Reference no: EM131004134

Practical Exercise: Topographic analysis

Introduction and Objectives

Elevation above sea level and terrain slope exert significant effects on the appearance of Earth surface features in satellite images and on the way that radiation scatters from the Earth's surface. Hence the availability of Digital Elevation models (DEMs) can help us to interpret and process remote sensing data.

This practical is designed to introduce basic analysis tools for DEM data and allows you to visualize remotely-sensed data using a pseudo-3D or 2.5D view.

Starting Up

Assuming you have logged into one of the Open access computers in Geography, start the Envi-4.2 software from the Geography

Applications Folder which is in the folder Specialist Apps / College of Science.

Using Open Image File from the Load Menu, load the file wales100.dem which you will find on the shared Common(L:) drive under:
L:/College of Science/Geography/GEGM10/prac1_topographic_analysis.

Meta-data for this image is included in the header file in the same directory and is loaded automatically.

The data

You will investigate a DEM of Wales at 100m pixel size derived from the Shuttle Radar Topography Mission (SRTM), namely file:

L:/College of Science/Geography/GEGM10/prac1_topographic_analysis/wales100.dem

and a Landsat TM mosaic of Wales from the Geocover project (https://zulu.ssc.nasa.gov/mrsid/) resampled to a pixel size of 50m and presented as only 3 of the spectral bands, namely:
wales50_bng_landsat.img
from the same folder.

Exercise

1. Open the file wales100.dem into a Grey Scale viewer as indicated in the instructions above for starting up. Notice how it is difficult to get a feel for the elevations using a grey-scale image, except for perhaps the visibility of river valleys.

2. From the Tools menu choose Colour Mapping-> ENVI Colour Tables. Investigate the colour tables available within ENVI and consider how using colour makes the elevation data easier to interpret. When you have finished, set the colour table back to B-W linear (grey- scale).

3. From the Topographic Menu, choose to apply Topographic Modelling to the DEM Band 1. Choose only to model the Slope, Aspect and Shaded Relief (use the Ctrl key to select more than one). Give a Sun Elevation Angle of 45 degrees and Azimuth Angle of 180 degrees (South), and set the X and Y Pixel size to be 100m. You will be prompted to choose filenames for the outputs but you may instead choose to place the outputs only in memory.

4. Have a look at the Slope and Aspect images and consider what these tell you and how they may be important in understanding both what might be visible on the land surface and how radiation from the sun may be scattered by it.

5. Open the Shaded Relief into the Gray Scale viewer. Investigate how and why this image gives a more visually appealing view of surface topography (https://www.reliefshading.com/).

6. Open the file wales50_bng_landsat.img and investigate this (limited spectral band) mosaic of Landsat data from the year 2000. Notice how the land surface topography is visually apparent in this image because of the relatively low sun angle.

7. From the Topographic Menu choose 3D Surface View. You may first be asked for the display image to drape over the DEM (choose the Display with the Landsat image in it), and then for the DEM over which to drape this image (wales100.dem). In 3D SurfaceView Parameters, choose a DEM Resolution of 512 and the Full Image Resolution.

8. The result is not unlike Google Earth but in this case we have complete control over the images displayed. Have fun ‘flying' around Wales. The left mouse button controls the rotation of the view, the middle button controls the view point and the right hand button controls the zoom factor. This takes some practice. Consider how this kind of pseudo-3D perspective view (also known as 2.5D) can make it easier to interpret the satellite image with respect to different land covers and types of terrain. Investigate Surface Controls under the Options Menu and try changing the DEM Resolution, the Vertical Exaggeration and the Perspective Controls.

9. You may also have a look at a dataset labelled Swansea100.dem and Swansea50_bng_landsat.img. These are merely a subset of the Wales dataset but they will allow you to visualize the landscape at a much higher spatial resolution without grinding the computer to a halt.

Practical Exercise: Feature Space and Classification

Introduction and Objectives

This exercise will introduce you to image analysis with remotely-sensed data, allow you to become familiar with two-dimensional feature space and show you how to make a simple image classification. The exercise is based on a Landsat Enhanced Thematic Mapper (ETM+) sub-image from Norfolk using a software package called ENVI (Environment for Visualising Images). Be bold in your interactions with the software. Investigate all of the facilities available in the menus and experiment with the options available to you. Use the on-line Help to learn more about the capabilities of the software, as well as showing you how to proceed when you get stuck.

Starting Up
Assuming you have logged into one of the Open access computers in Geography, start the Envi-4.2 software from the Geography Applications Folder, as shown in the previous practical.

Using Open Image File from the Load Menu, load the file etm_norfolk_19950509.img which you will find on the Common(L:) drive under:
L:/CollegeofScience/Geography/GEGM10/prac2_feature_space.

Meta-data for this image is included in the header file in the same directory and is loaded automatically.

When the image is located you should be presented with the Available Bands List.

