Multi-Sensor Land Surface Temperature (2018–2024)

Multi-Sensor Land Surface Temperature (2018–2024)

This dataset provides multi-temporal geophysical and topographic variables derived from Landsat 8, Landsat 9 Collection 2 Tier 1 Level-2 data, and MODIS products (MOD11A2). Data are processed within Google Earth Engine (GEE) and cover Northern Italy, bounded by coordinates [8°E, 43°N, 15°E, 46°N].

The temporal coverage spans January 2018 to December 2024 with five aggregation levels: ~1-day, ~8-day, ~16-day, and ~monthly (dates depend on mutual availability of MODIS and LANDAT) and 1-dayb a single day aggregation (regardless LANDAST or MODIS for a specific date are not available). All products are resampled to a uniform 2.5 km spatial resolution. Landsat products are available in two groups:

Non cloud-masked — no cloud masking applied

Cloud-masked — cloud masking applied using a GEE-based Landsat cloud detection algorithm

MODIS LST products are provided without additional cloud masking, as MODIS LST is generated with internal quality flags and cloud screening.

Variables and Bands

Variable
Unit / Type
Description

LST_K
K
Landsat 8–9 Land Surface Temperature (emissivity-corrected, Planck-based)

NDVI
unitless
Landsat 8–9 Normalized Difference Vegetation Index

emissivity
unitless
Surface thermal emissivity derived from NDVI

cloud_mask
binary (0/1)
0 = clear, 1 = cloud (Landsat 8–9 only)

MODIS_LST_K
K
MODIS Land Surface Temperature

LST_diff
K
Difference between Landsat 8–9 LST and MODIS LST (LST_K - MODIS_LST_K)

DEM
m
Digital Elevation Model (SRTM 30 m resampled to 2.5 km)

lon, lat
degrees
Longitude and latitude coordinates

Land Surface Temperature (LST) Calculation

Landsat 8–9 LST is derived from the thermal infrared band ST_B10 using radiometric scaling and a Planck-based emissivity correction:

LST_K = BT / (1 + (λ * BT / ρ) * ln(ε))

BT — Brightness temperature from ST_B10 (Kelvin): BT = 0.00341802 × ST_B10 + 149.0

λ — Band 10 effective wavelength = 10.895 µm

ρ — Constant = 14388 µm·K

ε — Surface emissivity estimated from NDVI

Emissivity and NDVI are computed as:

NDVI = (SR_B5 - SR_B4) / (SR_B5 + SR_B4)
Fv = ((NDVI - 0.2) / 0.5)^2 bounded to [0, 1]
ε = 0.986 + 0.004 × Fv

MODIS LST is directly retrieved from MOD11A2 products with scale factor applied. Differences (LST_diff) allow inter-sensor comparison and evaluation of systematic biases between Landsat 8–9 and MODIS LST retrievals.

Usage Notes

All products are projected in EPSG:4326 and resampled to 2.5 km resolution for sensor consistency.

Landsat composites include both raw and cloud-filtered versions to support reproducible studies.

MODIS LST does not include a separate cloud mask; internal quality flags ensure reliable LST values