[WalterBright]

> you can eliminate pipe leaks easily

Famous last words. A better design is to assume leaks occur, and design to minimize the damage. (This is how airliners are designed.)

> with automatic block valves

That's compartmentalization, not elimination.

Another method is to use two smaller pipes instead of one larger one - it reduces the scale of any leak.

[benzofuran]

With proper monitoring and a well maintained internal corrosion control and external threat control systems you can stretch out the timeline for non "acts of god" (ie natural disasters, terrorist attacks, etc) to greater than the life of the structure. The methods for internal and external control are well understood and practiced to varying degrees of effectiveness.

Elimination of the hazards is impractical, so minimization of risk exposure and frequency is the best that can practically be done.

All the hubbub about stopping new pipeline construction means that more incident prone methods are used - be it trucking, rail transportation, or use of 50+ year old pipelines. The new lines are better engineered, understood, and have more stringent operational controls in place. I've worked on pipelines from the 1940's that are still safely operated, and lines from the 2010's that have already self excavated. These things can be controlled with the right engineering, organizational, and governmental oversight.

[Retric]

Compartmentalizing risk as acts of god is a rather pointless exercise. If you can’t eliminate risk you need to be intellectually honest and say why that risk is acceptable. The Midwest for example gets large earthquakes in seemingly random locations, that’s a known issue and needs to be accounted for in a risk assessment.

As to other transport methods being less safe that also comes down to building safe systems. Car gas tanks for example are designed to survive crashes without spilling their contents. You can also build tanker trucks/trains to appropriate standards.

[masonic]

  use two smaller pipes instead of one larger one - it reduces the scale of any leak.

... while doubling the exposure to failure in the first place.

[benzofuran]

This actually ends up increasing your likelihood of failure, as smaller parallel lines mean higher operational costs for smart pigs and other integrity assessment tools making internal issues harder to detect before failure, and increased risk for external failure due to cathodic protection system failures due to interference and shielding.

[WalterBright]

Of course it doubles the risk of failure, but each failure does half the damage.

[improbable22]

Which sounds like a terrible trade-off, harm isn't linear here. Two leaks, each of 1000 barrels, will badly damage two different valleys. One leak of 2000 barrels would be better.

And cost isn't linear either, two half-capacity pipes will probably be almost twice the price. You could spend that on more steel for the large pipe instead.

[MertsA]

Er... I don't think you've completely taken into account what you're proposing. Due to the square-cube law, splitting a pipeline into two smaller pipelines means that in order to have the same total cross-sectional area you would need sqrt(2) times as much surface area. Not only is the surface area that is subject to corrosion increased, I would imagine the pipe wall thickness would probably decrease so corrosion would compromise the integrity of the pipeline even sooner. It's also double the number of welds that need to be made, double the length of pipeline to maintain, and as to the notion that it would do half as much damage, I doubt this claim. Why would it actually substantially decrease the damage to the environment? It's not like pipeline leaks completely sever the pipeline in half and just start inundating the surrounding area with a flood of oil. It's usually a crack or pinhole leak, so the smaller cross-sectional area of the pipe probably won't do much good in terms of decreasing the volume of oil leaking out.

[WalterBright]

Actually, I'm not so sure about doubling the risk. Failures may appear random, but they aren't. A failure may indicate a particular defect that can be corrected in the rest of the system before another one fails.