Exercise

The ETM Image has 7 bands with the following spectral characteristics:

SENSOR BAND	DESIGNATION	WAVELENGTH RANGE (μm)
Band 1	Blue	0.45 - 0.52
Band 2	Green	0.53 - 0.61
Band 3	Red	0.63 - 0.69
Band 4	NIR	0.78 -0.90
Band 5	MIR	1.55 -1.75
Band 6	TIR	10.4 - 12.5
Band 7	MIR	2.09 - 2.35

Have a look at each band in turn by loading it into a grey-scale image viewer. Make the image viewing window as large as possible. Use the scroll window to move around the image and the zoom window to look at detail. You should be able to identify various features including the Wash (with mudflats and water), clouds (and their shadows), agricultural fields, towns, roads and patches of woodland. Comment on how the different spectral bands compare in their ability to distinguish these different features.

Go back to the Available Bands List, choose an RGB-Colour image viewer, and load a true- colour-composite image into this viewer (red-band to R, green-band to G and blue-band to B). Discuss how this image appears in comparison to how you would expect the landscape to look to you directly from an aircraft. Why might you expect differences?

Load a false-colour-composite image into the viewer (e.g. band 4 [NIR] on red, band 5 [MIR] on green, band 3 [red] on blue). What is immediately apparent about this image, which includes both the visible and Near Infrared parts of the spectrum? Consider the normal spectral response curve for vegetation (below) and comment on your ability to discriminate different land-cover types using this method of display.

From the Functions menu choose Tools->2-D Scatter Plots. Pick band 3 (red) for the X-axis and band 4 (NIR) for the Y-axis. The resulting plot is of the two-dimensional feature space that you should be familiar with from the lecture. The density of points on this graph may be visualised by choosing Density Slice from the Options menu. Consider on the distribution of points within this scatterplot - are there any natural spectral classes?

From the Scatter Plot Class menu, choose a colour from Items 1:20. Using the left mouse button, draw a polygon within the plot window over an area of the plot that you would expect to be associated with a particular land cover class, finishing it with a right-mouse click. The corresponding land-cover type should immediately be highlighted on the image. Experiment with different colours and regions of the feature space and try to interactively discriminate at least the following image classes:

1. cloud
2. water
3. forest
4. mudflats
5. bare fields
6. fields with crops

This is a rudimentary image classification and is very useful for understanding the relationship between image space and feature space.

Practical Exercise: Exploring time-series data

Introduction and Objectives

Remote sensing gives us the capability of monitoring land surface features and processes through time. Global monitoring initiatives, in particular, result in datasets that demonstrate the time varying nature of the Earth's surface in its response to seasonal cycles, climate oscillations (such as El Nino) and longer term trends. This practical is designed to illustrate how global datasets can be visualized and allows you to explore the seasonality in land surface vegetation.

As normal, we will use the remote sensing software Envi-4.2 which you will find in the folder

Specialist Apps/College of Science/Geography on the desktop.

The data

You will explore a global vegetation index dataset generated by Dr. Sietse Los of the Geography Department. The Fourier-Adjusted, Sensor and Solar zenith angle corrected, Interpolated, Reconstructed (FASIR) Normalized Difference Vegetation Index (NDVI) data sets were generated from Advanced Very High Resolution Radiometer (AVHRR) data to provide a 17-year, satellite record of monthly changes in the photosynthetic activity of terrestrial vegetation.

This dataset is presented as a single file consisting of 204 bands of 360 columns by 180 lines of data. Each band represents a month between January 1982 and December 1998 and each pixel covers a grid square of one degree longitude by 1 degree latitude.

Values in these images represent the NDVI which is an index of vegetation concentration between 0 and 1. There are also other values in these images: -99=water, -88=missing data (persistent cloud) and -77=permanent ice.

On the Common(L:) drive, open the file ‘fasir_1982-1998.bin' from the folder:
L:/College of Science/Geography/GEGM10/prac3_time_series
When the image is loaded you should be presented with the Available Bands List.

Exercise

1. Load the first band (198201 = January 1992) into a grey-scale viewer. From the Enhance menu on the viewer, choose interactive stretching to adjust the contrast on this image. Under Options->Histogram Parameters choose Histogram Min and Max of 0 and 1 to give a value from black to white for the NDVI. Apply these parameters to view the global NDVI values for this time period.

2. Repeat this operation for the July image of the same year but open it into a New Display to allow you to compare NDVI between the Northern hemisphere winter and summer. Describe the patterns that you see, the differences between January and July and the overall distribution of NDVI between these time periods. Is the seasonality around the globe always clear?

3. From the Tools menu on one of the viewers, choose Colour Mapping->Density Slice. This dialogue allows you to map different ranges of the NDVI images onto different colours. Choose to apply the Density Slice to the January 1982 image. From the Density Slice dialogue choose to Restore ranges from the file ‘fasir_1982-1998.dsr' from the folder L:/College of Science/Geography/GEGM10/prac3_time_series. I have set up this colour mapping to clearly show the water and ice classes and to illustrate NDVI in shades of Green. Try adjusting the colours using Edit Range to improve the view. Think about how this helps to visualize the data. To change the band to which the Density Slice applies, use Options-
>Change Density Slice Band.

