We might need some clarification. I am assuming the following things:
[1] You have a regression model with DBH and canopy/tree height as predictors (from a published paper or other analysis)
[2] You have data on crown area and canopy/tree height at your study site
[3] You model to apply published regression model to your data.
Successfully doing this all depends on how generalizable the original regression model is to your data. If the original data were taken at your study site, then it seems that you would need a good relationship between crown height and DBH to make this work. In that case, you would definitely need to account for the measurement error and/or unexplained variance between crown height and DBH. I can imagine cases where doing this might be necessary (i.e. there was no way to measure DBH due to rabid wolves chained the base of the tree), but the introduction of error might reduce the explanatory power of your model.
If the original regression used taken from, say a tropical forest, and you are studying a fragmented, temperate zone forest, then there may the obvious issue that DBH and crown height might have a different relationship altogether.
Anyway, let us know more details. It sounds like a case where remote-sensing data might need a bit of ground-truthing!
There are numerous publications on using lidar to inventory forests that use the crown area to predict dbh and volume. I have not personally researched this question but you can search for lidar and forest inventory and find out more.
@Kathy Roche. Models that use LiDAR data to relate crown area to DBH still require some level of on-the-ground validation of the area-DBH relationship for particular tree species or in different forests. I only guessed that the original poster (Nicolas) was using some type of remote sensing. @Nicolas, perhaps you can describe your data a bit more?
Your assumptions are right. Crown area as a predictor of biomass is my missing link between my ground based measurements and remotely sensed information.
I used point-centred quarter method to estimated trees/ha, and have following data of 367 samples from an area of about 2 km²: DBH, height, crown area and crown height.
The regression model of Populus euphratica is from the same area (Xinjiang) - the authors are Chen and Li 1984 (in Chinese). So I do have something like "Volume = log 0.0382 + 0.8837 log D² H"(D²=dbh²; H=height).