Skip to main content

​
Overview

The dvnl_io module provides utilities for reading Defense Meteorological Satellite Program (DMSP) Visible Near-infrared (DVNL) GeoTIFF files with proper NoData handling.

​
read_raster_metadata

Reads basic metadata from a raster file without loading the full array. 
def read_raster_metadata(path: str) -> dict

​
Parameters

​
path
str
required
Path to the GeoTIFF file

​
Returns

​
metadata
dict
Dictionary containing raster metadata including:
  • width: Raster width in pixels
  • height: Raster height in pixels
  • crs: Coordinate reference system
  • transform: Affine transformation matrix
  • dtype: Data type of the raster
  • nodata: NoData value

​
Example

from TerraLab.light_pollution.dvnl_io import read_raster_metadata

meta = read_raster_metadata("C_DVNL_2022.tif")
print(f"Dimensions: {meta['width']} x {meta['height']}")
print(f"CRS: {meta['crs']}")
print(f"NoData value: {meta['nodata']}")

​
read_raster_window_filtered

Reads a specific window from a raster with NoData filtering. 
def read_raster_window_filtered(
    path: str,
    window: Window
) -> np.ndarray

​
Parameters

​
path
str
required
Path to the GeoTIFF file
​
window
rasterio.windows.Window
required
Window object specifying the region to read (row offset, column offset, height, width)

​
Returns

​
array
np.ndarray
Float32 array with NoData values replaced by np.nan

​
NoData Handling

This function handles NoData in two ways:
  1. Metadata NoData: If the raster has a NoData value in its metadata, those pixels are converted to np.nan
  2. Outlier Detection: Values exceeding 1e10 are also converted to np.nan (DVNL data sometimes has extreme values without proper metadata)

​
Example

import rasterio
from rasterio.windows import Window
from TerraLab.light_pollution.dvnl_io import read_raster_window_filtered

# Read a 1024x1024 window starting at (0, 0)
window = Window(col_off=0, row_off=0, width=1024, height=1024)
data = read_raster_window_filtered("C_DVNL_2022.tif", window)

# Count valid pixels
valid_pixels = np.sum(~np.isnan(data))
print(f"Valid pixels: {valid_pixels} / {data.size}")

​
Use Cases

  • Memory-efficient processing: Read only the region of interest
  • Tiled processing: Process large rasters in smaller chunks
  • Convolution operations: Read with halo overlap for kernel operations

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us