NDVI

---

Introduction: The Data Problem Every Agricultural Researcher Faces

In Part 1 of this series, I showed you how to calculate NDVI from Sentinel-2 satellite data using Python. The most common question I received after that post was:

“Where do I actually get the Sentinel-2 data files?”

This is the practical gap that most remote sensing tutorials skip. They show you the analysis but assume you already have the data sitting on your hard drive.

In this post I fill that gap completely. By the end you will know how to:

• Create a free Copernicus account and understand the data portal
• Find Sentinel-2 images for any location in India — your study area, your district, your field
• Filter images by date, cloud cover, and tile
• Download data automatically using Python and the sentinelsat library
• Understand the folder structure of downloaded Sentinel-2 data
• Extract exactly the bands you need (Band 4 and Band 8 for NDVI)

Everything here is free. No paid software. No institutional login required. Just a free Copernicus account and Python.

---

What is Copernicus Hub and Why is the Data Free?

The Copernicus Programme is the European Union’s Earth observation programme. It operates a fleet of Sentinel satellites that continuously image the entire Earth’s surface. The data collected is made available free of charge to anyone in the world — researchers, students, government agencies, farmers, and developers.

For agricultural scientists in India this is a remarkable resource:

• Sentinel-2 revisits any location every 5 days (with two satellites combined)
• 10-metre resolution for key bands — fine enough to distinguish individual farm plots
• 13 spectral bands covering visible, near-infrared, and shortwave infrared wavelengths
• Free archive going back to 2015 — over 10 years of historical data for time series analysis
• Covers all of India — every district, every state, every season

For context: equivalent commercial satellite data would cost ₹50,000–₹5,00,000 per image. You get it free.

---

Sentinel-2 Bands Quick Reference

Before downloading, it helps to know which bands you actually need:

Band	Name	Wavelength	Resolution	Key Use
Band 2	Blue	490 nm	10 m	True colour composite
Band 3	Green	560 nm	10 m	True colour composite
Band 4	Red	665 nm	10 m	NDVI, crop stress
Band 8	NIR	842 nm	10 m	NDVI, vegetation health
Band 8A	Red Edge	865 nm	20 m	Canopy chlorophyll
Band 11	SWIR-1	1610 nm	20 m	Soil moisture, NDWI
Band 12	SWIR-2	2190 nm	20 m	Mineral mapping

For NDVI: You need Band 4 (Red) and Band 8 (NIR) — both at 10m resolution.
For soil moisture: You additionally need Band 11 (SWIR-1) at 20m.

---

Part 1 — Manual Download via Copernicus Browser (No Code Required)

Before learning the Python approach, let me show you the manual method. This is useful when you need one or two images quickly.

Step 1: Create Your Free Copernicus Account

1. Go to https://browser.dataspace.copernicus.eu/
2. Click Register in the top right
3. Fill in your details — use your personal email
4. Verify your email
5. Log in

Note: The old Copernicus Open Access Hub (scihub.copernicus.eu) was retired in 2023. The new portal is Copernicus Data Space Ecosystem at dataspace.copernicus.eu. All links and API endpoints have changed. Make sure you are using the new portal.

Step 2: Find Your Study Area

1. On the map, navigate to your study area in India
2. Use the search box — type your district name (e.g., “Kamrup, Assam”)
3. The map will zoom to your location
4. You can draw a polygon around your exact area of interest using the draw tool

Step 3: Search for Images

On the left panel:

• Data source: Select Sentinel-2 L2A (atmospherically corrected — use this, not L1C)
• Date range: Enter your start and end date
• Cloud cover: Set maximum to 20% (anything above 20% clouds makes agricultural analysis unreliable)
• Click Search

A list of available images appears. Each entry shows:

• Acquisition date and time
• Cloud cover percentage
• Tile ID (more on this below)
• Thumbnail preview

Step 4: Download

Click on any image → click the download icon → select which files to download.

What to download:

• For NDVI only: Download Band 4 and Band 8 files individually (saves bandwidth)
• For full analysis: Download the complete product (2–8 GB per image)

---

Understanding Sentinel-2 Tile System

India is divided into Sentinel-2 Military Grid Reference System (MGRS) tiles. Each tile is 100km × 100km. Knowing your tile code saves a lot of search time.

