This article is a very interesting read on nature documentaries.
It seems that yes, generally, a nature documentary does at least some scripting first. For example, on the BBC's documentary series 'The Hunt':
The Hunt took three years to film; the beautiful sequence of a blue whale eating krill took two years. The first year, the water was too murky for any of the footage to be usable.
From this source on the making of Planet Earth and Planet Earth II:
Anderson, a part-time mountaineer, had been one of the crew who, 10 years ago in the mountains of Pakistan, captured on video a snow leopard hunting — the first ever footage of its kind — for the first Planet Earth. For the sequel, he was expected to pull it off again, this time tracking down those rarest of central Asian predators in the mountains of Ladakh in India.
Also, for less big productions (from the first article again):
Small budgets and limited time mean that filmmakers use captive animals for hunts, chum waters to send sharks into feeding frenzies, and otherwise sensationalize footage, giving audiences a false impression of animal behavior.
I also once watched 'Wilde dieren in Wildlands', a making of a nature documentary made by a Dutch Zoo. Here, they had a very clear idea of what they wanted to film: leaf-cutter ants transporting leafs and naked mole-rats huddling together for warmth amongst others. It was interesting to see how the cameramen would go about staging such shots. Again, this confirms that the general idea about which shots to include comes first, and the actual trying to film those shots latesr.