We've had very good success with random forests (randomForest package in R) when you are trying to fit a dependent variable (presence/absence or abundance, for example) to many independent/environmental variables. It works quite well with spatial modeling of habitat suitability.

These are two different questions. Why are you interested in canopy shadow? Do you want estimates of light penetration to the forest floor? Are you using the canopy to estimate total canopy biomass? You can answer these questions using difference pieces of equipment. If you are interested in estimating canopy biomass, you can use instruments like the Licor LAI2000 or take destructive measurements. If you are interested in knowing how much light is reaching the understory, you should use some PAR sensors.

the goal of question is not clear. do you take the canopy shadow equal to spatial distribution of plant species? as Mr Naguy-Robertson wrote, if you interest on canopy shadow biology then may be this reference help you to clear your purpose : [http://onlinelibrary.wiley.com/doi/10.1046/j.1365-3040.2001.00694.x/full] but if you want to find spatial distribution then depend to the scale of your case study region, there are many methodology from using plot in field work to using satellite images processing may be applied.

Thanks of all your reply. I don't know if It suitable to talk about Canopy shadow in woodland. I'm specially interesting to land cover study in the first time and spatial distribution of plant species in the second time. My field study is Mozogo Gokoro National park in Cameroon with 1723 ha

then if you want to study on land cover change in such big area. at first you must be test suitability of different satellite images for your purpose, i suggest "Landsat TM" and "IRS Awifs" . then i suggest you to check USGS archive (through Earth explorer) for past and present time series Landsat images in your region. then produce time series land cover map for change detection. i suggest land sat because it is available free and rectified (ready for classification).

Thanks for your help