California NHD Stream Import Chaos

Hello. I’ve been doing some cleaning up and remapping of old imports above Sacramento for awhile now. Up until now its been mostly confined to farmlands, but in the process of research new areas to improve, I stumbled on a large area west of Red Bluff that is covered in streams due to an NHD import in 2010. Looking over the area, it appears a lot of the areas where there is supposedly streams are either not visible as streams on the maps, are covered with shrubs, or are small valleys that appear to have been dry for a very longtime.

Although I am aware that some of the streams can be labeled as intermittent, it seems as though a lot of them, if any geographical feature exists at all, are just places where there is a valley of which there might of been a stream at some point, even long in the past, but there is not one there now. So adding intermittent seems rather misleading in those cases. Since it might have been a couple of thousand years since water was there. I think I rather just delete those ones if its appropriate and leave the streams that are actually visible on the map and that someone will actually come across.

Here’s a link to the area

Any advice on how to approach it would be much appreciated.


Define “now.”

As I understand it, that part of the world can occasionally have very heavy rainfall. Maybe only every 5 years. Or 10 years. Or 50 years. Heavy enough that those valleys, which appear dry at this moment, can flood. It might (as you suggest) have been a thousand years since some of them had water, or it could have been 100 years, or 10 years. NHD mapped them for a reason, and I suspect they know a lot more about the hydrography of the area than you or I. In your position I’d check with NHD before I deleted them.

In southern California it seems the NHD imported streams follow the topography shown by USGS maps fairly well. The major exception being in flat areas, especially those that have been built on, where the water course may have meandered naturally or, more likely, been engineered into a different location.

My major problem with the import is that it made no distinction between year round, seasonal, intermittent or ephemeral flow so the maps look overly wet. OSM does not seem to distinguish between seasonal, intermittent and ephemeral flows so I’ve simply tagged those as intermittent when I clean up the geometry to follow the usually obvious (sand in desert, line of trees in scrub) alignment.

I don’t think that that’s especially helpful here, for a couple of reasons.

One is that there’s not necessarily a match between what the NHD were recording and what was subsequently added to OSM (in fact initially, there absolutely wasn’t - at first these were just shovelled in as “waterway=stream”).

The second is that the NHD likely know nothing whatsoever about OSM tagging.

What the NHD appear to be recording here is “somewhere that water might flow”. If you’re downhill from it that’s an extremely useful thing to know, but it doesn’t necessarily mean that “waterway=stream” (even with “intermittent=yes”) is an appropriate tag. In Derbyshire where I live I can think of plenty of places that, after very heavy rain, will have water running down all sorts of gullies and paths, but in OSM terms they’re not “enough like” a stream (even an intermittent one) most of the time to justify tagging them as that. A quick look at the imagery* here suggests something similar - annually as the snow melts there will be water, and some of the larger gullies will have water in them predictably enough to justify tagging as an intermittent waterways. However nearer the top, even though the features are caused by water erosion, I suspect they’re not worth being labelled as “streams” of any sort.

Unfortunately going through imports like this after the event is really a shit-shovelling job - in some cases the best thing really is just to revert it (but perhaps leaving features edited by subsequent mappers), in others it might be possible to do some sort of tidy-up exercise (Maproulette or similar). Maybe there’s some data in the NHD dataset that could be used to say where things are “most like” intermittent streams and which aren’t and create some OSM tagging to match? “intermittent” and “seasonal” have been discussed on and off on OSM mailing lists and elsewhere for a while; it might be worth reading back through those for some clues. Alternatively maybe based on local experience you could say "well the top 50m (or whatever) of these “streams” could be tagged one way that renderers can treat as “not a stream” if they wish and lower down (where there’s more predictable flow) “as an intermittent stream”.

Unfortunately I’d write off OSM’s “standard” rendering of doing a decent job here because it has to try and do the best it can everywhere else on the planel too. You’ll likely need to create something that does a decent job with the data locally.

  • unfortunately I’m not really familiar with this bit west of the I5, I’ve only ever seen it through the trees from the west.

Indeed. If you look at a map and it shows no sign of water flow, that means it’s safe to be there even though there was a storm in the distant mountains several hours ago. Right? No, it’s not right. There are places where it’s better to know that water could flow and see no water there right now than to be taken by surprise.

As I understand it, California has things called “arroyos” which are dry creeks, stream beds or gulches that can suffer flash floods after thunderstorms. Vehicles can be swept away. I’d be hesitant to remove those waterways until I was sure they weren’t prone to flash flooding.

