{"id":16111,"date":"2025-12-15T14:06:36","date_gmt":"2025-12-15T13:06:36","guid":{"rendered":"https:\/\/www.sigterritoires.fr\/?p=16111"},"modified":"2025-12-15T14:06:36","modified_gmt":"2025-12-15T13:06:36","slug":"sentinel-2-with-1-m-resolution-towards-a-more-detailed-analysis-of-coastal-environments","status":"publish","type":"post","link":"https:\/\/www.sigterritoires.fr\/index.php\/en\/sentinel-2-with-1-m-resolution-towards-a-more-detailed-analysis-of-coastal-environments\/","title":{"rendered":"Sentinel-2 with 1 m resolution: towards a more detailed analysis of coastal environments"},"content":{"rendered":"\n

<\/p>\n\n\n\n

\n\n\n\n

\n

\n

Contenu <\/p>\nToggle<\/span>

The calculated indices<\/a>

Cloud mask for Sentinel-2 at 10 m<\/a>

How we proceeded in the Colab script<\/a>

Conclusion: towards integrated processing of Sentinel-2 images<\/a><\/li><\/ul><\/nav><\/div>\n\n\n\n\n

In two previous articles, we explored how to convert standard Sentinel-2 images (10 m resolution) to a super-resolution 1 m version using the S2DR3 module, developed for coral research in Mauritius. This step provides an unprecedented level of detail from open data, revealing fine structures that are often invisible at 10 m: seagrass beds, reef flats, sandy areas, and subtle transitions between vegetation and reef.<\/p>\n\n\n\n

Traditionally, spectral indices (NDVI, NDWI, MNDWI, BSI, NDBI, etc.) is done in QGIS, once the images have been downloaded and processed. However, it is often more efficient to generate these indices directly at the source, i.e., as soon as the super-resolution Sentinel-2 images are produced.<\/p>\n\n\n\n

This approach has several advantages:<\/p>\n\n\n\n

Complete automation: the indices are calculated as soon as the dataset is created, without additional manual processing in QGIS.<\/li>\n\n\n\n
Spatial consistency: the indices use exactly the same radiometric corrections and super-resolution as the original bands.<\/li>\n\n\n\n
Time savings: no intermediate exports or manual configuration of raster algorithms.<\/li>\n\n\n\n

Preparation for analysis: the results can be used immediately in QGIS or any other GIS tool.<\/li>\n<\/ul>\n\n\n\n

The script presented below automates the entire workflow:<\/p>\n\n\n\n

downloading and super-resolution of the Sentinel-2 scene,<\/li>\n\n\n\n
generation of an intelligent cloud mask (excluding wave-overlap areas),<\/li>\n\n\n\n
direct calculation of the main spectral indices,<\/li>\n\n\n\n

creation of a multiband file ready for analysis.<\/li>\n<\/ol>\n\n\n\n

This integration greatly simplifies the transition from raw image to interpretation maps, while ensuring the reproducibility of scientific processing.<\/p>\n\n\n\n

<\/span>The calculated indices<\/span><\/h2>\n\n\n\n

<\/span>NDVI \u2014 Normalized Difference Vegetation Index<\/span><\/h3>\n\n\n\n

Purpose: to measure vegetation density and health.<\/li>\n\n\n\n

Formula:<\/li>\n<\/ul>\n\n\n\n

.<\/p>\n\n\n\n\\(NDVI = \\frac{NIR – Red}{NIR + Red}\\)\n\n\n\n

.<\/p>\n\n\n\n

NIR = near-infrared band (B08)<\/li>\n\n\n\n
Red = red band (B04)<\/li>\n\n\n\n
Values: -1 to 1; >0.2 vegetation, <0.1 bare soil or water.<\/li>\n\n\n\n

Use: vegetation monitoring, biomass estimation, agriculture.<\/li>\n<\/ul>\n\n\n\n

