Exercises: Flood Analysis with False Color Composites and NDWI

Last updated on 2026-06-28 | Edit this page

Estimated time: 45 minutes

Overview

Questions

  • What is a false color composite and what does it reveal?
  • How do I create a false color composite from Sentinel-2 bands in QGIS?
  • What is NDWI and how does it differ from NDVI?
  • How can NDWI help identify flood extent more accurately than a false color composite alone?

Objectives

  • Understand how false color composites remap spectral bands to RGB channels to highlight landscape features
  • Build a virtual raster in QGIS to create a false color composite from Sentinel-2 bands
  • Calculate NDWI using the Raster Calculator to delineate water from non-water
  • Compare FCC, NDWI, and satellite basemap imagery to assess flood extent

Introduction

In the previous episode we used NDVI to identify vegetation. In this episode we shift focus to water — specifically, how to identify the extent of flooding using two complementary techniques: False Color Composites (FCC) and the Normalized Difference Water Index (NDWI).

What Is a False Color Composite?

A true color image maps the red, green, and blue bands to their corresponding RGB screen channels, producing an image that looks like a normal photograph. A false color composite swaps one of those bands — typically replacing blue with near-infrared — so that features invisible to the human eye become visible on screen.

In a standard NRG false color composite (NIR → Red channel, Red → Green channel, Green → Blue channel):

  • Vegetation appears red — because healthy vegetation strongly reflects near-infrared light, which is now displayed in the red channel
  • Water appears dark blue to black — because water absorbs most near-infrared light
  • Bare soil and infrastructure appear grey to white
Diagram showing how true color and false color composites map spectral bands to RGB screen channels differently.
Diagram showing how true color and false color composites map spectral bands to RGB screen channels differently.
Chart showing reflectance values across wavelengths for different surface types, illustrating why band substitution reveals hidden features.
Chart showing reflectance values across wavelengths for different surface types, illustrating why band substitution reveals hidden features.
A false color composite example from Sentinel-2 imagery.
A false color composite example from Sentinel-2 imagery.

What Is NDWI?

The Normalized Difference Water Index (NDWI) operates similarly to NDVI, but targets water rather than vegetation. It uses Sentinel-2 Bands 3 (Green) and 8 (Near-Infrared):

NDWI = (Green − NIR) / (Green + NIR)

Water absorbs near-infrared light but reflects green light, so water features produce positive NDWI values while non-water features produce negative values. This makes NDWI particularly useful for mapping flood extent, where muddy floodwater can appear ambiguous in both true and false color images.

Step 1: Set Up Your Project

  1. Create a folder on your desktop called Flood_Analysis.
  2. Open QGIS, close any pop-ups, and go to Project → Save As. Save the project as Flood_Project inside your Flood_Analysis folder.

Step 2: Download and Extract the Sentinel-2 Image

  1. Navigate to the workshop’s shared resources Google Drive. Under Day 1_Session 3a: Basic raster functions, download the ZIP file: S2A_MSIL1C_20220831T055651_N0510_R091_T42RVR_20240719T213641.SAFE.zip

  2. Save the ZIP file to your Flood_Analysis folder and extract it:

    • Windows: Right-click → Extract All
    • Mac: Double-click the ZIP file

This image captures the Sindh province of Pakistan during the devastating 2022 floods.

Step 3: Load the Required Bands

To create both the FCC and the NDWI, we need three bands:

Band Wavelength File name
Band 3 (Green) ~560 nm T42RVR_20220831T055651_B03.jp2
Band 4 (Red) ~665 nm T42RVR_20220831T055651_B04.jp2
Band 8 (NIR) ~842 nm T42RVR_20220831T055651_B08.jp2

In the Browser Panel, navigate to: Project Home → S2A_MSIL1C…SAFE → GRANULE → L1C_T42RVR_… → IMG_DATA

Drag B03, B04, and B08 into the Layers Panel.

Bands 3, 4, and 8 loaded into QGIS.
Bands 3, 4, and 8 loaded into QGIS.

Step 4: Create the False Color Composite

To combine three single-band rasters into one multi-band image, we use a virtual raster.

  1. From the menu bar, select Raster → Miscellaneous → Build Virtual Raster.
  2. Click the three dots next to Input Layers and arrange the bands by dragging them into this order:
    • B08 (NIR — will map to the red channel)
    • B04 (Red — will map to the green channel)
    • B03 (Green — will map to the blue channel)
  3. Click Select All, then click the back arrow in the top-left corner of the pop-up.
The Build Virtual Raster dialog with bands arranged in NIR-Red-Green order.
The Build Virtual Raster dialog with bands arranged in NIR-Red-Green order.
  1. Check the box Place each input file into a separate band.
  2. Under Virtual, click the three dots and select Save to File. Save it as Flood_Output in your Flood_Analysis folder.
  3. Click Run.