Common tiles for India:

Region	Tile Codes
Assam / Northeast India	T46RBN, T46RBM, T46QBL, T46QBK
West Bengal	T45RYL, T45RYK, T45QYL
Punjab / Haryana	T43RFQ, T43RGQ, T43SGQ
Maharashtra	T43PBR, T43PCR, T44PMS
Andhra Pradesh / Telangana	T44QMF, T44QNF, T44QLF
Tamil Nadu / Karnataka	T43PGR, T44PKS, T43QHB
Uttar Pradesh	T44RKQ, T44RLQ, T44RMQ
Rajasthan	T43RGP, T43RHP, T43RFP
Odisha	T45QXE, T45QYE, T44QQE
Gujarat	T42RUS, T43RBP, T43RCP

For Assam specifically: if your study area is in Kamrup or nearby districts, your tile is likely T46RBN.

---

Part 2 — Automated Download Using Python (sentinelsat)

The manual approach works for occasional downloads. But if you need:

• Multiple dates for time series analysis
• Multiple tiles for district or state-level analysis
• Regular automated downloads (monthly crop monitoring)

…then Python automation is the right approach.

Step 1: Install Required Libraries

pip install sentinelsat geopandas shapely requests

Note on sentinelsat: The library has been updated to work with the new Copernicus Data Space API. Make sure you install version 1.0 or later.

pip install sentinelsat --upgrade

Step 2: Set Up Your Credentials

Never hardcode your password in a script. Use environment variables or a config file.

import os

# Option 1: Set as environment variables (recommended)
# Run these in your terminal before running the script:
# export COPERNICUS_USER="your_username"
# export COPERNICUS_PASSWORD="your_password"

username = os.environ.get('COPERNICUS_USER', 'your_username_here')
password = os.environ.get('COPERNICUS_PASSWORD', 'your_password_here')

Step 3: Connect to Copernicus Data Space

from sentinelsat import SentinelAPI, read_geojson, geojson_to_wkt
from datetime import date
import geopandas as gpd
from shapely.geometry import box
import pandas as pd
import os

# Connect to new Copernicus Data Space API
api = SentinelAPI(
    username,
    password,
    'https://apihub.copernicus.eu/apihub'  # Updated endpoint
)

print("Successfully connected to Copernicus Data Space")

Step 4: Define Your Area of Interest

You can define your study area in three ways:

Method A — Bounding Box (simplest)

# Define bounding box for your study area
# Format: (min_longitude, min_latitude, max_longitude, max_latitude)

# Example: Kamrup district, Assam
min_lon, min_lat = 91.2, 25.8   # Southwest corner
max_lon, max_lat = 91.9, 26.3   # Northeast corner

# Create footprint from bounding box
from shapely.geometry import box
footprint = box(min_lon, min_lat, max_lon, max_lat).wkt

print(f"Study area defined: {min_lon}°E to {max_lon}°E, "
      f"{min_lat}°N to {max_lat}°N")

Method B — From a Shapefile (for district/block boundaries)

# If you have a shapefile of your study area
import geopandas as gpd

# Load shapefile (e.g., district boundary)
study_area = gpd.read_file("kamrup_district.shp")

# Convert to WGS84 if needed
study_area = study_area.to_crs("EPSG:4326")

# Get footprint as WKT
footprint = study_area.geometry.union_all().wkt
print("Study area loaded from shapefile")

Method C — Draw on Map and Export GeoJSON

# If you drew your area in QGIS or geojson.io and saved as GeoJSON
from sentinelsat import read_geojson, geojson_to_wkt

footprint = geojson_to_wkt(read_geojson('my_study_area.geojson'))
print("Study area loaded from GeoJSON")

Step 5: Search for Available Images

# Search for Sentinel-2 images
products = api.query(
    footprint,
    date=('20231001', '20231115'),    # Date range: 1 Oct to 15 Nov 2023
    platformname='Sentinel-2',
    producttype='S2MSI2A',             # L2A = atmospherically corrected
    cloudcoverpercentage=(0, 20)       # Maximum 20% cloud cover
)

# Convert to DataFrame for easy viewing
products_df = api.to_dataframe(products)

print(f"\nFound {len(products_df)} images matching your criteria")
print("\nAvailable images:")
print(products_df[['title', 'cloudcoverpercentage',
                    'size', 'beginposition']].to_string())

Sample output:

