Have you built that map yet? Really, concentrate on smaller things first, and don't already plan your "Bigger than GTA V" game without even having a working game prototype and a modest map size built yet.
Chances are good that you underestimate the work needed to build all of that by several orders of magnitude.
You ever worked with volumetric clouds? What is your plan to make your map mesh seamlessly into the horizon if people can see it from high up? I could go on and on... but really, it doesn't matter. Don't worry about those things until you have the foundations in place.
"But I worry we go into the wrong direction and have to throw away parts of our work!" - Yes, we hear that a lot on these forums. Don't worry. You will throw away way more than just parts... its part of a normal learning process which all of you will have to go through. Which all of US go through on a daily basis. Getting too attached to your work is quite a big problem.
Chances are good your first few 3D Models suck. They might even look good in the modelling tool. Once imported you might find out they do not work for your game, period. Throw them away. Don't try to fix things that take ages to fix. Mark the time down as part of the learning process, start over, do it better this time.
Your first 1 KM^2 most probably will never get finished. You might still try to hand populate it with assets and find out that it literally takes you days to only arrange all the assets that make up a single house, its interiors and surroundings. Throw the map away. Either rework the concepts of your game (do you NEED to be able to enter every house?), or come up with more clever ideas (have pre assembled asset-bundles you can just place in your map in a single drag and drop operation... yes, every house made from the same asset bundle now has the same interior... does that matter? Write a procedural routine that places assets for you).
Your first few stabs at the actual game logic will suck. Nobody gets their game logic right the first try, and there is a good chance yours turn out to be non-salvageable. Make sure you don't spend to much time on it, if it does not work no matter what you do, throw it away. Come up with a better game design, and try again.
Don't plan for a great future that will most probably never come... try to do what you can do here and now, and learn from it. You cannot climb the mount everest with zero mountain climbing expierience and lacking equipment. Get expierience first on that small rock wall in your neighbour hood, then try the steeper one some miles away, and so on.
Last thing: Don't worry about which engine will allow you to build X when you haven't built ANYTHING yet. Just start out in any of them. Pick one.
As for MY recommendations, scratch CryEngine off the list. The runtime engine might producing quite good graphics, I am not sure about the performance though (given the games made with it tend to be used as testcases for now hardware BECAUSE they are extremly demanding).
The tools provided with the stock CryEngine where crap 2 years ago... haven't heard that they got improved in the meantime. Given that its mostly used by bigger studios that have the money to pay programmers to write better tools for the engine for their own projects, even though the engine is available for free makes me believe the tools are still as dated as they were back then.
You could give Lumberyard, amazons fork of CryEngine a spin. Yet I would still be catious as I am not sure if amazone really has improved those tools yet.
Unreal Engine 4 and Unity 5 are both good recommendations. I personally would go with Unity for total beginners as it just has better documentation and an easier to grasp interface, but then UE4 is close even in my opinion, and some people found it easier to get into than Unity.
Keep in mind that Unity 5 has some quite dated stock tools, terrain system for example is not that high-end... you can upgrade that by purchasing third party assets from the store at actually quite good prices, which often are better than what Unreal Engine 4 gives you. But of course, it costs something.
Unreal Engine 4 comes with way better stock tools, though I found it throws too much of it at you at the same time, making the editor quite cluttered and hard to grok.
Really, at your stage, you cannot go wrong with either. And changing engines later is way less of a hassle than you might think.
Dude its a golden words, yes our first ever test map will be blank, later we modiffy it, but we want to put some mechanics that forced us to think about current games, yes from beggining we only concentrate on basics and prototypes, also we decided after we have very very very pre-alfa of game to show of some gameplay and ask here or where big community is do we should continue our development or just stop it. the main reason for big map is for planes, and dynamic gameplay. Anyways thanks for smart ideas.