Land Use / Land Cover (LULC) Classification
**LULC Classification** is the process of categorizing each pixel of a satellite image into meaningful classes — like urban, water, vegetation, or bare land — turning raw imagery into an actionable thematic map. ---
📋 Prerequisites
- QGIS Desktop 3.22+ installed on your computer.
- Basic understanding of GIS data concepts.
Step-by-Step Instructions
Define Your Classification Scheme
Before classifying, decide what classes you need. A basic scheme might include: - Urban / Built-up - Vegetation (forest, cropland) - Water bodies - Bare/Agricultural land Keep the number of classes reasonable — too many increases confusion between spectrally similar classes. ---
Prepare Your Satellite Image
1. Download a cloud-free multispectral image (Landsat or Sentinel-2) covering your study area 2. Load it into QGIS: **Layer → Add Layer → Add Raster Layer** 3. Clip it to your area of interest using **Raster → Extraction → Clip Raster by Mask Layer** ---
Choose a Classification Method
There are two main approaches — covered in detail in the next tutorial: - **Supervised Classification:** You provide training samples for each class - **Unsupervised Classification:** The software automatically groups pixels into clusters, which you then label This tutorial focuses on the supervised approach, which is more common for accurate LULC mapping. ---
Install the Classification Plugin
1. **Plugins → Manage and Install Plugins** 2. Search for **"Semi-Automatic Classification Plugin (SCP)"** 3. Click **Install** ---
Collect Training Samples
1. Open the SCP dock panel 2. Create a **Training Input** shapefile 3. For each class, digitize several polygons over representative areas (e.g., draw polygons over clearly urban areas for the "Urban" class) 4. Repeat for all classes — aim for multiple samples per class, spread across the image ---
Run the Classification
1. In SCP, go to the **Classification** tab 2. Choose an algorithm (Maximum Likelihood is a solid default) 3. Select your input bands and training samples 4. Click **Run Classification** 5. A new raster is generated, with each pixel assigned to one of your defined classes ---
Assess Accuracy
1. Collect independent reference/validation points (not used in training) 2. Use SCP's **Accuracy Assessment** tool to generate a confusion matrix 3. Check the **Overall Accuracy** and **Kappa Coefficient** — values above 80% and 0.8 respectively are generally considered good ---
🧠 Quick Recap & Practice Questions
## Quick Recap
- LULC classification assigns each pixel to a meaningful land cover class
- Define your class scheme before starting
- Supervised classification uses training samples you provide
- Always validate results using an accuracy assessment
---
## Practice Questions
1. Why is it important to keep your classification scheme simple?
2. What is a confusion matrix used for?
3. Name one classification algorithm available in the SCP plugin.
---
*Next Tutorial: Supervised vs Unsupervised Classification*