To create something like rainforests. I assume its not that easy. Much can go wrong. It is probably not just add seeds and water.
So when in doubt ?assume a machine learning network model.
I wonder if you can. Inspired by the old GIMP Heal Section function. Where you could magically replace a marked area with something that looked like the surrounding background.
So inspired by this it would be possible for a machine learning model patch = model(surrounding area). To replace the patch in the image or photo with a prediction of the patch area.
That is. The model would have learned from many examples what plants a rainforest would have at each position in the photos.
Then this prediction would be some inspirational data for a manually grown and watered successful rainforest.