On all three of the float creeks nearest Louisville and the Bluegrass, the median 8 September flow is below the flow the Kentucky Department of Fish and Wildlife Resources itself labels "good." Not marginally on two of them. Elkhorn Creek runs at roughly two-thirds of its floor. Floyds Fork runs at about a fourteenth of its.
Both halves of that comparison are published, by two different agencies, and as far as we can tell nobody puts them side by side. KDFWR's stream pages carry a low/good/high band in feet and cubic feet per second for each water it profiles. USGS publishes a median daily discharge for every calendar date on every gauge those pages point to. One is a recommendation; the other is 24 to 85 years of arithmetic. Together they answer the only question that matters in the first week of September: is the float I have in mind a float, or a walk with a boat attached?
The numbers
Every figure in this table is taken from the source named in it. The right-hand column is our arithmetic on their numbers.
| Water | Gauge | KDFWR "good" flow | KDFWR "good" stage | USGS median, 8 Sep | Record | Median as % of floor |
|---|---|---|---|---|---|---|
| Elkhorn Creek | Near Frankfort, 03289500 | 100–600 cfs | 2.5–4.0 ft | 64 cfs | 85 water years | 64% |
| Upper Salt River | At Glensboro, 03295400 | 130–400 cfs | 3.7–4.5 ft | 17 cfs | 24 water years | 13% |
| Floyds Fork | At Fisherville, 03298000 | 50–300 cfs | 1.3–2.5 ft | 3.6 cfs | 80 water years | 7% |
Read the last column slowly. To reach the bottom of its own recommended window on a median 8 September, the Upper Salt needs 7.6 times the water it typically has, and Floyds Fork needs 13.9 times.
This is a statistic about a typical September, not a forecast for this one. A wet week moves any of these gauges an order of magnitude. The point is not that the creeks are unfishable today — it is that the default condition of early September on Kentucky's profiled float streams is below the agency's own floatable threshold, and that planning a September float as though it were an October float is planning against the median.
Elkhorn is the outlier in the useful direction. At 64% of its floor it is the only one of the three that is arguably in the game on an ordinary September day, and that is worth knowing if you have a Saturday and a two-hour driving radius.
Why Elkhorn holds water and Floyds Fork doesn't
Raw discharge flatters big watersheds. Normalise it and the spread gets wider, not narrower.
Our arithmetic, dividing each median above by the drainage area USGS publishes for that gauge:
| Water | Drainage area | Median 8 Sep | Unit yield |
|---|---|---|---|
| Elkhorn Creek | 473 sq mi | 64 cfs | 0.135 cfs per sq mi |
| Upper Salt River | 172 sq mi | 17 cfs | 0.099 cfs per sq mi |
| Floyds Fork | 138 sq mi | 3.6 cfs | 0.026 cfs per sq mi |
Per square mile of watershed, Elkhorn delivers about 5.2 times the early-September baseflow Floyds Fork does. Floyds Fork is not simply a smaller creek having a smaller September. It is a materially drier one.
Here I am reasoning rather than reporting: that gap is what baseflow looks like when it has somewhere to come from. September flow in a Kentucky creek is not runoff — there has been no meaningful runoff for weeks — it is groundwater draining back out of the rock. Elkhorn drains Inner Bluegrass limestone with the conduit-and-spring plumbing that goes with it. Floyds Fork drains ground that holds far less and gives it back faster. We have found no published statement attributing the difference to that, and we are not going to pretend otherwise, but the mechanism is the obvious candidate and the ratio is not subtle.
The practical version: in a dry September, Elkhorn degrades gracefully and Floyds Fork does not. If you are choosing between them on the strength of a rain event three days ago, that difference decides it.
KDFWR's windows are not set by runoff
One more thing falls out of normalising, and it inverts the intuition.
Expressed as unit yield, the floors of KDFWR's own windows are wildly different: Elkhorn's 100 cfs is 0.21 cfs per square mile, Floyds Fork's 50 cfs is 0.36, and the Upper Salt's 130 cfs is 0.76 — about 3.6 times Elkhorn's requirement, from a watershed less than half the size.
That is not an inconsistency on KDFWR's part. Read as reasoning rather than reported fact, it is a description of channel shape. A wide, shallow, gravel-floored reach needs a great deal of water spread thin before it floats a loaded boat. A narrower channel that steps between bedrock pools floats at far less, because most of the water is already in the pools. The Upper Salt is the first kind of river and Elkhorn behaves more like the second.
The stage bands say the same thing from the other direction. The width of KDFWR's "good" window, in feet of gauge height:
- Elkhorn Creek — 1.5 ft (2.5 to 4.0)
- Floyds Fork — 1.2 ft (1.3 to 2.5)
- Upper Salt River — 0.8 ft (3.7 to 4.5)
Less than ten inches of stage separates "too low to float" from "too high to fish" on the Upper Salt. That is a fine target to hit on a river that responds to rain, and it is the strongest argument on this page for checking the gauge the morning of rather than the night before.
