[R-sig-Geo] Gaussian grids in R

2010-07-01 Thread Matthew Landis
Dear R-sig-geo members, I wonder if any of you have suggestions about how to import and use rasters with a spectral Gaussian T62 CRS in R. I'm not very familiar with this coordinate system (commonly used for climate models), and I haven't been able to discover tools for dealing with them in

[R-sig-Geo] Nugget from variogram.fit vs GLS

2010-07-01 Thread Zev Ross
Hi All, I'm verifying coefficients generated with GSTAT kriging with external drift using GLS. The coefficients match absolutely perfectly. The range also matches perfectly. But there is a big difference in what GLS gives as the output for the nugget. Can anyone explain why this is? Zev

[R-sig-Geo] spTransform changes the class

2010-07-01 Thread Agustin Lobo
Hi! I use spTransform to reproject: class(eugrd025) [1] SpatialGrid attr(,package) [1] sp projection(eugrd025) [1] +proj=longlat +ellps=WGS84 eugrd025EFDC - spTransform(eugrd025, CRSobj=CRS(projection(eugrd025EFDC_SPDF))) projection(eugrd025EFDC) [1] +proj=laea +lat_0=52 +lon_0=10

Re: [R-sig-Geo] spTransform changes the class

2010-07-01 Thread Michael Sumner
spTransform cannot reproject a grid - this will (usually) requires destructive resampling of the data to a completely new grid. There are functions in the raster package to do this: projectRaster. I think this would work, but there may be important details that you will need to investigate:

Re: [R-sig-Geo] spTransform changes the class

2010-07-01 Thread Michael Sumner
Ah, sorry that was completely wrong. I think this is right, but still untested: raster.eugrd025 - raster(eugrd025) pr - projectExtent(raster.eugrd025, projection(eugrd025EFDC_SPDF)) eugrd025EFDC - projectRaster(raster.eugrd025, pr) On Fri, Jul 2, 2010 at 1:46 AM, Michael Sumner

Re: [R-sig-Geo] spTransform changes the class

2010-07-01 Thread isabelle boulangeat
Hi! Are you sure that the *spTransform *method can be applied to a SpatialGrid object? I think it only works for SpatialPoints, SpatialLines and SpatialPolygons. If you are working with rasters, perhaps you should use the function *projectRaster *from the package raster. I hope it helps... Isa.

Re: [R-sig-Geo] spTransform changes the class

2010-07-01 Thread Michael Sumner
I forgot to mention the first time, spTransform will not reproject a grid but it will coerce the grid to points and then return reprojected versions of those. That's why Agustin sees the class change from grid to points. You can see this in the method definitions for SpatialGrid/PixelDataFrame,

Re: [R-sig-Geo] spTransform changes the class

2010-07-01 Thread Robert J. Hijmans
The general idea is that for changing the projection of a raster you provide the input RasterLayer and a Raster object with the spatial parameters (projection, extent, resolution) to which it should be transformed. For the latter, you can use the RasterLayer that you want to project the input data

Re: [R-sig-Geo] Nugget from variogram.fit vs GLS

2010-07-01 Thread Edzer Pebesma
Hi Zev, it seems that gls returns the relative nugget rather than the absolute returned by gstat: 1.30231/(1.413246+1.30231) [1] 0.479574 Could that be? I'm quite surprised results are that similar, doesn't gls use REML to fit variograms? Best wishes from a vivid geostat 2010, On 07/01/2010

Re: [R-sig-Geo] spTransform changes the class

2010-07-01 Thread Agustin Lobo
Thank you all for your ideas. Please note though that the goal is to create a raster as a reference object for running resample(). The original data is a high resolution raster in EPSG 3035 (Br) that I want to resample to a lon,lat grid of 0.25 deg resolution (Frankly, I'm astonished that such a