Harmonized Landsat Sentinel (HLS): One Consistent Time Series from Two Sensors

What Is HLS, and Why Does It Exist?

If you monitor a single place over time with optical satellites, you hit a wall fast: clouds. The Sentinel-2 constellation revisits every 5 days at the Equator (SentiWiki); each Landsat satellite repeats every 16 days, with Landsat 8 and 9 offset by 8 days (USGS) — but subtract the cloudy passes and you may get only a handful of usable scenes per month. Miss the two-week window when a crop emerges, a field floods, or a forest is cleared, and the event is gone before your next clear image.

Harmonized Landsat Sentinel (HLS) solves this by merging two satellite families into one consistent time series. Landsat 8/9 and Sentinel-2 observe the same wavelengths at similar resolution, but their data isn't directly comparable out of the box — different atmospheric processing, slightly different band wavelengths, and different grids. HLS is a NASA product that reconciles all of these so that a Landsat image and a Sentinel-2 image taken three days apart can be stacked and compared as if they came from the same instrument. It began as research at NASA Goddard; the algorithm comes from the HLS Science Team and production is run by the IMPACT team at Marshall Space Flight Center (NASA Earthdata).

The payoff: NASA describes the combined record as observing the land every 1.6 days on average at 30m, covering all land except Antarctica, open to everyone, with data typically available two to three days after acquisition (NASA Earthdata).

The Two HLS Products: HLSL30 and HLSS30

HLS ships as two parallel products on a shared 30m grid:

Product	Source sensor	Record starts	Native resolution	Resampled to	Bands
HLSL30 (Earthdata)	Landsat 8/9 OLI (+TIRS)	2013-04-11	30m	30m	Coastal, Blue, Green, Red, NIR, SWIR1, SWIR2, Cirrus, + two thermal bands
HLSS30 (Earthdata)	Sentinel-2A/2B/2C MSI	2015-11-28	10-60m	30m	Coastal, Blue, Green, Red, Red-Edge ×3, NIR broad, NIR narrow, Water vapor, Cirrus, SWIR1, SWIR2

Both are delivered as Cloud-Optimized GeoTIFFs, one file per band, tiled to the Sentinel-2 Military Grid Reference System (MGRS) tiles, with surface reflectance values, sun and view angle bands, and a quality-assessment (Fmask) layer for clouds, cloud shadow, snow, and water. Because Sentinel-2 carries red-edge bands that Landsat lacks, HLSS30 has a few extra bands — but the common bands are harmonized to be interchangeable. Since February 2025 NASA also publishes HLS vegetation index products (NASA Earthdata).

How Harmonization Actually Works

Turning two sensor archives into one usable dataset takes four processing steps, all described in the HLS v2.0 user guide (HLS User Guide, LP DAAC):

1. Atmospheric correction. Both archives are converted to surface reflectance (removing haze and atmospheric scattering) with the same code — a USGS C version of LaSRC v3.5.5 — so a "0.3 NIR" means the same thing in both.
2. Cloud and shadow masking. In v2.0 the cloud and shadow masks come exclusively from Fmask 4.7, which flags clouds, shadows, snow, and water in every scene, so you can filter contaminated pixels the same way regardless of source.
3. Spectral bandpass adjustment. Landsat and Sentinel-2 bands are centered on slightly different wavelengths. HLS v2.0 adjusts the Sentinel-2 bands to the Landsat OLI bandpasses with a linear fit, derived from hyperspectral spectra over diverse land cover, so the values line up.
4. Common gridding and BRDF normalization. Landsat data are gridded to the Sentinel-2 MGRS tiles, both are resampled to 30m, and reflectance is normalized to a nadir view (NBAR) using BRDF coefficients derived from the MODIS BRDF product — removing brightness differences caused by how the sensor was looking at the scene.

The result is a stack where the satellite that took each image no longer matters for analysis — only the date does.

When HLS Beats a Single Sensor

HLS is not always the right choice. Here's where the fused time series earns its keep:

Use case	Why HLS helps
Crop & phenology monitoring	Capturing green-up, peak, and senescence needs frequent observations; doubling cadence catches transitions a single sensor misses.
Flood & rapid disaster response	More passes means a higher chance of a clear image soon after the event.
Cloud-prone regions (tropics, monsoon)	More satellites mean more chances of a clear view through persistent cloud cover.
Consistent multi-year change detection	A harmonized record avoids artificial "jumps" when you switch sensors mid-series.

When you should not reach for HLS:

• You need sub-30m detail — use native Sentinel-2 (10m) instead.
• You need all-weather coverage — of the sensors discussed here, only SAR (Sentinel-1) sees through clouds; see SAR vs Optical: Choosing the Right Sensor.
• You need a multi-decadal baseline before 2013 — the native Landsat archive reaches back to 1972 (USGS), while HLSL30 starts in April 2013 with Landsat 8.

HLS, NDVI, and Index Time Series

Because HLS bands are harmonized, spectral indices computed from them are far more comparable across sensors than indices from the native products. An NDVI of 0.7 from an HLSL30 scene and an NDVI of 0.7 from an HLSS30 scene three days later should describe the same vegetation state rather than an artifact of different instruments. The harmonization is not perfect: the user guide notes that the bandpass fit minimizes the OLI–MSI difference on the land covers it was derived from, and the difference may be greater over other distinctive land cover types (HLS User Guide). That consistency is exactly what makes HLS valuable for the kind of vegetation index time series and anomaly detection that single-sensor records struggle to support cleanly.

For a deeper look at why combining sensors outperforms any one of them, see Multi-Index Satellite Monitoring and SAR-Optical Data Fusion Techniques.

Where HLS Appears in Off-Nadir Delta

The live map on the Sentinel-2 viewer shows Sentinel-2 passes in natural color from the HLS S30 product (served through NASA's GIBS) for a date you pick — one day, or five days stacked newest on top — so you can see where clear views exist. Area monitoring, by contrast, runs on native Sentinel-2 Level-2A (and Sentinel-1, NISAR and VIIRS), not on HLS: you draw a polygon, pick an index, and each new scene's value is checked against the area's previous observations. To see how continuous monitoring works end to end, read How to Monitor Any Area with Satellite Time Series.

Key Takeaways

• HLS = Landsat 8/9 + Sentinel-2A/2B/2C reconciled into one consistent 30m surface-reflectance dataset.
• HLSL30 comes from Landsat, HLSS30 from Sentinel-2; both share the Sentinel-2 MGRS grid and a common quality mask.
• Harmonization means atmospheric correction, cloud masking, bandpass adjustment, and BRDF normalization — so indices are comparable across sensors.
• The real benefit is temporal density: an observation every 1.6 days on average, which improves cloud-prone and time-sensitive monitoring.
• Reach for native Sentinel-2 when you need 10m detail, and SAR when you need to see through clouds.

HLS data is produced by NASA and distributed by the LP DAAC under NASA's open data policy. Landsat is a joint USGS/NASA program; Sentinel-2 is part of the EU Copernicus programme (ESA).