Streams of Tennessee

Longest Stream Segments Between Confluences

All of the longest stream segments between confluences occur in areas of karst topography, interestingly enough. The implication is that there are actually other streams and confluences, but are underground as cave streams and springs, neither of which would appear in the National Hydrographic Dataset. Part of this

1) Calfkiller River†, White County, 12.12 miles
2) Collins River, Warren County, 10.69 miles
3) Falling Water River, Putnam & White County, 10.59 miles
4) Blackwater Creek, Hancock County, 10.23 miles
5) Duck River, Maury & Marshall County, 10.12 miles
6) Big Sycamore Creek, Claiborne & Hancock County, 10.03 miles
7) Hurricane Creek, Rutherford & Wilson County, 10.02 miles
8) Happy Creek, Sevier County, 9.53 miles
9) East Rock Creek, Marshall County, 8.82 miles
10) War Creek, Hawkins County, 8.56 miles

†Calfkiller River is more likely 13.23 miles since the "confluence" measured is actually an "island" where the flow is shown to branch around either side. The Calfkiller River is pictured above.

Longest Named Rivers in Tennessee

Who decides what water course is named what for what stretch? Early geographers? Modern geographers? The people? Some combination of all of the above. Whatever the reason, here's the data. The length is the clipped length within the state of Tennessee and may not reflect the full length of the river.

1) Tennessee River, 374.14 miles
2) Cumberland River, 310.01 miles
3) Duck River, 276.94 miles
4) Hatchie River, 196.58 miles
5) Clinch River, 176.81 miles
6) Elk River, 167.50 miles
7) Caney Fork, 143.05 miles
8) Holston River, 136.08 miles
9) Buffalo River, 129.00 miles
10) Harpeth River, 117.15 miles

Shortest Named Rivers in Tennessee

Someone somewhere thought it was a good idea to name these short streams. I provide this list as a curiosity.

1) Barley Creek, Campbell County, 373'
2) Sal Branch, Jackson County, 449.37'
3) Middle Fork Gulf Fork Big Creek, Cocke County, 802'
4) Bible Creek, Blount County, 976'
5) Carringer Branch, Monroe County, 1185'
6) Walnut Log Ditch, Obion County, 1696'
7) Lyles Creek, Trousdale County, 1945'
8) Arcadia Mills Bayou, Shelby County, 2347'
9) Cemetery Bayou, Shelby County, 2595'
10) Green Island Cutoff, Lake County, 2786'

Loading Data into Locus Pro

Every TAG caver knows that Locus Pro is the best GPS program for you mobile device (sorry iPhone users it's only for Android). Locus Pro easily handles the ten thousand plus data points for the Tennessee Cave Survey's dataset of cave entrances, and can simultaneously display several other hefty datasets atop it. I'm running a Pixel 3 and it runs quite smoothly.



So you've got a dataset, maybe it's the new TCS distribution, or maybe it's some other KML or GPX file and you want to load it to Locus Pro. Here's how you do that. We start with the screen above. My layout may be different than yours, so if it's unfamiliar don't sweat it. We're looking for the Data menu. Above I've circled its icon in red. You can also get to this menu by pressing the three bars in the upper left and going to Points.



Your Data menu should look something like the above, perhaps without all the lines and the weird mosaic blur (don't ask).



I can see an already existing TCS dataset. Let's see what year it is by hitting the three vertical dots and selecting edit.



This is 2018 cave data. We can just delete this since the dataset we're about to load is new and improved. Hit cancel and return to the previous menu.



Hit the three vertical dots next to the folder and choose delete.



Let's load our data now. From the Data menu hit the blue plus sign in the bottom right.



Choose Import.



Wherever your data is (I don't know, don't ask me!), navigate to it. Mine is in a folder on my Google Drive which I have linked to Locus Pro. I've found the 2019 TCS KMZ file and I will select it.



It's asking you where you want this data to go. I like to put different data into different folders. This lets me toggle off and on the data that I need, when I need it. This keeps my map free of unnecessary clutter, and keeps my data organized. I am going to select Add New Folder.



Above are the settings I use for import.



This step is super important for TCS data users. Be sure to check Apply to all and hit ignore. This will allow multiple entries to be loaded with the same name. In other words, Indian Cave won't be overwritten as each county is loaded leaving you with only the last entry.

As a final step I would suggest you update your icons so your symbology doesn't clash on your maps. Hope this worked alright for you!

Spring Creek Gorge

Spring Creek Gorge begins at Upper Waterloo Falls (36.302529,-85.475847) and ends at the confluence of Spring Creek and Roaring River. It's a rugged stretch of river best negotiated via canoe or kayak when the water is high enough to run. In other parts of the year, people access the gorge by alternating hiking and floating on inner tubes. Numerous cascades and waterfalls can be found along the course of Spring Creek as it drops down into the Nashville Basin.



