qgis plugins - Creating raster layer from numpy array using pyqgis?

I am working on a plugin for Qgis to calculate spatial Kernel density maps. I have all the calculations working, all I am missing is a way to turn a Numpy Array, with density values into a multiband raster layer.

Do I have to create a geotiff on a temp file using Gdal and then load it?

Or is there a direct way to create the layer from data in memory?

if so, how to do it?

Answer

Here is the code that I use to convert an array to gdal raster saving it to the disk, "param" is a dicitionary containing gdal parameters (check the gdal documentation) and "array" is a numpy array. Than you can instantiate a QgsMapLayer with your file as source. You have to create the geotiff in the disk.

    from osgeo import gdal as osgdal  # Adapt the import to fit yor environement.

    driver = osgdal.GetDriverByName(param['out_format'])

    dataset = driver.Create(
            param['dst_filename'],

            param['x_pixels'],
            param['y_pixels'],
            1,
            osgdal.GDT_Float32,
            )

    dataset.SetGeoTransform((
            param['xmin'],           #0
            param['pixel_size'],     #1
            0,                       #2

            param['ymin'],           #3
            0,                       #4
            param['pixel_size']))    #5

    out_srs = osr.SpatialReference()
    out_srs.ImportFromEPSG(param['SRID'])

    dataset.SetProjection(out_srs.ExportToWkt())
    dataset.GetRasterBand(1).WriteArray(array.T)  # Remove "T" if it's inverted.
    dataset = None