A scene is an intuitive unit of storytelling that's difficult to describe, but fairly easy to recognize where one ends and the next begins.  When the story needs to skip boring bits like narratively uninteresting travel, or needs to switch viewpoint characters, or needs an abrupt change in setting, the current scene ends and a new one begins in the new location, at the new time, as the new character.  If a scene changes mid-chapter there's a typographical convention like a blank line or a centered trio of asterisks to inform the reader of the break in continuity.  

Inside a scene the reader sticks with a given character who's trying to get something accomplished.  The scene shows (not tells) this attempt to the reader, and the seasoned reader knows this task won't be easy, otherwise the author wouldn't have made a scene out of it to begin with.  So that's what a scene is, and you already know what a planning tree is, but what's the connection?  I'm getting to that, but first an example will come in handy. 

Let's say the two main characters need a pick-up truck mid-story, and one tells the other she'll just borrow her ex-boyfriend's.  Because this ex has never been mentioned in the story it's unlikely he'll be a major character, so the reader might assume that the truck-owning ex is just the author dodging the problem of getting the large, mysterious artifact to where the plot requires.  But then the next scene opens with the character arriving at her ex-boyfriend's place.  Immediately the reader knows there's going to be a complication here, because if there wasn't, the author wouldn't be showing the mundane detail of borrowing a truck in the first place.

How does this resemble a planning tree for that character?  A node marked BORROW TRUCK will initially fail for some reason.  Several alternatives will crop up, like APOLOGIZE FOR PAST EVENTS, ACCUSE OF RECIPROCAL WRONGDOING, and OFFER FAVOR IN TRADE.  Some of those may also fail at first, prompting further negotiations and whatnot, until finally BORROW TRUCK either succeeds or exhausts all tactics.  Then the character leaves her ex's place with or without a truck, a favor owed, and a nosy tagalong ex-boyfriend or other plot complication, and the scene ends.

Both a scene and a tree's node (with its sub-nodes) have a single character, a goal, some obstacles with corresponding workarounds, and a final outcome.  Of course not just any action in the tree can drive a scene.  All of a storyteller's reasons for what can constitute a scene apply, though keeping in mind games with a sharp gameplay/story divide may contain "scenes" of pure gameplay.  Tree-wise, the node will lie neither too high or too low in the tree.  And a scene could even be tied to a node in the author-character's tree rather than an in-world character's.   

I'm unsure if this connection is strong from a coding perspective, or if code would work this way in practice (since it seems to require a bit of prescience regarding which nodes will be difficult to accomplish), or if it's just a handy analogy for AI programmers to understand stories.  Either way, it seems a useful idea at least for certain parts of prose.