I would say that intermittent is not wrong. It says that sometimes there is water there, it doesn’t say how often or for how long. It makes no promises other than that sometimes you may see water there and other times you will not. And if it’s a stream when the water does happen to be there then waterway=stream (or even waterway=river if it happens to be a river when water is flowing) isn’t wrong either.

That said, I would have no objection to inventing other values for intermittent than yes or no, such as ephemeral. Or even an ephemeral=yes tag. But intermittent=yes doesn’t actually misrepresent the situation. The stream may be a lot more intermittent than you were expecting, but that’s a problem with your expectations, not the tag.

Incidentally, even if the waterway is ephemeral to the point of only having water every hundred years, having it on the map indicates a valley of some sort. An indication, in the absence of contour lines, that the ground isn’t completely flat. Something that may be important in some situations.

So render intermittent waterways with dotted lines or a lighter shade of blue or something. You’re proposing omitting tags for features that are actually there (even if only intermittently) so that the map doesn’t look too blue. I believe that’s called tagging for the renderer… :slight_smile:

Based on spending a lot of time in the south of Spain, where we also have “arroyos”, I would be inclined to agree. They are an identifiable feature of the landscape, often identified by name in the same way as a river or stream even if is is years since water has been seen there. I feel like there should be some way of identifying them. If there is a feature of the landscape that a local can point to and say “that is Arroyo Toquero, don’t park your car there if heavy rain is forecast”, it seems like we should be able to map that feature in OpenStreetMap. (To be clear, I don’t know the area of California originally mentioned so I don’t know if we are talking about exactly the same kind of feature).

Perhaps tagging these features as intermittent waterways is not ideal. But what is the alternative? As far as I can reconstruct past tagging discussions and wiki edits, it seems that some mappers tried using “waterway=wadi” to distinguish this kind of feature. The terms “arroyo” and “wash” were also mentioned. But that tag seems to have been shot down, ending up being marked as deprecated in favour of intermittent=yes. As far as I can see, more granular values of intermittent= (to distinguish say “rare flash floods” from “water here every winter”) don’t seem to have gained much traction. Overall I would prefer to leave these mapped with a less-than-ideal tag rather than remove them completely.

Me neither, but arroyos are apparently fairly common in that area. So I don’t know for sure that any of them are arroyos, but I also don’t know for sure that none of them are.

As far as I can tell (from Wikipedia, so I don’t entirely trust it) arroyo, wadi, wash, gulch and rambla are terms used in different parts of the world for essentially the same thing. With the usual complications, so in South America arroyo means any small stream even if it’s never dry (which makes sense, because arroyo.was originally the equivalent of the English brook).

Possibly because it’s hard to tell from a quick survey which level applies. Seasonal does seem to have gained some ground. As does intermittent. Whether or not we need more granularity around intermittent is a matter for discussion (which may go nowhere).

I’d go a little further. Leave them there with a less-than-ideal tag unless you can be certain they’re piddling little streams that never flood, don’t have steep banks and even a cat could cross them without getting wet. Steep banks are a hazard if you’re on a BMX bike and may present insurmountable obstacles to hikers. Flash floods are just plain dangerous to anyone. Change them to match OSM definitions of stream and river if necessary, tag them as intermittent if they’re dry right now, but otherwise leave in place until absolutely certain they’re not a hazard.

My gut feeling is still that NHD wouldn’t have mapped them as waterways if they’d been completely dry for as long as we have records of them. Which means that if they’re dry now they’re intermittent. For small values of intermittent.

Er, not really. Further south, and further east - yes. Here however we’re talking about gullies at the tops of hills. Have a look at the link that was provided and look at the cycle map view there to see the topography - a wash in a flat desert floor it is not.

My sense of things is that “gulch” or “gully” is more apt to be used in northern California where there was large Anglo immigration in the middle of the 1800s. Whilst “arroyo” or “barranca” is more likely to be found in southern California which retained a Spanish/Mexican majority population a bit until a bit later. Gulches, gullies, arroyos and barrancas may or may not have a definite stream bed in the bottom. And if they have a stream bed is may or may not be a seasonal/intermittent/ephemera flow.

And “wash” seems to be used more in the deserts of southeastern California and in Arizona for the actual course that water is likely to take when it is actually flowing. Washes are generally visible in aerial imagery by the sand surface and lack of vegetation such as it exists in the Mohave, Sonoran or whatever desert (the Sonoran Desert of Southern Arizona and Sonora Mexico are actually semi-arid rather than arid and have a fairly extensive plant covering).

