USGS Multimedia Gallery: (Video)--Positional Accuracy of the National Hydrography Dataset- Part II

Let’s continue our study in Montana. Looking at Skalkaho Creek. Here we can see the NHD in the blue line. The orange dots you see are the sample points. This is where we took a lateral measurement of how far we are in the NHD from where the stream is today. 84 samples selected here, a mean of 27 feet, a standard deviation of 40 feet, a 90th percentile of 70 feet, and a maximum of 227 feet.

Here we can see where that 227 foot maximum is. The measurement that is shown is 161 feet. The other 227 foot measurement is the offset you see.

Here is another good example of Eightmile Creek. We’re zoomed in pretty tight here. These orange dots are 100 feet apart. This is a good example of generalization of the NHD. Here the stream is slightly meandering, but the NHD is represented by more-or-less a straight line. Here 100 samples were taken at the orange dots. A mean offset of 32 feet, a standard deviation of 33 feet, the 90th percentile is within 82 feet, and a maximum deviation of 160 feet.

Here is another section of the Eightmile Creek. You can see the NHD, more-or-less a generalized straighter line against the imagery which shows a little bit more meandering. You can see the 100 foot intervals where the offsets were measured at the orange dots.

Here’s the Musselshell River in Montana, a larger stream. Here 100 samples were taken at the orange dots. There is a mean of 15 feet, and a standard deviation of 22 feet, a 90th percentile of 42 feet, and a maximum of 127 feet. So in this river the results are very good, very tight control. The mean being 15 feet is one of lower means that we see. So this is an example of a section of the Musselshell River where the NHD lines up very well to the contemporary position we can see in the imagery.

Here is a spot in the Musselshell River where it’s not lined up very well. This river has changed course quite a bit from the NHD. So one of the issues we need to address is when we do our samples to make sure we are looking at average situations and not just looking at the bad situations or good situations, but getting a good representative sample of the average.

Another approach to this whole study is to look at points in the country where we have created new NHD typically at 1 to 4,800 scale that we call local resolution NHD. Comparing that against the original NHD that was collected from topographic maps that may be 20, 30, or 40 years old. And comparing that NHD line to the new NHD line.
So here we’ve looked at 560 miles of stream in Arkansas using this technique. What we do is create a buffer around the stream, a 40 feet buffer, and we can find out how much of our original NHD falls within the 40 foot buffer from the new NHD. So here in these 560 miles of streams, 82% of the total length is within 40 feet, 58% is within 20 feet, and 35% within 10 feet.

So what do these buffers look like. Here we see that the blue line we see there is the new NHD at local resolution. And the shaded area blue is the 40 foot buffer.

Here is the original NHD in orange. The question is how much of that orange stream is in the buffer. GIS is very effective in doing this kind of analysis. So we just let the GIS tell us how much of the stream is within the 40 foot buffer and how much of it isn’t.

Here’s an example of a 20 foot buffer in the darker blue area.

So Brush Creek in Arkansas is in the northwest corner of the state.

The red line you see there is Brush Creek amongst the NHD.

Here we see Brush Creek in green against the imagery.

Here we are looking at Brush Creek where the yellow indicates those sections of the river that are within 10 feet.

The purple are those sections of the river that are within 20 feet. So you can see most of the original NHD is within 20 feet of where the new NHD is. Indicating that the original stuff is not too bad.

Here are the sections within 40 feet. Virtually all of the river, except for a few spots, are within 40 feet. You can see that one green spot there on the left

Here is a zoom-in of that. This is where we are off by 79 feet. And so the orange indicates those sections of the stream that are within the 40 foot buffer. The green indicates we are outside of 40 feet and we can see an example of the offset of about 80 feet.

Here is another section of the river that was off by 48 feet. We’re outside of 40 feet so it’s indicated by the green line that we’re outside of the 40 foot buffer.

Here’s what Brush Creek looks like in the original topographic mapping. Here we’re looking at the topographic map and the darker blue line is the NHD on top of the topographic map. We can see that the NHD as it was digitized fits the topographic map fairly well. And that’s another issue in positional accuracy is how far off the stream is just due to digitizing error. Typically it’s not really significant. It’s just more-or-less noise and not a major factor here. But as an example of topographic mapping from which the NHD was taken, the photographs were taken in 1968, and the stream was field checked in 1970.

Here’s Brush Creek showing the blue which is the new NHD and the yellow is the original NHD. One of the things we can see here is that additional tributaries were added to the new NHD that we see in blue, and we don’t see those same lines shown in yellow.

We looked at Iowa at 391 miles of streams using this same technique of comparing the new contemporary NHD against the original. We found that 86% of the total length was within 40 feet, 64% was within 20 feet, and 38% within 10 feet.

Here is an example of a stream in Iowa. This is an unnamed stream. Showing in green the original NHD.

Yellow indicates those sections of the stream that were within 10 feet. Green indicates we are outside of ten feet.

Here are the sections in purple that are within 20 feet so green indicates we are outside of 20 feet.

Here these orange sections indicate we are within 40 feet. So we see no green sections, so essentially the entire stream is within 40 feet.

Here is an example of the new contemporary stream in light blue. We can see those sections in orange in the original NHD. You can see from the blue line that we are further away from the stream indicating we are 40 feet. As the light blue line gets close to the multi-colored line you can see that in some sections we are within 20 feet, and in yellow we are within 10 feet.

We are looking at New Jersey. Using the same technique we looked at 591 miles of streams. We found that 70% of this was within 40 feet, 40% within 20 feet, and 23% within 10 feet. So in other words, 30% of the steams were outside of the 40 foot buffer that we created.

So here’s the original NHD in green. We overlaid this on a contemporary photograph.

The orange indicated those sections where we are within 40 feet. The green indicates therefore that we are outside of 40 feet. So you can see how much of the stream is outside of 40 feet.

Purple indicates those sections that are within 20 feet.

And furthermore the yellow indicates those sections of the streams that were within 10 feet.

And there you can see the new blue revised stream against the original NHD stream, the multicolored line with those green segments. You can see that where the river is green it’s particularly far away from the new NHD, and more than 40 feet.

Here in Vermont we looked at 7,500 miles of streams, and found that 62% was within 40 feet. In other words 38% is outside of 40 feet. 38% within 20 feet, and 20% within 10 feet.

Here is an example of those streams in Vermont.
Yellow shows the new local resolution NHD and you can see there is a new tributary added to the network.
Red indicates the original NHD.
The blue segments are where we are within ten feet.

Furthermore the green indicates where we are within 20 feet. And orange indicates where we are within 40 feet. So those sections were we have deviations you will see in yellow.

Another issue that we have is dealing with Alaska.
The mapping in Alaska is a 1;63,360-scale and we have yet to carefully analyze the statistics in Alaska.
But for the rest of the United States, the conterminous United States, we have found that the typical offset of the NHD is somewhere with a mean of about 25 feet, a standard deviation of about 25 feet, a 90th percentile of about 50 feet, meaning that about 90 percent of the measurements fall within 50 feet. That varies a bit, but 50 feet is a general rule. The maximums are on the order of say 70, 80, 90, 100, 130 feet. But very few of those offsets. So that’s the conclusion of our study. We obviously need to do more work. We’ve looked at about 15,000 miles of streams to collect these statistics. There are seven and a half million miles of streams in the United States so we have just taken a sample. We need to look at more of the United States. More of the geomorphology conditions and see how the positional accuracy varies due to geomorphology. Also in different climatic conditions, streamflows, things like that. But this is an initial study that gives us some good initial results.
Thank you.