<\/span>NDWI \u2014 Normalized Difference Water Index<\/span><\/h3>\n\n\n\n

Purpose: detect the presence of water in the landscape.<\/li>\n\n\n\n

Formula:<\/li>\n<\/ul>\n\n\n\n

.<\/p>\n\n\n\n\\(NDWI = \\frac{Green – NIR}{Green + NIR}\\)\n\n\n\n

.<\/p>\n\n\n\n

Green = green band (B03)<\/li>\n\n\n\n
NIR = near-infrared band (B08)<\/li>\n\n\n\n
Values: positive = water, negative = soil\/vegetation.<\/li>\n\n\n\n

Use: mapping lakes, rivers, wetlands.<\/li>\n<\/ul>\n\n\n\n

<\/span>MNDWI \u2014 Modified NDWI<\/span><\/h3>\n\n\n\n

Purpose: to better separate water from urban or built-up areas.<\/li>\n\n\n\n

Formula:<\/li>\n<\/ul>\n\n\n\n

.<\/p>\n\n\n\n\\(MNDWI = \\frac{Green – SWIR1}{Green + SWIR1}\\)\n\n\n\n

.<\/p>\n\n\n\n

SWIR1 = short wavelength band (B11)<\/li>\n\n\n\n
Values: positive = water; improves detection compared to NDWI.<\/li>\n\n\n\n

Use: monitoring water bodies in urban or coastal areas.<\/li>\n<\/ul>\n\n\n\n

<\/span>BSI \u2014 Bare Soil Index<\/span><\/h3>\n\n\n\n

Purpose: to detect bare soil and denuded areas.<\/li>\n\n\n\n

Formula:<\/li>\n<\/ul>\n\n\n\n

.<\/p>\n\n\n\n\\(BSI = \\frac{(SWIR1 + Red) – (NIR + Blue)}{(SWIR1 + Red) + (NIR + Blue)}\\)\n\n\n\n

.<\/p>\n\n\n\n

Blue = blue band (B02)<\/li>\n\n\n\n
Red = B04, NIR = B08, SWIR1 = B11<\/li>\n\n\n\n
Values: the higher the BSI, the more open the surface.<\/li>\n\n\n\n

Use: monitoring bare soil, urban areas, mineral areas.<\/li>\n<\/ul>\n\n\n\n

<\/span>NDBI \u2014 Normalized Difference Built-up Index<\/span><\/h3>\n\n\n\n

Purpose: to detect built-up or impervious areas.<\/li>\n\n\n\n

Formula:<\/li>\n<\/ul>\n\n\n\n

.<\/p>\n\n\n\n\\(NDBI = \\frac{SWIR1 – NIR}{SWIR1 + NIR}\\)\n\n\n\n

.<\/p>\n\n\n\n

Values: positive = built-up, negative = vegetation or water.<\/li>\n\n\n\n

Use: urban mapping, monitoring of urban sprawl.<\/li>\n<\/ul>\n\n\n\n

<\/span>EVI \u2014 Enhanced Vegetation Index<\/span><\/h3>\n\n\n\n

Purpose: to improve the sensitivity of NDVI in areas with very dense vegetation and correct for atmospheric effects.<\/p>\n\n\n\n

Formula:<\/p>\n\n\n\n

.<\/p>\n\n\n\n\\(EVI = 2.5 \\cdot \\frac{NIR – Red}{NIR + 6 \\cdot Red – 7.5 \\cdot Blue + 1}\\)\n\n\n\n

.<\/p>\n\n\n\n

Blue = B02, Red = B04, NIR = B08<\/li>\n\n\n\n
Values: -1 to 1, similar to NDVI but more robust in dense areas.<\/li>\n\n\n\n

Usage: forests, tropical areas, monitoring vegetation health.<\/li>\n<\/ul>\n\n\n\n

