I am trying to merge and reproject MODIS 250m NDVI product (https://lpdaac.usgs.gov/dataset_discovery/modis/modis_products_table/myd13a3_v006) to EASE2.0 grid. I am interested in projecting these to 9km EASE2.0 grid (defined by NSIDC for SMAP mission), but I have tried with the 36km EASE2.0 grid as well and I have the same problem as below.
I first use
pymodis library to merge all the tiles from each month of MODIS data and generate a mosaic version of the data.
Then I use
gdalwarp with the following command to project to EASE2.0 grid:
gdalwarp -overwrite -tr 9008.055210146 9008.055210146 -t_srs '+proj=cea +lat_0=0 +lon_0=0 +lat_ts=30 +x_0=0 +y_0=0 +ellps=WGS84 +datum=WGS84 +units=m' -r average mosaic_input.tif output.tif
The issue is that the output.tif file does not have the same lat/lon in each pixel as the EASE2.0 9km grid. The number of columns in the longitudinal direction are the same, but the corresponding longitudes are not the same. These are the longitudes of the center of pixels from columns 1 to 10:
179.895906749369 179.989345967666 -179.917214814038 -179.823775595741 -179.730336377444 -179.636897159148 -179.543457940851 -179.450018722554 -179.356579504258 -179.263140285961
The same issue happens with latitudes. The upper latitude in the output.tif file is
86.9278324335268 (center of the pixel), while the 9km EASE2.0 grid has an upper latitude of
I should add that I am sure the merging with
pymodis is working fine because when I use the mosaic file with
gdalwarp and reproject it to EPSG 4326 (rectangular lat/lon grid) the output file has the the correct lat lon info.