OpenStreetMap

Lakes, dams, and where to find them

Posted by Matt_ on 12 October 2023 in English. Last updated on 18 October 2023.

I wanted to share my experience finding and fixing a misplaced lake and dam as an example of what I’ve been working on in case it’s of interest to anyone else. I think I’ve picked a good example of what I’ve been finding. I also tried to make it as short and to the point as possible and then made this too long anyway. Here we go!

Finding a problem

Rowe Lake as a node in a forest

Rowe Lake (gnis:feature_id=1640193) (node) and Rowe Lake Dam (gnis:feature_id=1640194)(node) were two nodes on the side of a mountain next to a road in Monroe County, Tennessee. There were two problems with that:

  1. On satellite there was no lake or dam visible in the area. Which isn’t proof this is the wrong location, but it looks like solid trees.
  2. The GNIS tags on both nodes had them in county 157: Shelby County, over 300 miles away on the other side of the state.

Not finding a solution

Not great search results for Rowe Lake, TN

My first attempt to find the location for this lake and dam didn’t work so well but I wanted to mention this. I searched the web for Rowe Lake, and Rowe Lake, Tennessee. There are a bunch of results, but they all seem to have been generated off of those same coordinates from GNIS! Fishing, weather, hiking, and lodging, all for a patch of forest with no easy foot access! Even the ones that describe it as being in Shelby County still give the coordinates in Monroe County. So much anti-helpful content from such a small piece of information.

Actually finding it

The lake's position over 300 miles east of Shelby County

So where is it? The county Rowe Lake is supposed to be in (Shelby) is due West of the coordinates given. What if the latitude was right, but the longitude was the problem?

In JOSM I began searching along the line at 35.325° North for anything lake-like. Shelby County is 25 miles wide however. I’m scrolling while zoomed all the way in and not finding it, at least not very quickly.

What if only part of the longitude is wrong? Shelby County extends from about -89.63°W to about -90.09°W. GNIS’s longitude was -84.0499°W. What if we try -90.0499°W at the same latitude? A whole number of degrees difference.

A possible solution as seen in JOSM

Hey, that’s a lake with a flat, dam-like side! But is it right? Enter the hero of this story: the United States Army Corp of Engineers’ National Inventory of Dams!

The National Inventory of Dams entry for a dam by the name of Rowe

Yup, found it. So I moved the lake and dam nodes, used them when tracing the outline of the lake, and tagged the dam with the nice id number the NID provides (TN15714). I’ve also sent the correction for the lake’s coordinates to GNIS, since they don’t track dams any more. I haven’t contacted them before, hopefully they’ll make an update. Either way, this was fun for me. Thanks for reading!

The final result

Extra info

This is one example of the edits I’ve been working on for the past few weeks. There have been a bunch of lakes and dams I was able to successfully (re)locate the same way, but other patterns emerged as well. I also wanted to share that perspective (briefly-ish):

  • There were many lakes and dams where a one digit change in its coordinates gets it into the county specified by GNIS, and the new location matches what is in the National Inventory of Dams (NID).
  • Almost as many cases where the coordinates were right and the GNIS county was simply wrong. Again, the NID came to the rescue for confirmation.
  • A bunch where the coordinates weren’t wrong they were just very close to the county border and the dam really is in both counties.
  • Sometimes the coordinates were wrong but GNIS already fixed them by creating new entries that were also already imported into OSM. So you move the lake and dam nodes to the correct location but another lake and dam with the same name is already there.
  • Finally, there are times where the coordinates were wrong, GNIS made new entries with the correct info, and a body of water near the incorrect location somehow picked up the same name as the one a hundred miles away. An example is Tabernacle Camp Lake at the Tabernacle Campground in Monroe County, Georgia. And then Tabernacle Camp Lake, almost exactly one degree longitude to the west, in the middle of somebody’s home or farm. So you have the NID showing a dam at the place the campground actually is, but the NHD is showing a lake with that name at the other location! So for now, I guess both exist.

Discussion

Comment from gvwaal on 12 October 2023 at 17:06

Great work, that’s some neat detective work. How come the NID wasn’t your first resource to find the dam? Are erroneous dam points typically not that far away from the correct location?

Comment from watmildon on 12 October 2023 at 21:11

The original points were almost certainly from an import from GNIS long ago. The precision on many features back then was… not optimal. Cleanup continues.

It turns out that Rowe Lake is also not correctly named in the NHD dataset (!!) so I’m sure the GNIS folks will be very happy to get this corrected for everyone.

Great sleuthing!!

Comment from Matt_ on 12 October 2023 at 22:29

Thank you both! And thank you for your question. I didn’t use the NID at first because I didn’t know it existed yet. I am a total novice when it comes to any of this

But once I found out about the NID, I went looking for more dams like this one. As for how many of these errors exist, yeah I guess that depends on how GNIS entered and maintained this information over the decades. A lot of what I found seemed like they just mistyped a digit or two when entering the coordinates.

I went looking through my changesets, this example is the farthest edit I made at 542km from where the nodes were to where I used them to draw the lake. Then for other reservoirs and dams that I found the correct positions for, there were a total of 4 in the 300-400km range, 17 between 200 and 300km, 118 between 100 and 200km, 175 between 10 and 100km, and finally 56 between 1 and 10km.

Looking at the longest distance ones, they all seem to be in large or wide states like Texas, Tennessee, and Oklahoma. Maybe if a coordinate error was enough to push something into another state GNIS actually noticed and fixed it themselves? At least before this data was imported into OSM.

Comment from Skunkman56 on 13 October 2023 at 16:04

Excellent work, there are many problems with GNIS- data entry errors like these, obsolete entries, etc. I particularly like MapRoulette for organizing and working through GNIS nodes that have not been addressed yet (reviewed, converted to ways, etc.). Another place to update the bad coordinates is Wikidata which has many of the GNIS entries imported as well.

Comment from InfosReseaux on 22 October 2023 at 17:02

Hello

It is a great work, thank you to have published this here.

I second your post and mention this wiki page for anyone interested in mapping hydropower facilities. https://wiki.openstreetmap.org/wiki/Power_generation/Hydropower

Many additional features between lakes and power house generally remain to be mapped.

All the best

Login to leave a comment