One thing to note, in this post I will try to avoid suggesting things that would require a drastic change in game mechanics or physics and keep the suggestions as close to the current state of affairs as possible. Namely, if not stated otherwise, I suppose all landscape static and permanent (biter bases, trees and ore deposits do not count as landscape) and flat.
Also, if possible and does not contradict the abovementioned, I will try to stick to real-world examples.
1. The mining.
I totally agree with the suggestion made in https://forums.factorio.com/forum/vie ... p?f=9&t=10 that underground mining is a way to go. So, I would suggest trimming down current ore fields and/or making large ones extremely uncommon. From the "lore" point of view these deposits would then represent close-to-surface veins of metal, that are easily accessible through straight open-cast mining (essentially digging a pit in the landscape). The present mines, either electric or coal-powered, would represent the infrastructure supporting these mines. The point then is that a player must either constantly expand, searching for more and more ores, or deeg deeper.
Next come the subterranean deposits. One of them may be discovered in a number of indirect ways or a direct one - building a mine and bumping into a deposit. The latter though should be comparable with building a mine at a random spot at normal settings in present version (0.8.8) and hitting ore. The indirect ways are the following. For both metals (copper and iron) some sort of a field scanner should be the way. We may have a stationary one, something radar-like in operation, and a less powerful handheld unit. The sensitivity may or may not be later upgraded with research, but either way, a mine plopped down in the middle of the resultant blob on the map will hit iron 100% (supposing the ore field shapes stay unchanged). The mine building itself should have a much wider reach than an electrical mine now has, but a smaller size of the actual building. Also the mining rate should be somewhat higher, because it represents an entire mineshaft network. Also because of this the mine would take some not inconsiderable time to reach maximum mining rate, representing the time required to build the mineshafts. However (both from common sense and game balance points of view) an overlap of those areas-of-effect should decrease the efficiency of both overlapping mines. Ideally, these mines would have to be connected to a "mini-railway" type of logistic system, that would use chestlike objects for transport (see below). For coal a different type of scanner would have to be used, a chemical scanner. This comes only as a handheld and even then has a lower range than a field scanner. To compensate, coal deposites would be on average noticeably bigger than metal ones (AFAIK, this is true as of real life, and the different range of scanners is also easily explained). Another method, that will map out a large portion of the subterrain, would be a seismic probe. It would be a building that ("lore") sends down explosive charges (maybe rockets) and several separate buildings that record the resultant echoes. The more shots it manages to pull off, the better, with the following drawbacks: only one such device may be active at a given time, time betweem bombings must be rather large (on the order of minutes) and, finally, every last biter in the affected area would totally LOVE to know, what has made such a boom.
Finally, the third "tier" of mines: the deep core mine. It is a large structure (really large, on the order of tens of tiles in width and length), it uses an insane amount of energy ("Lore": nitrogen liquification for coolant), it has an area-of-effect similar to a roboport (or bigger - an issue of balance) that cannot have any "capital" structures in it (excluding rail network, belt network, inserters and chests; "lore": local quakes from so deep a structure) and it produces a largeish amount of pollution. It does not produce any coal or copper, but rapidly mines iron. It also should not be infinite, although the amount should be great. And yes, this thing is a primary attack target. On the point of balance: copper deposits on average would have to be bigger, both the surface and the subterranean ones, in order to compensate for their absence in a deep core mine.
2. The ores.
I totally think that the chemical table has enough elements without having to resort to fiction. Also I am totally opposed to "minecrafting" Factorio. While I do really enjoy Minecraft and its gameplay, it is rather different from the way Factorio does things. Based on these, I propose the following additions to the current list:
- Uranium. Goes without saying, if we are to have nuclear power. Detectable by chem sensor, found underground or very rarely as a single tile above ground. Its processing should of course be a multistep process.
- Bauxite, aka ore of aluminium. For simplicity's sake, let it be simply smeltable, but only in an electric furnace (IRL its refining requires elecrtrolysis). It should be used instead of iron plates primarily in flyers and not-so-armoured objects. It should appear at the surface and in subterranean levels, chem scanner detectable. Also, a deep core mine should produce aluminium at about half the rate of iron (real world numbers, no relation to balance).
- Gold. Appears uncommonly on the surface and in the subterranean (field sensor). Smeltable. Not required as is, but any assembly machine (not the player!) that uses Electronic Circuits or Advanced Circuits as a part gets a bonus, if supplied with gold, representing a better quality of electrical contacts: with present concepts, either a speed increase, or a productivity increase. Also may be extracted from bodies of water by a special building at an abismal rate.