The map above shows in detail the path of Spring Creek, a designated State Scenic River, and the locations of its many waterfalls.

Access to the gorge is through private property. An access easement appears to still exist, though access to Upper Waterloo Falls is now prohibited unless one comes in from Waterloo Rd before the bridge. The access easement hinges on respect of the land. Periodically it's necessary to clean up after the idiots who party there and trash the parking area and around Waterloo Falls.

Heading downstream from Waterloo Road along Spring Creek, one first encounters Upper Waterloo Falls. This feature is smaller than several unnamed features along the course of the river, but earned a name given its proximity to the road.



Waterloo Falls (36.302807,-85.467413) is next. It's a wide 35' drop at the contact of the Fort Payne and Chattanooga Shale (like many other local waterfalls).



A small 5' drop (36.304800,-85.466451) is encountered just before a sharp oxbow bend. From this vantage one can see up to Waterloo Falls, and down the gorge a short distance.



The next drop is at Meat Grinder Falls (or if you're a kayaker, just Meatgrinder; 36.304448,-85.468762). It is a nice cascade which follows a joint at 45 degrees to the river.



Mill Creek Falls (36.302529,-85.475847) pours in on the left dropping 54'. It's a stunning, and challenging waterfall to get to. When it is flowing best, one must negotiate whitewater in a kayak, canoe, or other such device.



More waterfalls await the daring adventurer, but many are wet weather features best photographed during high water conditions. It is strongly inadvisable that one visit the gorge during high water without the right gear (a boat), and experience. Nearby Blackburn Fork and Cummins Falls have had several fatalities as a result of water rising quickly in a narrow canyon just like that of Spring Creek Gorge.

Creating a Noise Profile and using it as a Mask in Photoshop

I have a bunch of old photographs which are very noisy, but are still of interest to me. In the reprocessing of these images with better tools and a better eye I've developed a simple way to create a mask to apply noise smoothing to a portion of the image.

My goal is to create the following image. Click the image to load its Flickr page and click the image again to magnify it. It's not a perfect image, but the noisy parts of the image have been smoothed, and the parts of the image which aren't noisy haven't been, which allows them to retain their detail.

What I am calling the noise profile is simply the difference between the original image and a noise reduced version of the image with a little bit of modification. If you would like to follow along, find an image of your own with noise in it. I'm using the latest version of Adobe Photoshop CS for the PC (10/25/2019).

First, let's load our image. Here's what mine looks like.

I usually start by hitting the auto button and then fine tuning the image from there. Let's see what auto does.

Let's take a look around our images and find some places with lots of noise. In the lower right of my image, in the dim portions

That's an issue. I'm going to make note of where this is and come back to it later. For now I'm going to do some spot exposure adjustments to darken some areas and maybe lighten a few other areas a wee bit. Here's what I've got now.

Go ahead an open your image so you can use the standard Adobe Photoshop interface. I always do an auto-contrast to snap my highs and lows into place and balance out the middle luminosity values. Next, let's duplicate the original layer by clicking and dragging Background down to the Create New Layer button in the Layer toolbar.

You should have two layers now. Make sure Background copy is active, and let's bring it back into the Camera Raw dialog by clicking the Filter menu and selecting Camera Raw Filter. Choose the details menu in the right pane and crank Luminance to 100 and Color to 50. Click OK.

Now your Background copy layer is noise smoothed, and your original isn't. The next step is to figure out where the noise is programmatically. Fortunately some basic raster math options are available in Photoshop. Change the overlay method on the right to difference and it should make your image black. This is what we want to happen. Right now we're asking "What's the difference between the regular image and the noise smoothed image?" It turns out that the answer at this resolution is difficult to see. We'll fix that in just a moment.

Choose the square marque tool from the toolbar on the left and hit ctrl-a to select the entire image. You should have a marquee around the entire image. Let's use the copy merged command from the edit toolbar to copy the combination of these layers into the clipboard. Immediately paste the resulting layer into the workspace. It will appear as Layer 1.

Navigate to Image > Adjustments > Desaturate. The image > auto-contrast. Now you should see something like this.

Next, navigate to Filter > Blur > Gaussian Blur and set the radius to 12 pixels. Hit OK, and do another auto-contrast.

From this image we can create our mask. Make sure your primary color is set to black, then navigate to Select > Color Range and and set the fuzziness to 200.

Select Background copy now as your active layer. Toggle off Layer 1, and return the overlay method of Background copy to Normal. Your marquee selection should still be active.

Hit ctrl-x to remove the selected marquee. If we turn off the background layer then you can see what your noise mask looks like on the image.

Toggle the visibility back on for the background layer and flatten the image. Finish with another auto-contrast and you have your noise smoothed image. Oh, let's check that original area's noise now.

