How to Identify Land Cover Types in Built-Up Areas

If you are not provided with sample training sites for supervised image classification, you can identify the correct land cover types for training sites by following these steps: 1. Acquire knowledge of the study area: Familiarize yourself with the study area by researching the geography, climate, vegetation types, land use patterns, and any existing land cover maps or reports. This will help you recognize potential land cover types. 2. Interpretation of high-resolution imagery: Use high-resolution imagery, such as Google Earth or available satellite imagery, to study the landscape features and land cover types present. Visually identify training sites by cross-referencing color, texture, shape, and patterns to associate them with specific land cover classes. 3. Ground truthing (if possible): Conduct field visits to verify and collect information about land cover types in different locations within the study area. This information can be used to validate your interpretation of the high-resolution imagery and select accurate training sites. 4. Ancillary data: Use auxiliary resources like topographic maps, land use maps, existing land cover maps, vegetation maps, and aerial photographs to help identify land cover types and potential training sites accurately. 5. Expert consultation: Consult local experts or stakeholders knowledgeable about the region, who can provide valuable insights and guidance on identifying the correct land cover classes for the study area. 6. Stratified random sampling: Divide the study area into smaller, homogeneous regions based on characteristics such as elevation, slope, aspect or land use patterns, and collect training sites within these smaller regions. This approach ensures an even distribution of the training sites and reduces the impact of spectral variability. After carefully identifying the training sites, create training samples in your image classification software by digitizing these locations and assigning appropriate land cover class labels, which will be used for training the supervised classification algorithm.