4. So far we have viewed only a small proportion of the available data. Most of the information resides in the time-varying differences in NDVI throughout the 17 year sequence. To visualize this signal, choose Profiles->Z Profile from the Tools menu on the viewer. Click the mouse over a part of the image to draw a graph of NDVI as it varies through time. From the graph, choose Edit->Plot Parameters and set the Y-axis range between 0 and 1 and turn off Options->Autoscale Y axis to always view the NDVI between zero and one.

5. Using Tools->Cursor Location/Value, find out where in the world the mouse is pointing and convert the X and Y pixel positions to Latitude/Longitude. NB: 0,0 will have a Latitude of +90° and a Longitude of -180°; Longitude and Latitude of 0°,0° will have pixel coordinates of (180,90), etc.

Seek parts of the world that have the following time-varying NDVI characteristics:
a. Annual greening up in Apr/May/Jun and senescence in Sep/Oct/Nov (e.g. Europe).
b. Clear southern hemisphere reversal of this seasonal trend (e.g. South Africa).
c. Bi-annual greening signal (e.g. Central Africa).
d. Consistently LOW NDVI (e.g. the Sahel in central North Africa).
e. Consistently HIGH NDVI (e.g. the Amazon Rainforest in Brazil).
f. Occasional departure from a consistent signal (e.g. parts of central Australia). These may be due to the El Nino Southern Oscillation or ENSO events.

By understanding the nature of NDVI, and exploring the different time-signatures of NDVI around the world, it will become clear how valuable quantitative remote sensing of vegetation has been to understanding vegetation patterns and seasonality.

Practical Exercise: Comparing data from different parts of the spectrum

Introduction and Objectives
Our own visual system leaves us familiar with only the visible part of the electromagnetic spectrum which is narrow compared to the range of wavelengths that remote sensing can employ. This practical is designed to help you become familiar with the microwave part of the spectrum by comparing radar images, and interferometric coherence, with a Landsat image.

As normal, we will use the remote sensing software Envi-4.2 which you will find in the folder

Specialist Apps/College of Science/Geography on the desktop.

The data

The data covers part of South Wales and England between Cardiff and Bristol and has been resampled to a pixel size of 50m. All files have the same spatial sampling and dimensions to allow for easy comparisons.
There are 3 ENVI data files on the Common drive:
L:/College of Science/Geography/GEGM10/prac4_compare_spectrum

• landsat_20020911: A 7-band landsat image from September 2002 resampled to 50m pixels.
o Band 1
o Band 2 Blue Green 0.45 - 0.52 μm
0.53 - 0.61 μm
o Band 3 Red 0.63 - 0.69 μm
o Band 4 NIR 0.78 -0.90 μm
o Band 5 MIR 1.55 -1.75 μm
o Band 6 TIR 10.4 - 12.5 μm
o Band 7 MIR 2.09 - 2.35 μm

• ers_and_jers: ERS (European Remote Sensing Satellite) and JERS (Japanese Earth Resources Satellite) images resampled to 50m
o band names gives dates of the images. ERS is a C-band sensor (5.6 cm wavelength) and JERS is an L-band sensor (23 cm wavelength)
o ERS includes ascending (satellite headed North looking East into the scene) and descending data (satellite heading south and looking West into the scene)
o The one JERS scene covers Cardiff only (not used in this practical).

• ers_coherence: ERS interferometric coherence (phase correlation between two images)
o one image of tandem (1 day) coherence
o one image of coherence between images separated by 13 months
To load each image use Open Image File from the Load Menu. When the image is located you should be presented with the Available Bands List. You can choose to look at each image separately as a grey-scale image, or load three bands at once into a colour composite.

Exercise

1. Open the file landsat_20020911 and re-familiarise yourself with the wavebands from Blue through Near-InfraRed (NIR) to Thermal InfraRed (TIR). There are two False Colour Composites which are of particular value in discriminating vegetation from other land covers. These are 4,5,3 on R,G,B and 7,4,2 on R,G,B. Consider the following:

a. What land covers can you identify in this part of the UK?
b. What proportion of the scene is urbanised?
c. Can you identify the major Cities (Cardiff, Newport, Bristol)?
d. Can you see the two Motorway crossings over the Severn Estuary?
e. How does the false-colour composite aid image interpretation?
f. How does feature space compare to East Anglia (Practical 2)?

2. Open the file ers_and_jers. For this practical ignore the JERS (L-band) image, as it has neither the coverage nor the quality for a reasonable comparison. Investigate the various ERS bands on their own and as multi-temporal colour composites. Think about the following issues:
a. What dominates the level of backscatter in the SAR images (think of several features of the landscape and a hierarchy of influence)?
b. What changes most between SAR images from different dates and why?
c. How do the SAR images compare to the Landsat scene?
d. In what way do the ascending and descending images appear different, and why?
e. Why do the SAR images appear much more ‘noisy' than the optical one?

3. Open the file ers_coherence. Compare the phase coherence for a pair of images separated by one day to that between a pair of images separated by over a year. Consider the following:
a. Can you explain the differences between these images?
b. What does short-term coherence discriminate between?
c. What does long-term coherence discriminate between?
d. How does the information content compare to that of the individual SAR images?

Use these questions as a guide to understanding the images but avoid following them as a structure for your write-up. Think about how answers to these questions will help you in your preparation of the coursework.