Once complete, uncheck B03, B04, and B08 in the Layers Panel to see the false color composite.

False color composite of the Sindh province during the 2022 floods.
False color composite of the Sindh province during the 2022 floods.

The red areas are vegetation, dark blue and black areas are water, and grey-to-white areas are bare soil and infrastructure.

Step 5: Compare Against a Satellite Basemap

Add a basemap to compare the FCC against post-flood satellite imagery.

  1. If you have not already installed QuickMapServices, go to Plugins → Manage and Install Plugins, search for QuickMapServices, and install it.
  2. Open the QMS panel (the globe-with-magnifying-glass icon in the toolbar), search for imagery, and add Esri Satellite (ArcGIS/World_Imagery).

Toggle the Flood_Output layer on and off to compare the flooded landscape to post-flood imagery.

Discussion

Exercise 1: Interpret the False Color Composite

  1. How accurate is the FCC at showing flood extent? Can you identify areas that were clearly underwater?
  2. What does the FCC struggle to distinguish? Look carefully at lighter blue areas — are they floodwater, or bare soil?
  3. What was the approximate extent of flooding in this region?

Step 6: Calculate NDWI

The FCC provides a useful overview, but it has a limitation: muddy floodwater often appears as a lighter blue-grey that is difficult to distinguish from bare soil or infrastructure. NDWI solves this by isolating water based on its spectral signature rather than relying on visual color interpretation.

  1. From the menu bar, select Raster → Raster Calculator.
  2. Enter the NDWI formula in the expression box (double-click band names to insert them):
( "T42RVR_20220831T055651_B03@1" - "T42RVR_20220831T055651_B08@1" ) / ( "T42RVR_20220831T055651_B03@1" + "T42RVR_20220831T055651_B08@1" )
The Raster Calculator with the NDWI expression entered.
The Raster Calculator with the NDWI expression entered.
  1. Click the three-dot button next to Output layer and save it as NDWI_Output in your Flood_Analysis folder.
  2. Click OK to run the calculation.

Step 7: Apply Symbology to the NDWI Output

  1. Right-click the NDWI_Output layer → Properties → Symbology tab.
  2. Change the Render type to Singleband pseudocolor.
  3. Set Interpolation to Discrete.
  4. Set the Color ramp to Blues.
  5. Set the Mode to Equal Interval and reduce Classes to 2.
  6. In the Value column, set the first value to 0 — this creates a clean split between water (positive values, blue) and non-water (negative values).
  7. Click Apply.
NDWI results with two-class blue symbology applied.
NDWI results with two-class blue symbology applied.

Step 8: Overlay NDWI on the False Color Composite

To directly compare how much water the NDWI detected versus what the FCC shows, we can make the non-water class transparent so only detected water is visible over the FCC.

  1. Reopen the Symbology tab for NDWI_Output.
  2. Right-click the color swatch for the first value (the non-water class) → Change Opacity → set to 0.
  3. Click Apply.
  4. In the Layers Panel, make sure Flood_Output (the FCC) is turned on and positioned below NDWI_Output.
NDWI water detection overlaid on the false color composite, showing detected flood extent in blue over the FCC.
NDWI water detection overlaid on the false color composite, showing detected flood extent in blue over the FCC.

The blue areas are pixels the NDWI identified as water. You can now see exactly where the NDWI detected flooding that may have been ambiguous in the FCC alone.

Discussion

Exercise 2: Compare All Three Layers

Toggle between the NDWI overlay, the FCC, and the Esri satellite basemap to answer the following:

  1. Does the NDWI detect flood extent that the FCC missed? Where?
  2. Are there areas where the NDWI may have incorrectly classified non-water as water (false positives)?
  3. What are the strengths and limitations of each method? When would you use FCC versus NDWI?
Key Points
  • False color composites remap spectral bands to RGB channels, making vegetation (red), water (dark blue), and bare soil (grey/white) visually distinct.
  • FCC is a quick visual tool but struggles with muddy or shallow floodwater that appears similar to bare soil.
  • NDWI uses the ratio of green and near-infrared reflectance to isolate water features, providing a more reliable delineation of flood extent.
  • Overlaying NDWI on an FCC combines the strengths of both methods — spectral precision from NDWI and spatial context from the FCC.