[R-sig-Geo] adapting spatial points and wrld_smpl to a reference system implicit in a .nc file
Agus Camacho
agus.camacho at gmail.com
Tue Feb 23 23:41:17 CET 2016
Thanks again, but with gdalinfo and gdal_translate giving NULL after
running to the end, i just cant do that..
2016-02-23 15:36 GMT-07:00 Chris Reudenbach <reudenbach at uni-marburg.de>:
> Agus
>
> sorry I did miss the crucial part:
>
> if you are doing as suggestest you have to define manually the Lambertian
> ymin xmax ymin ymax using the header information of the nc file.
>
> here an example for the unstaggered data
>
> library(gdalUtils)
> library(raster)
> library(proj4)
> library(mapview)
>
> r<-gdal_translate('NETCDF:"geo_em.d01.nc":LANDMASK', 'landmask.tif',
> of="GTiff",
> ot="Byte",
> output_Raster=TRUE,
> verbose=TRUE)
>
> finfo <- gdalinfo("geo_em.d01.nc")
> ##extract parameters
> lat_1=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#TRUELAT1",
> finfo))],"=")[[1]][2])
> lat_2=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#TRUELAT2",
> finfo))],"=")[[1]][2])
> lat0=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#CEN_LAT",
> finfo))],"=")[[1]][2])
> lon0=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#CEN_LON",
> finfo))],"=")[[1]][2])
> dx=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#DX",
> finfo))],"=")[[1]][2])
> dy=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#DY",
> finfo))],"=")[[1]][2])
> y=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#SOUTH-NORTH_PATCH_END_UNSTAG",
> finfo))],"=")[[1]][2])
> x=as.numeric(strsplit(finfo[which(grepl("NC_GLOBAL#WEST-EAST_PATCH_END_UNSTAG",
> finfo))],"=")[[1]][2])
> x_0=0
> y_0=0
>
> ## generate compliant prj4 string
> projLcc=paste("+proj=lcc +lat_1=",lat_1," +lat_2=",lat_2,
> " +lat_0=",lat0," +lon_0=",lon0, " +x_0=",x_0,"
> +y_0=",y_0,sep="")
>
> ## project centre coordinates
> tr <- ptransform(cbind(lon0, lat0)/180*pi,'+proj=longlat +datum=WGS84
> +no_defs',proj)
>
> ## calculate extent using the Lambertian units (m)
> xmin=as.integer(tr[1,1]-(x/2*dx-dx))
> xmax=as.integer(tr[1,1]+(x/2*dx-dx))
> ymin=as.integer(tr[1,2]-(y/2*dy-dy))
> ymax=as.integer(tr[1,2]+(y/2*dy-dy))
> wrfLccExt<-extent(xmin,xmax,ymin,ymax)
> extent(r) <- extent(wrfLccExt)
> projection(r) <- projLcc
> mapview(r)
>
>
> cheers Chris
>
>
>
> Am 23.02.2016 um 22:22 schrieb Chris Reudenbach:
>
>> Augustin
>>
>> just quick and dirty if you run gdalinfo("geo_em.d01.nc") your are
>> getting the information about the corner coordinates the subdatasets and so
>> on. Together with Dominiks suggestion you can do something like this:
>>
>> library(gdalUtils)
>> library(raster)
>> Sys.setenv(GDAL_NETCDF_BOTTOMUP="YES")
>> wrffake<- "+proj=longlat +ellps=sphere +a=6370000 +b=6370000 +units=m
>> +no_defs"
>> x<-gdal_translate('NETCDF:"geo_em.d01.nc":LANDMASK', 'landmask.tif',
>> of="GTiff",
>> ot="Byte",
>> output_Raster=TRUE,
>> verbose=TRUE)
>> wrfExt<-extent(-151.29639,-48.703613,12.355667,50.26619)
>> extent(x) <- extent(wrfExt)
>> projection(x) <- wrffake
>> plot(x)
>>
>> Some remarks:
>> (1) I just took the first pair of coordinates as derived from gdalinfo("
>> geo_em.d01.nc")
>> you will find 4 different coordinate pairs (i did not proof which one is
>> right
>>
>> The data is staggered (as outlined by Dominik) So some of the corner
>> coordinates belongs to the staggered data and the others coordinates to
>> the unstaggered ones. You will find them marked
>>
>> If you have installed the netcdf libs you easily can use ncview
>> geo_em.d01.nc or ncdump -h geo_em.d01.nc to view the data or get more
>> information of the header.