<\/span>SAVI \u2014 Soil Adjusted Vegetation Index<\/span><\/h3>\n\n\n\n

Purpose: to measure vegetation by correcting for the effect of soil (useful in sparse areas).<\/li>\n\n\n\n

Formula:<\/li>\n<\/ul>\n\n\n\n

.<\/p>\n\n\n\n\\(SAVI = \\frac{(NIR – Red) \\cdot (1 + L)}{NIR + Red + L}\\)\n\n\n\n

.<\/p>\n\n\n\n

where:<\/p>\n\n\n\n

NIR = near-infrared band (Sentinel-2 B08)<\/li>\n\n\n\n
Red = red band (B04)<\/li>\n\n\n\n
L = soil correction factor, often 0.5<\/li>\n\n\n\n
Values: -1 to 1, like NDVI, but less sensitive to bare soil.<\/li>\n\n\n\n

Use: sparse or low vegetation (grasslands, savannas, coastal areas with low vegetation cover).<\/li>\n<\/ul>\n\n\n\n

<\/span>UI \u2014 Urban Index (or Urbanization Index)<\/span><\/h3>\n\n\n\n

Purpose: to detect and quantify urban areas.<\/li>\n\n\n\n

Simple formula:<\/li>\n<\/ul>\n\n\n\n

.<\/p>\n\n\n\n

\n\nUI = \\frac{SWIR – NIR}{SWIR + NIR}\n<\/annotation><\/semantics><\/math><\/div>\n\n\n\n.<\/p>\n\n\n\n\nSWIR = short wavelength band (Sentinel-2 B11 or B12)<\/li>\n\n\n\nNIR = near infrared (B08)<\/li>\n\n\n\nValues: positive for impervious or built-up areas, low\/negative for vegetation or water.<\/li>\n\n\n\nUse: monitoring urbanization, detecting buildings or artificial surfaces.<\/li>\n<\/ul>\n\n\n\n\n\n\n\n<\/span>RDI \u2014 Redness Difference Index<\/span><\/h3>\n\n\n\n\nPurpose: to detect bare soil or areas rich in iron oxide (\u201cred\u201d) in the landscape.<\/li>\n\n\n\nFormula:<\/li>\n<\/ul>\n\n\n\n.<\/p>\n\n\n\n\\(RDI = Red – Green\\)\n\n\n\n.<\/p>\n\n\n\n\nRed = red band (B04), Green = green band (B03)<\/li>\n\n\n\nValues: the more \u201cred\u201d the soil, the higher the RDI.<\/li>\n\n\n\nUse: monitoring erosion, bare soil, mineral or arid areas.<\/li>\n<\/ul>\n\n\n\n\n\n\n\n<\/span>Complete summary table<\/span><\/h3>\n\n\n\nIndex<\/th>Objective<\/th>Formula<\/th>Typical use<\/th><\/tr><\/thead>NDVI<\/td>Vegetation<\/td>(NIR\u2212Red)\/(NIR+Red)<\/td>Monitoring vegetation, agriculture<\/td><\/tr>EVI<\/td>Vegetation enhanced<\/td>2.5*(NIR-Red)\/(NIR+6Red-7.5<\/em>Blue+1)<\/td>Dense areas, tropical forests<\/td><\/tr>SAVI<\/td>Vegetation<\/td>(NIR\u2212Red)*(1+L)\/(NIR+Red+L)<\/td>Sparse areas, visible soil<\/td><\/tr>NDWI<\/td>Water<\/td>(Green\u2212NIR)\/(Green+NIR)<\/td>Mapping lakes, rivers<\/td><\/tr>MNDWI<\/td>water modified<\/td>(Green\u2212SWIR1)\/(Green+SWIR1)<\/td>Water in urban or coastal areas<\/td><\/tr>BSI<\/td>Bare soil<\/td>((SWIR1+Red)-(NIR+Blue))\/((SWIR1+Red)+(NIR+Blue))<\/td>Bare areas, bare soil<\/td><\/tr>NDBI<\/td>Urban<\/td>(SWIR1\u2212NIR)\/(SWIR1+NIR)<\/td>Building detection, cities<\/td><\/tr>UI<\/td>Urban<\/td>(SWIR1\u2212NIR)\/(SWIR1+NIR)<\/td>Urban mapping<\/td><\/tr>RDI<\/td>Soil\/red<\/td>Red\u2212Green<\/td>Bare soil or iron oxide monitoring<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\n<\/span>Sentinel Index Palette<\/span><\/h3>\n\n\n\nEach index is presented with:<\/p>\n\n\n\n\nits name and acronym,<\/li>\n\n\n\ntarget surface type (vegetation, water, soil, urban),<\/li>\n\n\n\ntypical values\/scale,<\/li>\n\n\n\nrepresentative color for map visualization (e.g., NDVI \u2192 green, NDWI \u2192 blue, BSI \u2192 light brown, etc.).