What this changes about the float you pick
Nothing above tells you where the fish are. It tells you how far you can reasonably expect to travel, and that is the variable September actually controls.
KDFWR publishes the access-to-access mileages, which makes low-water planning arithmetic rather than guesswork. From their own tables:
Elkhorn Creek — 4.8 mi (AMA on the Creek → VPA #3), 2.6 mi (VPA #3 → American Whitewater), 1.3 mi (American Whitewater → T.N. Sullivan WMA), 5.8 mi (T.N. Sullivan → Peter W. Pfeiffer Hatchery), 3.5 mi (Pfeiffer Hatchery → Still Water Campground).
Floyds Fork — 2.1 mi (North Beckley → Creekside), 4.3 mi (Creekside → Fisherville), 4.2 mi (Fisherville → Cane Run), 3.1 mi (Cane Run → Seaton Valley), 3.4 mi (Seaton Valley → Broad Run Valley), 2.1 mi (Broad Run Valley → Cliffside).
Upper Salt River — 3.1 mi (Dry Branch Road → Salt River Ramp 3), 30.5 mi (Salt River Ramp 3 → Four Roses Distillery), 3.8 mi (Four Roses → Rice Road Ford), 7.5 mi (Rice Road Ford → Drydock Road).
The 30.5-mile leg on the Upper Salt is the number to sit with. At a median 8 September flow of 17 cfs, that is not a long day — it is a committed multi-day trip through the shallowest conditions of the year with no intermediate exit. Below the floatable window, the shortest legs are the only honest options: 1.3 miles on Elkhorn, 2.1 on Floyds Fork, 3.1 on the Upper Salt. A one-mile creek reach fished properly at low water, wading between pools and covering each one from below, will out-produce five miles of dragging.
KDFWR's own caveat on the Elkhorn page is worth quoting because it is the correct framing: "These ratings are only recommendations. Skill and experience will dictate levels appropriate for each angler." The bands are a boatability guide, not a fishing forecast. Plenty of the best wade fishing on these creeks happens in water KDFWR calls "low." What you cannot do in water KDFWR calls low is cover distance.
The hazard that low water makes worse
Two low-head dams on the Upper Salt, both documented by KDFWR on its own access-site descriptions: one directly downstream of the Salt River Ramp 3 access, and one half a mile downstream of the Four Roses Distillery access.
Low water does not make a low-head dam safer. It makes it quieter, shorter and easier to misread from upstream, right up until the hydraulic on the far side is the one thing on the river still holding a full head of water. Know which side of each one you are launching on, and portage on sight rather than on assessment.
Then, on 15 October, the state splits in two
The other half of the Kentucky fall water story runs on a schedule instead of a weather forecast.
The Corps of Engineers' Louisville District holds every one of its Kentucky lakes at summer pool through early October, then draws them down between 15 October and 1 December — Buckhorn from 782.0 to 757.0 ft, Barren River from 552.0 to 528.0, Nolin River from 515.0 to 492.0, Rough River from its restricted 490.0 to 470.0. Taylorsville is the exception, holding until 15 November and then dropping just two feet, from 547.0 to 545.0. We covered the full schedule and what it does to each lake in the fall drawdown article.
That storage leaves through the dams. Which means that from the middle of October, Kentucky offers two entirely different kinds of moving water:
- Free-flowing creeks — Elkhorn, Floyds Fork, the Upper Salt above the lake. Flow is a rain lottery. September's median is the floor; October's is whatever the weather decides.
- Corps tailwaters — below Barren, Nolin, Rough River, Buckhorn. Six weeks of sustained, scheduled discharge, published in advance, largely indifferent to rainfall.
For six weeks a year, the tailwaters are the more predictable water and the creeks are the gamble. That is the reverse of the rest of the calendar, and it is worth building an October around. The Upper Salt is the interesting hybrid: free-flowing above Taylorsville Lake, which is the one Corps lake on the list that spends October at full summer pool.
Gear, briefly
Low water is a hull problem before it is a tackle problem.
Hull and protection. Every yard of gravel you drag over comes out of the keel. A skid plate or keel guard on the stern is cheap insurance in September and irrelevant in April. If your boat has a rudder, plan on lifting it for the whole float rather than remembering to lift it.
Stakeout over anchor. In water shallow enough to make a 3.1-mile leg a project, an anchor and rope is the wrong tool. A stakeout pole holds you off a pool tail without dragging a chain across the exact fish you came for, and doubles as the pole you push off gravel with.
Footwear, and expect to use it. At these flows you will be out of the boat repeatedly and standing on wet limestone and algae-slicked gravel. Studded soles, and a PFD you are willing to wear while wading rather than one you leave clipped to the seat.
Downsize the terminal end, not the rod. Low, clear September water on a smallmouth creek is a leader-and-presentation problem. What it is not is a reason to leave the rod that can turn a fish before it reaches a logjam — the fish get smaller-minded at low water, the wood does not.
