Large scale aquifer removal technique.
This describes the techniques employed in the aquifer removal for the Dwarf Fortress Masters Challenge 2 - Muckgrotto. This was a two-layer aquifer that dropped a layer for half the map (aquifer for half of one layer, all of the layer below, and the other half of the layer below that). The chosen project was to eliminate the majority of the aquifer on the map by creating a perimeter magma pierce near the outside edge of the map and another near the center of the map and eliminating all aquifer tiles between the two pierces by channeling and then draining off the resulting water, repeating for each of the aquifer layers, and the collapsing most of the top layer of the map.
Because it was a timed challenge (8 years) an efficient channeling (or equivalent) system needed to be developed. Fortunately there was plenty of time to experiment. The effort started with the complete excavation of the layer directly above the aquifer so that every water-producing tile could be removed from above, but the process could not strand dwarves. Access was from the center of the map, so dwarves would have to work from the outside toward the center.
Method 1: Strip Channeling
Working from the outer perimeter of the map inward, 1-tile wide channels were designated. This created several problems:
1) Assuming the ideal case of one miner per designated tile, dwarves would on average walk the width of the map for each tile to excavate. Some more, others less, but they would walk the distance out and back. Since miners couldn't be stranded, the next strip was difficult to designate until the first was completed. 2) Because dwarves prefer to channel from the west, horizontal designations would routinely cause dwarves to cancel channeling because another dwarf was standing on the tile they were channeling. They would choose another tile, walk to it, and then someone else would come back and channel the first tile. This would increase the amount of walking even more. 3) Some dwarves might be down in the bowels of the fortress and pick up the channeling job, delaying its completion due to the long walk. Again, because dwarves can't be stranded, dwarves that had completed their jobs would wander off. Moving the meeting area close to the channeling area helps, but it's still a very slow process. This could be sped up by removing the designation and redesignating, causing close-by dwarves to pick up the job (the closest dwarf takes the job) but that's a massive burden on the player.
This process would never have allowed the task to be completed in time.
Method 2: Checkerboard Channeling
Instead of designating strips, designating a checkboard pattern over a large area (ideally using a macro), and then following up with strip channeling addresses a number of the problems above:
1) Dwarves can take jobs in two dimensions, massively reducing the amount of walking relative to mining. Dwarves appear to work in vertical strips 16 tiles wide - they'll move from top to bottom of the designated area and 16 tiles wide on a pre-determined grid. Reasonably efficient, and easy to work with. 2) Because of the checkerboard, dwarves never cancel jobs due to another dwarf on the tile - either in the checkerboard effort or the strip channel effort. This further speeds up the effort. 3) The checkerboard allows access to the entire map at all times, so theres no risk of stranding in the first pass, although the strip channeling needs to work as it did in the first method.
The downside is that the player still needs to designate all of those strips at the end, and the dwarves still do a huge amount of walking. It's much faster than 1), but not fast enough.
Method 3: Staircases
Aquifer tiles that have staircases dug into them do not produce water. This method requires 2 passes to eliminate water production, but 3 passes if you want to fully collapse the layer down. For multiple layers, one pass can be eliminated by good planning. The three passes are: down staircase from the dry layer above, up (or up/down) staircase from the aquifer layer, remove up staircase from the aquifer layer. Advantages:
1) The entire layer can be designated as one block for each pass. This provides the best digging/walking ratio, even with the extra passes. 2) Minimal effort for the player. No need to designate all of those strips. 3) If planned well, the 3rd pass of removing the up staircases can be done while the layer is draining minimizing the amount of additional time needed. By using up/down staircases on multi-layer aquifers, the first down staircase pass can be eliminated after the first layer and the free down effort provides a very rapid way to drain the layer. 4) There is no possibility of stranding dwarves.
1) The first set up up staircases dug will be very slow as miners won't have swimming skill and will cause an incredible number of job cancellations. Fortunately, ever tile allows the dwarves to escape so there should be no drownings, but it'll start off slow. Once they reach 'Swimmer' that'll end and they'll mine much faster. This can be helped a bit by designating a checkerboard pattern of up staircases. Dwarves usually won't jump down into the new staircase to dig the one next to it due to the pattern, and when you designate the entire layer after the first half of the checkerboard is done, they won't jump down because the tile will be 7/7 water. But it is one extra walk across the map for everyone. It might be faster if you don't expect your miners to become swimmers, but if you do just get that over with and designate the entire area as one block. 2) The removal of up staircases needs to be well choreographed. 4/7 or deeper water will cause cancellations, so start draining one part of the map after eliminating water sources so that the water is shallow enough for miners to work after other areas are prepared. If collapsing is the goal, you'll need to collapse each aquifer layer in turn since eliminating the up staircases doesn't eliminate the floor of the layer above. If done carelessly it's easy to collapse the entire region on your miners, likely killing them all.
The map this was tested in was 5x5 embark tiles and almost completely excavated. 45,000 tiles per layer and 4 layers were excavated (180,000 tiles total). Half of the tiles were aquifer tiles (90,000) which were removed with the methods above. The rest were dry soil tiles removed either by standard digging (top layer) or ramps (dry portions of aquifer layers). Method 1 is estimated would have taken approximately 2 years for 40-ish legendary miners to excavate one aquifer layer. Method 2 cuts that time almost in half to 4-5 seasons. Method 3 cuts it down to approximately 3-4 seasons to do all 3 passes, but 3 or a bit less if it had been employed from the outset and could have allowed for the elimination of one pass due to up/down staircases and the very fast draining into the layer below.