README.md
In lightson: Download and Analyse Nighttime Lights Satellite Data

lightson

Download and analyse nighttime lights satellite data from R.

lightson supports two data sources:

ISRO Bhuvan NTL: pre-aggregated state radiance for India (2012-2024); WMS raster tiles for any region.
NASA VIIRS Black Marble: annual (VNP46A4), monthly (VNP46A3), and QA-filtered daily (VNP46A2) calibrated radiance from 2012.

The ISRO Bhuvan NTL source does not require any authentication. VIIRS requires a free NASA Earthdata account.

India at night, 2012-2024

pak::pak("saketlab/lightson")

library(lightson)

df <- bhuvan_stats()          # all states, 2012-2024
head(df)
#>        state year mean_radiance
#> 1 Andhra Pradesh 2012      22.4

bihar <- bhuvan_stats(state = "BIHAR")
plot_ntl_trend(bihar, region = "Bihar", caption = "Source: ISRO Bhuvan")

states <- c("Bihar", "Maharashtra", "Tamil Nadu", "Uttar Pradesh", "Kerala")
plot_ntl_trend(
  df[df$state %in% states, ],
  region   = states,
  subtitle = "Bhuvan state mean radiance (nW/cm^2/sr)",
  caption  = "Source: ISRO Bhuvan NTL portal"
)

library(sf)
library(terra)

states_sf <- get_india_admin("state")   # bundled LGD/SoI boundaries, full J&K

rasters <- bhuvan_raster("IND", years = 2022, bbox_pad = c(0, 0, 0, 3.5))
plot_ntl_map(rasters[["2022"]], polygons = states_sf, dark = TRUE)
plot_ntl_map(rasters[["2022"]], polygons = states_sf, palette = "bhuvan")

districts <- sf::read_sf("https://bharatviz.org/India-bhuvan-districts.geojson")

rasters <- bhuvan_raster(districts, years = 2018:2023)
panel   <- extract_panel(rasters, districts, id_col = "district_name")

head(panel)
#>   region_id year mean_radiance n_pixels
#> 1  Adilabad 2018          14.2     3821

plot_ntl_trend(panel, region = c("Mumbai", "Delhi", "Chennai", "Bengaluru"))

bhuvan_raster() accepts any sf object (including your own shapefile/geojson).

VIIRS gives physical radiance in nW/cm^2/sr, which is useful when you need calibrated values or coverage outside India.

Getting a token:

Create a free account at https://urs.earthdata.nasa.gov
Accept the MERIS and Sentinel EULAs under Profile > EULAs
Go to Profile > Generate Token and copy the EDL-... string
Add it to ~/.Renviron: EARTHDATA_TOKEN=EDL-xxxxxxxxxxxxxxxxxxxx

# read EARTHDATA_TOKEN from environment
token <- earthdata_token()   
rasters <- ntl_download(region = "IND", years = 2020:2025, token = token)

# Monthly and near-current daily Black Marble
monthly <- ntl_download("IND", 2025, token, product = "monthly")
daily <- ntl_download("IND", token = token, product = "daily",
                      dates = as.Date("2025-11-12"))

panel <- extract_panel(rasters, get_india_admin("state"), id_col = "state_name")
plot_ntl_trend(panel, region = c("Bihar", "Maharashtra"))

You can also set EARTHDATA_USERNAME and EARTHDATA_PASSWORD in ~/.Renviron instead and earthdata_token() will fetch the token for you.

Bhuvan WMS rasters are classed maps with four colours (< 5, 5-25, 26-200 and > 200 nW/cm^2/sr). bhuvan_raster() converts each class to a representative radiance: the midpoints 2.5, 15 and 112.5, and 200 (the floor) for the open-ended top class. These are banded values, so the brightest areas are underestimated. Use VIIRS when you need continuous calibrated radiance, and don't pool the two sources in a regression without normalising first.

Bhuvan NTL data from 2024 onwards uses VNP46A4 Collection 2.0; 2012-2023 uses Collection 1.0. The two collections use different calibration algorithms, so estimates aren't directly comparable across that boundary. lightson warns automatically when a request spans both.