Not perfect, but a far cry better than the original. I hope you found this useful. If you use this, let me know in the comments below. :)

Big Tennessee Sinkholes

Tennessee has perhaps millions of sinkholes, most of which are very small features. Based on data analysis from a 2013 sinkhole study (Dunigan/Sutherland 2013) I provide here a list of some of the largest features by dimension in the state. Below you'll find an interactive map, a table with this information, and photographs of some of the sinkholes.









* No official place name, named for nearest GNIS feature if nearby.
** Multiple possible names for feature. Made best guess.

Excluded from these lists is Montlake in Hamilton County. It may fit on the volume or depth list, but I cannot easily get at that information since it is flooded. It is a truly spectacular feature. More about Montlake here.

Delineating Sinkholes in GIS

A sinkhole, also known as a cenote, sink, sink-hole, swallet, swallow hole, or doline (the different terms for sinkholes are often used interchangeably), is a depression or hole in the ground caused by some form of natural collapse of the surface layer. Most are caused by karst processes – the chemical dissolution of carbonate rocks or suffosion processes. Sinkholes vary in size from 1 to 600 m (3.3 to 2,000 ft) both in diameter and depth, and vary in form from soil-lined bowls to bedrock-edged chasms. Sinkholes may form gradually or suddenly, and are found worldwide.

Above language is borrowed and slightly modified from the Wikipedia page for Sinkholes.

In order to have a tutorial showing my methods for delineating sinkholes, I'll need to define some vocabulary. This vocabulary isn't intended to replace any existing definitions so much as to clarify the context in which I'll be using them.
Sinkhole: A closed depression through which water can only escape through the subsurface whose area is defined as the upper most closed elevation contour of the feature.
Swallet: The point where water disappears into the subsurface. This could also be a cave entrance.

For the purpose of this tutorial I have chosen a portion of Warren County, Tennessee. If you would like to follow along, I've created an area of interest feature class and have set it to be the extent which I will be working in. This will only allow my commands to affect that region, thus saving time by not processing the entire county. The area is centered at 35.688251, -85.818811 (nearly at the confluence of Oakland Branch and Willow Branch) and my scale is set to 1:8000. LiDAR elevation data can be found here under the folder WarrenCo_2015_16_DEM_by_Tiles.

Conceptually, it's pretty easy to figure out how to create a layer with sinkholes. It should be the difference between the DEM and a filled DEM, right? This is a good start, but as you will see, it requires some fine tuning.

First, let's look at what our end goal is. We want to produce a map that looks like this.


Let's begin by loading the elevation into the data frame and using the fill tool on it.


We will get a filled sink version of the DEM.


Setup your raster calculation to look like this and execute the command.


Your output will be the difference between the filled and regular DEM. This layer is useful for a few steps later on, so be sure you've named it something meaningful where you can get back to it easily should you need it. This is what mine looks like.


For the next step we'll simplify the raster layer so that we are just showing where the sinkholes are. To do this I'll use the command in raster calculator Con("sinkhole01" > 0,1).


The resultant layer which shows the frightening number of sinkholes present should look like this.


Clearly that's a ridiculous amount of data, most of which is probably noise, and/or simply isn't useful. We'll want to filter that. But first, let's convert the data to vector using the Raster to Polygon tool.


We end up with a polygon layer that looks like this.


Let's first filter out all the sinkholes that are less than 500 square feet.


That certainly removes a lot of smaller, trivial features. You should have something that looks like this.


There are still lots of false positives in this batch. We could manually edit them with as few as we have, or we could prepare ourselves for a larger dataset that cannot be easily manually filtered (that said, you're going to manually filter these no matter what). You'll note that natural sinkholes tend to be more round. Many features visible here are long and thin, or have a strait line associated with them. Some sinkholes have been bisected or crossed by roads, so we have to be careful to not remove real sinkholes. This is where it gets tricky.

Let's invoke a third party Add-in for ArcMap called Easy Calculate 10. It will help us easily calculate thickness - which is the ratio of area of the polygon to the area of a minimum bounding circle (which can be done in ArcMap, but is a multi-step process). As thickness approaches 1 the feature approaches a circular shape.


There's no telling what the right thickness is, but with a little play I came up with the below number. The goal is to not remove any real sinkholes.


As you can see there are still a number of false positives. Other filters can be used by collecting data from the original DEM or from sinkholes01 via Zonal Statistics, and applying filters as shown in the previous steps. Other filter possibilities could include depth (range in Zonal Statistics).

I'll choose to manually edit the sinkhole layer from this point. Clearly the highlighted features following roads and man made features like the pond won't make the cut.


If you look carefully you can eliminate most of the features down to just the ones I have shown. Once you're here, you're ready to make a map.