Rasterio misidentifying all pixels in band as Nodata

Question

Main goal: Create pandas DataFrame from raster image - one row per pixel

I have been using this resource

Main problem: Rasterio is improperly marking every pixel as NoData instead of pulling the proper values

I am working with .tif files from SoilGrids.org which give soil attributes such as sand, silt, clay, ph, etc.. Each .tif file has only one band. When I open the .tif files in ArcMap, I clearly see the values of the raster and everything displays properly.

When I display in rasterio using

img=rasterio.open("C:\\Users\\Research\\Documents\\Tara_Fall_2019\\Senegal_LCC\\SG_data\\Senegal_cec_0-5cm_mean_clipped.tif")
show(img,0)

I get the following image which shows the correct raster

but when I try to pull the actual data using

#read image 
band=img.read(1)
band

I get an array full of nodata values

For reference, here are the properties of the raster pulled from arcmap

I have tried opening other .tif rasters from the same source with all the same properties and have the same problem. My guess is it is something wonky about the properties of the .tif files.

Answer 1

What makes you think it's all nodata? Just when you print out the array?

When you're printing out the array you only see three row and column values from each corner, which in this case are nodata.

Try printing out some stats instead:

tif = "mydata.tif"

with rasterio.open(tif) as img:
    band = img.read(1, masked=True)  # read band 1, return a nodata MaskedArray
    print(band)
    print(band.min(), band.max())

For my data, this outputs the following. Note the -- values are how a MaskedArray represents nodata values

[[-- -- -- ... -- -- --]
 [-- -- -- ... -- -- --]
 [-- -- -- ... -- -- --]
 ...
 [-- -- -- ... -- -- --]
 [-- -- -- ... -- -- --]
 [-- -- -- ... -- -- --]]  <-- Can only see nodata
211 351                    <-- But there are other values

If I don't ask for a MaskedArray, I can still see that it is not returning all nodata values as the min is still the same:

with rasterio.open(tif) as img:
    band = img.read(1)  # read band 1, return a numpy ndarray
    print(band)
    print(band.min(), band.max())

Output:

[[32767 32767 32767 ... 32767 32767 32767]
 [32767 32767 32767 ... 32767 32767 32767]
 [32767 32767 32767 ... 32767 32767 32767]
 ...
 [32767 32767 32767 ... 32767 32767 32767]
 [32767 32767 32767 ... 32767 32767 32767]
 [32767 32767 32767 ... 32767 32767 32767]]  <-- Can only see nodata
211 32767                                    <-- But there are other values