>>
>> Hope this helps
>>
>> cheers Chris
>>
>>
>>
>> Am 23.02.2016 um 21:11 schrieb Agus Camacho:
>>
>>> Thanks for that Dominik,
>>>
>>> Giving that projection to either the locations, the raster layer
>>> generated
>>> from the .nc file, or both, still did not work. I keep having locations
>>> that should be on land falling far on the sea. Might this be a problem
>>> derived from using raster with a file whose original grid distances are
>>> not
>>> constant?
>>>
>>> Here is a link with the original file which has the original coordinate
>>> data.
>>>
>>> https://www.dropbox.com/s/qpt5twtunhy3x3x/geo_em.d01.nc?dl=0
>>>
>>>
>>> 2016-02-23 12:07 GMT-07:00 Dominik Schneider <
>>> Dominik.Schneider at colorado.edu
>>>
>>>> :
>>>> This looks like WRF <http://www.wrf-model.org/index.php> data. I just
>>>> dealt with this.
>>>> The data is on a sphere as opposed to WGS84 so you need +ellps=sphere
>>>> +a=6370000 +b=6370000 +units=m
>>>>
>>>> +proj=lcc which is usually what wrf is run with.
>>>> The tricky part is:
>>>> +lat_1=25.0 +lat_2=45.0 +lat_0=38.0 +lon_0=-100.0
>>>> because every WRF run is different (the WRF Preprocessing System
>>>> optimizes
>>>> the projection for the domain).
>>>> and then there is probably no shift so you need(?) +x_0=0 +y_0=0
>>>>
>>>> This gives:
>>>> +proj=lcc +lat_1=25.0 +lat_2=45.0 +lat_0=38.0 +lon_0=-100.0
>>>> +ellps=sphere
>>>> +a=6370000 +b=6370000 +units=m +no_defs
>>>>
>>>> But, wrf users like to give out lat and long so you need to assign it:
>>>> +proj=longlat +ellps=sphere +a=6370000 +b=6370000 +units=m +no_defs
>>>>
>>>> and then reproject the lat/long to lcc coordinates using this string:
>>>> +proj=lcc +lat_1=25.0 +lat_2=45.0 +lat_0=38.0 +lon_0=-100.0
>>>> +ellps=sphere
>>>> +a=6370000 +b=6370000 +units=m +no_defs
>>>>
>>>> One word of caution, make sure you received the correct coordinates.
>>>> Some
>>>> variables are run cell center while some are run at cell edge. It looks
>>>> like from your .nc file it was made by your collaborator so I assume
>>>> they
>>>> are right.
>>>>
>>>> That said, another word of caution, I found that the XLAT and XLONG
>>>> variables from WRF output aren't very precise. There is a "geogrid" file
>>>> from the preprocessing system that has the domain corners, resolution,
>>>> nrow
>>>> and ncol from which you can make a better grid using the native
>>>> projection
>>>> system (in my case it was a 4km grid). I suggest you try to get those.
>>>>
>>>> I hope this helps... I have to run but wanted to save people too much
>>>> head
>>>> scratching. I can get you running with more help tonight if you need.
>>>> Dominik
>>>>
>>>>
>>>> On Tue, Feb 23, 2016 at 11:27 AM, Agus Camacho <agus.camacho at gmail.com>
>>>> wrote:
>>>>
>>>> Thanks heaps to all for your effort. If I go to another GEOSTAT ill
>>>>> bring
>>>>> more giant crab this time.
