PROJECT 5 : NORMALIZED DIFFERENCE VEGETATION INDEX (NDVI) IMAGE PROCESSING

Normalized Differentiation Vegetation Index, widely known as NDVI, was introduced in 1979 by Tucker as an index that measured the health and density of vegetation. 

HOW?

Conceptually, NDVI measures reflectance difference between Near Infra-Red or NIR and Red bands. NDVI is a good indicator of green biomass, leaf area index and production patterns and is the most commonly used vegetation indices. Theoretically, NDVI values are represented as a ratio ranging in value from -1 to 1. A low NDVI value usually shows areas of barren rock, sand, or snow (0.1 or less). In contrast, a moderate value of NDVI shows sparse vegetation like shrubs and grasslands (0.2 to 0.5). Dense vegetation like temperate and tropical forests will reflect high NDVI values of approximately 0.6 to 0.9. However, this classification depends on the sources of the satellite images used and the aims of the study. NDVI imagery is useful for various applications, including monitoring vegetation health, assessing vegetation stress, mapping land cover, studying land-use changes, and analyzing the effects of environmental factors on vegetation growth. It helps in identifying areas of dense vegetation, detecting drought conditions, monitoring deforestation, and managing agricultural practices. To generate an NDVI image, the reflectance values in the near-infrared and red wavelengths are collected using remote sensing platforms. These reflectance values are then processed using the NDVI formula to create a grayscale or color-coded image, where different shades or colors represent the range of NDVI values.

NDVI image processing provides a valuable tool for understanding vegetation patterns, dynamics, and changes across large areas, making it a widely used technique in environmental monitoring, agriculture, forestry, and land management applications. NDVI can be used in land cover classification due to its ability in differentiating type of the land cover by the percentage of red light reflected back into the space. Water or snow have a low NDVI value since it does not absorb the visible or red light while forest have high NDVI value since most of the visible light will be absorbed during photosynthesis. 

Table below show the classification of NDVI value:

For this project, the datasets was obtained from United States Geological Survey (USGS) official website which was retrieved from  Landsat 8-9 OLI/TIRS. The data was categorized into spectrum bands. And the suitable band must be achieved in order to display the image according to the type of analysis that done. For this project I specifically choose a band combination 4,3,2 displays a natural color image which will be suitable for NDVI process. Below are numbers of band combination according to their usage : 

The final result of my NDVI Image Processing project is shown below :

Based on the NDVI classification table above, The Selangor area has wide amount of agricultural forests land cover as shown in above map. The green spectrum represents agricultural forests and the purple spectrum represents water, bare soil or rock which is also represented as buildings. The major buildings are concentrated at Federal Territory (Wilayah Persekutuan ) Kuala Lumpur and Putrajaya.

There are huge difference of using remote sensing as data source in analyzing land use change compared to other map types 

Remote sensing data provides up-to-date and frequently updated information on land use. Satellite imagery can be collected on a regular basis, allowing for near-real-time monitoring and assessment of land use changes. In contrast, other maps, such as administrative or cadastral maps, may not be as promptly updated, and their availability can vary depending on the region or jurisdiction. Remote sensing data covers large geographic areas, making it suitable for regional or global land use analysis. Satellites can capture imagery that spans vast regions in a single acquisition, providing a comprehensive view of land cover and land use patterns. Other maps, such as detailed land use plans or zoning maps, often focus on smaller areas and may not capture the broader context of land use dynamics. Remote sensing data is collected using standardized and objective methods, ensuring consistency and reducing potential biases in land use analysis. In contrast, other maps may be influenced by subjective interpretations or outdated information. Remote sensing imagery allows for an unbiased assessment of land cover and land use, providing a reliable and consistent source of data. Remote sensing data provides multispectral information captured at different wavelengths, enabling the discrimination of various land cover types. By analyzing different spectral bands, such as near-infrared and red, it becomes possible to identify and differentiate vegetation, water bodies, urban areas, and other land use categories. Additionally, the availability of historical satellite imagery allows for the analysis of land use changes over time, providing valuable insights into trends and patterns. Remote sensing data collection does not require physical presence on the ground, making it non-intrusive and cost-effective for large-scale land use analysis. It eliminates the need for extensive field surveys or manual data collection, reducing both time and financial resources required. This makes remote sensing a practical and efficient tool for assessing land use on a broad scale. Despite these advantages, it's important to note that remote sensing data also has limitations. Cloud cover, atmospheric conditions, and the spatial resolution of the imagery can affect the accuracy and interpretability of land use information. Additionally, remote sensing data alone may not provide detailed contextual information or capture nuances that can be found in other maps, such as local land use plans or cadastral maps. Integrating remote sensing data with other sources can enhance the accuracy and completeness of land use analysis, providing a more comprehensive understanding of the landscape.

For additional references and datasets visit my data repository page in the blog