Machine learning human footprint index (ml-HFI)

Summary: This dataset introduces a novel machine learning-based Human
Footprint Index (ml-HFI) with 300-meter spatial resolution, with values
ranging from 0 to 100, where 0 represents intact natural areas and higher
values indicate increasing human pressure. Method: The ml-HFI is developed
using a convolutional neural network (CNN) trained on an existing Human
Footprint Index (HFI) dataset, with Landsat imagery as input features.
This approach builds upon the approach by Keys et al. (2021) and removes
dependencies on externally processed datasets, making it a fully
self-sufficient index that only requires Landsat data for
calculation.  Landsat imagery serves as the input data,
pre-processed using Google Earth Engine to remove cloud contamination and
ensure consistent quality, including cloud, snow, and shadow masking, and
annual median composites to reduce noise.