>>>>>
>>>>> The creator of the .nc file also looked at this webpage:
>>>>> http://www.pkrc.net/wrf-lambert.html
>>>>> It seemed like the right proj4 string might be this one:
>>>>>
>>>>> +proj=lcc +lat_1=25.0 +lat_2=45.0 +lat_0=38.0
>>>>> +lon_0=-100.0 +a=6370 +b=6370 +towgs84=0,0,0 +no_defs
>>>>>
>>>>> However this projection also does not allow me to adequately plot the
>>>>> locations on the raster.
>>>>>
>>>>> Here is the .nc file. it contains several layers.
>>>>>
>>>>>
>>>>>
>>>>> https://www.dropbox.com/s/yto3linsgom3zi7/results_us_future_output_none_0.nc?dl=0
>>>>>
>>>>>
>>>>>
>>>>> 2016-02-23 2:25 GMT-07:00 Michael Sumner <mdsumner at gmail.com>:
>>>>>
>>>>> On Tue, 23 Feb 2016 at 20:09 Roger Bivand <Roger.Bivand at nhh.no> wrote:
>>>>>>
>>>>>> On Tue, 23 Feb 2016, Alex Mandel wrote:
>>>>>>>
>>>>>>> I made an attempt at it too. Investigating the original data, I'm
>>>>>>>>
>>>>>>> not
>>>>>
>>>>>> sure that the projection information supplied is correct for the
>>>>>>>>
>>>>>>> data
>>>>>
>>>>>> linked. When I load up the data in a unprojected space, the
>>>>>>>>
>>>>>>> coordinates
>>>>>
>>>>>> don't look at all similar to any Lambert projected data I have, they
>>>>>>>> actually look like Lat/Lon in some unprojected coordinate system,
>>>>>>>> perhaps a different spheroid than expected.
>>>>>>>>
>>>>>>> Does anyone have a link to the original data? Is is possible that
>>>>>>>
>>>>>> this is
>>>>>
>>>>>> the General Oblique Transformation used by modellers - that is
>>>>>>>
>>>>>> something
>>>>>
>>>>>> that feels like longlat but is recentred and oblique? Example at the
>>>>>>>
>>>>>> very
>>>>>
>>>>>> end of my GEOSTAT talk last year (slides 81-83):
>>>>>>>
>>>>>>> http://geostat-course.org/system/files/geostat_talk_150817.pdf
>>>>>>>
>>>>>>> Roger
>>>>>>>
>>>>>>>
>>>>>>> For what it is worth, the General Oblique Transformation is not the
>>>>>> only
>>>>>> example - it's very common for modellers to have a mesh that has the
>>>>>> "mostly-properties" of a projection, but is not actually describable
>>>>>>
>>>>> with
>>>>>
>>>>>> standard transform + affine parameters. The main cases that I've seen
>>>>>>
>>>>> are
>>>>>
>>>>>> polar stereographic, equal area or oblique Mercator. Often they really
>>>>>>
>>>>> are
>>>>>
>>>>>> simple transforms and you can reconstruct without loss, but it's not
>>>>>> usually possible to tell without exploration. It's an interesting
>>>>>> dis-connect to see code that builds a mesh with certain properties,
>>>>>> then
>>>>>> only stores longitudes and latitudes - when it could be done with
>>>>>>
>>>>> standard
>>>>>
>>>>>> tools and be stored and used much more efficiently.
>>>>>>
>>>>>> (I've seen Lambert Conformal Conic and Lambert Azimuthal Equal Area
>>>>>> terminology conflated in this context too. )
>>>>>>
>>>>>> I'm also interested to explore the original data.
>>>>>>
>>>>>> Cheers, Mike.
>>>>>>
>>>>>>
>>>>>>
>>>>>> -Alex
>>>>>>>>
>>>>>>>> On 02/22/2016 10:17 PM, Frede Aakmann Tøgersen wrote:
>>>>>>>>
>>>>>>>>> Hi
>>>>>>>>>
>>>>>>>>> I tried to make it work but I had to give up. I wanted to reproject
>>>>>>>>>
>>>>>>>> the
>>>>>>
>>>>>>> Lamberth conformal conic coordinates to long-lat but it didn't work.