<\/li>\n<\/ul>\n\n\n\nIndex<\/th>Type<\/th>Values<\/th>Representative color<\/th><\/tr><\/thead>NDVI<\/td>Vegetation<\/td>-1 \u2192 1 (dark green = dense, yellow = sparse) <\/td>Green<\/td><\/tr>EVI<\/td>Vegetation<\/td>-1 \u2192 1 (similar to NDVI, more sensitive)<\/td>Light green<\/td><\/tr>SAVI<\/td>Vegetation<\/td>-1 \u2192 1 (soil corrected)<\/td>Green-yellow<\/td><\/tr>NDWI<\/td>water<\/td>-1 \u2192 1 (positive = water)<\/td>Light blue<\/td><\/tr>MNDWI<\/td>Modified water<\/td>-1 \u2192 1<\/td>Turquoise blue <\/td><\/tr>BSI<\/td>Bare soil<\/td>-1 \u2192 1 (high = light soil)<\/td>Beige\/Brown <\/td><\/tr>NDBI<\/td>Urban<\/td>-1 \u2192 1 (positive = built-up) <\/td>Light gray <\/td><\/tr>UI<\/td>Urban<\/td>-1 \u2192 1<\/td>Medium gray<\/td><\/tr>RDI<\/td>Soil\/red<\/td>variable, higher = red<\/td>Brick red<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\n<\/span>Why use a cloud mask?<\/span><\/h2>\n\n\n\nClouds and their shadows pose problems in satellite imagery because:<\/p>\n\n\n\n\nThey obscure surfaces: soil, vegetation, water, coral.<\/li>\n\n\n\nThey distort indices (NDVI, NDWI, etc.) because the radiometric values do not correspond to the actual surface.<\/li>\n\n\n\nAreas covered by clouds can generate outliers: very high or very low values, skewing trend calculations.<\/li>\n<\/ol>\n\n\n\nTherefore, for index analysis and the detection of features on land or in water, it is important to mask these pixels.<\/p>\n\n\n\n\n\n\n\n<\/span>Cloud mask for Sentinel-2 at 10 m<\/span><\/h2>\n\n\n\nSentinel-2 offers several options for detecting clouds at its native resolution (10 m, 20 m, or 60 m depending on the band):<\/p>\n\n\n\n<\/span>a) Official mask provided by ESA<\/span><\/h3>\n\n\n\n\nQA60 file in Sentinel-2 L2A products: 10 m binary mask indicating clouds and cirrus.<\/li>\n\n\n\nAdvantages: reliable and easy to apply.<\/li>\n\n\n\nDisadvantages: coarse mask \u2192 does not always differentiate well between thin clouds and shadows.<\/li>\n<\/ul>\n\n\n\n<\/span>b) Methods based on spectral indices<\/span><\/h3>\n\n\n\n\nSWIR and NIR bands: clouds are highly reflective in the visible spectrum but less so in SWIR.<\/li>\n\n\n\nExample of a simple index:<\/li>\n<\/ul>\n\n\n\n.<\/p>\n\n\n\n\\(CI = \\frac{(B02 + B03 + B04)}{3} – B11\\)\n\n\n\n.<\/p>\n\n\n\n\nBright pixels in the visible spectrum and faint pixels in SWIR \u2192 probably clouds.<\/li>\n\n\n\nAllows you to create a cloud mask adapted to the context, useful if you want to detect local clouds and cirrus clouds.<\/li>\n<\/ul>\n\n\n\n\n\n\n\n<\/span>c) Automatic algorithms<\/span><\/h3>\n\n\n\n\nFmask or s2cloudless: sophisticated detection using machine learning and multispectral spectra.<\/li>\n\n\n\nUseful for large time series.