As some of you know, most of my life for the last two years has revolved around predicting the locations of small headwater streams that are not captured by the venerable National Hydrography Dataset. The NHD is a fantastic resource overall, but it does have its shortcomings, particularly in small, ephemeral channels. Usually, the limitations are fairly predictable, as they result from the intersection of limits of scale, subjective cartographic choices, and deliberate design rules that make a map good for one purpose but not necessarily others. Sometimes, those shortomings manifest themselves in unexpected ways.
While preparing my final channels, I noticed an interesting problem. In some areas, there seemed to be short channel segments that did not connect downstream. While this can happen in nature, this should not be possible with the particular process I used to create channels.
Since I was using a somewhat convoluted procedure to reconcile my predicted streams with the existing NHD data (we wanted to extend it, not replace it), my first thought was that I had made some sort of mistake. Just to be sure, I looked at the existing NHD channels, and the gaps were still there. To get to the bottom of this, I pulled in some georeferenced USGS topographic maps for the area. These maps are the source of the NHD data, so I was curious to see if maybe some lines had not been digitize properly or if there may be an on the ground explanation. However, these discontinuous channels were there on the original topo maps. This was getting stranger. Why would a cartographer draw a channel, then fail to connect it downstream? It was while pondering this that I noticed the seam on the map. These channels are stopping right where two topographic maps come together.
To really understand what’s going on here, we need understand where topo maps come from and the rules the USGS cartographers use to draw these streams. Topo maps are drawn by hand using paired, stereo aerial images. The cartographer can’t draw anything on the map if it’s not visible on the imagery. In particular, a channel is only drawn if there is water in it at the time the image was taken or if it is obviously a channel. So if the imagery used to draw each map was taken on different dates, the channels may not have had water in them in one image whereas they may have been flowing in another. Additionaly, if different cartographers worked on each map, one cartographer may have noticed the channel or deemed it worthy of inclusion, whereas the other may not have. Even within the limits of the accuracy standards and the thorough list of decision rules that cartographers must use when mapping, there is still some room for subjectivity. It’s not too surprising that the variation in this subjectivity is going to show up where you try to match the edges of two maps. What I find particularly surprising is how this was not corrected when digitizing the NHD from the topographic maps.
While these errors/discrepancies are a little jarring at times, on the whole USGS topographic maps and the NHD are an excellent resource. While they may not capture every feature and the occasional inaccuracy may crop up, the comprehensive coverage of the continental US at 1:24,000 scale is an incredible achievement. When creating a project of this size, it’s inevitable that errors such as this make it in and are missed in QA/QC. It’s all part of the process of modeling and mapping our world, one step at a time. Just remember to double-check your data.