Cemetery Heights got me to thinking. While I replied in my comments that site mechanics force the park-and-ride approach in this instance, the larger question that he asks--when are they valuable and useful--still must be addressed. The direct criticism that "[c]oncentrated parking lot stations with express service, such as Ronkonkoma and Metropark, can get relatively high ridership, and this can mislead planners into thinking this can work throughout the system" is valid and relevant--and there is too great a tendency among transportation planners to convert every commuter rail station into park-and-rides--yet I would suggest that there are places in the metropolitan network where they are useful.
My analogy with Cornwells Heights, the largest park-and-ride station on SEPTA's network, was no accident. Cornwells Heights is a large 1,600 car lot, with 1,104 daily boardings, is only accessible from I-95 and Woodhaven Road (itself a highway), and sits at the city border. Philadelphia's Planning Commission's new city plan and zoning code both codify the conversion of Eastwick station into a large park-and-ride. Like Cornwells Heights, Eastwick sits with excellent highway access (I-95 via Bartram Avenue) near the city edge. Eastwick also has a major knock against upzoning due to sitting smack dab between the airport and the largest active industrial area in the city, one of the reasons that the far end of Southwest Philadelphia remains undeveloped, even though it also sits next to one of the crown jewels of the region, Heinz Wildlife Refuge. Cemetery Heights, similarly, sits on I-76 just west of the city limits, has severe access limitations, and sits on a site which, due to its grade and its surroundings (it's halfway up a hill, bounded on two sides by cemeteries, and on the third by a highway and an active and busy freight line), makes the idea of upzoning questionable.
I would also like to suggest that, while TOD and upzoning may, in the main, be the most useful and desirable land-use plan surrounding a station (most stations are, or should be, after all, suburban centers), the twin and conflicting desire to maximize transit access while acting under the constraint that access can only be guaranteed where the line actually is, do necessitate purpose-built park-and-ride stations in some locations.
But then what, exactly, are the locations? An exurban park-and-ride terminus does little more than justify further exurban sprawl around the transit connection, just like how new highway exits at the exurban fringe justify further exurban sprawl (see Strong Towns' Rogers Interchange case study here), in a certain sense. Of course, market upheavals and the death of sprawl may render that argument moot...but the fact remains that a commuter rail system, like Rail Runner, that is designed as a park-and-ride centric system does not promote densification, as the most valuable land closest to the station is purposefully given over to parking, thereby (ironically enough) using the cars' own competition to enforce autocentrism. Yet at the same time the metropolitan framework we have now, by being so excessively suburban in scope, requires the use of park-and-rides to capture ridership outside of the rail line's own natural catchment. So, while we should not place park-and-rides at all suburban train stations--nor even at the majority of them--there are places were we can put them where they will significantly impact ridership patterns and, hopefully, driving patterns.
My suggestion is: tie park-and-rides into highway-rail junctions. Highways are already the most important arterials in the suburban system, are often crowded unto gridlock close to the key metropolitan employment centers, and often decrease local land values to a point where they make upzoning and TOD unfeasible. But it shouldn't be used at all such junctions: would one want to build a park-and ride at every time a parallel highway and rail line cross? No: it would bring about the same effect of park-and-ride exclusivity, namely, co-opting cars' own competition into reinforcing autocentrism. But at just the right places it can be used to expand a rail line's catchment far beyond pedestrian and bicycling radii.
Every urban system has its own idiosyncrasies, of course: odd places where seemingly natural park-and-rides aren't, or a seemingly odd place for a park-and-ride makes a great deal of sense on the ground, but there does seem to be a rule of thumb related to park-and-ride placement, and this rule of thumb grows out of the urban highway network. Urban highway networks are characterized by two types or roads: axials or intercity highways (i.e. I-95 and I-76 in Pennsylvania are both axial roads), and radials, roads which loop around the city, normally at a certain average distance. Philadelphia has half-radials in the Blue Route/Pennsylvania Turnpike (I-476 south of Plymouth Meeting and I-276 east) and U.S. 202; the Roosevelt Expressway and Betsy Ross Bridge represent the only built sections of a very incomplete inner radial--one about as far from the city core as Paris' Boulevard Périphérique, that city's inmost expressway.
Park-and-rides should be placed on axials just before the areas of maximum congestion--that is, right before they enter the metropole's most urban sector. (Paris continues to be urban well outside the Périphérique, as an example; the banlieues bordering the city proper are just as urban.) Congestion is a strong deterrent against further driving, and so placing park-and-rides there are a natural market tap. Nor do they justify further sprawl because these places are already so deep into the metropole--the urbanized area--that patrons of these places will have come from a wide variety of suburbs, of which both transit-accessible suburbs and super-autocentric exurbs are both minorities. In the Philadelphia area, Cornwells Heights, Eastwick, and Cemetery Heights are just these points of maximum access.
Junctions where commuter rail lines, with their naturally axial nature, cross middle radials (that is, further into the suburbs than the Boulevard Périphérique, but not primarily within the exurban fringe), are likewise natural park-and-ride locations. Just like with axials, the key is trying to extend catchment into the existing development pattern but not going so far as to justify it on the urban fringe. If there's an axial-radial junction nearby, better still. Fort Washington is an excellent example of this: a park-and-ride just off the junction between the Turnpike (I-276) and the Fort Washington Expressway (PA 309)*.
Now, by saying highway access is necessary for the consideration of park-and-rides, we are also saying where they shouldn't be built. They shouldn't be built in the middle of older towns, and they definitely shouldn't be built by displacing the station from the town center to the fringes just so a park-and-ride can be built by the bypass. Bypasses aren't very good highways for park-and-rides anyway; they are usually both minor and disconnected. Places like Ambler, Jenkintown, Lansdale, and Warminster are poor locations for park-and-rides, due to their either (a) being in the center of a built-up area or (b) having poor highway access. The single best way to ensure ridership is to place stations in such a way to best capture it; park-and-rides, like TOD and upzoning, is just a tool in the transportation planner's toolkit to maximize mode capture. Like any tool, it can be used or misused; and like any tool involving built form in the past half-century, it has been misused far more often than used.
* This is not to justify Fort Washington's current built form, however. Upzoning along Bethlehem Pike and Pennsylvania Avenue, and extending a TOD heart from the village peak land-value intersection, that of Bethlehem and Pennsylvania, to the station proper would be an investment complementary to investing in park-and-ride facilities here.