I'm a masters student at the University of Washington without any background in geo, so if this question seems naive or unclear, I apologize. Please let me know if I can add any information for clarity.

I'm working on a data visualization project and my subject is streamflow data. I'm attempting to display streamflow over a map produced with NHDPlus flowline data, which shows the location of all the rivers and streams in a given region. I've chosen HUC 17, the PNW, because I'm trying to render in the browser and working with data at the national level looked unwieldy.

I've come around to a vector tile approach as the best option for performance and interactivity. The shapefile, when converted to GeoJSON, weighs in at 438MB. To try and break that up I've found and am trying to use Tippecanoe, a tool for producing vector tiles from huge datasets. It works as advertised. It's awesome. Very easy if you're comfortable on the command line. The trick is that to upload these tiles to mapbox.com, where I could, ostensibly, view and style the tiles, each tile must be less than 500k and contain no more than 200,000 features.

I would like to know how to slice and dice the data in such a way that I can get it up to mapbox, or, how to work with the tiles locally using something like tilelive?

2 Answers 2


Glad to hear that you like tippecanoe! I really love that tool, too, but have been through exactly what you're facing, though with point data. In that case, the answer is to adjust the minzoom, maxzoom and droprate parameters. If you specify a higher minzoom, the average number of features per tile will decrease. If you specify a higher droprate, more features will be discarded at low zoomlevels.

But this brings me to the next point, which is that tippecanoe might not be the best tool for this data. Have a look at the work that Nelson Minar did with NHDPlus data. He prioritizes which features are included by their importance at low zoomlevels. This, rather than tippecanoe's randomized feature dropping, might be closer to what you want.

I hope this provides some clues, but if they're not enough, get in touch ([email protected] should work) and we'll connect you to team members who have more experience than me at using lines with tippecanoe.

EDIT: if it's not already obvious, I work at Mapbox. Hi!

  • 1
    Thanks, Chris. I've edited my answer to make this more obvious.
    – Tom Lee
    Commented Jun 1, 2015 at 18:43
  • Thanks for the advice! I've seen Nelson's work. In fact, the famous Wind Map from Wattenburg and Viegas, Nelson Minar's River Map and Mike Bostock's subsequent reinterpretation were inspirations for what I would like to do. Your point about lines vs points is good. I was hoping to avoid using PostGIS, because I know nothing about back ends, but if that's my way forward so be it. I sent a message to Mapbox at the same time I posted here. Maybe they'll share some wisdom.
    – bwstud
    Commented Jun 2, 2015 at 6:42

Simplify your geojson with Mike Bostock's topojson library

It can reduce a lot. You can also test online.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.