>>>>>>>
>>>>>>>> Perhaps someone can see what I did wrong. Here is what I did (data
>>>>>>>>>
>>>>>>>> in
>>>>>
>>>>>> R
>>>>>>
>>>>>>> binary format and figure in png format both attached):
>>>>>>>
>>>>>>>> library(raster)
>>>>>>>>> library(maptools)
>>>>>>>>> data(wrld_simpl)
>>>>>>>>>
>>>>>>>>> r <- raster("raster.grd")
>>>>>>>>> projection(r)
>>>>>>>>> ## > NA
>>>>>>>>>
>>>>>>>>> uro <- read.table("clean urosaurus records.csv", h = TRUE, sep =
>>>>>>>>>
>>>>>>>> ",")
>>>>>
>>>>>> coordinates(uro) <- ~lon+lat
>>>>>>>>>
>>>>>>>>> ## Set projections for the 3 data sets
>>>>>>>>>
>>>>>>>>> ## Lamberth's confocal conic projection with given parameters
>>>>>>>>> crs(r) <- "+proj=lcc +lat_0=38.0 +lon_0=-100 +lat_1=25.0
>>>>>>>>>
>>>>>>>> +lat_2=45.0
>>>>>
>>>>>> +ellps=WGS84"
>>>>>>>
>>>>>>>> projection(r)
>>>>>>>>>
>>>>>>>>> ## Assume that lon, lat are geographical coordinates (degrees
>>>>>>>>>
>>>>>>>> decimal)
>>>>>
>>>>>> proj4string(uro) <- CRS("+proj=longlat +datum=WGS84")
>>>>>>>>>
>>>>>>>>> ## wrld_simpl is in geographical coordinates
>>>>>>>>> proj4string(wrld_simpl)
>>>>>>>>>
>>>>>>>>> ## Make figure in png format
>>>>>>>>> ## Of course plotting data with 2 different projections will give
>>>>>>>>> ## some distortions
>>>>>>>>> pdf("uro.png")
>>>>>>>>>
>>>>>>>>> plot(r)
>>>>>>>>> points(uro)
>>>>>>>>> plot(wrld_simpl, add = TRUE) # World will be clipped to extent of
>>>>>>>>>
>>>>>>>> 'r'
>>>>>
>>>>>> dev.off()
>>>>>>>>>
>>>>>>>>> extent(r)
>>>>>>>>> ## class : Extent
>>>>>>>>> ## xmin : -131.4368
>>>>>>>>> ## xmax : -68.56323
>>>>>>>>> ## ymin : 12.35567
>>>>>>>>> ## ymax : 50.26619
>>>>>>>>>
>>>>>>>>> ## Reproject the raster to long-lat
>>>>>>>>> ## This doesn't work (collapsed domain)
>>>>>>>>> rp <- projectRaster(r, crs = "+proj=longlat +datum=WGS84 +no_defs
>>>>>>>>>
>>>>>>>> +ellps=WGS84 +towgs84=0,0,0")
>>>>>>>
>>>>>>>> ## Because
>>>>>>>>>
>>>>>>>>>> extent(rp)
>>>>>>>>>>
>>>>>>>>> ## class : Extent
>>>>>>>>> ## xmin : -100.0015
>>>>>>>>> ## xmax : -99.68557
>>>>>>>>> ## ymin : 37.70658
>>>>>>>>> ## ymax : 38.00046
>>>>>>>>>
>>>>>>>>> ## Save data in R binary format
>>>>>>>>> save(list = c("r", "uro", "wrld_simpl"), file = "uro.RData")
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Yours sincerely / Med venlig hilsen
>>>>>>>>>
>>>>>>>>> Frede Aakmann Tøgersen
>>>>>>>>> Specialist, M.Sc., Ph.D.
>>>>>>>>> Plant Performance & Modeling
>>>>>>>>>
>>>>>>>>> Technology & Service Solutions
>>>>>>>>> T +45 9730 5135
>>>>>>>>> M +45 2547 6050
>>>>>>>>> frtog at vestas.com
>>>>>>>>> http://www.vestas.com
>>>>>>>>>
>>>>>>>>> Company reg. name: Vestas Wind Systems A/S
>>>>>>>>> This e-mail is subject to our e-mail disclaimer statement.
