Sentinel-2 L2A

Available now Collection id sentinel-2-l2a. Aliases include sentinel-2-c1-l2a (Earth Search), S2L2A (Sentinel Hub) and SENTINEL2_L2A (CDSE openEO), matched without regard to case.

Overview

Sentinel-2 is the Copernicus programme’s optical mission. Each satellite carries the MultiSpectral Instrument (MSI), which records 13 bands from visible light to shortwave infrared at 10, 20 and 60 m. Level-2A is bottom-of-atmosphere reflectance, produced by ESA’s Sen2Cor processor, which also writes a scene classification layer and cloud and snow probabilities.

satellitedata.net serves Collection 1, ESA’s consistent reprocessing of the archive, as Element 84 indexes it in Earth Search under sentinel-2-c1-l2a. The files are cloud-optimized GeoTIFFs in public S3 buckets in AWS us-west-2, the same region the API runs in. Item ids stay native, for example S2A_T31TCF_20260829T105044_L2A.

Mission references: ESA SentiWiki on the Sentinel-2 mission, its processing levels and products; the AWS registry entry for Sentinel-2 COGs.

Basic facts

Satellites	Sentinel-2A (launched 2015), 2B (2017) and 2C (2024); item platform is sentinel-2a, sentinel-2b or sentinel-2c
Spatial resolution	10 m (B02, B03, B04, B08), 20 m (red edge, B8A, SWIR, SCL), 60 m (B01, B09)
Revisit	5 days at the equator with two satellites in orbit, more often at higher latitudes
Swath	290 km
Tiling	MGRS grid; each item covers one 110 × 110 km tile, such as 31TCF (item property grid:code = MGRS-31TCF)
Processing level	Level-2A, bottom-of-atmosphere reflectance (Sen2Cor), Collection 1
Temporal extent in the API	From 27 June 2015, as the live collection document reports; see data availability for gaps
Zoom range for tiles	8 to 15
Credit line	Contains modified Copernicus Sentinel data {years}

Bands

Spectral bands, with the wavelengths and bandwidths Earth Search publishes in each item’s eo:bands. ESA publishes slightly different spectral response figures per satellite; see SentiWiki for those.

Band	Common name	Earth Search asset	Also accepted	Centre (nm)	Bandwidth (nm)	Resolution	Type	Scale	Offset	Nodata
B01	coastal	coastal	CoastalAerosol	443	27	60 m	uint16	0.0001	−0.1	0
B02	blue	blue	B2	490	98	10 m	uint16	0.0001	−0.1	0
B03	green	green	B3	560	45	10 m	uint16	0.0001	−0.1	0
B04	red	red	B4	665	38	10 m	uint16	0.0001	−0.1	0
B05	rededge071	rededge1	RedEdge1	704	19	20 m	uint16	0.0001	−0.1	0
B06	rededge075	rededge2	RedEdge2	740	18	20 m	uint16	0.0001	−0.1	0
B07	rededge078	rededge3	RedEdge3	783	28	20 m	uint16	0.0001	−0.1	0
B08	nir	nir	NIR_Broad, B8	842	145	10 m	uint16	0.0001	−0.1	0
B8A	nir08	nir08	NIR_Narrow	865	33	20 m	uint16	0.0001	−0.1	0
B09	nir09	nir09	WaterVapor	945	26	60 m	uint16	0.0001	−0.1	0
B11	swir16	swir16	SWIR1	1610	143	20 m	uint16	0.0001	−0.1	0
B12	swir22	swir22	SWIR2	2190	242	20 m	uint16	0.0001	−0.1	0

Earth Search’s own common name for B05, B06 and B07 is rededge; the API uses the STAC 1.1 names rededge071, rededge075 and rededge078 so that each one is unique.

Class, probability and atmosphere layers:

Layer	Earth Search asset	Also accepted	Content	Resolution	Type	Scale	Offset	Nodata
SCL	scl		Scene classification, classes 0 to 11	20 m	uint8	none	0	0
CLD	cloud	MSK_CLDPRB	Cloud probability, 0 to 100 %	20 m	uint8	none	0	0
SNW	snow	MSK_SNWPRB	Snow probability, 0 to 100 %	20 m	uint8	none	0	0
AOT	aot		Aerosol optical thickness at 550 nm	20 m	uint16	0.001	0	0
WVP	wvp		Water vapour column in cm	20 m	uint16	0.001	0	0

Every name in the tables resolves in POST /v1/process: the native name, the common name, the asset name and the aliases, without regard to case. The GeoTIFF’s band descriptions are the Earth Search asset names.

