Creating GeoTIFF from Pandas DataFrame

Question

I have a Pandas DataFrame that looks like this:

SIF_743,latitude,longitude
0.7626782059669495,-85.90831756591797,22.40107536315918
0.3139636516571045,-85.9532470703125,22.630943298339844
0.17063966393470764,-86.00765228271484,21.14249610900879
0.3484233617782593,-85.9980697631836,22.866853713989258
0.20155811309814453,-86.05291748046875,21.364904403686523
-0.17634174227714539,-86.10398864746094,19.85099220275879
-0.3291313946247101,-86.15137481689453,18.326879501342773
0.10919605940580368,-86.04275512695312,23.10919952392578
0.3178038001060486,-86.09805297851562,21.59358787536621

Is it possible to convert the data where "latitude" and "longitude" will represent a pixel and the "SIF_743" value will represent the DN value of that pixel with EPSG:4326 CRS?

Answer 1

Here is a possible solution:

import numpy as np
import pandas as pd
import geopandas as gpd
from geocube.api.core import make_geocube


A = np.array([
    [0.7626782059669495,-85.90831756591797,22.40107536315918],
    [0.3139636516571045,-85.9532470703125,22.630943298339844],
    [0.17063966393470764,-86.00765228271484,21.14249610900879],
    [0.3484233617782593,-85.9980697631836,22.866853713989258],
    [0.20155811309814453,-86.05291748046875,21.364904403686523],
    [-0.17634174227714539,-86.10398864746094,19.85099220275879],
    [-0.3291313946247101,-86.15137481689453,18.326879501342773],
    [0.10919605940580368,-86.04275512695312,23.10919952392578],
    [0.3178038001060486,-86.09805297851562,21.59358787536621]
])

xmin = np.min(A[:,2])
xmax = np.max(A[:,2])
ymin = np.min(A[:,1])
ymax = np.max(A[:,1])
deltax = (xmax - xmin)/10
deltay = (ymax - ymin)/10
res = min([deltay, deltay])

df = pd.DataFrame(A)
df.columns = ['SIF_743','latitude','longitude',]

gdf = gpd.GeoDataFrame(
    df,
    geometry=gpd.points_from_xy(df.longitude, df.latitude)
)

gdf.plot() # first image hereunder

geotif_file = "/tmp/raster.tif"
out_grd = make_geocube(
    vector_data=gdf,
    measurements=["SIF_743"],
    resolution=(-res, res)
)

out_grd["SIF_743"].rio.to_raster(geotif_file)

Your data doesn't seem to be regular so it will not be straightforward to build an image out of this, unless you do some interpolation:

You will end up with this raster close to the south pole:

Documentation is here: https://corteva.github.io/geocube/html/geocube.html

Answer 2

A great tool for this is geocube. It will rasterize each of the columns in your geopackage (or a subset if you specify it).

Some examples of usage are listed here

To convert your csv file to geopandas: https://github.com/geopandas/geopandas/issues/2221

It can be as simple to use as:

import geopandas
from geocube.api.core import make_geocube

gdf = geopandas.read_file(
    "test_shp.csv",
    driver="CSV",
    x_possible_names="x,lon,longitude",
    y_possible_names="y,lat,latitude"
)

out_grid = make_geocube(
    vector_data=gdf,
    measurements=["SIF_743"],
    resolution=(-0.0001, 0.0001),
)

out_grid["SIF_743"].rio.to_raster("my_rasterized_column.tif")

Answer 3

You could convert the dataframe to geodataframe with geopandas and export it to vector format supported by OGR and then use gdal_rasterize.