Image Calibration and Indices

• Import Landsat 8 image (“LC08_L1TP_232066_20220921_20220929_02_T1_MTL.txt”) into ArcGIS Pro

• First visualise the image in natural color and take a moment to look at it and understand what it shows.

Band 1 captures deep blue and purple. Bands 2, 3, and 4 are blue, green, and red. Band 5 measures near infrared, or NIR. Bands 6 and 7 cover different parts of the short-wave infrared, or SWIR.Band 8 is the panchromatic, or simply panoramic, band. It works in the same way as black and white film: instead of collecting visible colours separately, it combines them into one channel.

First, I visualised the image in natural colour

Using the base map, I compare the image with the base map. In the photo, you can see residential infrastructure, agricultural infrastructure, smoke and clouds, but the river is lighter in the photo

• Next, zoom all the way in to some areas where the smoke is originating and change the visualisation tothe Vegetation Analysis band combination

Change the band combination to Vegetation Analysis. Use the Swipe tool "Raster Image Layer tab"

The water changed colour to blue, the green areas remained green, in places where there was heavy smoke it was almost invisible, and in places where it was almost invisible it disappeared completely

1. Q1: How does this band combination help us get additional information about what is happening(compared to the natural color band combination)?

Based on the name "Vegetation Analysis" and the fact that it has been tested in practice, this combination of colour bands is well suited for plants and areas with brighter colours. It is suitable if you need to capture images without smoke and, depending on the situation, without clouds. Areas with residential infrastructure are brown, so this band combination is not suitable for their display. I think it will be good for comparing the area after natural disasters

Corrected the image using a raster function "Apparent Reflectance“ ( Imagery tab >Raster Functions > Correction)

• When all parameters are set RUN the tool to create a new layer.

• Тo ensure that new layer is properly saved, export the new layer as a Raster

To make sure that the new layer was saved correctly, I exported it as a raster by "right click on the layer > Data > Export Raster"

• Visualise the two images (the uncorrected one and the corrected one you just created) in the same bandcombination (e.g. vegetation analysis)

I've visualised two images (the uncorrected image on the top and the corrected image on the bottom, which you just created) in the Vegetation Analysis strip combination after applying the stretch type "None".

1. Q2: Use the swipe tool: Are the two images visually different from one another? Provide a screenshotand shortly answer the question

Comparing the two photos, you can see a slight difference between the photos. When viewed on a large scale, there is no visual difference as such, but differences can be seen when looking at individual elements in the photo. However, you can see some difference in the data of these images, as the calibrated ones have more accurate and reliable measurements

Created a spectral profile for comparing images using the provided point (poi) shapefile

Сreate the spectral profiles "Click on the image layer to select it > Data Tab > Create Chart > Spectral Profile"

Mark the points that you set to "poi" according to the terrain type.Resample the image to "nearest neighbour" so that you can identify individual pixels. Then I zoomed in on the "poi" points in the shapefile to see the pixels they are located on

I selected the same pixels in both images to compare their spectral profiles.

Set the Y-axis for both spectral profiles to 0 - 65500 (for comparison).

Set the chart names to "Multispectral" and "Apparent reflectance"

For a clearer visualisation of the spectral profile, I changed the Graph Type to Inserts and Midlines

1. Q3: What are the differences? Are the differences greater in some bands than others? If so, what reason mightthis have

The differences between spectral profiles of calibrated and uncalibrated images can be significant and can vary across different bands. These differences arise due to several factors( Calibration Process,Sensor Response,Atmospheric Effects,Radiometric Calibration)

The extent of differences between calibrated and uncalibrated spectral profiles can vary across different bands. This variation can be influenced by factors such as the sensor's spectral sensitivity, the specific calibration techniques employed, and the presence of atmospheric or surface effects that affect certain bands more than others. Additionally, the magnitude of differences may also depend on the quality and precision of the calibration process applied to the images.

In conclusion, calibrated images provide more accurate and reliable spectral profiles by accounting for sensor response, atmospheric effects, and radiometric calibration. Uncalibrated images, lacking these calibration corrections, may exhibit deviations, distortions, and variations in their spectral profiles. The differences in spectral profiles between calibrated and uncalibrated images highlight the importance of calibration for obtaining accurate and meaningful remote sensing data.

