Safety and ease. Landing in the ocean they don't have to boost back which saves them fuel (so it's essentially trivial to just add a test onto an existing launc

Safety and ease. Landing in the ocean they don't have to boost back which saves them fuel (so it's essentially trivial to just add a test onto an existing launch at no significant payload penalty), plus they get to run through most of the flight profile (especially the most important parts) without having any potential of hitting anything on land and with the expectation that the vehicle is going to be destroyed anyway. They have several (at least 4) more flights within the next year with sufficient payload margin to do more of these tests so they can spend time gaining confidence (at almost zero cost to SpaceX) and exploring the boost back trajectory (coming closer to land) before actually attempting a landing.