3. The caves and underground in general.
The bulk of the underground, IMO, should be more or less packed solid. However, on occasions there might be a cave connected to the surface and thus visible. Also, while using the second-tier mines (see above) the player might stumble across an unaccessible cave system in the deep. The first purpose of these would be to show exact positions of ore deposits. However, another use presents itself. I propose to make the biter bases interconnected with subterranean tunnels. A biter base may (or may not) be connected to another one underground. The passageway would most likely be guarded, but no biters would actually spawn in them. A biter base with such a tunnel would have a cave in its midst, that the biters would use if alerted to a player's presence. The spawners should have an underground part that may be attacked, just like the aboveground part.
Also, while establishing new bases the biter will dig a tunnel from the parent base. This may be detected by seismic detection arrays and probably attacked from above by using the depth charges (see above). If terrain modification is implemented, maybe it may also collapse the tunnel, forceing the biters to restart or even abandon the attempt.
Also, the second-tier mines should create artificial caves under them, accessible through the mine building. These should be lit and straight. In case of them crossing a cave accessible to biters, it may become a backdoor into base defences.
Second and third tier mines should not be buildable underground; a first-tier mine build underground mines ore in the wall instead of under it (maybe a model change). No electric poles can be placed, instead, cable must be run to places needing power. Cable runs "on the ceiling" thus obstructing no constructions and creates a 1 wide powered area (i.e. directly under the cable). Roboports can not be placed (or require special research). Seismic detectors work better, depth charge deployer can not be placed. Solar panels can be placed, but produce no power (naturally). An underground pool of water, on the other hand, is entirely possible. The place that corresponds to a deep core mine above is inaccessible below, surrounded by walls. If those are broken, the mine disassembles, leaving only half of the materials ("lore": the rest being molten by extreme heat of the core). The same happens if a wall is destroyed by biters.
And of course the player himself can mine walls.
4. High-throughput short-range logistics.
From my point of view, Factorio at present lacks a logistics system that would fill a role in-between the rails and the belts. This addition has not been considered balance-wise, it is more of "it would be cool if" concept. This system should, on the one hand, be as compact as the belts and on the other be able to benefit from inserter stack bonus. A concept of some sort of minecarts comes into consideration almost immediately. A note on the word "minecart" - this is only a shorthand version for "mobile pathed inventory". It may be better represented in the game as a line of metal bowls moving suspended under a cable. It should most definitely require power. It should have the same dimensions as a conveyor belt.
5. Building construction.
Also a topic that has been mentioned quite a few times. From my point of view, the player should be able to place a rather small amount of buildings personally. These most definitely include belts, inserters, walls, small and medium electric poles and, generally speaking, all the buildings that are one tile wide and long. All the others should have to be constructed on site. For this, three ways of construction are possible.
The first is the most manual: the player places a building designation, and, providing he has all the items in his inventory, the construction begins. Construction time should be similar to the present one. If a building is dismantled, the components return to the inventory. This approach has a major drawback from the point of game design: it requires part of the buildings to have an object, and the other half not to have. A solution to it may be to create "construction packages" that are created instantly from required resources, but take time on deployment.
The second is the one already planned and discussed: construction robots and blueprints. Instead of buildings either components or "construction packages" as mentioned above should be delivered to the marked place and then assembled.
The third one closely relates to a later suggestion. Instead of building something in the field, a structure is assembled in a hangar on the cargo plate of a vehicle. Then the player drives it to a spot where he wants the building to stand and deploys it. Some buildings may be stacked (rails and large electric poles a prime example), others shouldn't (factories, steam engines, other large buildings). The player must have a way to control several vehicles at once, the simpliest being to force them all to follow a preset path. At the end of the path the player would have to place the structures manually. Another way (possibly a later research) would be to create fully automatic vehicle trains, marking a way with some sort of a guideline. And these vehicles should totally be transportable by rail by attaching them to a locomotive. Perhaps excessive, but a rail-only variation may also be possible, placing rails along a preset path (has been already brought up, I believe).
Large structures, namely the deep core mine I've mentioned above, would have to be constructed out of several component parts.
6. Building the vehicles.
This suggestion is not entirely my idea, it has been most certainly brought up, however, I would like to elaborate on it. A new type of building should be introduced, the hangar. It would not be a building as is, but rather a designation (depicted by a paved tile, for example). When the hangar interface is accessed, the player can select a vehicle from a list. What is and is not in there depends on current research and the size of the hangar. If a rail-based vehicle is selected, the hangar automatically adds rails to itself (thus its recipe should contain steel). Then a number of inserters would have to be placed around the hangar and supplied with the required materials. As the materials are collected, there should be a sprite of a partially constructed vehicle in the hangar. Finally, after construction is complete, it has to be moved out. Perhaps it should be done by the player, perhaps by a special extension of the hangar, or both as alternatives. And by the way, the train totally should be a bit more expensive!
I thought my postings are too long, but... 
tell this? 
Thats my problem IRL also.