Shaun,

That was a really interesting read! Nice work and good writing! I
haven't checked what you put on github, but it should prove useful to
the OP at the conceptual if not the practical level.

Chris

On 11/11/2014 2:08 PM, Shaun Walbridge wrote:
> I’ve done some work on global modeling of ship movement, and can
> confirm this is a tricky problem. The ultimate solution that I ended
> up using was a weighted graph structure to connect adjacent ‘cells’ of
> the ocean. This has the advantage of not requiring arbitrary new
> calculations for any new pair of points, and can vary in resolution.
> The edges are then weighted by their geodesic distance, so that cost
> calculations are ‘true’ and not fixed to any single projection. Some
> details here <https://4326.us/thesis/>.
> 
> cheers,
> Shaun
> -- 
> 
> *Shaun Walbridge | GIS Engineer*
> 
> 
> From: a brody <[log in to unmask] <mailto:[log in to unmask]>>
> Reply-To: a brody <[log in to unmask] <mailto:[log in to unmask]>>
> Date: Tuesday, November 11, 2014 at 10:59 AM
> To: "[log in to unmask] <mailto:[log in to unmask]>"
> <[log in to unmask] <mailto:[log in to unmask]>>
> Subject: Re: Finding Distance between points over water
> 
> FYI,
> Mercator projection is a very poor measure of distances.  Its
> advantages are that the latitude and longitude lines come at right
> angles, make it easy for people plotting their routes, and determining
> the distance travelled in latitude.  Longitude varies though based on
> latitude. The best distance measurement for shortest distance is great
> circle routes when you can get it, but the problem remains many of
> those routes run into the arctic regions and iceburgs.  So the best
> compromise may vary by season for the hemisphere travelled.
> 
> It really depends on your end users needs.
> 
> Sincerely,
> Abraham Brody
> On Nov 11, 2014, at 12:14 PM, Don Cooke <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
> 
>> A friend at GDT suggested a projection-free way to calculate
>> distances between lat/long pairs:  calculate the spherical angle
>> between the two points relative to the center of the earth, then
>> convert the angle to a distance using an assumed radius of the
>> earth.  I guess you need to assume a spherical earth….
>>
>>  
>>
>> Interesting way to think about it.  We do tend to get the rope wound
>> around the axle given our comfort with projections.
>>
>>  
>>
>> ****** Don
>>
>> *From:*Northeast Arc Users Group [mailto:[log in to unmask]]
>> *On Behalf Of *Andy Anderson
>> *Sent:* Tuesday, November 11, 2014 8:29 AM
>> *To:* [log in to unmask] <mailto:[log in to unmask]>
>> *Subject:* Re: Finding Distance between points over water
>>
>>  
>>
>> No single projection will work, you must use different ones over
>> subglobal distances, which prevents a single calculation. But one
>> approach would be to use the plate carrée projection (your attached
>> example) to generate a rough route, then choose an appropriate
>> projection over each leg, e.g. two-point equidistant, and then
>> project to a new raster and redetermine the legs and then calculate
>> the distance. Still won’t be the best possible distance because
>> you’re fixing points, but it should be a better estimate.
>>
>>  
>>
>> For the resolution you are using (~70 Km/pixel) I don’t think a land
>> buffer is necessary.
>>
>>  
>>
>> — Andy
>>
>>  
>>
>> On Nov 11, 2014, at 10:56 AM, Milan Budhathoki
>> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>>
>>
>>
>> Hi Chris and Andy,
>>
>>  
>>
>> Thank you for your suggestions. I was given just Lat/Long of ship's
>> starting port and destination port in .CSV. Beside these I don't I
>> have any additional information. Yes, prevailing winds and ocean
>> currents will play a role in ships actual path but I think my
>> collaborator is looking for the shortest distance just over water
>> regardless of any other assumption that might affect on actual
>> route. I haven't research at published route documents but this is
>> worth trying. Projection is another issue since dataset has
>> thousands of ports around the globe. I am thinking of  the
>> "Merrcator Projection ?". I did a quick cost-path analysis for one
