values of bands are not the same in QGIS about rasterio HOT 4 CLOSED

@IAGO1215 without reproducible example nor data nor screenshots it's pretty hard for use to tell what's going on. I can simply make a guess that your data is a COG with internal overview and in QGIS it's the overview that you are seeing

from rasterio.

Dear Sarago,

I have uploaded the raster I am using here. (200MB)

As for the reproducible example:
First, we can open the raster in QGIS, and check the "Information" tab where we can find the max, mean, min... of each band.
Then we can read the first band by rasterio in VSCode. And check the max, mean, min... They should be identical to those shown in QGIS.
a = rio.open('Band') a = a.read(1).reshape(-1) a = a[~np.isnan(a)] a.max()

from rasterio.

With GDAL

GDAL_DISABLE_READDIR_ON_OPEN=FALSE gdalinfo OutputPCA_30_PCs -stats -json | jq '.bands[0]'
{
  "band": 1,
  "block": [
    1137,
    1
  ],
  "type": "Float32",
  "colorInterpretation": "Undefined",
  "description": "PC 1",
  "minimum": -71.904,
  "maximum": 42.271,
  "mean": 0.001,
  "stdDev": 12.35,
  "metadata": {
    "": {
      "STATISTICS_MAXIMUM": "42.271053314209",
      "STATISTICS_MEAN": "0.0011011385352394",
      "STATISTICS_MINIMUM": "-71.90389251709",
      "STATISTICS_STDDEV": "12.349624688556",
      "STATISTICS_VALID_PERCENT": "8.865"
    }
  }
}

With Rasterio

with rasterio.Env(GDAL_DISABLE_READDIR_ON_OPEN=False):
    with rasterio.open("OutputPCA_30_PCs") as src:
        arr = src.read(indexes=1)
        non_nan = arr[numpy.isfinite(arr)]
        print(non_nan.min(), non_nan.max())

-71.90389 42.271053

In QGIS

from rasterio.

Dear Sarago,

thanks for the prompt reply and the issue has been fixed, and it is not even related to rasterio, which shows the correct max/min values of the raster.

Instead the issue is related to the settings in QGIS on my side. To fix it, I have set "Accuracy" from "Estimate" to "Actual" (as shown in your screenshot), and now it is showing correctly in QGIS as well.

Thanks again for the help.

Best

from rasterio.