Found 4 images matching your criteria

Available images:
                                    title  cloudcoverpercentage  size  beginposition
S2A_MSIL2A_20231005T043701_N0509...        3.24  1.2 GB  2023-10-05
S2B_MSIL2A_20231010T043659_N0509...        8.91  1.1 GB  2023-10-10
S2A_MSIL2A_20231020T043701_N0509...       12.45  1.3 GB  2023-10-20
S2B_MSIL2A_20231115T043659_N0509...        5.67  1.2 GB  2023-11-15

Step 6: Select the Best Image

# Sort by cloud cover — lowest first
products_df = products_df.sort_values('cloudcoverpercentage')

# Select the clearest image
best_product = products_df.iloc[0]
best_product_id = products_df.index[0]

print(f"\nBest image selected:")
print(f"  Date: {best_product['beginposition'].date()}")
print(f"  Cloud cover: {best_product['cloudcoverpercentage']:.1f}%")
print(f"  Size: {best_product['size']}")
print(f"  Product ID: {best_product_id}")

Step 7: Download the Image

import os

# Create download directory
download_dir = "sentinel2_data"
os.makedirs(download_dir, exist_ok=True)

print(f"\nDownloading image to '{download_dir}' folder...")
print("This may take 10–30 minutes depending on your internet speed.")
print("File size is approximately 1–1.5 GB\n")

# Download the selected product
api.download(best_product_id, directory_path=download_dir)

print(f"\n✓ Download complete!")
print(f"  Check the '{download_dir}' folder for your data.")

---

Understanding the Downloaded Data Structure

After downloading, you will find a .SAFE folder. Here is what is inside:

S2A_MSIL2A_20231005T043701_N0509_R031_T46RBN_20231005T221503.SAFE/
│
├── GRANULE/
│   └── L2A_T46RBN_A043082_20231005T044459/
│       └── IMG_DATA/
│           ├── R10m/          ← 10-metre resolution bands
│           │   ├── T46RBN_20231005T043701_B02_10m.tif  (Blue)
│           │   ├── T46RBN_20231005T043701_B03_10m.tif  (Green)
│           │   ├── T46RBN_20231005T043701_B04_10m.tif  ← RED (for NDVI)
│           │   └── T46RBN_20231005T043701_B08_10m.tif  ← NIR (for NDVI)
│           │
│           ├── R20m/          ← 20-metre resolution bands
│           │   ├── T46RBN_20231005T043701_B05_20m.tif
│           │   ├── T46RBN_20231005T043701_B06_20m.tif
│           │   ├── T46RBN_20231005T043701_B8A_20m.tif
│           │   ├── T46RBN_20231005T043701_B11_20m.tif
│           │   └── T46RBN_20231005T043701_B12_20m.tif
│           │
│           └── R60m/          ← 60-metre resolution bands
│               ├── T46RBN_20231005T043701_B01_60m.tif
│               └── T46RBN_20231005T043701_B09_60m.tif
│
├── MTD_MSIL2A.xml             ← Metadata file
└── INSPIRE.xml

For NDVI analysis, you need only these two files from R10m folder:

• B04_10m.tif — Red band
• B08_10m.tif — NIR band

Automatically Extract Band Paths

import glob
import os
import zipfile

# The download may come as a .zip file — extract it first
zip_files = glob.glob(os.path.join(download_dir, "*.zip"))

if zip_files:
    zip_path = zip_files[0]
    print(f"Extracting {os.path.basename(zip_path)}...")
    with zipfile.ZipFile(zip_path, 'r') as zip_ref:
        zip_ref.extractall(download_dir)
    print("Extraction complete.")

# Find the .SAFE folder
safe_folders = glob.glob(os.path.join(download_dir, "*.SAFE"))
safe_folder = safe_folders[0]
print(f"\nSAFE folder: {os.path.basename(safe_folder)}")

# Find Band 4 (Red) and Band 8 (NIR) at 10m resolution
red_band_path = glob.glob(
    os.path.join(safe_folder, "GRANULE/*/IMG_DATA/R10m/*B04_10m.tif")
)[0]

nir_band_path = glob.glob(
    os.path.join(safe_folder, "GRANULE/*/IMG_DATA/R10m/*B08_10m.tif")
)[0]

print(f"\nBand paths found:")
print(f"  Red (B04): {os.path.basename(red_band_path)}")
print(f"  NIR (B08): {os.path.basename(nir_band_path)}")
print("\n✓ Ready for NDVI calculation!")
print("  Feed these paths into the NDVI code from Part 1.")