>>>>>>>>> Please refer to www.vestas.com/legal/notice
>>>>>>>>> If you have received this e-mail in error please contact the
>>>>>>>>>
>>>>>>>> sender.
>>>>>
>>>>>>
>>>>>>>>>
>>>>>>>>> -----Original Message-----
>>>>>>>>> From: R-sig-Geo [mailto:r-sig-geo-bounces at r-project.org] On
>>>>>>>>>
>>>>>>>> Behalf Of
>>>>>
>>>>>> Agus Camacho
>>>>>>>
>>>>>>>> Sent: 22. februar 2016 19:20
>>>>>>>>> To: tech at wildintellect.com
>>>>>>>>> Cc: r-sig-geo
>>>>>>>>> Subject: Re: [R-sig-Geo] adapting spatial points and wrld_smpl to a
>>>>>>>>>
>>>>>>>> reference system implicit in a .nc file
>>>>>>>
>>>>>>>> Thanks Alex, but the locations still fall in the sea when i plot
>>>>>>>>>
>>>>>>>> them
>>>>>
>>>>>> using
>>>>>>>
>>>>>>>> your recommended Solution. I looked at the sites you proposed and
>>>>>>>>>
>>>>>>>> they
>>>>>
>>>>>> have
>>>>>>>
>>>>>>>> other values for lat_1, lat_0, etc..
>>>>>>>>>
>>>>>>>>> 2016-02-22 11:04 GMT-07:00 Alex M <tech_dev at wildintellect.com>:
>>>>>>>>>
>>>>>>>>> On 02/22/2016 09:50 AM, Agus Camacho wrote:
>>>>>>>>>>
>>>>>>>>>>> Dear all,
>>>>>>>>>>>
>>>>>>>>>>> Im trying to overlap these points:
>>>>>>>>>>>
>>>>>>>>>>>
>>>>> https://www.dropbox.com/s/awdclg4cvsdngej/clean%20urosaurus%20records.csv?dl=0
>>>>>
>>>>>> and a wrld_simpl object:
>>>>>>>>>>> library(maptools)
>>>>>>>>>>> data(wrld_simpl)
>>>>>>>>>>>
>>>>>>>>>>> Over this raster layer
>>>>>>>>>>>
>>>>>>>>>>>
>>>>> https://www.dropbox.com/sh/qcw174tgogpnz7s/AAByDc3TeyFe3W4nEqTFix6Oa?dl=0
>>>>>
>>>>>> This rastr comes from a .nc file without a reference system. The
>>>>>>>>>>>
>>>>>>>>>> author
>>>>>>>
>>>>>>>> of
>>>>>>>>>>
>>>>>>>>>>> that .nc file gave me the following data about the .nc.
>>>>>>>>>>>
>>>>>>>>>>> The projection is *Lambert conformal conic* projection
>>>>>>>>>>> CEN_LAT = 38.0
>>>>>>>>>>> CEN_LON = -100.0
>>>>>>>>>>> TRUELAT1 = 25.
>>>>>>>>>>> TRUELAT2 = 45.
>>>>>>>>>>>
>>>>>>>>>>> However, despite i have gone through many sites in the internet,
>>>>>>>>>>>
>>>>>>>>>> i
>>>>>
>>>>>> cant
>>>>>>>
>>>>>>>> figure it out:
>>>>>>>>>>>
>>>>>>>>>>> a) if that is all the data i need to set a reference system for
>>>>>>>>>>>
>>>>>>>>>> my
>>>>>
>>>>>> points
>>>>>>>
>>>>>>>> and the wrld_simp object.
>>>>>>>>>>>
>>>>>>>>>>> b) how to change a typical CRS object with such data
>>>>>>>>>>>
>>>>>>>>>>> Ex.CRS ("+proj=lcc+lat_0=38.0+lon0_2=-100+ellps=WGS84")
>>>>>>>>>>>
>>>>>>>>>>> Where do i enter the TRUELAT and CENLAT values?
