This guide walks you through comparing satellite-derived vegetation indices (like NDVI) with local ground measurements (like percent green cover or biomass). You’ll learn how to align the data in time and space, extract satellite values at your sample points, and run a simple correlation to see how well they agree. Before starting, make sure you have: satellite imagery (e.g., Landsat, Sentinel-2) and ground data collected within a few days of each other, plus free software like QGIS or Google Earth Engine (GEE) and a spreadsheet.
What You’ll Need
- Satellite imagery (Landsat 8/9 or Sentinel-2) covering your study area
- Ground truth data: GPS points with vegetation measurements (e.g., % cover, biomass, leaf area index)
- QGIS (free) or Google Earth Engine account for spatial extraction
- A spreadsheet program (Excel, Google Sheets) for statistical analysis
- Optional: R or Python for advanced correlation tests
Step-by-Step
- Choose your satellite product and index. For vegetation, NDVI (Normalized Difference Vegetation Index) is the most common. Compute it from red and near-infrared bands: NDVI = (NIR – Red) / (NIR + Red). For Landsat 8/9, use bands 4 (red) and 5 (NIR); for Sentinel-2, bands 4 and 8.
- Select imagery close to your ground survey date. Ideally within ±3 days, but up to ±10 days can be acceptable if vegetation is stable. Avoid images with clouds over your sample points.
- Preprocess the satellite image. If using raw data, apply atmospheric correction (e.g., DOS1 or Sen2Cor) to get surface reflectance. Many platforms (GEE, USGS EarthExplorer) offer surface reflectance products already.
- Calculate NDVI. In QGIS, use the Raster Calculator or the Semi-Automatic Classification Plugin. In GEE, use the expression: (B8 – B4) / (B8 + B4).
- Extract NDVI values at your ground sample points. In QGIS, load your ground points as a vector layer and the NDVI raster. Use the tool “Sample Raster Values” (in the Processing toolbox) to add the NDVI value to each point’s attribute table. In GEE, use
sampleRegions()to extract values. - Export the data to a spreadsheet. You should have columns: sample ID, ground measurement (e.g., % green cover), and NDVI value.
- Plot the data. Make a scatter plot with ground measurement on the x-axis and NDVI on the y-axis. Add a trendline to visually inspect the relationship.
- Run a correlation analysis. Calculate Pearson’s correlation coefficient (r) and the coefficient of determination (R²) in your spreadsheet or stats software. A positive r (e.g., 0.7) indicates a strong positive relationship.
- Interpret the results. If r is high (>0.6), the satellite index aligns well with ground data. If r is low, check for spatial misalignment, time lag, or saturation of NDVI in dense vegetation.
Common Mistakes
- Mismatched time windows: Comparing ground data from one month with satellite imagery from another can introduce error. Always note the exact dates.
- Ignoring spatial resolution: Each satellite pixel covers a large area (e.g., 30 m for Landsat, 10 m for Sentinel-2). Your ground plot should be at least as large as the pixel, or you should average several pixels around the point.
- Using uncalibrated NDVI: Raw digital numbers (DN) are not comparable across dates. Use surface reflectance products.
- Cloud and shadow contamination: Even a small cloud over a sample point can ruin the NDVI value. Always mask clouds or exclude those points.
- Assuming linearity: NDVI saturates in dense vegetation (NDVI > 0.8). The relationship may be non-linear; consider using other indices like EVI or soil-adjusted SAVI.
Tips
- Use a handheld GPS with 3–5 m accuracy to record sample points. Mark the center of each plot.
- Collect ground data within ±2 hours of solar noon to match satellite overpass conditions (sun angle).
- If you have multiple ground measurements per pixel, average them to reduce local variability.
- Check the NDVI range: healthy vegetation is 0.3–0.8, bare soil is around 0.1, water is negative. If you see values outside these, suspect error.
- Use a consistent coordinate system (UTM) for all data layers to avoid spatial offset.
Troubleshooting
- NDVI values are all negative or zero: Check if the image is in reflectance or DN. Also verify that the red and NIR bands are correctly selected (bands 4 and 5 for Landsat, not 5 and 4).
- Correlation is very low (r < 0.3): Re-examine the time difference. Vegetation may have changed (e.g., drought, harvest). Also check for GPS errors: plot your points on the image to see if they fall on roads or buildings.
- Some sample points have no NDVI value: This often happens if the point falls outside the image extent or on a masked cloud. In QGIS, check the attribute table for NULL values; remove those points or use a buffer to get the nearest pixel.
- NDVI values are suspiciously high (e.g., >0.9): This could be due to atmospheric haze or sensor noise. Apply a stricter cloud mask or use a different atmospheric correction method.
- Ground data is in biomass (kg/m²) but NDVI is unitless: You can still correlate, but the relationship may be exponential. Try log-transforming the biomass data.
Frequently Asked Questions
What is the best satellite for vegetation comparison?
Sentinel-2 (10 m resolution) is ideal for small plots; Landsat (30 m) is good for larger areas with longer historical records. Choose based on your plot size and revisit frequency.
How many ground samples do I need?
At least 30 samples for a reliable correlation, but more is better. Aim for 50–100 if you want to detect small effects.
Can I compare NDVI with plant height data?
Yes, but height is not directly related to greenness. Use a vegetation index like NDVI for biomass or cover; for height, consider radar or LiDAR data.
What if my ground data is collected over a week?
Use the satellite image closest to the majority of your samples. If vegetation is changing fast, split the analysis by date or use a time series.
Do I need to correct for sun angle?
If you use surface reflectance products, sun angle is already corrected. For raw data, you can apply a cosine correction, but it’s rarely necessary for NDVI.
What does an R² of 0.5 mean?
It means 50% of the variation in ground data is explained by NDVI. The other 50% is due to other factors (soil, species, moisture). R² > 0.6 is generally considered good for vegetation studies.
Related
Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.
More cards in this drawer
- Microsoft Surface Duo: The Foldable Phone Story
- Samsung SM-T813 Hard Reset: Step-by-Step Guide
- Mafia Clash Codes: Real Meanings & Usage
- Remove Programs from Mac: Full Uninstall Guide
- Project Sentinel Morning Digest: Data Stack Failure Alerts
- Best KVM VPS Hosting: Top Picks & Guide