Probably not a set formal definitions and probably nearly as many exceptions as meets these rules. They are just my observations from having been raised in Arizona and having lived in various parts of California for many decades.

SomeoneElse, that’s my thing. In that particular area on top of the hill it seems like the NHD just excessively tagged every indent in the mountain as a stream even if it wasn’t one. Other areas in California, like the Sierra Nevada mountain range or the Mt. Shasta area, do not have galleys mapped that excessively as streams although I am sure water might have flowed through them at some point in the past to. So I get the idea it might something particular this area. Its doubtful to me that it is because all the streams mapped are valid ones. In other areas of California there are streams imported that are either way off. go through buildings, or have no clear sign that there is a stream there at all. Some of the ones that go through buildings could be underground, but it still shows there was no quality checks or review of the streams before they were imported.

I also don’t see how not mapping them or using another tag would be mapping for the render as Brain De Ford suggests. It seems like it would be the other way around. Since they are not even geographically there most of the time. To me, mapping something as a stream should indicate there is a flow of water there more often then not. Otherwise, why not map snowy areas as surface=snow since there is snow covering them 5 months out of the year? Maybe just mapping the cliff sides to indicate there is a gully would be a better way to go? I don’t think intermittent is appropriate here since there’s no indication that the streams are there seasonally. They aren’t created by a melting glacier or an underground spring and it go for years without snowing. So I don’t think it fits.

As far as the flash flood discussion goes, The argument could also be made that mapping none existent streams could be a hazard to someone out hiking that is looking for a water source or in a situation where someone might avoid the area because they think there is an impassible stream and get off track. There’s a few places in the Red Bluff area where a streams creek bed was added by the NHD import as a water way area and it looks like they are practically rivers, but 99% of the year you could just hop across them. Plus, I’m pretty sure flood zones are vastly larger areas then the actual creek path. So it would be misleading to leave them there simply for that. Maybe there should be a landuse=flood_zone tag. There’s a lot of places in my area where it would probably be useful.

To me, “seasonal” and “intermittent” are not equivalent. The wiki also indicates that they are not only different words, they have different meanings. Yes, seasonal implies intermittent, but intermittent with some degree of regularity and predictability. Whereas intermittent does not imply seasonal.

They’re rendered as dotted lines, which means intermittent. If you’re going on the sort of hike where you think there’s a possibility of needing water from a stream but you don’t bother to do the checking and planning to determine which sources are reliable and which aren’t, maybe a mishap would improve the gene pool.

Frustrating, perhaps, to learn you could have taken a different route but not lethal. After all, you do plan your hike, don’t you? So if avoiding the intermittent stream means a far longer hike than you’re capable of, you look for somewhere else to hike, right?

I realize you probably wrote “99%” as a rough guess, it could be 90% or 99.9%. But let’s go with 99%. 99% of the time it’s a dry bed you can cross. 1% of the time it’s a river you can’t cross. Wouldn’t you like to know that in advance, rather than be taken by surprise? I think if you arrived there planning to cross what shows as dry land on the map and found a river there you’d be somewhat annoyed and wondering why nobody mapped it.

I’m having difficulty grasping your reasoning. If there was nothing on the map there I could understand you not bothering to add intermittent streams and rivers as being too much work - OSM data is incomplete and imperfect but slowly improves. If there are intermittent waterways on the map that are now blocked by construction or have been diverted into culverts, I could understand you wanting to fix things. What I can’t understand is wanting to remove an intermittent feature simply because it has no water right now and which you have decided never does have water based, not upon years of observation or records, but because you think it will never have water.

Landuse would be silly. “This area was industrial in the past, but we decided to use it for flooding instead.” There is, however, flood_prone=yes and the proposed natural=floodplain. Floodplains may have no water 99% of the time, or even 99.9% of the time, but some consider it useful to know of their existence (like before buying a house, for example). But if streams that are dry 99% of the time shouldn’t be mapped then floodplains that are dry 99% of the time shouldn’t be mapped either.

To be honest I wasn’t thinking about language differences here, just the physical feaure differences between a wadi/wash/arroyo in the middle of a desert and a gully near the top of a mountain.

