Change Detection Exercise:

Image Differencing

Introduction

One of the simplest techniques for detecting change between two images is that of image differencing. With this, each pixel from an image is subtracted from its corresponding pixel in another image. All raster systems have some form of image subtraction routine. In this exercise, we will use this technique to see change across two time slices.

The Data Set

The data used in this exercise consists of two images of Normalized Difference Vegetation Index from Southern Africa for March of 1991 and March of 1992. Most of the crops are in their maturity stage around this time of the year. Hence, the vegetation index should be at its highest around this time. This data set was produced by USAID/FEWS from compositing dekadal (10 day) images into a monthly maximum value composite, where the maximum value over the three dekads in each month for each pixel is used to represent that pixel's monthly vegetation index.

Procedure

These are the NDVI images for southern Africa for March of 1991 and 1992. The values range from 0 to 255 with higher values indicating higher vegetation levels.

Q1 Can you visually detect areas of negative change? Use the inquiry option in your system to check the values for corresponding pixels in both images. See if you can find areas of change based on the NDVI values.

P2 Use the Map Algebra procedures of your software to subtract SMAR91 image from SMAR92 image. Call the output image SDIF9291. Display it with a 256 color palette. If there is a possibility to overlay a vector layer and a text layer of country boundaries and names, display COUNTRY (vector file) and NAMETEXT (name file) on the SDIF9291 image. Use the inquiry option in your system to check the values in this output image.

Is it relatively easier to detect areas of negative and positive change from the SDIF9291 output image. Areas showing negative values are those that show lower levels of vegetation in March of 1992 as compared to that of March 1991 and vice versa for those with positive values.

P3 Use the inquiry mode in Northern Mozambique (around the area of column 330, row 160) and south eastern Zaire (around column 238, row 43) and note down the values from the SDIF9291 image. To get a better idea of the change areas, we could reclass them into a smaller number of categories. The classes could be determined by some previously agreed threshold (based on expert opinion or empirical analysis, etc.) or spatial standard deviation units (showing relative levels of change across the image) or temporal standard deviation units (each pixel expressed in standard deviation units based on its own long term mean and standard deviation).

P4 We will first do a categorical classification of SDIF9192 with a predetermined threshold. Use the reclassification procedure in your software to assign the following values:

Assign a new value of	to all old values
Assign a new value of	ranging from	to
1	-241	-31
2	-30	0
3	1	29
4	30	242
0	0	1

Call the output image SD9192RC. Display it with a qualitative (four color) palette.

Now it is easier to identify areas showing lower NDVI in 1992 as compared to 1991.

Q2. If you overlay the country name file on this image, could you draw conclusions on which parts of certain countries are showing high negative deviations?

Observations

Simple differencing is one of the most common techniques to detect change across time. An evaluation of change is usually followed by overlaying it with socio-economic data on population, their access to markets, etc. Extraction of numbers of affected population in this area and a sketch of their vulnerability profile to crop losses can explain the NDVI deviations in terms of human/economic impact. Later exercises will highlight some of these interactions.

Back to Module 8 - Digital Change Detection

Credits

This exercise was written by Mahadevan Ramachandran, Research Fellow at the IDRISI Project/Clark University. A major portion of this exercise was culled from the NDVI workbook and the Change and Time series Analysis workbook (Ramachandran et al, 1996 and Eastman et al, 1995).

References

Eastman et al., (1995) "Change and Time Series Analysis" United Nations Institute for Training and Research/Clark Labs for Cartographic Technology and Geographic Analysis, Clark University, Worcester, MA 01610, USA.

Ramachandran et al., (2996) "An African NDVI Archive, 1982-1996: Techniques for Environmental Monitoring and Analysis" United Nations Institute for Training and Research/Clark Labs for Cartographic Technology and Geographic Analysis, Clark University, Worcester, MA 01610, USA.