• Calculate an index for both images

I calculated the NDVI index for both images using the raster function "Strip Arithmetic" to perform the necessary calculations for the indices.

• Use the Raster Function “Band Arithmetic” to do the required band math for the indices

When using indices, the symbology can be a crucial element to help us interpret the result.Seted the symbology to “Classify” and then change the settings to create 3 classes with theupper bounds: < 0.0, < 0.45, < 1.0

I gave the symbols meaningful colours, applied the symbols that I created for the first NNDVI layer to the other one using the "Import from Layer" function in the Symbology tab

1. Q4: What does the NDVI show (in general)?

The Normalized Difference Vegetation Index (NDVI) is a commonly used vegetation index derived from remote sensing data, particularly satellite imagery. It provides valuable information about the presence and vigor of vegetation in a given area.

In general, the NDVI quantifies the difference between the reflectance of near-infrared (NIR) and red light regions of the electromagnetic spectrum. It is calculated using the following formula:

NDVI = (NIR - Red) / (NIR + Red)

By analyzing the NDVI over time and space, researchers, ecologists, and agricultural experts can gain insights into vegetation dynamics, monitor changes in land cover, identify areas affected by drought or deforestation, and assess the overall health and productivity of vegetation.

It is important to note that while the NDVI is a widely used index for vegetation assessment, it provides a general indication of vegetation presence and health. It does not provide information about specific plant species, individual plant health, or other factors that may affect vegetation condition. Therefore, other data sources and indices may be combined with NDVI for a more comprehensive analysis of vegetation characteristics.

1. Q5: How can the 3 classes be interpreted, that we created in the symbology? Are they different fromeach other? If so, how? Please be explicit in your answer

Class 1: NDVI Range: -1 to 0 This class represents the lowest values of NDVI, ranging from -1 to 0. It generally corresponds to non-vegetated or sparsely vegetated areas.

Class 2: NDVI Range: 0 to 0.45 This class represents moderate NDVI values, ranging from 0 to 0.45. It generally represents areas with varying levels of vegetation density and health.

Class 3: NDVI Range: 0.45 to 1 This class represents the highest NDVI values, ranging from 0.45 to 1. It generally corresponds to areas with dense and healthy vegetation.

The three classes created based on the NDVI range differ from each other in terms of the level of vegetation presence and health they represent:

• Class 1 (NDVI: -1 to 0): Represents non-vegetated or sparsely vegetated areas.
• Class 2 (NDVI: 0 to 0.45): Represents areas with moderate vegetation density and health.
• Class 3 (NDVI: 0.45 to 1): Represents areas with dense and healthy vegetation.

These classes provide a way to categorize and distinguish different levels of vegetation cover and vigor based on the NDVI values. By visually representing the data with different symbology, these classes allow for a clear differentiation of the extent and health of vegetation across the observed landscape.

1. Q6: Could there be a risk in using an uncalibrated image when trying to find out how muchdeforestation there is? What conclusions can you draw about the usefulness/necessity of performingimage calibration as an image pre-processing step?

Using uncalibrated images when trying to determine the extent of deforestation can indeed introduce risks and affect the accuracy of the analysis. Image calibration is an important preprocessing step in remote sensing and image analysis, and its necessity depends on the specific application and objectives.

Uncalibrated images can have various problems, such as radiometric and geometric distortions, variations in lighting conditions, sensor artifacts, or inconsistencies in color representation. These factors can significantly affect the accuracy and reliability of the analysis, especially when trying to quantify the extent of deforestation.

By performing image calibration as a pre-processing step, several benefits can be achieved:

Radiometric Consistency: Calibration ensures that the pixel values in the image match the valid radiometric values recorded by the sensor.

Geometric accuracy: Calibration corrects geometric distortions caused by the characteristics of the sensor or imaging platform.

Standardization: Calibration helps standardize images by removing sensor-specific biases and artifacts.

Accuracy of derived indices: Image calibration enhances the accuracy of derived indices, such as vegetation indices (e.g., NDVI) commonly used in deforestation analysis

In conclusion, image calibration is highly useful and often necessary when analyzing deforestation or any remote sensing application. It improves the accuracy, reliability, and comparability of image data, enabling more precise measurements, change detection, and monitoring over time.