Point Nemo of Interstates?

[Zzonkmiles]

Just a question that's always been bugging me.

Point Nemo is the single place in the ocean that is the farthest away from land on all sides.  There is a place in the southeastern Pacific Ocean that is like 1500 miles north of Antarctica, 1500 miles west of Chile, 1500 miles south of Easter Island and 1500 miles east of some other islands.

So my question is, which point in the US (lower 48) is the farthest away from the Interstate Highway System?  I'm guessing this would be somewhere in Nevada or maybe south-central Oregon?

[getemngo]

There's already a thread about this.
Also, have an Atlantic Cities article about different ways to determine it.

But there are still unanswered questions. What's the farthest point on an Interstate from another Interstate?

Or, what's the farthest point from an Interstate within each state?

[Brandon]

But there are still unanswered questions.

Or, what's the farthest point from an Interstate within each state?

In Illinois, there are a couple that could qualify:

Lawrenceville
Macomb

I'd say that Macomb and the area around it is the furthest from an interstate within Illinois.  Prior to the extension of I-72, there was nothing interstate-wise west/north of I-55 and I-155, and south of I-74 in the state.

[corco]

Using ArcGIS and an Albers projection, I get 48.934, -108.7067 in northeast Montana.

If you consider the border to be an interstate to get the biggest interior hole, it is at 42.87177, -119.8173 in  central Oregon

[bzakharin]

In NJ, the southern tip of Cape May is over 77 miles away from the nearest Interstate (295). This is more than 1/3 the longest distance you can travel (in a relatively straight line) while still remaining in NJ. Delaware is much worse, proportionally speaking, as most of it is not served by Interstates at all. The furthest point is presumable in Fenwick Island near the MD border.

[hotdogPi]

For the mainland* of Massachusetts, it's somewhere in Cape Cod.

*Islands are kind of cheating.

[vdeane]

For Vermont I'm guessing somewhere west of Fair Haven.  NY depends on whether you're including ON 401 or not.  If not, somewhere around Brasher Falls.  Otherwise, try the Tupper Lake area.

[vtk]

Using ArcGIS and an Albers projection,

You're telling me the industry-leading GIS suite can't do this kind of analysis without first bending the data into a 2D, Euclidean approximation?

[corco]

Using ArcGIS and an Albers projection,

You're telling me the industry-leading GIS suite can't do this kind of analysis without first bending the data into a 2D, Euclidean approximation?

Well yeah, that's how GIS data tends to work- the raw TIGER shapefiles are in straight up lat-long, which would be less accurate than bending it for this sort of analysis, even when you spit the results back out in lat-long.

I mean, yeah, I could have done it in straight up lat-long, but that's even more two-dimensional and less accurate than projecting it.

[vtk]

Using ArcGIS and an Albers projection,

You're telling me the industry-leading GIS suite can't do this kind of analysis without first bending the data into a 2D, Euclidean approximation?

Well yeah, that's how GIS data tends to work- the raw TIGER shapefiles are in straight up lat-long, which would be less accurate than bending it for this sort of analysis, even when you spit the results back out in lat-long.

I mean, yeah, I could have done it in straight up lat-long, but that's even more two-dimensional and less accurate than projecting it.

Yes but in my idealistic mind I always imagined ArgGIS would do one of two things: 1) project from lat-lon to a 3D data space and perform analysis that way, or 2) keep the data in lat-lon internally but not treat it as a 2D Euclidean space, performing all calculations using functions that already know how to correctly calculate directions, distances, and other measures from latitude and longitude.  The first method may suffer from weird over/under problems due to introducing a third degree of freedom in the data, while the second method would certainly be more computationally expensive, but both would be able to perform arbitrary analyses on even a global dataset without any distortions inherent in any map projection.

[roadman]

For the mainland* of Massachusetts, it's somewhere in Cape Cod.

*Islands are kind of cheating.

Not just 'somewhere' on Cape Cod.  Provincetown.

[hotdogPi]

For the mainland* of Massachusetts, it's somewhere in Cape Cod.

*Islands are kind of cheating.

Not just 'somewhere' on Cape Cod.  Provincetown.

Provincetown by driving distance, but not as the crow flies.

[Thing 342]

Virginia's is probably somewhere south of Appomattox, if not somewhere on the Eastern Shore.
North Carolina's is likely on Cape Hatteras.

[tdindy88]

Before I-69 was built in Southern Indiana, the area around the Crane Naval Warfare Center probably would have been the state's Point Nemo. Now with a freeway there, I'd have to say Rochester would be the new Point Nemo...at least until Interstate 67 is built.

[agentsteel53]

California?  probably the north end of CA-266 at NV-264.  237 miles to I-15 just south of Barstow.  I-80 in Nevada may be reached in 208 miles, and I-580 (Carson City) in 212. 

to reach the Nevada/California state line on I-80 is 247. 