>>>>>>>>>>> Are there any site that explains easily what the fields in the
>>>>>>>>>>>
>>>>>>>>>> CRS
>>>>>
>>>>>> mean
>>>>>>>
>>>>>>>> and
>>>>>>>>>>
>>>>>>>>>>> how to change them?
>>>>>>>>>>>
>>>>>>>>>>> Thanks in advance.
>>>>>>>>>>>
>>>>>>>>>>> https://github.com/OSGeo/proj.4/wiki/GenParms
>>>>>>>>>> https://trac.osgeo.org/proj/wiki/GenParms
>>>>>>>>>>
>>>>>>>>>> I believe:
>>>>>>>>>> +lat_0 = CEN_LAT Latitude of origin
>>>>>>>>>> +lat_1 = TRUELAT1 Latitude of first standard parallel
>>>>>>>>>> +lat_2 = TRUELAT2 Latitude of second standard parallel
>>>>>>>>>> +lon_0 = CEN_LON Central meridian
>>>>>>>>>>
>>>>>>>>>> proj strings are defined by the proj4 libary. It's website listed
>>>>>>>>>>
>>>>>>>>> above
>>>>>>
>>>>>>> and the associated mailing lists or gis stackexchange would be the
>>>>>>>>>> places to get help on it.
>>>>>>>>>> https://lists.osgeo.org/mailman/listinfo/metacrs
>>>>>>>>>>
>>>>>>>>>> It often helps to browse similar projections on
>>>>>>>>>> http://spatialreference.org/
>>>>>>>>>> http://epsg.io/
>>>>>>>>>>
>>>>>>>>>> Enjoy,
>>>>>>>>>> Alex
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>> R-sig-Geo mailing list
>>>>>>>> R-sig-Geo at r-project.org
>>>>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>>>>>>
>>>>>>> --
>>>>>>> Roger Bivand
>>>>>>> Department of Economics, Norwegian School of Economics,
>>>>>>> Helleveien 30, N-5045 Bergen, Norway.
>>>>>>> voice: +47 55 95 93 55; fax +47 55 95 91 00
>>>>>>> e-mail: Roger.Bivand at nhh.no
>>>>>>> http://orcid.org/0000-0003-2392-6140
>>>>>>> https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en
>>>>>>> http://depsy.org/person/434412
>>>>>>> _______________________________________________
>>>>>>> R-sig-Geo mailing list
>>>>>>> R-sig-Geo at r-project.org
>>>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>>>>>
>>>>>> --
>>>>>> Dr. Michael Sumner
>>>>>> Software and Database Engineer
>>>>>> Australian Antarctic Division
>>>>>> 203 Channel Highway
>>>>>> Kingston Tasmania 7050 Australia
>>>>>>
>>>>>> [[alternative HTML version deleted]]
>>>>>>
>>>>>> _______________________________________________
>>>>>> R-sig-Geo mailing list
>>>>>> R-sig-Geo at r-project.org
>>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>>>>
>>>>>>
>>>>>
>>>>> --
>>>>> Agustín Camacho Guerrero.
>>>>> Doutor em Zoologia.
>>>>> Laboratório de Herpetologia, Departamento de Zoologia, Instituto de
>>>>> Biociências, USP.
>>>>> Rua do Matão, trav. 14, nº 321, Cidade Universitária,
>>>>> São Paulo - SP, CEP: 05508-090, Brasil.
>>>>>
>>>>> [[alternative HTML version deleted]]
>>>>>
>>>>> _______________________________________________
>>>>> R-sig-Geo mailing list
>>>>> R-sig-Geo at r-project.org
>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>>>
>>>>>
>>>>
>>>
>> _______________________________________________
>> R-sig-Geo mailing list
>> R-sig-Geo at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>
>>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>
--
Agustín Camacho Guerrero.
Doutor em Zoologia.
Laboratório de Herpetologia, Departamento de Zoologia, Instituto de
Biociências, USP.
Rua do Matão, trav. 14, nº 321, Cidade Universitária,
São Paulo - SP, CEP: 05508-090, Brasil.
[[alternative HTML version deleted]]
More information about the R-sig-Geo
mailing list