svenmeijboom/geospatially_enriched_ndvi

Geospatially Enriched NDVI (16-Day Terra/MODIS)

This dataset transforms raw 16-day MODIS NDVI grids into a per-pixel time series enriched with hierarchical administrative boundaries. It covers every 0.1°×0.1° land pixel worldwide from 2000 onward and is partitioned for efficient bulk download and selective access.

Dataset Contents

• Partitioned Parquet files
  Stored under:

ndvi/
├── year=YYYY/
│ ├── country=Netherlands/
│ │ └── data_0.parquet
│ └── country=India/
│ │ └── data_0.parquet
│ │ └── data_1.parquet
│ └── ...
└── …

• manifest.json
  A JSON array listing each file’s path and the 16-day dates it contains.

Schema

Column	Type	Description
latitude	decimal(4,2)	Pixel center latitude (–89.95 to 89.95, step 0.1°)
longitude	decimal(5,2)	Pixel center longitude (–179.95 to 179.95, step 0.1°)
date	date	Snapshot date (YYYY-MM-DD) for each 16-day interval
vegetation_index	decimal(3,3)	NDVI value (–0.1 to 0.9) or NULL for ocean/cloud pixels
country	varchar	ISO-normalized country name
level_1…level_5	varchar	Nested administrative levels (province, municipality, etc.)

Loading & Usage

from datasets import load_dataset

ds = load_dataset("svenmeijboom/geospatially_enriched_ndvi", split="train")