In cases like these, graphic models of standard GIS software are good solution, and today we will see how to solve this problem easily with ArcMap’s Model Builder. Not to mention that often you will look for a quick an easy solution in a graphic environment and the time of building a custom script (which includes searching for documentation, debugging, etc.) does not seem “worth it”. Although I strongly encourage other young professionals like me, as well as students and young graduates, to start building confidence with programming as soon as possible and become independent from commercial software, I understand that not everyone is on the same level. This is all well-known standard Remote Sensing theory and the automatic reclassification of a raster is rather easy to pull off, using programming (with Python, R, or a similar language).

We can, therefore, translate this knowledge to classify the values of a change image into discrete values: Toews, CC BY 2.5, MediaWiki CommonsĪs we can see, in a population that follows (or approximates) the normal distribution, 64.2% of values range from (mean-1 standard deviation) to (mean+1 standard deviation). But how do we decide for all the in-between values? Traditionally, and considering that the pixel values in the image follow the normal distribution, the change severity is decided based on the raster statistics: Chart of a ‘normal’ distribution with the classic bell curve shape. Obviously, a change value of 0 corresponds to an area that has not been burnt, and a value of 1 to an area that has been burnt severely. For example, to study a burned area, the difference of Normalized Burn Ratio (NBR) is a good indicator for the burn severity:

This is often performed in change detection studies, where the change of value between the same pixels in different dates corresponds to the level of change. In Remote Sensing, raster reclassification is a useful translation from continuous reflectance or index values to discrete classes that correspond to specific information. Recently, I found what seemed to be a good start: Automatic raster reclassification with ArcMap’s Model builder based on the statistics of the raster. I’ve often considered starting a series of tutorials regarding common GIS and Remote Sensing problems that we face in various software and might be somewhat tricky to work around. This article offers a step-by-step guide with relevant screenshots and explanations, as well as a short overview of the theory behind this problem. Alexandros Voukenas writes about the problem of automatic raster reclassification based on its statistics, using ArcMap’s Model Builder.

Arcmap model builder