<\/li>\n<\/ul>\n\n\n\n\n\n\n\n<\/span>How we proceeded in the Colab script<\/span><\/h2>\n\n\n\nIn the script, we did not use the official QA60, but rather detection by spectral indices and local texture, adapted to coastal areas and 1 m super-resolution:<\/p>\n\n\n\n<\/span>Main steps:<\/span><\/h3>\n\n\n\n<\/p>\n\n\n\n\nBand normalization<\/strong> : blue_n = blue \/ max <\/code> green_n = green \/ max <\/code> red_n = red \/ max <\/code> nir_n = nir \/ max<\/code>swir1_n = swir1 \/ max<\/code> \u2192 all values between 0 and 1 for homogeneity.<\/li>\n\n\n\n\u201cInitial cloud\u201d condition based on 4 criteria :TVery clear in the visible (albedo_vis > 0.35) High SWIR1 (swir1_n > 0.15) High blue\/red ratio (blue_n \/ red_n > 1.2)High NIR (nir_n > 0.25)<\/li><\/ul>\u2192 pixels meeting at least 3\/4 criteria are considered clouds..<\/li>\n\n\n\n\u201cFoam\u201d mask for marine areas<\/strong> :\n\nWe detect foam\/breaking waves via the local standard deviation on the blue band and visible albedo.<\/li>\n\n\n\nThese pixels are removed from the cloud mask : cloud_final = cloud_init & (~foam_mask)<\/code><\/li>\n\n\n\nThis avoids masking coral reefs or rough seas..<\/li>\n<\/ul>\n<\/li>\n\n\n\nFiltering small components:<\/strong> :\n\nRemoves very small clouds that could be noise \u2192 only objects > 500 pixels are kept.<\/li>\n<\/ul>\n<\/li>\n\n\n\nFinal result<\/strong> :\n\ncloud_mask<\/code> = cleaned cloud mask, ready to be used to exclude cloudy pixels when calculating indicess<\/strong>.<\/li>\n<\/ul>\n<\/li>\n<\/ol>\n\n\n\n\n\n\n\n<\/span>Why this method?<\/span><\/h3>\n\n\n\n\nThe scene is super-resolved (1 m) \u2192 the official QA60 does not exist at this resolution.<\/li>\n\n\n\nThe area is coastal\/coral \u2192 foam and breaking waves must be distinguished from clouds.<\/li>\n\n\n\nThe spectral + local texture approach allows us to:\n\ndetect clouds<\/li>\n\n\n\navoid masking water or foam<\/li>\n\n\n\nwork on super-resolved images generated by S2DR3<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n\n\n\n\n<\/span>Proceduree<\/span><\/h2>\n\n\n\nOpen Google Colab: https:\/\/colab.research.google.com<\/a><\/p>\n\n\n\nIn Collab, link your notebook to your Google Drive<\/p>\n\n\n\n<\/a><\/figure>\n\n\n\nNow, to ensure that processing does not take too long, you need to enable a GPU in the runtime environment. Look in the lower right corner of your notebook, where you will see something like \u201cPython 3\u201d with a chip icon. Click on it and select Change Runtime Environment Type. A window will open where you will need to search for Hardware Accelerator and select GPU T4. Save the changes and the notebook will automatically restart with the new configuration.<\/p>\n\n\n\n<\/a><\/figure>\n\n\n\nCopy this code into a new cell in your notebook and run it:<\/p>\n\n\n\n<\/div>script Colab<\/div><\/div><\/div><\/strong><\/p>\n\n\n\n# =========================================================\n# S2DR3 + Calcul indices Sentinel-2 + Masque nuages\n# \u00c9tude des coraux \u00e0 Blue Bay \/ \u00cele aux Aigrettes\n# =========================================================\n\n# --- 1. Monter Google Drive ---\nfrom google.colab import drive\ndrive.mount('\/content\/drive')\n\n# Cr\u00e9e un dossier pour les