Pre-processing Nighttime lights data

This repo contains code to scale and normalize nighttime lights data from multiple sensors or datasets. Processes data for use in RGIS hydrological model, which uses a particular raster format. Needs GHAAS and RGIS code installed. Alternatively, edit SConstruct file to not run RGIS commands and leave processed data in AsciiGrid format, or use GDAL to convert to another raster format.

Processing code runs as an SCons build. SConstruct contains code to define a computational graph with input dependencies for each output file. Run with scons to read/execute SConstruct file, which then calls shell commands or functions in lib.py.

First download data using get_*.py scripts.

Edit hard-coded data directory paths in SConstruct. Then run with scons $STNres where STNres is resolution of STN digital river network for final output.

Sensors:

DMSP-OLS

• And older data set, but available for a relatively long time frame (1992-2013)
• Calibration is uneven, varies between sensor (flew on multiple satellites), and there is drift over sensor lifetime
• Sensor saturation challenges

DMSP-OLS Radiance Calibrated

• Apply limited radiance calibration data.
• Not available for full DMSP-OLS data set

VIIRS

• Higher resolution that DMSP-OLS, radiance calibrated
• Data comes from the Earth Observations Group at NOAA/NCEI
• Monthly composites available from 2012-2018
• Annual composites under development
• Cleaned, but still substaintial noise/erros from sun, moon glint, clouds,...

VIIRS Black Marble VNP46

• Roman et al., 2018, Remote Sens of Envi.
• More complete cleaning, best product in terms of data quality, currently available only for 2012 and 2016
• More soon, to be available on LADAAC