---

Part 3 — Download Multiple Dates for Time Series

For crop monitoring, you often need images across an entire season. Here is how to download multiple dates efficiently:

# Search for entire Kharif season: June to November 2023
products_season = api.query(
    footprint,
    date=('20230601', '20231130'),
    platformname='Sentinel-2',
    producttype='S2MSI2A',
    cloudcoverpercentage=(0, 25)
)

products_season_df = api.to_dataframe(products_season)
products_season_df = products_season_df.sort_values('beginposition')

print(f"Found {len(products_season_df)} images for Kharif 2023 season")
print("\nImage dates available:")
for idx, row in products_season_df.iterrows():
    print(f"  {row['beginposition'].date()} — "
          f"Cloud cover: {row['cloudcoverpercentage']:.1f}%")

Sample output:

Found 18 images for Kharif 2023 season

Image dates available:
  2023-06-07 — Cloud cover: 45.2%   ← Too cloudy — skip
  2023-06-12 — Cloud cover: 18.9%   ← Acceptable
  2023-06-17 — Cloud cover: 67.3%   ← Too cloudy — skip
  2023-06-22 — Cloud cover:  8.1%   ← Good
  ...
  2023-10-05 — Cloud cover:  3.2%   ← Excellent
  2023-10-20 — Cloud cover: 12.4%   ← Good
  2023-11-15 — Cloud cover:  5.7%   ← Excellent

# Filter to only low-cloud images (under 20%)
clear_images = products_season_df[
    products_season_df['cloudcoverpercentage'] < 20
].copy()

print(f"\n{len(clear_images)} clear images selected for download")
print(f"Estimated total download size: "
      f"~{len(clear_images) * 1.2:.0f} GB")
print("\nConfirm before downloading large datasets!")

# Download all clear images
# WARNING: This can be 5–20 GB total — ensure you have disk space
confirm = input("\nProceed with download? (yes/no): ")

if confirm.lower() == 'yes':
    for i, (product_id, row) in enumerate(clear_images.iterrows(), 1):
        print(f"\nDownloading image {i}/{len(clear_images)}: "
              f"{row['beginposition'].date()}")
        try:
            api.download(product_id, directory_path=download_dir)
            print(f"  ✓ Downloaded successfully")
        except Exception as e:
            print(f"  ✗ Download failed: {e}")
            continue

    print(f"\n✓ All downloads complete!")
    print(f"  {len(clear_images)} images saved to '{download_dir}'")

---

Part 4 — Lightweight Alternative: Download Only Specific Bands

Full Sentinel-2 products are 1–1.5 GB each. If you only need Band 4 and Band 8 for NDVI, you can save significant bandwidth by downloading only those files.

import requests
from requests.auth import HTTPBasicAuth

def download_specific_band(product_id, band_name, output_dir,
                            username, password):
    """
    Download a specific band file from a Sentinel-2 product.

    Args:
        product_id: Product UUID from sentinelsat search
        band_name: e.g., 'B04_10m' or 'B08_10m'
        output_dir: Where to save the file
        username: Copernicus username
        password: Copernicus password
    """
    # Get product info
    product_info = api.get_product_odata(product_id)

    # Construct band file URL
    # Note: This requires knowing the internal file path
    # The full download approach above is more reliable for beginners
    pass  # Implementation varies by API version

# For most use cases, downloading the full product
# and extracting bands locally is more reliable.
# The full product approach in Part 2 is recommended.

Practical tip: If bandwidth is a concern, use Copernicus Browser (browser.dataspace.copernicus.eu) to visually browse and download individual band files manually. Select your product → click “Download individual files” → select only B04 and B08.

---

Alternative: Google Earth Engine (For Large Areas)

If your study covers an entire state or multiple districts, downloading Sentinel-2 data locally may not be practical. Google Earth Engine (GEE) is a better option for large-scale analysis — it processes data in the cloud without downloading anything.

GEE is free for researchers. Here is a preview of how simple it is:

// Google Earth Engine JavaScript API
// Run this in code.earthengine.google.com

// Load Sentinel-2 collection for Assam, October 2023
var collection = ee.ImageCollection('COPERNICUS/S2_SR_HARMONIZED')
  .filterDate('2023-10-01', '2023-10-31')
  .filterBounds(ee.Geometry.Rectangle([90.2, 24.1, 96.0, 28.2]))
  .filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 20));

// Calculate NDVI for the least cloudy image
var image = collection.sort('CLOUDY_PIXEL_PERCENTAGE').first();
var ndvi = image.normalizedDifference(['B8', 'B4']).rename('NDVI');

// Visualise
Map.centerObject(ndvi, 8);
Map.addLayer(ndvi, {min: 0, max: 0.8, palette: ['red','yellow','green']},
             'NDVI Assam Oct 2023');

I will cover Google Earth Engine with Python in a separate post in this series.