r\u00e9sultats\n!mkdir -p \/content\/drive\/MyDrive\/Sentinel2_Coraux_S2DR3\/output\n!ln -s \/content\/drive\/MyDrive\/Sentinel2_Coraux_S2DR3\/output \/content\/output\n\n# --- 2. Installer le paquet S2DR3 ---\n!pip -q install https:\/\/storage.googleapis.com\/0x7ff601307fa5\/s2dr3-20250905.1-cp312-cp312-linux_x86_64.whl\n\n# --- 3. Importer le module principal ---\nimport s2dr3.inferutils\nimport os, glob, numpy as np, rasterio, scipy.ndimage as ndi\nfrom skimage.transform import resize\n\n# --- 4. D\u00e9finir la zone d'int\u00e9r\u00eat et la date ---\nlonlat = (57.73, -20.44)\ndate = '2025-10-09'\n\n# --- 5. Lancer le traitement S2DR3 ---\ns2dr3.inferutils.test(lonlat, date)\n\n# =========================================================\n# 6. Lecture du fichier _MS.tif\n# =========================================================\nroot_dir = '\/content\/output'\nfiles = sorted(glob.glob(os.path.join(root_dir, '**\/*_MS.tif'), recursive=True),\n key=os.path.getmtime, reverse=True)\nif not files:\n raise FileNotFoundError(\"❌ Aucun fichier *_MS.tif trouv\u00e9 dans \/content\/output\")\nms_path = files[0]\noutdir = os.path.dirname(ms_path)\nprint(f\" Fichier trouv\u00e9 : {os.path.basename(ms_path)}\")\nprint(f\" Dossier : {outdir}\")\n\n# Lecture des bandes\nwith rasterio.open(ms_path) as src<\/a>:\n profile = src.profile\n n_bands = src.count\n print(f\"Nombre de bandes d\u00e9tect\u00e9es : {n_bands}\")\n bands = [src.read(i).astype('float32') for i in range(1, n_bands+1)]\n\ndef get_band(index, name):\n try:\n return bands[index-1]\n except IndexError:\n print(f\" Bande {name} manquante, remplac\u00e9e par z\u00e9ro.\")\n return np.zeros_like(bands[0])\n\n# Attribution bandes principales Sentinel-2\nblue = get_band(1, \"B02\")\ngreen = get_band(2, \"B03\")\nred = get_band(3, \"B04\")\nnir = get_band(4, \"B08\")\nswir1 = get_band(9, \"B11\")\nswir2 = get_band(10, \"B12\")\n\n# =========================================================\n# 7. Calcul des indices\n# =========================================================\ndef safe_div(a, b, mask=None):\n res = np.where((b!=0) & (~np.isnan(a)) & (~np.isnan(b)), a\/b, 0)\n if mask is not None:\n res[mask] = np.nan\n return res\n\nL = 0.5 # facteur SAVI\n\nindices = {\n 'NDVI' : safe_div(nir - red, nir + red),\n 'NDWI' : safe_div(green - nir, green + nir),\n 'MNDWI': safe_div(green - swir1, green + swir1),\n 'BSI' : safe_div((swir1 + red) - (nir + blue), (swir1 + red) + (nir + blue)),\n 'NDBI' : safe_div(swir1 - nir, swir1 + nir),\n 'EVI' : 2.5 * safe_div((nir - red), (nir + 6*red - 7.5*blue + 1)),\n 'SAVI' : safe_div((nir - red) * (1 + L), (nir + red + L)),\n 'UI' : safe_div(swir1 - nir, swir1 + nir),\n 'RDI' : safe_div(red - green, red + green)\n}\n\n# =========================================================\n# 8. Masque nuages l\u00e9ger adapt\u00e9 1 m\n# =========================================================\neps = 1e-8\nblue_n, green_n, red_n, nir_n, swir1_n = [x\/(np.nanmax(x)+eps) for x in [blue, green, red, nir, swir1]]\nalbedo_vis = (blue_n + green_n + red_n)\/3\n\ncond_bright = albedo_vis > 0.35\ncond_swir = swir1_n > 0.15\ncond_ratio = (blue_n\/(red_n+1e-6)) > 1.2\ncond_nir = nir_n > 0.25\ncloud_init = (cond_bright.astype(int) + cond_swir.astype(int) + cond_ratio.astype(int) + cond_nir.astype(int)) >= 3\n\nMNDWI_calc = np.where((green + swir1)!