>> of the ports pair. Due to lack of information to make a cost-raster
>> ( I used just land and water mask as a cost raster) path tends to
>> grip through land as below;
>>
>>  
>>
>> <image84 Nov. 11.jpg>
>> ​
>>
>>  
>>
>>  
>>
>>
>> ​
>>
>> Here voyages tends to pass very closely to land (mostly touches land
>> mask). I am thinking to buffer land mask outward with X miles so
>> that ships path will be X miles way from land. Also I am considering
>> to bring ocean depth as another cost-raster if data is publicly
>> available. Eventually, I am write a python script or run model
>> builder once I figure out better way.  
>>
>>  
>>
>> I would like to hear more if anyone has any thoughts ! 
>>
>>  
>>
>> Thank you ! 
>>
>>  
>>
>>  
>>
>>  
>>
>>  
>>
>> On Mon, Nov 10, 2014 at 11:38 AM, Andy Anderson
>> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>>
>> To remove the tedium, write a Python script.
>>
>>  
>>
>> Remember, though, that distance calculations depend on the
>> projection you use. If you want the cost-path from, say, New York to
>> Sydney, there is no single projection that will give you an accurate
>> measure.
>>
>>  
>>
>> A better approach might be to determine the standard shipping lanes,
>> calculate the distances (if you can’t find them in a table), and
>> piece together routes.
>>
>>  
>>
>> More generally, you could set up an iterative algorithm to calculate
>> distances using spheroid-based angular calculations with
>> restrictions based on open water (e.g. at 40° north latitude,
>> longitude will be restricted to roughly –74° to –9° and 128° to 140°
>> and 142° to -124°).
>>
>>  
>>
>> — Andy
>>
>>  
>>
>> On Nov 10, 2014, at 9:13 AM, Milan Budhathoki
>> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>>
>>  
>>
>>     Looks like email that I sent yesterday didn't go through. 
>>
>>     Here it is again: 
>>
>>
>>
>>     Hello Listserv,
>>
>>
>>
>>     I have point dataset of ship trips from one port to another. I
>>     want to calculate the shortest distance between each port pair
>>      *over water*. There are thousands of voyages, and 5,000 unique
>>     ports from all over the world. One of the approach I can use in
>>     ArcGIS  is to run the Cost-Path tool having water/land as a cost
>>     raster to make a path only on water. But I assume that the
>>     Cost-Path approach would be little tedious for a large dataset.
>>     I wonder if anyone in this forum has a suggestion to calculate a
>>     shortest distance between two points having restricted path.
>>
>>
>>
>>     I will highly appreciate your feedback.
>>
>>
>>
>>
>>
>>     -- 
>>
>>     Milan Budhathoki
>>
>>      
>>
>>     -------------------------------------------------------------------------
>>     This list (NEARC-L) is an unmoderated discussion list for all
>>     NEARC Users.
>>
>>     If you no longer wish to receive e-mail from this list, you can
>>     remove yourself by going to http://listserv.uconn.edu/nearc-l.html.
>>
>>  
>>
>> -------------------------------------------------------------------------
>> This list (NEARC-L) is an unmoderated discussion list for all NEARC
>> Users.
>>
>> If you no longer wish to receive e-mail from this list, you can
>> remove yourself by going to http://listserv.uconn.edu/nearc-l.html.
>>
>>
>>
>>  
>>
>> -- 
>>
>> Milan Budhathoki
>>
>>  
>>
> -------------------------------------------------------------------------
> This list (NEARC-L) is an unmoderated discussion list for all NEARC
> Users.
> 
> If you no longer wish to receive e-mail from this list, you can remove
> yourself by going to http://listserv.uconn.edu/nearc-l.html.
> 

-- 

Chris Duncan, Ph.D.
President, GISmatters
[log in to unmask]

http://www.gismatters.com/
http://www.allterraindigital.com/
http://www.plugandplaymaps.com/

Tel: +1 413.549.2052
Fax: +1 413.658.0346

1 Tuckerman Ln
Amherst, MA 01002

------------------------------------------------------------------------- This list (NEARC-L) is an unmoderated discussion list for all NEARC Users.

If you no longer wish to receive e-mail from this list, you can remove yourself by going to http://listserv.uconn.edu/nearc-l.html.