Skip to content

GEE ACOLITE

ACOLITE atmospheric correction adapted for Google Earth Engine with Sentinel-2 imagery.

GEE ACOLITE integrates selected components of the ACOLITE processor — specifically its Look-Up Tables (LUTs) and Dark Spectrum Fitting (DSF) algorithm — into Google Earth Engine workflows. The key design decision is that AOT (aerosol optical thickness) is estimated as a single fixed value per image, making the correction computationally efficient while remaining scientifically rigorous for most aquatic remote sensing applications.

Key Features

  • Dark Spectrum Fitting (DSF): Estimates a single global AOT per image using ACOLITE's LUTs and atmospheric models
  • GEE-native processing: All image operations run server-side on GEE infrastructure; only AOT estimation runs client-side
  • Sentinel-2 support: Full 13-band support (10m, 20m, 60m), including resampling and geometry extraction
  • Water quality products: SPM, turbidity, chlorophyll-a (OC2, OC3, NDCI), and remote sensing reflectance (Rrs)
  • Satellite-derived bathymetry: pSDB indices, optical filtering, and calibration against in-situ data
  • Flexible masking: Water, cirrus, cloud, and cloud-shadow masks with configurable thresholds

System Architecture

graph TB subgraph Client["Client (Local Python)"] USER[User Script] AC[ACOLITE Package] LUT[LUT Tables] DSF[DSF / AOT Estimation] end subgraph Server["Server (Google Earth Engine)"] S2[Sentinel-2 L1C Catalog] L1CONV[L1 Conversion & Resampling] MASK[Masking] RHOS[Surface Reflectance] WQ[Water Quality Products] SDB[Bathymetry] end USER --> AC AC --> LUT USER -->|search / ImageCollection| S2 S2 -->|TOA DN| L1CONV L1CONV -->|dark spectrum getInfo| DSF LUT --> DSF DSF -->|fixed AOT + LUT params| RHOS L1CONV --> RHOS MASK --> RHOS RHOS --> WQ RHOS --> SDB style Client fill:#dbeafe,stroke:#3b82f6 style Server fill:#dcfce7,stroke:#16a34a

Processing Pipeline

sequenceDiagram participant User participant Local as Local (ACOLITE + numpy/scipy) participant GEE as GEE Server User->>GEE: search(roi, start, end) GEE-->>User: ee.ImageCollection (L1C) User->>GEE: l1_to_rrs(images, scale) GEE-->>GEE: DN → TOA reflectance + geometry angles loop For each image GEE-->>Local: compute_pdark() → dark spectrum (getInfo) Local->>Local: estimate_aot_per_lut() — LUT interpolation Local->>Local: select_best_model() — model selection Local-->>GEE: fixed AOT + atmospheric params GEE->>GEE: compute_rhos() — surface reflectance GEE->>GEE: compute_water_bands() — WQ products end GEE-->>User: ee.ImageCollection (L2 + products)

Main Modules

Module Description
gee_acolite.correction ACOLITE class — DSF atmospheric correction
gee_acolite.water_quality Water quality parameter computation (SPM, turbidity, Chl-a, Rrs)
gee_acolite.bathymetry Satellite-derived bathymetry (pSDB, calibration, mosaicking)
gee_acolite.utils.search Sentinel-2 image search and cloud probability joining
gee_acolite.utils.l1_convert L1C DN → TOA reflectance + geometry extraction
gee_acolite.utils.masks Water, cloud, cirrus, and shadow masking
gee_acolite.sensors.sentinel2 Sentinel-2 band definitions and resolution mapping

Quick Start

import ee
import acolite as ac
from gee_acolite import ACOLITE
from gee_acolite.utils.search import search

# Initialize GEE
ee.Initialize(project='your-project-id')

# Define region of interest
roi = ee.Geometry.Rectangle([-0.40, 39.43, -0.30, 39.50])

# Search Sentinel-2 L1C images
images = search(roi, '2023-06-01', '2023-06-30', tile='30SYJ')
images = images.filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 10))

# Configure and run atmospheric correction
settings = {
    'dsf_spectrum_option': 'darkest',
    'dsf_model_selection': 'min_drmsd',
    's2_target_res': 10,
    'l2w_parameters': ['spm_nechad2016', 'chl_oc3', 'pSDB_green'],
}
ac_gee = ACOLITE(ac, settings)
corrected, final_settings = ac_gee.correct(images)

# Export to Google Drive
task = ee.batch.Export.image.toDrive(
    image=corrected.first(),
    description='valencia_corrected',
    folder='GEE_ACOLITE',
    region=roi,
    scale=10,
    maxPixels=1e9,
)
task.start()

Requirements

  • Python >= 3.11
  • earthengine-api >= 0.1.350
  • numpy >= 1.20, scipy >= 1.7, netcdf4 >= 1.7
  • ACOLITE (bundled as git submodule)
  • Google Earth Engine account

See Installation for full setup instructions.

License

GPL-3.0-or-later

This software was developed at the Institute of Marine Sciences of Andalusia (ICMAN-CSIC), Spanish National Research Council (CSIC), Puerto Real, Spain.