Do you have a measure of crown size for example? there are several approaches you could use considering that you have some ground measurements or information on allometric relationship (crown area, dbh, etc) for tree species mapped.

Once you have some sort of known allometric relationship you can easily derive estimates for number of trees, using a simple deterministic approach.

http://www2.geog.ucl.ac.uk/~mdisney/fieldwork/misc/tree_measurement.pdf

http://www.isprs.org/proceedings/XXXVIII/part8/pdf/W07P04_20100307064753.pdf

http://www.fao.org/docrep/018/i3058e/i3058e.pdf

You can as well get a fairly accurate  estimate of number of trees on a high resolution satellite image by using segmentation techniques. By using multiresolution segmentation techniques for instance a software like ecognition and using template matching algorithm or others, you can count the number of trees on the satellite imagery in a savannah region.

It is not as simple as previous answers suggest. We found that the errors by using segmentation on high resolution imagery as well as high density lidar are large, even for the simplest of trees by crown shape (conifers) and extremely homogeneous forests (single conifer species plantation forest). Multi-layered equatorial forest or multi-stemmed-tree,multi-species savanna around the tropic of Capricorn would seem even more of a challenge. 

My answer to Marcus is therefore no. I have not come across a general statistical relationship between tree crown/canopy cover and number of trunks/stems per unit area. from first principle is unlikely that such relationship could exist over a larger area.

For what purpose is tree density information required? Maybe there is another way/method to answer your research question.

I have indeed some ground measurements and allometric relations. But my question is, as Neils has noted, if there is some general statistical relationship between tree crown/canopy cover.

I have a high resolution land cover map. This can be considered a brief indicator of canopy cover.

Field observations provide tables with densities and basal area, but I´ve never seen any table providing both densities and crown cover.

The final purpose is to provide a brief estimate of the woody biomass contained in tree/wood land cover cells to map the potential biomass supply at a landscape-regional level.

My idea is to resample the initial 5m cellsize LULC map to one ha, with four-five woody classes based on their density (grass savanna, shrub savanna, savanna woodland, open forest...).

Hi Marcos,

Tree cover is not a good indicator of density particularly in the savanna areas. You can have 100% cover from 50cm high seedlings or 5m high trees. Density may vary with soil and particularly moisture condition. With increased tree density you will also find a change in tree form - this may affect your biomass estimate substantially. Unless you have already found the ref below you might want to follow up on the work of Prof. Nico Smit in South Africa, particularly about his work on the BECVOL program. (Smit, N. (2014). BECVOL 3: an expansion of the aboveground biomass quantification model for trees and shrubs to include the wood component. African Journal of Range & Forage Science, 31(2), 179–186. http://doi.org/10.2989/10220119.2013.866161).

I agree with Hein that you are unlikely to find a simple relationship. Do you have other info that might help you with a model?

Dear Friedrich,

Unfortunately, I don´t much more info. Only a few measurements in sampling plots, but it is a small sample. Further, the gps points I recorded on the tree location, and the pixels classified as trees in the land use map doesn´t match, which difficults its association.

The only work I´ve reviewed in which there is some data on both tree density and canopy cover is "Intermediate tree cover can maximize groundwater recharge in the seasonally dry tropics" from Ilstedt et al. 2016, taking in Burkina Faso, not far from my research region in northern Ghana. But I also know I shall be cautious to extrapolate this relationship to my region. Even though, accuracy is not the most important because my main aim is to conduct landscape level assessment of biomass provision (agricultural and forest land use types). But I am aware of your observations, and conscious of all the uncertainties I´m facing.

I´ll dig into BECVOL, if I can get it. I don´t have access to that publication from my institution.

Thanks,

Marcos