Using Google Earth Engine in RStudio to Export and Visualize Sentinel-2 NDVI

Updated: August 2025 | Categories: Remote Sensing, NDVI, R, Python, Earth Engine

This tutorial demonstrates how to:

• Set up Google Earth Engine (GEE) Python API in RStudio using reticulate
  
  

• Download monthly Sentinel-2 NDVI GeoTIFFs to Google Drive
  
  

• Visualize NDVI .tif files using R's raster and viridis packages
  
  

Perfect for researchers analyzing vegetation trends or land degradation across time.

🔧 1. Set Up Google Earth Engine in Conda

We'll use a dedicated Conda environment to manage dependencies.

💻 Step-by-Step: Run these in the Conda Command Prompt

bash

# Create new environment

conda create --name gee-demo python=3.10

# Activate environment

conda activate gee-demo

# Install GEE Python API and supporting libraries

conda install -c conda-forge earthengine-api pandas numpy

# Authenticate Earth Engine (opens browser window)

earthengine authenticate

📦 2. Enable Python in R with reticulate

In RStudio:

install.packages("reticulate")     # Install reticulate if needed

library(reticulate)

# Point R to the correct Conda environment

use_condaenv("gee-demo", conda = "auto", required = TRUE)

🌐 3. Import GEE and Initialize

Load Python packages and initialize Earth Engine:

ee <- import("ee")      # Earth Engine

np <- import("numpy")   # Numpy

pd <- import("pandas")  # Pandas

time <- import("time")  # Time (for delays)

# Initialize Earth Engine (optionally specify a GCP project)

ee$Initialize(project = "smart-caster-359617")

📍 4. Define Region of Interest: Kenya

r

CopyEdit

kenya_poly <- ee$Geometry$Polygon(list(

  list(

    c(33.79791, -4.82957),

    c(42.01568, -4.82957),

    c(42.01568, 5.04587),

    c(33.79791, 5.04587),

    c(33.79791, -4.82957)

  )

))

🧹 5. Define Processing Functions

# Cloud mask using QA60 band

maskClouds <- function(image) {

  cloudMask <- image$select('QA60')$bitwiseAnd(py_eval("1 << 10"))$eq(0L)$

    And(image$select('QA60')$bitwiseAnd(py_eval("1 << 11"))$eq(0L))

  image$updateMask(cloudMask)

}

# NDVI computation

computeNDVI <- function(image) {

  image$normalizedDifference(list('B8', 'B4'))$

    rename('NDVI')$

    copyProperties(image, list('system:time_start'))

}

📤 6. Export Monthly Mean NDVI for 2023

for (m in 1:12) {

  month_start <- ee$Date$fromYMD(2023, m, 1)

  month_end <- month_start$advance(1, 'month')

  month_str <- sprintf("%02d", m)

  monthly_collection <- ee$ImageCollection('COPERNICUS/S2_SR_HARMONIZED')$

    filterDate(month_start, month_end)$

    filterBounds(kenya_poly)$

    filter(ee$Filter$lt('CLOUDY_PIXEL_PERCENTAGE', 20))$

    map(maskClouds)$

    map(computeNDVI)

  monthly_mean <- monthly_collection$mean()$clip(kenya_poly)$

    set('system:time_start', month_start$millis())

  count <- monthly_collection$size()$getInfo()

  if (count > 0) {

    task <- ee$batch$Export$image$toDrive(

      image = monthly_mean,

      description = paste0('NDVI_2023_', month_str),

      folder = 'ndvi',

      fileNamePrefix = paste0('NDVI_2023_', month_str),

      region = kenya_poly,

      scale = 50,

      crs = 'EPSG:4326',

      maxPixels = 1e13

    )

    task$start()

    cat(sprintf('Export task started for NDVI_2023_%s\n', month_str))

  } else {

    cat(sprintf('No images for month %s, skipping export.\n', month_str))

  }

  time$sleep(1)

}

cat('All export tasks submitted. Check your Drive under the "ndvi" folder.\n')

🖼️ 7. Visualize the Downloaded NDVI .tif in R

After downloading from Google Drive to your local system:

library(raster)

library(viridis)

# Load and rescale NDVI

ndvi <- raster("C:/Users/your_user/Downloads/NDVI_2022_02.tif")

ndvi <- ndvi / 10000  # Convert to standard NDVI range [-1, 1]

# Plot

plot(ndvi, col = viridis(100), main = "NDVI February 2022")

💡 Use terra::rast() instead of raster() if you prefer the terra package.

🧠 Additional Tips

• ☁️ QA60 masking removes clouds and cirrus for clean NDVI
  
  

• 📆 You can automate for any year by changing 2023
  
  

• 📂 Exports are saved as .tif in your Google Drive's ndvi folder
  
  

✅ Conclusion

By integrating Google Earth Engine with R using reticulate, you can automate powerful remote sensing pipelines—from cloud-masked NDVI extraction to export and local visualization. This approach scales efficiently, is reproducible, and bridges the power of GEE with the flexibility of R.