B08 or B8A

Sentinel-2 has two near-infrared bands. B08 (nir) is the broad 10 m band. B8A (nir08) is the narrow 20 m band that matches Landsat OLI band 5. For anything that compares with Landsat, such as NDVI over both, ask for nir08. The map presets use B08, as the web map does. See the B08 trap.

B10, the 1,375 nm cirrus band, is not part of Level-2A. ESA uses it during atmospheric correction and keeps it only in Level-1C, which is coming later.

Units and values

Stored values are integers. Collection 1 reflectance is:

reflectance = DN × 0.0001 − 0.1

The −0.1 offset is ESA’s BOA_ADD_OFFSET of −1,000, introduced with processing baseline 04.00 in January 2022. Collection 1 reprocessed the older scenes to the same convention, so every item in this collection carries the offset, and its raster:bands metadata says so. Do not choose the offset from the acquisition date.

What you get depends on three processing options in POST /v1/process:

units	harmonize	Spectral bands come back as	Sample type
REFLECTANCE (default)	either	DN × 0.0001 − 0.1	FLOAT32, nodata NaN
DN	offset_removed (default)	DN − 1,000, so DN = 10,000 × reflectance	INT16
DN	none	stored DN, with the scale and offset written into the GeoTIFF	UINT16, nodata 0

• clipNegative: true sets converted values below zero to zero, which is what Sentinel Hub’s harmonizeValues does. On the older download API the same switch is clamp: true.
• output.sampleType is optional. If you give it, it must match the table. INT16 on its own selects DN with offset_removed.
• With REFLECTANCE, AOT and WVP get their scale too (unitless and cm). SCL, CLD and SNW always come back as stored.

See Values, scale and offset and Harmonization levels.

Cloud and quality layers

Each item has eo:cloud_cover, the percentage of the scene Sen2Cor flags as cloud. Filter scenes with it in search (query), tiles (maxCloudCoverage) and process (dataFilter.maxCloudCoverage).

For pixels, request SCL as a band:

SCL	Class
0	No data
1	Saturated or defective pixel
2	Dark area or topographic shadow
3	Cloud shadow
4	Vegetation
5	Not vegetated
6	Water
7	Unclassified
8	Cloud, medium probability
9	Cloud, high probability
10	Thin cirrus
11	Snow or ice

A common clear-pixel rule keeps classes 4, 5 and 6. Class 7 is a judgement call; Sentinel Hub and Earth Engine treat it as cloud. Sen2Cor draws no buffer around clouds, so growing the cloud and shadow areas by a few pixels helps at the edges.

CLD and SNW give Sen2Cor’s cloud and snow probability per pixel, 0 to 100, if you want your own threshold instead of SCL’s classes.

Lean v1 applies no pixel cloud mask: tiles and previews show every pixel of the scenes they mosaic, and processing.cloudMask is rejected. The unified cloud mask, built from SCL with optional dilation, is planned.

Data availability

Earth Search rebuilt its Sentinel-2 index as sentinel-2-c1-l2a from ESA’s Collection 1. On 16 October 2026 we compared it with Earth Search’s original sentinel-2-l2a over two test areas, in Borneo (about 117.9 to 118.2 E, 1.5 to 1.7 N) and Sulawesi (about 120.0 to 120.2 E, 3.2 to 3.4 S). Collection 1 was complete for 2020 to 2021 and 2023 to 2026. It was missing these scenes:

Year	Borneo: missing of original	Sulawesi: missing of original
2017	31 of 31	20 of 21
2018	75 of 115	39 of 59
2019	30 of 144	15 of 73
2022	132 of 144	67 of 73

These are gaps observed in the source collection over two small areas, not a global figure. The gaps may be smaller elsewhere, and they may close as reprocessing continues. Before relying on a period, run /v1/search for your area and dates and count what comes back.

Values before 2022. Scenes acquired before baseline 04.00 are ESA’s reprocessed product, with newer atmospheric correction and geometry than the products first published for those dates. Their values can differ from originally processed products, such as Sentinel Hub’s, by up to about 0.02 reflectance. From baseline 04.00 (2022) on, the two match.

Collection-specific parameters

Search

Lean v1 search takes the STAC query extension, which Earth Search evaluates. CQL2 filter answers invalid-filter. These properties work in query:

Property	Example	Notes
eo:cloud_cover	{"eo:cloud_cover": {"lte": 20}}	Scene cloud percentage
platform	{"platform": {"in": ["sentinel-2b", "sentinel-2c"]}}	One satellite or several
grid:code	{"grid:code": {"eq": "MGRS-31TCF"}}	One MGRS tile. Earth Search’s Collection 1 items have no s2:mgrs_tile; they carry grid:code and mgrs:utm_zone, mgrs:latitude_band, mgrs:grid_square

Each item also carries s2:processing_baseline, such as 05.12.

