17.2, 17.4, 17.6, 17.8, 18.0, 19.0, 20.0, 21.0, 22.0, 23.0, 24.0, 25.0, 26.0, 27.0, 28.0, 29.0, 30.0, 31.0, 32.0, 33.0, 34.0, 35.0, 36.0, 37.0, 38.0, 39.0, 40.0, 41.0, 42.0, 43.0, 44.0, 45.0, 46.0, 47.0, 48.0, 49.0, 50.0, 51.0, 52.0, 53.0, 54.0, 55.0, 56.0, 57.0, 58.0, 59.0, 60.0, /
&convert_cosmic_ionosphere_nml input_file = '' input_file_list = 'input_file_list.txt' output_file = 'obs_seq.out' observation_error_file = 'none' observation_error_method = 'scaled_lookup' locations_only = .false. obs_error_factor = 1.0 verbose = 0 obs_levels = -1.0 /
|input_file||character(len=256)||The input filename when converting a single profile. Only one of the input_file or input_file_list items can have a valid value, so to use a single filename set input_file_list = ''|
|input_file_list||character(len=256)||To convert a series of profiles in a single execution create a text file which contains one filename per line. Set this item to the name of that file, and set input_file = ''|
|output_file||character(len=256)||The output file to be created. If this file already exists the new data will be added to that file. DART observation sequences are linked lists. When the list is traversed, the observations are in ascending time order. The order they appear in the file is completely irrelevant.|
|observation_error_file||character(len=256)||This specifies a lookup table. The table created by I-Te Lee and Xinan Yue is called f3coerr.nc.|
|observation_error_method||character(len=128)||There are multiple ways of specifying the observation error variance. This character string allows you to select the method. The selection is not case-sensitive. Allowable values are: 'constant', 'scaled', 'lookup', or 'scaled_lookup'. Anything else will result in an error. Look in the electron_density_error() routine for specifics.|
|locations_only||logical||If locations_only = .true. then the actual observation values are not written to the output observation sequence file. This is useful for designing an OSSE that has a realistic observation sampling pattern. Keep in mind that some methods of specifying the observation error variance require knowledge of the observation value. If the observation value is bad or the entire profile is bad, this profile is rejected - even if locations_only = .true.|
|obs_error_factor||real(r8)||This is the scalar that is used in several of the methods specifying the observation error variance.|
|verbose||integer||controls the amount of run-time output echoed to the screen. 0 is nearly silent, higher values write out more. The filenames of the profiles that are skipped are ALWAYS printed.|
|obs_levels||integer(200)||A series of heights, in kilometers, where observations from this profile should be interpolated. (Note that the other distances and heights in the namelist are specified in meters.) The values must be listed in increasing height order.|
A more useful example follows:
&convert_cosmic_ionosphere_nml input_file = '' input_file_list = 'file_list.txt' output_file = 'obs_seq.out' observation_error_file = 'f3coeff.dat' observation_error_method = 'scaled' locations_only = .false. obs_error_factor = 0.01 verbose = 1 obs_levels = 160.0, 170.0, 180.0, 190.0, 200.0, 210.0, 220.0, 230.0, 240.0, 250.0, 260.0, 270.0, 280.0, 290.0, 300.0, 310.0, 320.0, 330.0, 340.0, 350.0, 360.0, 370.0, 380.0, 390.0, 400.0, 410.0, 420.0, 430.0, 440.0, 450.0 /
convert_cosmic_gps_cdf and convert_cosmic_ionosphere almost use the same set of modules.
adaptive_inflate/adaptive_inflate_mod.f90 assim_model/assim_model_mod.f90 assim_tools/assim_tools_mod.f90 common/types_mod.f90 cov_cutoff/cov_cutoff_mod.f90 ensemble_manager/ensemble_manager_mod.f90 location/threed_sphere/location_mod.f90 models/template/model_mod.f90 mpi_utilities/null_mpi_utilities_mod.f90 obs_def/obs_def_mod.f90 obs_kind/obs_kind_mod.f90 obs_model/obs_model_mod.f90 obs_sequence/obs_sequence_mod.f90 observations/gps/convert_cosmic_ionosphere.f90 observations/utilities/obs_utilities_mod.f90 random_seq/random_seq_mod.f90 reg_factor/reg_factor_mod.f90 sort/sort_mod.f90 time_manager/time_manager_mod.f90 utilities/utilities_mod.f90
convert_cosmic_ionosphere also uses utilities/netcdf_utilities_mod.f90
Some COSMIC files seem to have internal times which differ from the
times encoded in the filenames by as much as 2-3 minutes. If it is
important to get all the observations within a particular time window
files with filenames from a few minutes before and after the window
should be converted.
Times really outside the window can be excluded in a separate step using the
convert_cosmic_ionosphere has a problem using the lookup table for observation error variances. Turn on array bounds checking and things get references outside array bounds. At this point, using the lookup table is unsupported until we can verify that portion of the code is working correctly. Since this code requires the netCDF file f3coerr.nc, we are intentionally not distributing the file until the code is verified to be working correctly.
The converters have a parameter declaring the maximum number of desired levels as 200. If more than 200 levels are entered as input (to obs_levels), a rather uninformative run-time error is generated:
ERROR FROM: routine: check_namelist_read message: INVALID NAMELIST ENTRY: / in namelist convert_cosmic_ionosphere_nml
Your error may be different if obs_levels is not the last namelist item before the slash '/'
convert_cosmic_gps_cdf currently bins the data
into 200 vertical bins as a first step, and then reinterpolates again
onto the requested vertical heights.
The original intent was to smooth out any noise in the dense profiles
before interpolating onto the final heights, but plots have shown that
the original data is smooth and can be used directly. There are plans
to remove the intermediate binning step.
Write a script that will take a start date, an end date, an assimilation time step, and a desired time 'window' - and strip out the unwanted observations from a series of observation sequence files. This will be useful for windowing the COSMIC_ELECTRON_DENSITY observations from convert_cosmic_ionosphere.
Fix the lookup table interpolation in convert_cosmic_ionosphere.
|Revision:||$Revision: 12139 $|
|Source:||$URL: https://svn-dares-dart.cgd.ucar.edu/DART/releases/classic/observations/gps/gps.html $|
|Change Date:||$Date: 2017-11-21 09:58:45 -0700 (Tue, 21 Nov 2017) $|
|Change history:||try "svn log" or "svn diff"|