Hi F-N,

You probably mean fire probability maps? Have a look at my previous answer on your previous question with regard to natural fires (if I am not mistaken). There is alink in that answer, which leads you to a reserach group which produces (vegetation) fire probability maps (for Europe).

Cheers,

Frank

Hi F,

No I don't mean fire probability maps. I really mean vegetation maps, but without the relationship to previous fires. Some sort of potential vegetation.

The goal is to avoid the obvious relationship between actual post-fire vegetation and recent fire perimeters, which will lead to a strong probability of fire whereas this area is less fire-prone since this fire => The fuel load is far lower than before the fire.

The routine way is to combine three groups of indices: first is an index to show the amount of green vegetation (such as NDVI); second it an index to demonstrate the amount of water content of the vegetation and the last index is something to bring out non-photosynthetic vegetation (dry carbon). If I comprehend your question correctly, you found high probability in some regions that have already been fired?! Well, what indices have you been utilized? If you exactly made use of three mentioned groups of indices; please let me know to read some relevant scientific paper or find a solution to resolve the matter.

Thanks Amir for the proposed method.

Yes, when you compute a statistical model of area burned, you will have a correlation between post-fire vegetation and fire perimeters, because they are spatially overlayed, so the model "thinks" that they are related.

Actually they are, but in a misleading way, because it is the pre-vegetation which lead to the fire, not the post.

Does it make sense ?

Okay, I do not know exactly what approach you've performed for your case study. But I think you should alter your method basically, since biomass have already been consumed and it can not be regarded as a influential factor for future fires, can it?

Please make use of a dry or descendent carbon index too. I should also mention that using passive remotely sensed imageries is a little tricky for fire fuel mapping, since you are not aware of some dry organic matters beneath tall trees. The better approach is to use LiDAR and SAR to model biomass and subsequently use it for fire fuel mapping.

If you need data on the pre-fire vegetation you should check for pollens and ashes in the lakes sediments or caves sediments if you are in a karst area.

@ Eric : Unfortunately, the major part of Alberta is not a karstic area...but I could check for existing studies for pollen and past vegetation in here.

@Amir : I work at regional scale, so RADAR would be a good approach for fuel mapping. Althought, it won't fix my problem.

@Alessandro : Your approach has been used by Marc-André Parisien and Max Moritz several times with good results. It is actually a good one, and I wan't to try it. But it is still a statistical approach, and the problem will remain : the model will strongly associate presence of fire with recent post-fire vegetation. As fires in Canada are so big, it creates a bias in the results.

So what I'm probably going to do is taking the oldest vegetation data I can get (around 1980 probably) and consider it as a base layer for pre-fire vegetation, as I'm working on fire occurrence in a recent period.

Thanks for all your tips.