---

Complete Script: Search, Select, and Download

Here is the complete workflow in one clean script:

"""
Sentinel-2 Data Download — Complete Workflow
Author: Data Science with DEB | dibyendudeb.com
Part 2 of Agricultural Data Science with Python Series

Downloads Sentinel-2 L2A data from Copernicus Data Space
for a specified study area and date range.

Requirements:
    pip install sentinelsat geopandas shapely
"""

import os
import glob
import zipfile
from datetime import datetime
from sentinelsat import SentinelAPI
from shapely.geometry import box

# ─── CONFIGURATION ─────────────────────────────────────────────────────────
USERNAME     = os.environ.get('COPERNICUS_USER', 'your_username')
PASSWORD     = os.environ.get('COPERNICUS_PASSWORD', 'your_password')
DOWNLOAD_DIR = "sentinel2_downloads"

# Study area: Kamrup district, Assam
# Replace with your study area coordinates
MIN_LON, MIN_LAT = 91.2, 25.8
MAX_LON, MAX_LAT = 91.9, 26.3

# Date range
START_DATE = '20231001'
END_DATE   = '20231115'

# Maximum acceptable cloud cover (%)
MAX_CLOUD  = 20

# ─── CONNECT ───────────────────────────────────────────────────────────────
print("Connecting to Copernicus Data Space...")
api = SentinelAPI(USERNAME, PASSWORD,
                  'https://apihub.copernicus.eu/apihub')
print("✓ Connected\n")

# ─── DEFINE STUDY AREA ─────────────────────────────────────────────────────
footprint = box(MIN_LON, MIN_LAT, MAX_LON, MAX_LAT).wkt
print(f"Study area: {MIN_LON}°E–{MAX_LON}°E, {MIN_LAT}°N–{MAX_LAT}°N")

# ─── SEARCH ────────────────────────────────────────────────────────────────
print(f"\nSearching for images: {START_DATE} to {END_DATE}")
print(f"Maximum cloud cover: {MAX_CLOUD}%\n")

products = api.query(
    footprint,
    date=(START_DATE, END_DATE),
    platformname='Sentinel-2',
    producttype='S2MSI2A',
    cloudcoverpercentage=(0, MAX_CLOUD)
)

df = api.to_dataframe(products).sort_values('cloudcoverpercentage')

print(f"Found {len(df)} suitable images:")
for _, row in df.iterrows():
    print(f"  {row['beginposition'].date()} | "
          f"Cloud: {row['cloudcoverpercentage']:.1f}% | "
          f"Size: {row['size']}")

# ─── SELECT BEST ───────────────────────────────────────────────────────────
if len(df) == 0:
    print("\n✗ No images found. Try:")
    print("  - Widening the date range")
    print("  - Increasing maximum cloud cover")
    print("  - Checking your bounding box coordinates")
    exit()

best_id  = df.index[0]
best_row = df.iloc[0]
print(f"\nBest image: {best_row['beginposition'].date()} "
      f"({best_row['cloudcoverpercentage']:.1f}% cloud)")

# ─── DOWNLOAD ──────────────────────────────────────────────────────────────
os.makedirs(DOWNLOAD_DIR, exist_ok=True)
print(f"\nDownloading to '{DOWNLOAD_DIR}'...")
print("Please wait — this may take 10–30 minutes.\n")

api.download(best_id, directory_path=DOWNLOAD_DIR)

# ─── EXTRACT ───────────────────────────────────────────────────────────────
zip_files = glob.glob(os.path.join(DOWNLOAD_DIR, "*.zip"))
if zip_files:
    print("Extracting zip file...")
    with zipfile.ZipFile(zip_files[0], 'r') as z:
        z.extractall(DOWNLOAD_DIR)

# ─── FIND BAND PATHS ───────────────────────────────────────────────────────
safe_dir  = glob.glob(os.path.join(DOWNLOAD_DIR, "*.SAFE"))[0]
red_path  = glob.glob(
    os.path.join(safe_dir, "GRANULE/*/IMG_DATA/R10m/*B04_10m.tif"))[0]
nir_path  = glob.glob(
    os.path.join(safe_dir, "GRANULE/*/IMG_DATA/R10m/*B08_10m.tif"))[0]

print("\n" + "=" * 55)
print("✓ DOWNLOAD COMPLETE")
print("=" * 55)
print(f"Red band (B04): {os.path.basename(red_path)}")
print(f"NIR band (B08): {os.path.basename(nir_path)}")
print("\nNext step: Feed these paths into the NDVI")
print("calculation script from Part 1 of this series.")
print("=" * 55)