A thermometer earns its place all fall. Flow tells you whether you can float. Temperature tells you what the fish are doing when you get there, and it is the one variable on this page you can measure yourself for less than the cost of a pack of jigheads.
What we don't know yet
Five genuinely open items, and we would rather list them than paper over them.
1. Our medians are for a single date. USGS publishes the median for the current calendar day on each gauge's conditions page; the full daily-statistics table would not load for us. Every median above is specifically 8 September — not a September average, and not a whole-month figure. The seasonal comparison we would like to publish, September against April on the same gauge, is status pending until we can retrieve the statistics series.
2. The records are not the same length. Elkhorn's median rests on 85 water years, Floyds Fork's on 80, and the Upper Salt's on only 24. A 24-year record ending in 2026 is not the same statistical object as an 85-year record starting in 1915, and the Salt figure should carry proportionally more doubt.
3. One gauge cannot describe 44.9 river miles. KDFWR ties the Upper Salt's window to the Glensboro gauge, but the access sites it lists span 44.9 river miles with tributary inflow between them. A reading that is accurate at Glensboro is an inference everywhere else on that run.
4. KDFWR's Elkhorn mileages don't reconcile with its own summary. The page states the creek "provides 13 miles of adventure" from the Forks to the Kentucky River, and then lists five access-to-access legs totalling 18.0 miles. We have published the individual legs, because those are the operational numbers, and we are not confident enough to tell you which figure is wrong.
5. The Corps schedule is still labelled 2024. The Louisville District's fill and drawdown page remains titled "Louisville District Reservoirs 2024 Proposed Fill and Drawdown Schedules", and still carries a banner saying the site was to be decommissioned on 15 July 2024. The dates and elevations above are the most recent published figures, not a commitment for this year.
We have deliberately not put current flows on this page. A live reading is wrong the day after it is written, and the whole argument here is about medians rather than mornings. The gauges are below — check yours.
Check it yourself
- Elkhorn Creek near Frankfort (03289500): https://waterdata.usgs.gov/monitoring-location/USGS-03289500/
- Salt River at Glensboro (03295400): https://waterdata.usgs.gov/monitoring-location/USGS-03295400/
- Floyds Fork at Fisherville (03298000): https://waterdata.usgs.gov/monitoring-location/USGS-03298000/
- USGS Kentucky current streamflow, all gauges: https://waterdata.usgs.gov/ky/nwis/current/?type=flow
- USACE Louisville District daily lake report: https://www.lrl-wc.usace.army.mil/reports/lkreport.html
Sources
Recommended river levels, access sites and mileages
- KDFWR, Elkhorn Creek — the 2.5–4.0 ft / 100–600 cfs bands, the access-site table, the 13-mile statement and the "skill and experience" quotation: fw.ky.gov/Fish/pages/elkhorn-creek.aspx
- KDFWR, Upper Salt River — the 3.7–4.5 ft / 130–400 cfs bands, the Glensboro gauge reference, the access mileages and both low-head dam locations: fw.ky.gov/Fish/Pages/Upper_Salt_River.aspx
- KDFWR, Floyds Fork — the 1.3–2.5 ft / 50–300 cfs bands and the access mileages: fw.ky.gov/Fish/pages/floyds-fork.aspx
- KDFWR, Kentucky's Stream Fisheries — the index of the 21 streams the department profiles this way: fw.ky.gov/Fish/Pages/Kentucky's-Stream-Fisheries.aspx
Median flows, drainage areas and periods of record
- USGS, Elkhorn Creek near Frankfort, KY (03289500) — median daily discharge for 8 September of 64 cfs on 85 water years; drainage area 473 sq mi: waterdata.usgs.gov/monitoring-location/USGS-03289500/
- USGS, Salt River at Glensboro, KY (03295400) — median daily discharge for 8 September of 17 cfs on 24 water years; drainage area 172 sq mi: waterdata.usgs.gov/monitoring-location/USGS-03295400/
- USGS, Floyds Fork at Fisherville, KY (03298000) — median daily discharge for 8 September of 3.6 cfs on 80 water years; drainage area 138 sq mi: waterdata.usgs.gov/monitoring-location/USGS-03298000/
Drawdown schedule
- USACE Louisville District, Louisville District Reservoirs 2024 Proposed Fill and Drawdown Schedules — the 15 October to 1 December window and the summer and winter pool elevations quoted: lrl.usace.army.mil/Missions/Civil-Works/Water-Information/Fill-and-Drawdown/
Unit yields, percentages and multiples in this article are our arithmetic on the agencies' published figures, not numbers either agency publishes. KDFWR's level bands are recommendations, not rules, and USGS medians describe a typical year rather than this one. Low-head dams are the single most lethal feature on any of these three rivers and they do not become safer at low water. Check the gauge the morning you go.