Presets

From GET /v1/presets, the web map’s recipes for this collection:

Preset	Assets	Stretch
true-color	visual (the 8-bit TCI image)	as stored
false-color	nir (B08), red, green	DN 1,000 to 5,000, reflectance 0 to 0.4
ndvi	nir (B08), red	(b1 - b2) / (b1 + b2 - 2000), −0.2 to 0.8 on rdylgn

The NDVI expression subtracts the two 1,000 offsets in the denominator; the offsets cancel in the numerator. Default cloud ceiling 30 % (60 % for ndvi), least cloudy scene first, a 30-day window.

Zoom

Tiles exist from zoom 8 to 15. Outside that range a tile answers 400 zoom-out-of-range.

Examples

Routes marked anonymous need no key. The others send your key in X-API-Key; see Authentication. The examples use the area west of Tarragona, Spain, and August 2026.

Collection document

Anonymous.

curl -s https://api.satellitedata.net/v1/collections/sentinel-2-l2a

The document lists the bands from schema v1, sdn:aliases, sdn:attribution and sdn:zoom. /v1/collections/S2L2A returns the same document with a Content-Location header.

Search one area and month

curl

curl -s -X POST https://api.satellitedata.net/v1/search \
  -H "X-API-Key: $SDN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "collections": ["sentinel-2-l2a"],
    "bbox": [0.7, 40.7, 0.8, 40.8],
    "datetime": "2026-08-01T00:00:00Z/2026-08-31T23:59:59Z",
    "limit": 10,
    "query": {"eo:cloud_cover": {"lte": 20}}
  }'

Python

import os
import requests

API = "https://api.satellitedata.net/v1"
headers = {"X-API-Key": os.environ["SDN_API_KEY"]}

body = {
    "collections": ["sentinel-2-l2a"],
    "bbox": [0.7, 40.7, 0.8, 40.8],
    "datetime": "2026-08-01T00:00:00Z/2026-08-31T23:59:59Z",
    "limit": 10,
    "query": {"eo:cloud_cover": {"lte": 20}},
}
page = requests.post(f"{API}/search", json=body, headers=headers).json()
for item in page["features"]:
    p = item["properties"]
    print(item["id"], p["datetime"], p["eo:cloud_cover"], p["s2:processing_baseline"])

The next link of a POST search has method: POST and a body holding only token; merge it into your body to get the next page. pystac-client does that for you; see Search.

One item

Anonymous.

curl -s https://api.satellitedata.net/v1/collections/sentinel-2-l2a/items/S2A_T31TCF_20260829T105044_L2A

Asset href values point at the public COGs, which you can read directly without a key.

Preview image

curl -s -o preview.png \
  -H "X-API-Key: $SDN_API_KEY" \
  "https://api.satellitedata.net/v1/collections/sentinel-2-l2a/items/S2A_T31TCF_20260829T105044_L2A/preview.png?preset=true-color&max_size=256"

preset is true-color, false-color or ndvi; max_size is the long edge in pixels, 64 to 1,024. The preset’s formats in /v1/presets lists the image formats it serves (ndvi has no JPEG).

TileJSON and a tile URL for MapLibre

The TileJSON is anonymous; the tiles need the key.

curl -s "https://api.satellitedata.net/v1/tiles/true-color/WebMercatorQuad/tilejson.json?collection=sentinel-2-l2a&datetime=2026-08-01/2026-08-31&maxCloudCoverage=20"

{
  "tilejson": "3.0.0",
  "name": "true-color",
  "attribution": "Contains modified Copernicus Sentinel data 2026",
  "tiles": ["https://api.satellitedata.net/v1/tiles/true-color/WebMercatorQuad/{z}/{x}/{y}.png?collection=sentinel-2-l2a&datetime=2026-08-01%2F2026-08-31&maxCloudCoverage=20"],
  "minzoom": 8,
  "maxzoom": 15
}

(Shortened.) One tile from that template, at zoom 12 over Tarragona’s hinterland:

curl -s -o tile.png -H "X-API-Key: $SDN_API_KEY" \
  "https://api.satellitedata.net/v1/tiles/true-color/WebMercatorQuad/12/2055/1534.png?collection=sentinel-2-l2a&datetime=2026-08-01%2F2026-08-31&maxCloudCoverage=20"

In MapLibre, give the TileJSON URL as a raster source and add the key with transformRequest, as in the Quickstart and Map tiles.