---

Troubleshooting Common Issues

Issue: “Authentication failed”

• Double-check your Copernicus username and password
• Make sure you are using the new dataspace.copernicus.eu account — old scihub accounts may need migration

Issue: “No products found”

• Try widening your date range to 2–3 months
• Increase maximum cloud cover to 30% temporarily
• Verify your bounding box coordinates are correct (use Google Maps to check lat/lon)
• Northeast India gets heavy cloud cover June–September — try October–November for Kharif monitoring

Issue: “Product is offline”

• Older Sentinel-2 images (before 2020) may be in long-term archive storage
• Request the product to be brought online — Copernicus usually processes this within 24 hours
• Use the api.download() with checksum=False parameter if download is interrupted

Issue: Download very slow

• Copernicus servers can be slow during peak hours (European business hours = IST 13:30–21:30)
• Download early morning IST (before 13:00) for best speeds
• File sizes are 1–1.5 GB — expect 20–60 minutes on a typical Indian broadband connection

Issue: Disk space

• Each full Sentinel-2 product is 1–1.5 GB compressed, 3–5 GB extracted
• For time series (10 images), plan for 30–50 GB storage
• Delete the zip files after extraction to save space

---

What to Do With Your Downloaded Data

Once you have your Band 4 and Band 8 files, the next steps are:

1. Calculate NDVI — use the complete Python script from Part 1 of this series
2. Clip to your study area — use rasterio.mask to clip the 100km tile to your district or field
3. Stack multiple dates — for time series NDVI analysis across the crop season
4. Export maps — save as GeoTIFF for QGIS or as PNG for reports

---

Frequently Asked Questions

Q: Do I need to pay for Copernicus data? No. All Sentinel data is free to download and use for any purpose — research, commercial, educational.

Q: How often does Sentinel-2 revisit India? Every 5 days on average (combining Sentinel-2A and Sentinel-2B satellites). In practice you get 2–4 usable (low-cloud) images per month for most Indian locations.

Q: Can I download data for the entire state of Assam? Yes — but it will span multiple tiles and multiple gigabytes. For state-level analysis, Google Earth Engine (cloud processing) is more practical than local download.

Q: Is Sentinel-2 data available for the 1990s or 2000s? No. Sentinel-2 archive starts from 2015. For historical analysis before 2015, use Landsat data from USGS EarthExplorer (free, going back to 1972 but at 30m resolution).

Q: Can I use this data in my research publications? Yes — Copernicus data is freely usable for research. Cite as: “Copernicus Sentinel data [year], processed by [your name/institution].”

---

What’s Next in the Agricultural Data Science Series

• Part 3: SAVI vs NDVI vs EVI — which vegetation index should you use and when?
• Part 4: Random Forest for crop type classification using Sentinel-2 multispectral data
• Part 5: Time series NDVI analysis for paddy crop monitoring — a complete case study

---

Conclusion

In this post you learned how to:

• Navigate the Copernicus Data Space and find Sentinel-2 images for any location in India
• Use Python and sentinelsat to search, filter, and download images automatically
• Understand the downloaded folder structure and extract the exact bands you need
• Download multiple dates for seasonal time series analysis
• Troubleshoot the most common download problems

Combined with the NDVI calculation script from Part 1, you now have a complete free pipeline — from raw satellite to vegetation health map — entirely in Python.

---

Take the Next Step

📩 Get the next tutorial in your inbox — I publish one practical agricultural data science tutorial every week. Subscribe free below and never miss a post in this series.

👉 [Subscribe here — it’s free]

💬 Tell me your study area — drop a comment below with your district or state and I’ll confirm which Sentinel-2 tile code you need. I reply to every comment.

Want a quick reference guide?

Download the Remote Sensing Cheat Sheet (₹199)

Includes: Sentinel-2 bands, vegetation index formulas,
crop signatures, Python template, and common mistakes

---

Data Science with DEB — Practical tutorials on Python, remote sensing, and machine learning for agricultural researchers. Published weekly at dibyendudeb.com

---