=0, (green - swir1)\/(green + swir1), -1)\nwater_mask = MNDWI_calc > 0\n\nsize = 7\nmean = ndi.uniform_filter(blue_n, size=size)\nmean_sq = ndi.uniform_filter(blue_n**2, size=size)\nlocal_std = np.sqrt(np.maximum(0, mean_sq - mean**2))\nfoam_mask = water_mask & (albedo_vis>0.25) & (local_std>0.03)\n\ncloud_mask = cloud_init & (~foam_mask)\nlabel_im, nb = ndi.label(cloud_mask)\nsizes = ndi.sum(cloud_mask, label_im, range(1, nb+1))\nmask_keep = np.zeros(nb+1, dtype=bool)\nmask_keep[1:] = sizes >= 500\ncloud_mask = mask_keep[label_im].astype(bool)\n\nprint(f\" Nuages finaux : {cloud_mask.sum()} pixels\")\n\n# =========================================================\n# 9. Sauvegarde indices individuels\n# =========================================================\nfor name, data in indices.items():\n out_path = os.path.join(outdir, f\"{name}.tif\")\n profile_idx = profile.copy()\n profile_idx.update(dtype='float32', count=1, compress='lzw', nodata=None)\n with rasterio.open(out_path, 'w', **profile_idx) as dst:\n dst.write(data.astype('float32'), 1)\n print(f\" {name} enregistr\u00e9 \u2192 {out_path}\")\n\n# =========================================================\n# 10. Cr\u00e9ation du stack multibande robuste\n# =========================================================\nall_layers = list(indices.values()) + [cloud_mask.astype('float32')]\nband_names = list(indices.keys()) + ['CLOUD_MASK']\n\nref_shape = indices['NDVI'].shape\nfor i, layer in enumerate(all_layers):\n if layer.shape != ref_shape:\n print(f\" Redimensionnement couche {band_names[i]} de {layer.shape} \u2192 {ref_shape}\")\n all_layers[i] = resize(layer, ref_shape, order=0, preserve_range=True, anti_aliasing=False).astype('float32')\n\nprofile_stack = profile.copy()\nprofile_stack.update(\n count=len(all_layers),\n dtype='float32',\n compress='lzw',\n nodata=None,\n height=ref_shape[0],\n width=ref_shape[1]\n)\n\nstack_path = os.path.join(outdir, 'indices_stack.tif')\nwith rasterio.open(stack_path, 'w', **profile_stack) as dst:\n for i, layer in enumerate(all_layers, start=1):\n dst.write(layer.astype('float32'), i)\n dst.set_band_description(i, band_names[i-1])\n\nif os.path.exists(stack_path):\n print(f\"\\n Fichier multibande cr\u00e9\u00e9 avec succ\u00e8s : {stack_path}\")\n print(\" Bandes : \" + ', '.join(band_names))\nelse:\n print(f\"\\n \u00c9chec de cr\u00e9ation de {stack_path}\")\n\nprint(\"\\n Traitement complet termin\u00e9.\")<\/code><\/pre>\n\n\n\n<\/div><\/div><\/p>\n\n\n\n<\/a><\/figure>\n\n\n\nOnce executed, you will have the .tif files in your Google Drive.