True colour as a reflectance GeoTIFF

A 2 km box from the scenes of 29 August 2026, in UTM zone 31N. REFLECTANCE is the default; it is written out here for clarity.

curl

curl -s -L -X POST https://api.satellitedata.net/v1/process \
  -H "X-API-Key: $SDN_API_KEY" \
  -H "Content-Type: application/json" \
  -o truecolor.tif \
  -d '{
    "input": {
      "bounds": {"bbox": [0.73815, 40.71096, 0.76185, 40.72904]},
      "data": [{
        "type": "sentinel-2-l2a",
        "dataFilter": {"timeRange": {"from": "2026-08-29T00:00:00Z", "to": "2026-08-29T23:59:59Z"}},
        "processing": {"units": "REFLECTANCE"}
      }]
    },
    "bands": ["red", "green", "blue"],
    "output": {"width": 200, "height": 150, "crs": "EPSG:32631"}
  }'

Python

import io
import os
import rasterio
import requests

API = "https://api.satellitedata.net/v1"
headers = {"X-API-Key": os.environ["SDN_API_KEY"]}

body = {
    "input": {
        "bounds": {"bbox": [0.73815, 40.71096, 0.76185, 40.72904]},
        "data": [{
            "type": "sentinel-2-l2a",
            "dataFilter": {"timeRange": {"from": "2026-08-29T00:00:00Z", "to": "2026-08-29T23:59:59Z"}},
            "processing": {"units": "REFLECTANCE"},
        }],
    },
    "bands": ["red", "green", "blue"],
    "output": {"width": 200, "height": 150, "crs": "EPSG:32631"},
}
r = requests.post(f"{API}/process", json=body, headers=headers)  # follows a 303 for large outputs
r.raise_for_status()
with rasterio.open(io.BytesIO(r.content)) as ds:
    print(ds.descriptions, ds.dtypes[0], ds.crs)  # ('red', 'green', 'blue') float32 EPSG:32631

NDVI inputs as harmonized integers

red and nir08 as INT16, where DN = 10,000 × reflectance on every baseline. sampleType: "INT16" selects units: "DN" with harmonize: "offset_removed".

curl -s -L -X POST https://api.satellitedata.net/v1/process \
  -H "X-API-Key: $SDN_API_KEY" \
  -H "Content-Type: application/json" \
  -o ndvi-inputs.tif \
  -d '{
    "input": {
      "bounds": {"bbox": [0.73815, 40.71096, 0.76185, 40.72904]},
      "data": [{
        "type": "S2L2A",
        "dataFilter": {"timeRange": {"from": "2026-08-29T00:00:00Z", "to": "2026-08-29T23:59:59Z"}}
      }]
    },
    "bands": ["red", "nir08"],
    "output": {"width": 200, "height": 150, "crs": "EPSG:32631", "sampleType": "INT16"}
  }'

Then NDVI is (nir08 - red) / (nir08 + red) on the integers directly, because the offset is already gone. Add "processing": {"units": "DN", "clipNegative": true} to the data source for Sentinel Hub’s harmonized values.

A larger area as a job

Synchronous process stops at 2,500 pixels a side. For more, send the same body to /v1/process/jobs and poll. This one is a 20 km box at 5 m.

curl

curl -s -i -X POST https://api.satellitedata.net/v1/process/jobs \
  -H "X-API-Key: $SDN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": {
      "bounds": {"bbox": [0.63147, 40.62956, 0.86853, 40.81044]},
      "data": [{
        "type": "sentinel-2-l2a",
        "dataFilter": {"ids": ["S2A_T31TCF_20260829T105044_L2A"]}
      }]
    },
    "bands": ["blue", "green", "red", "nir08"],
    "output": {"width": 4000, "height": 4000, "crs": "EPSG:32631"}
  }'

# 202 Accepted, with Location: https://api.satellitedata.net/v1/process/jobs/{jobId}
curl -s -H "X-API-Key: $SDN_API_KEY" https://api.satellitedata.net/v1/process/jobs/$JOB_ID

Python

import os
import time
import requests

API = "https://api.satellitedata.net/v1"
headers = {"X-API-Key": os.environ["SDN_API_KEY"]}

body = {
    "input": {
        "bounds": {"bbox": [0.63147, 40.62956, 0.86853, 40.81044]},
        "data": [{"type": "sentinel-2-l2a", "dataFilter": {"ids": ["S2A_T31TCF_20260829T105044_L2A"]}}],
    },
    "bands": ["blue", "green", "red", "nir08"],
    "output": {"width": 4000, "height": 4000, "crs": "EPSG:32631"},
}
r = requests.post(f"{API}/process/jobs", json=body, headers=headers)
r.raise_for_status()
job_url = r.headers["Location"]

while True:
    job = requests.get(job_url, headers=headers).json()
    if job["status"] in ("done", "failed"):
        break
    time.sleep(5)

if job["status"] == "done":
    for f in job["files"]:
        print(f["url"])  # valid for one hour; poll again for fresh links
else:
    print(job["error"])

A job goes queued, running, then done or failed. See Process jobs for the limits.