[TheCatalyst31]

But there are still unanswered questions.

Or, what's the farthest point from an Interstate within each state?

In Illinois, there are a couple that could qualify:

Lawrenceville
Macomb

I'd say that Macomb and the area around it is the furthest from an interstate within Illinois.  Prior to the extension of I-72, there was nothing interstate-wise west/north of I-55 and I-155, and south of I-74 in the state.

For Illinois, I think it's East Dubuque, actually. East Dubuque is about 66 miles from I-80 in the Quad Cities (and farther from I-380 or I-39), while Macomb is 53 miles from I-74 in Galesburg (and 58 miles from the north end of I-172). Lawrenceville is less than 50 miles from both I-64 and I-70.

[roadman65]

In Florida Key West would be the furthest point away from an interstate.  Maybe Keys that are closer than that as well, but in Florida most places on the mainland are close to or near interstates.  Even in Sebring where that would probably be the mainland's furthest city from the interstate system at just over 55 miles from I-4, it is not as far as Key West or from the Middle Keys.

[Eth]

My best guess for the state of Georgia would be Fort Gaines, due west of Albany and SSE of Eufaula, AL. Looks like about the same distance to I-185 and I-10, and farther to I-65, I-75, or I-85. Either that or Alma (roughly equal distance to I-16, I-75, and I-95).

[Pete from Boston]

Using ArcGIS and an Albers projection, I get 48.934, -108.7067 in northeast Montana.

If you consider the border to be an interstate to get the biggest interior hole, it is at 42.87177, -119.8173 in  central Oregon

Please remind me what tool is used for this.  I'm rusty in GIS and it's been several years since I did a "most distant point from..." type of question.  Is it a "near"?

[froggie]

Thinking of playing with it today, time permitting (getting my household goods delivered this morning).  My first inkling is to use "proximity".

[vtk]

If a projection into 2D must be used, I think a conformal would give the most accurate results to this query.  Is Snyder's 50-state projection available?  If not, perhaps a standard conic conformal will suffice (for the lower 48 anyway).

[corco]

Euclidean distance. Vector tools like near or things in the proximity toolset I guess Could work, but that's a lot more steps. Much easier and more accurate just to create a raster masked by a vector of the lower 48.

As for projections, I always used an Albers equal area calibrated for the US in both undergrad and masters work, never had a problem with it. If it's good enough for USGS and the Census Bureau, it's good enough for me. Albers is almost perfect for an east west oriented country like the continental US- conformal conics are better for looking at smaller regions of the US, but for looking at the whole tamale its hard to beat an Albers for the continental United States. Its not that Albers is perfect, it's just that given the shape and location of the US on the globe, it's nearly perfect.

[vtk]

As for projections, I always used an Albers equal area calibrated for the US in both undergrad and masters work, never had a problem with it. If it's good enough for USGS and the Census Bureau, it's good enough for me.

Different projections are "good enough" for different purposes.  With the Albers, north-south distances are slightly extended and east-west distances slightly compressed in the more northern latitudes, while in the more southerly latitudes, the opposite is true.  When you're trying to find a point equidistant from things in 3 different directions*, such shape distortion will throw off the results.  Positions of local maxima will be very slightly offset, and if two distinct local maxima are at nearly equal distances from Interstates, there's a chance the shape distortion could actually influence which point wins.

*The point maximally distant from a set of objects surrounding it will be nearest to those objects at 3 points, and the distance to those points will be equal or they wouldn't all be the nearest points.

[corco]

Right, but when you use Albers in the continental USA, there is nearly no shape distorion . What you are saying is valid, but would be more applicable in Canada or Brazil-in the USA, there is nearly no shape distortion with Albers and you get the benefits of more accurate area.

There is a reason Albers is the gold standard for continental USA spatial analysis. Lambert is an excellent projection too, but it most useful when looking at regions of the US. Go do a Google scholar search for published papers with distance based spatial analysis in the continental US published in the last decade. I'll bet you $20 that at least 75% of them use an Albers Equal-Area.

[vtk]

Right, but when you use Albers in the continental USA, there is nearly no shape distorion . What you are saying is valid, but would be more applicable in Canada or Brazil-in the USA, there is nearly no shape distortion with Albers and you get the benefits of more accurate area.

How does accurate area benefit this particular query?  I don't believe it does.  And as there is "nearly no" shape distortion with Albers, there is "nearly no" area distortion with Lambert Conformal Conic, because it's a direct tradeoff between the two, with what little distortion exists appearing in the same extremities of the map.

So yes, both are nearly perfect.  I'm just saying, the conformal projection is more perfect for this query than the equal-area one.

I may not be an expert in GIS as a whole, but I understand math and map projections well enough for this argument.  Snyder's "Flattening The Earth" is one of my favorite books.