Hi, thank you. I'm not sure what level of detail that will produce, but, it sounds much better, and appropriate for remote regions. In OSM we usually speak of zoom level (e.g. 15), or meters of error on simplification (e.g. 0.1-1m). Perhaps by "40-50m" you mean the number shown on the scale bar in the upper-left of JOSM window.

In case you don't know, the setting for Simplification in JOSM is: Settings, Advanced Preferences, search for "simplify-way:max-error". For example, 0.8 is a reasonable value for hydrology, even in developed areas.

Hi, I ran simplify with a 90cm threshold. That is very conservative for a remote, wild area of mud and tundra. As for accuracy, I ran the JOSM validator and fixed many of the issues you left, with multipolygon ways intersecting each other, overlapping ways, and many other problems. It is fixed now.

velmyshanovnyi, PLEASE stop doing this. These features have way, way too many nodes - tens of thousands of colinear, inefficient, bad nodes. Please, please learn how to use Simplify to the correct level of detail. For these edits, I have to load them, delete most the points, and wait a long time for the deletion to upload.

Sorry i have no information on that highway. My own changes here are only on boundaries and waterways.

Diego, you seem to have added "natural=wood" polygons which are fairly arbitrary; they mostly contains forest but are more frequently contained inside larger continuous sections of forest, which makes them cartographically confusing (they do no correspond well). Do you mind if I replace them where possible with actual forest extent?

I have gone ahead and reverted this changeset and fixed all the geometry and topology it broke.

Dguillen, it looks like you accidentally "squared" and uploaded a very long tertiary route here, which destroyed its geometry. Please be careful! You should ask an experienced mapper how to undo your changeset, or manually fix the road you mangled.

Marek the starting point of your analysis is incorrect. It assumes that the existing, hyper-noded wood areas are precisely correct with no uncertainty or error bars. Consider the factors involved - imprecise criteria, human variability, seasonality, poor resolution, loose rectification, etc. it is absurd to claim certainty or accuracy within 20%, or for some parts of Nepal with even worse imagery, 50%. A change of 5% in the area only reflects the mathematical adjustment due to a larger number of convex than concave polygons in the digitization. It is completely dwarfed by the uncertainty.

Apply a strong simplification to these bad landuse polygons, or the community will have to do it for you (again).

OK, thanks! I am much reassured that quality is being cared for and you aren't just making work for me to clean up later :)

OK, it's possible that the places i checked were worse than average, but consider the fine-grained nature of these features, like at changeset/49780057#map=17/34.37226/104.93077 When the nodes are as close as 2.4m but the source data (Landsat) is 28m (!) then there is something terribly wrong here. I am worried I will have to come in here and do another massive simplification cleanup.

I spot-checked these imported forests and they bear no similarity to what is actually on the ground.

Could you please explain what these edits are? I truly hope it is not the "USGS treecover 2010" which is derived from low-resolution LandSat which is entirely inappropriate for mass-import into OSM. Or did you get this import discussed and approved?

For what it's worth, I wrote OSM messages to marek kleciak explaining these edits, and even described the detailed steps he can do in JOSM to do the cleanup himself.

For an example of BAG problems, see buildings like way/277899879 which has a "sliver" which is 18cm wide and 3m long; it is clearly a mechanical error. I have manually fixed hundreds of these problems.

For what it's worth, I recognize unlike other mass imports I'd had to clean up (Canvec, NHD, Tiger...) this data is much better in precision as well as accuracy. That's why i use thresholds not of 70-110 cm as with NHD, but as little as 6 cm for BAG, guaranteeing that cleanup stays within the accuracy of the source data and everything is improved, absolutely nothing is lost.

Hi, if you take a close look, you will see that no "details" were removed. As part of other cleanup (maproullette farm->farmland), I've been run the JOSM validator on some of the nearby buildings and landuse, and tidying up the results. The BAG buildings have numerous issues (including tiny slivers of geometry that are obviously artifacts) which i clean up along with other bad modelling. The is also overlapping identical landuse (which i fix) and some very poor overnoding (which i fix). I am an expert at what I'm doing and I'm happy to defend any node or way that anyone has an issue with. The final result is a map that loads and renders faster, is free of topological issues, and is easier to manually maintain.

Example: before: http://i.imgur.com/HuJTxvi.png
After: http://i.imgur.com/VDAa6uE.png

Notice how the nodes for the selected are reduced from 1832 to 241, without affecting the accuracy of the data at all.

Hi there, many of the ways in this changeset are EXTREMELY overnoded. When you upload, you need to check the data FIRST for this problem and solve it before uploading. Otherwise it creates 26903 nodes in the OSM database, then someone has to come along afterwards and clean it up by deleting ~90% of the nodes. This is really wasteful on the servers and also makes work for other.

To answer your last question, Yes.

Yes, it's NHD. Welcome to my world, Jeopardy. I've spent much of the last 6 months of my life cleaning up North Carolina's NHD (outdated wetlands just like these here in Florida). It's just a huge task, thanks for helping!