Right now, natural gas is $2/MMBtu, equivalent to $0.286/gal diesel.
The confined space issue isn't a big deal for trains as well.
Conversion to NG of a conventional engine does mainly affect peak power, as NG is more susceptible to Pre-ignition, as it, doesn't have the evaporative charge-cooling effect as gasoline or diesel and must run excessive ignition timing, lower compression, or other things. For this discussion, if you use a turbine, no issues.
The energy density is what AJ was referring to, energy density per unit VOLUME. Diesel is 38MJ/L, CNG is about 1/4 of that. However, by MASS it's actually higher. While volumetric density matters significantly for on-road vehicles, for locomotives, much less so, especially with the ease that a LNG tank could be coupled directly behind the locomotive as a "tender"--and then it's all about cost. A 6000hp locomotive would burn 2000gal per hour or more of diesel at full throttle, and tow let's say 5000 ton at 60mph. That's (at $4/gal diesel) $0.026/ton-mile. Natural gas would reduce the fuel cost to $0.002.
US freight rail costs about $0.03/ton-mile... Assuming an average of 25% throttle for conservatism, the fuel cost is 25% of that.
US freight rail moved about 1.7 trillion ton-miles last year, which likely consumed about 2.8 billion gallons of diesel ($12 billion dollars), and required 50 million barrels of oil (at least...I'm being really conservative). Switching to NG would save $8-10 billion in fuel per year. There are 24000 locomotives in the US, which means the current fuel cost alone is about $500,000 each per year. Hmm... Conversion costs are starting to look attractive.