adapting spatial points and wrld_smpl to a reference system implicit in a .nc file

Folks:

Just noticed this thread -- I see I didn't include a "--config" option
with any of the gdalUtils functions (it isn't one of the documented
parameters on the individual utility website, but it seems it would have
allowed you to run the GDAL_NETCDF_BOTTOMUP without setting a system
environment variable -- see
http://lists.osgeo.org/pipermail/gdal-dev/2014-July/039452.html).  If I
make this tweak and pushed it to r-forge, would one of you be willing to
see if it solves the problem?

I assume this will be something that would be needed for any gdal utility
that allows an "of" to be set?

--j

On Wed, Feb 24, 2016 at 2:37 AM Chris Reudenbach <
reudenbach at uni-marburg.de> wrote:

Agus,

sorry for the addon but I think I have to provide a correction of the
corner coordinates (e.g. the extent values):

In the example that you have posted below I did calculate the extent
using the domain center coordinate and the WRF grid resolution in meter
and the number of rows and cols.

Since Dominik provides a link to the file description of the netcdf file
I think it is more accurate to reproject the  corner coordinates  as
given by the netcdf header variables (NC_GLOBAL#corner_lats,
NC_GLOBAL#corner_lons). Assuming that your variable "dctmx" (which I can
not identfy it in the nc file)  is of type "Mass" (staggered = M) the
correct corner coordinates are stored as the first 4 entrys of the dump
snippet below:

lats<-"

NC_GLOBAL#corner_lats={12.355667,50.26619,50.26619,12.355667,12.308136,50.18787,50.18787,12.308136,12.210785,50.403816,50.403816,12.210785,12.163345,50.325382,50.325382,12.163345}"

lons<-"

NC_GLOBAL#corner_lons={-131.43678,-151.29639,-48.703613,-68.563232,-131.5851,-151.51157,-48.488434,-68.414917,-131.38828,-151.41891,-48.581085,-68.611725,-131.53641,-151.63441,-48.36557,-68.463593}"

after cleaning and converting the strings you may calculate the corner
coordinates:


library(proj4)
## project mass corner coordinates to lambertian
llMass <- ptransform(cbind(clon[1],clat[1])/180*pi,'+proj=longlat
+datum=WGS84 +no_defs',proj)
ulMass <- ptransform(cbind(clon[2],clat[2])/180*pi,'+proj=longlat
+datum=WGS84 +no_defs',proj)
lrMass <- ptransform(cbind(clon[3],clat[3])/180*pi,'+proj=longlat
+datum=WGS84 +no_defs',proj)
urMass <- ptransform(cbind(clon[4],clat[4])/180*pi,'+proj=longlat
+datum=WGS84 +no_defs',proj)
wrfLccExtMass<-extent(ulMass[1],lrMass[1],llMass[2],ulMass[2])


According to this the correct extent for mass variables should be:

extent(wrfLccExtMass)
class       : Extent
xmin        : -3575343
xmax        : 3575342
ymin        : -2293058
ymax        : 2306330


hope this is correct now

cheers Chris

Am 24.02.2016 um 05:12 schrieb Agus Camacho:

With the help of everybody i finally got this to work, so here is a

script

that does the job of reprojecting both, a raster layer obtained from a

.nc

and some locations in order to overplot them using plot.raster or

mapview.

Im using a combination of the advices of Dominik, Michael and Chris.


require(ncdf4)
require(raster)


setwd("C:/~")


r=raster("geo_em.d01.nc",
               varname="dctmx")# days of ctmax events

# Set extent and projections of rasters for plotting
# chris gave me the orig data fom the nc file because i could not

install

gdal
xmin= -3545999
xmax= 3546000
ymin = -2286000
ymax=2286000
pr<- "+proj=lcc +lat_1=25 +lat_2=45 +lat_0=38.000008 +lon_0=-100 +x_0=0
+y_0=0"

wrfLccExt<-extent(xmin,xmax,ymin,ymax)

extent(r) <- extent(wrfLccExt)
projection(r) <- pr

# get and prepare  urosaurus locations

x=read.csv("C:/Users/Agus/Dropbox/Functional Biogeography -
Copy/scripts/class 4 maxent model/clean urosaurus records.csv")
x=x[,1:3]
colnames(x)
coordinates(x)=~lon+lat
proj4string(x)<- CRS("+proj=longlat +datum=WGS84")
x=spTransform(x, pr)


# get prepare and plot wrld_simpl
require(maptools)
require(sp)
data(wrld_simpl)
namer <- spTransform(subset(wrld_simpl, NAME %in% c("United States",
"Canada", "Mexico")), prj)

#plot with raster
plot(r, cex.main=.7,legend=F)
points(x)
plot(namer, add = TRUE)

# plot with mapview (cool!)
m=mapview(r)
u=mapview(x)
m+u


Thanks to all!
Agus

2016-02-23 13:11 GMT-07:00 Agus Camacho <agus.camacho at gmail.com>:

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 T?gersen 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 T?gersen
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

--
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

--
Dr. Michael Sumner
Software and Database Engineer
Australian Antarctic Division
203 Channel Highway
Kingston Tasmania 7050 Australia

         [[alternative HTML version deleted]]

--
Agust?n Camacho Guerrero.
Doutor em Zoologia.
Laborat?rio de Herpetologia, Departamento de Zoologia, Instituto de
Bioci?ncias, USP.
Rua do Mat?o, trav. 14, n? 321, Cidade Universit?ria,
S?o Paulo - SP, CEP: 05508-090, Brasil.

         [[alternative HTML version deleted]]

--
Agust?n Camacho Guerrero.
Doutor em Zoologia.
Laborat?rio de Herpetologia, Departamento de Zoologia, Instituto de
Bioci?ncias, USP.
Rua do Mat?o, trav. 14, n? 321, Cidade Universit?ria,
S?o Paulo - SP, CEP: 05508-090, Brasil.