@Brian de Ford

OK. I don’t care about the semantics of it. The fact is some of the things tagged as streams probably aren’t actually streams. Either because they were a long time ago or they never were in the first place. I don’t know how you define what constants a stream but I define it much how Wikipedia does “A stream is a body of water with a current, confined within a bed and banks.” A lot of those “streams” don’t have a bed or banks. Take a look for yourself. A gully would not qualify under that definition either. It doesn’t help that what constitutes a stream is not defined on the OSM Wikipage. Neither is what constitutes a waterway. The fact that is called a water way in the first place insinuates to me that there should be water involved somewhere. There’s no “current” proof of that though. Since there is a tag for disused and historical things, I assume most things on the map have a life cycle. Why should streams be any different? If there visually a gorge there, it should be mapped as a gorged. It’s not a historical map is it? I tagged an usable water fountain the other day as disused. Its been out of commission for years, but who knows, it might get turned back on eventually. So according to your logic I should of left it alone?

Yeah, but like I’ve already said multiple times, some of them are not intermittent. I have no problem tagging the ones that are as such, but I don’t see the point leaving things on the map that don’t actually exist. And yeah, people should do planning and not rely singularly on OSM, but there is at least some level of presumed accuracy of the map or else why bother using it.

Yeah obviously I’m making a rough estimate based on the fact that none of the images showed it as full of water and also because it was mostly empty for the 2 years I use to drive by there. There is also places where there are pinch canals or whatever they are called and the stream takes a different course now but it was still mapped as going the other way. As far as if id like to know in if I can’t cross it, the exact opposite could just as likely true if I want to know to if I can cross it. Essentially nullifying your argument. Plus there is a natural=shingle tag that has been used like 14 thousand times to map river beds. So it seems to there are situations where just making it a body of water even though its not isn’t always the best way to go.

You are probably having a hard time understanding me because that’s not what I said. I never said I wanted to remove a feature because it doesn’t have water right now. What I said was that there is a freakishly large amount of streams in that area compared to the rest of California. So it is highly likely some of them are not streams. Once again, I have no problem tagging streams that actually exist as interment. I have an issue tagging streams that aren’t actually there that way. I never said it wouldn’t ever have water either. I just don’t think it having water once every hundred of years should qualify it to be mapped. If you bothered to look up the NHD and how they got their stream information, you would see they didn’t get it through years of observation either. They used LiDAR to survey soil temperatures in some areas and then combined that with geographical features maps, along with maps going back to the 50s in order to make a guess as to where the streams are located. Even they say on their website that it is speculative non-scientific way to find streams and that the usefulness of LiDAR for that purpose is still being researched. There’s a bunch of articles on the internet about it to that essentially say the same thing. A few samples include “High flow accumulation channels can be accurately predicted by LiDAR data, but lower flow channels were less accurately estimated.” “Use of LiDAR based data may be an effective way to identify channels that may be otherwise hidden from view.” "In mapping, the NHD is used with other data themes such as elevation, boundaries, and transportation to produce general reference maps. “the NHD has millions and millions of streams, and it would be extremely difficult for anyone to collect information like that for even a small portion of the stream network.” I think that pretty much proves my point. Sounds pretty iffy to me. Its not like I didn’t do research before I brought this to the attention of the forum. Plus I have geography classes under my belt that covered this subject, including the use of LiDAR. Thanks for the assumption that I was just pulling all this out of thin air though.

What would be silly about that exactly? There’s several parks where I live that use to be residential zones with houses but are now flood plain areas because of flooding due to excess water being released by the dam near by. The same is true a few cities south of us too. There are multiple parks in those towns that are flood plains. Benches get washed away every few years and they never build there. Its a specific landuse case. flood_prone=yes and natural=floodplain wouldn’t really work because the areas are not “prone” to flooding and they do not get flooded due to natural causes. And yeah, people should know about if there house is in a flood plain before they buy it, obviously. That wasn’t the scenario I was talking about though. If the water district here decides to demolish the dam, as does happen sometimes, it would essentially make the floodplain areas no longer flood prone. There are situations where things like that stop being the case. So I don’t even know what your trying to get at. Anyway, the flood thing was a pointless side conversation that has to do with the originally question since the streams are on top of a mountain like SomeoneElse has said several times.

Plus I think the whole usage of a percentage point is misleading because it doesn’t matter if it is dry %99 or %1 of the time. What matters is how much time has passed between the percentage points. If the 1% of the time it is wet only comes around every 100 years and only lasts for a few days, then I don’t that the stream should be mapped. Plus, there’s plenty of things in nature that stop and never come back. So ultimately your making an assumption that the water will flow through the channel again and it might not. Not just because it won’t rain again, but also because there is a pretty chance the water could take a different path next time due to plant growth, wind erosion, clear cutting, mining, etc etc…So really if its wet again or not eventually doesn’t really matter either.