My understanding of the question is this: You're asking how to compute differences in deforestation from 200-06, 2006-20, and 2010-12. And you have three raster layers, showing some measure of change in forest cover for a given area for those time periods.

Assuming this is correct, and you have raster maps covering the same region, QGIS raster calculator might help you out - it's a pretty quick and simple method for computing differences between raster maps:

http://docs.qgis.org/2.8/en/docs/user_manual/working_with_raster/raster_calculator.html

Hope that helps, apologies if I've misunderstood the question.

Regards

Just a step back: how large are your pixels? What is the underlying question you want to answer? What do you mean by a deforestation pixel? Is it a pixel which has been deforested in the period specified, so if the deforestation is older, it will not show up? In that case there should be no overlap in deforestation, since a forest cannot grow back between 2006 and 2012, an area cannot be deforested twice. So, what are the differences you are looking for? Size or pattern of the deforested area, whether it is in the same geographic region? An easy visual start could be to display the three deforestation maps in three colors on your screen (like you would display 3 bands of a satellite image), that would clearly show you (change in) deforestation pattern and give you clues on how to further your analysis.

Thanks for all your answers

The question is related to quantify how much is the similarity in the deforestation patterns through statistical analysis.

E. G. Moran index

Hi Miguel, 

I think the  reference attached might help.

Best, 

 Daniela

Article The Map Comparison Kit

There is another way. You can do a statistical comparison based on mutual information or other similarity measures between your latest dataset and the other two. 

Thank you Daniela! i will try to implement this tool!

best regards

Hi,

You can also think to use geostatistical tools, for example: compare the variogram of the two data set and also you could try to use cross-variograms.

Then you can think to use the fragstat approach and see how pattern metrics change between the 2 datasets (well, you need a categorical map in this case).

Bye,

Sebastiano

"similarity" is a useful idea but not a precise one, essentially the only way you can define it is to choose a particular way to quantify it. Perhaps you want to step back and ask what you will use the measure of "similarity" for, is it a biological question or an economic one or something else. Obviously you can compare pixel by pixel but it is unlikely that is good enough, you need to be able to "summarize"  those results in some way and there is no single best way. Note that Moran's I and variograms are based on rather different underlying statistical assumptions.

Kriging is an interpolation method. We use this method if we want to interpolate the "value" of a point based on some other points that have known values. If we are interested in location of the points, we could use point pattern analysis. However, it works for vector data.

Dear

you can do change detection you will reach your point

I note that the lengths of the different time periods are also different, any scheme for comparing deforestation that does not take this into account will be mis-leading. Observed differences or similarities may be a function of the difference in the lengths of the time periods, making the results of the analysis irrelevant. Obviously you can not go back in time to collect data differently but it does point out the importance of thinking about how you are going to use data before you spend the time, money and effort to collect it. You didn't say how the data sets were acquired and at what expense. If you or your group collected the data you may have wasted your time, money and effort. If you got them for free, i.e. someone else paid for collecting the data it is reasonable to ask what information you can glean from the data but it may still be erroneous or mis-leading.

You can devide intersection areas by union areas between two rasters