<\/p>\n\n\n\n<\/a><\/figure>\n\n\n\n\n\n\n\n<\/span>Conclusion: towards integrated processing of Sentinel-2 images<\/span><\/h2>\n\n\n\nThe integration of spectral index calculation directly into the super-resolution process represents a natural evolution in Sentinel-2 data processing.By automating both the generation of the 1 m image and the production of indices, we obtain a homogeneous, complete, and immediately usable dataset for spatial analysis.<\/p>\n\n\n\nApplied to the coastal environments of Mauritius and Rodrigues, this method opens up new perspectives:<\/p>\n\n\n\n\nmonitoring of coastal vegetation and mangroves,<\/li>\n\n\n\nobservation of reef flats and seagrass beds,<\/li>\n\n\n\ndetection of areas of turbidity or siltation,<\/li>\n\n\n\nmapping of land use changes in sensitive coastal areas.<\/li>\n<\/ul>\n\n\n\nBy combining all these steps into a single reproducible workflow, this script becomes a research and environmental monitoring tool particularly suited to tropical islands.It also demonstrates that open-source tools (Sentinel-2, Google Colab, QGIS) now make it possible to achieve a level of spatial analysis that was previously reserved for professional platforms.<\/p>\n\n\n\nNext step: integrate these indices into a multi-date chronological analysis to track seasonal or post-cyclonic changes in coastal habitats.<\/p>\n\n\n\n\n","protected":false},"excerpt":{"rendered":"In two previous articles, we explored how to convert standard Sentinel-2 images (10 m resolution) to a super-resolution 1 m version using the S2DR3 module, developed for coral research in Mauritius. This step provides an unprecedented…<\/p>\n","protected":false},"author":1,"featured_media":16031,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"give_campaign_id":0,"_bbp_topic_count":0,"_bbp_reply_count":0,"_bbp_total_topic_count":0,"_bbp_total_reply_count":0,"_bbp_voice_count":0,"_bbp_anonymous_reply_count":0,"_bbp_topic_count_hidden":0,"_bbp_reply_count_hidden":0,"_bbp_forum_subforum_count":0,"sfsi_plus_gutenberg_text_before_share":"","sfsi_plus_gutenberg_show_text_before_share":"","sfsi_plus_gutenberg_icon_type":"","sfsi_plus_gutenberg_icon_alignemt":"","sfsi_plus_gutenburg_max_per_row":"","_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[3857],"tags":[3875,3843],"class_list":["post-16111","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-iaen","tag-ia-en","tag-sentinel-2-en"],"aioseo_notices":[],"jetpack_featured_media_url":"https:\/\/i0.wp.com\/www.sigterritoires.fr\/wp-content\/uploads\/2025\/11\/Sentinel-2.jpg?fit=2126%2C1594&ssl=1","jetpack_shortlink":"https:\/\/wp.me\/p6XU0A-4bR","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts\/16111","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/comments?post=16111"}],"version-history":[{"count":0,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/posts\/16111\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/media\/16031"}],"wp:attachment":[{"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/media?parent=16111"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/categories?post=16111"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.sigterritoires.fr\/index.php\/wp-json\/wp\/v2\/tags?post=16111"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}