Gage Height Is Measured. The Streamflow Beside It Is Read Off a Curve.

Two numbers sit side by side on a United States Geological Survey streamgage page: gage height in feet, and discharge in cubic feet per second. On 19 March 2026 the gage on the Mississippi River at Baton Rouge published a daily mean gage height of 24.66 feet and a daily mean discharge of 601,000 cubic feet per second. On 12 May 2026, fifty-four days later, it published 24.66 feet and 601,000 cubic feet per second again.

In between, the river had dropped more than eleven feet and climbed back. The pair repeated anyway, and it repeated exactly, because only one of the two numbers is a reading. The other is what a stored relation returns when that reading is entered into it. USGS documents this plainly. It rarely survives the trip into how the numbers get read outside the agency.

The Chain

1. Water stands in the gage pool at the station. The agency's guidance on interpreting the two values says gage height "refers to the elevation of the water surface in the specific pool at the streamgaging station, not along the entire stream," and that it "also does not refer to the depth of the stream."
2. A recorder logs that surface against the gage datum. "Measurements of gage height are continually recorded by equipment inside a gagehouse on the streambank."
3. That recorded height is published as gage height. It is the measured column.
4. A shift may be added to the recorded height first. The shift is a correction to the height, applied before anything is computed from it.
5. The height is entered into a stage-discharge rating. "For each streamgaging station, a relation between gage height and streamflow is determined by simultaneous measurements of gage height and streamflow over the natural range of flows (from very low flows to floods). This relation provides the real-time streamflow data from that station. All real-time data are provisional and subject to revision."
6. What the rating returns is published as discharge. It is the computed column.
7. A crew measures discharge directly, about every six weeks. Each new measurement is folded back into the evaluation of the relation, and USGS states that this "may lead to changes in the provisional discharges that had been computed for previous months."

What the Height Is Measured Against

A gage height is not an elevation and it is not a depth. It is a distance above a reference point chosen for that one gage. USGS explains the choice in its answer to why gage height is not measured from the streambed: "Sometimes streams erode, becoming deeper, while other times streams deposit sediment, becoming shallower. To understand how those changes can affect streamflow, we need to measure from a reference point that does not change over time, which is why gage datums are chosen to be below the base of a streambed. This means that the gage datum is a uniquely selected reference point for each gage site."

Uniquely selected means the zero is local. The station page for USGS 01646500, the Potomac River near Washington at Little Falls, states that the "Datum of gage is 37.04 ft above North American Vertical Datum of 1988." A reading of 10.00 feet there is a water surface at 47.04 feet on that national datum. Ten feet at the next gage upstream is a different water surface at a different elevation, and the two readings are not a comparison of anything. The number answers one question only: how far above this gage's own zero the pool stood.

The second limit in that definition is the one a reader on a riverbank runs into. Gage height describes the pool at the station. It does not describe the reach a mile downstream where the channel narrows, or the tributary behind a neighbourhood, or the storm drain that backed up while the main stem sat well under its threshold. A gage reports the surface it can see, at the one place it stands, against a zero chosen for that place alone.

Where the Second Number Comes From

The network is large and jointly paid for. The USGS says the National Streamgaging Network "is a multipurpose network that comprises more than 12,165 streamgages," and that "The streamgages are primarily operated and maintained by the USGS, but most are funded in partnership with one or more of about 1,500 Federal, State, local, and Tribal agencies or organizations."

Discharge is a volume rate, and measuring it takes people in the water or over it. The same USGS guidance is specific about the interval: "Measurements of streamflow are made about every six weeks by USGS personnel; they wade into the stream to make the measurement or do so from a boat, bridge, or cableway over the stream."

Those visits do not produce the daily numbers. They produce the relation that produces the daily numbers. USGS Circular 1123 describes the construction: "Measurements made over the range in stage of the stream are plotted against the corresponding stages to define the stage-discharge relation that is used in conjunction with the recorded stage record to determine the discharges throughout the year." How often a station is revisited depends on whether that relation holds. Where the channel is stable, "periodic measurements are made every 6 to 8 weeks to verify that the relation has not undergone some unrecognized change." Where it is not, "frequent measurements (about weekly) are needed to define how the rating is changing and to define its present condition."

The relation itself is not permanent either. The USGS answer to "How is a rating curve used to convert gage height into streamflow?" says: "It's developed by making frequent manual discrete discharge measurements at stream gaging stations. The rating curve depends on the physical characteristics of the stream channel and floodplain, and will vary over time at almost every station." Between visits, every published discharge at that station is the rating applied to a recorded height.

What 119 Published Pairs Look Like

The Mississippi River at Baton Rouge is USGS site 07374000. Its daily values for 1 February through 31 May 2026, parameters 00065 for gage height and 00060 for discharge, come to 119 pairs; no daily value is published for 2 April. Gage height across the window runs from 6.64 to 25.79 feet and discharge from 207,000 to 625,000 cubic feet per second.

Five gage heights in that window occur on more than one date. In all five, the discharge published beside them is identical to the digit.

• 17.24 feet on 1 April and again on 13 April: 443,000 cubic feet per second both times.
• 18.96 feet on 15 April and again on 30 May, forty-five days apart: 480,000 both times.
• 19.34 feet on 20 April and again on 18 May: 488,000 both times.
• 22.01 feet on 4 May and again on 6 May: 545,000 both times.
• 24.66 feet on 19 March and again on 12 May: 601,000 both times.

The same height was entered into the same relation, and the same value came back out. The agreement is a property of the lookup, not evidence about the river.

How much water a foot of stage stands for is also readable from the same file, and it is not constant. Near the bottom of the window, between 6.64 feet on 17 February and 9.13 feet on 9 February, the published pairs move 61,000 cubic feet per second across 2.49 feet, about 24,500 per foot. Near the top, between 23.14 feet on 14 May and 25.79 feet on 23 March, they move 56,000 across 2.65 feet, about 21,100 per foot. The relation flattens slightly as the river rises here; at another station it may steepen sharply.

The point extends to the whole window. Fitting one straight line to all 119 pairs leaves 0.023 percent of the variation in the discharge column unexplained. That is not a fact about the Mississippi. It is a fact about the arithmetic that produced the column.

The Correction Is Applied to the Height

When the channel changes, the relation stops matching the river, and the first repair is not to the flow number. It is to the height that goes into the lookup. The USGS Wyoming-Montana Water Science Center's page on shift-adjusted ratings defines the correction: "Some measurements indicate a temporary change in the rating, often due to a change in the streambed (for example, erosion or deposition) or growth of riparian vegetation. Such changes are called shifts; they may indicate a short- or long-term change in the rating for the gage." It then states how the correction is used: "In normal usage, the measured shifts (or corrections) are applied mathematically to a defined rating."

The correction is not one number applied across the range. The published rating file for 01646500, retrieved 2026-08-12, carries a shift record beginning 11 August 2026 with three breakpoints: 0.00 feet of shift at a stage of 6.00 feet, 0.20 feet at 10.00 feet, and 0.00 feet again at 13.00 feet. The middle of the range is being corrected and the ends are not.

The same file records why the rating was rebuilt in the first place, in the operator's own words: "Eliminating crossover shift after gage pool filled in during rain event at the beginning of April 2024." Sediment moved into the pool the recorder sits in, and the relation between the height in that pool and the flow past the station stopped being the relation the table encoded. The file also carries its own warning: "Stage-discharge ratings change over time as the channel features that control the relation between stage and discharge vary. Users are cautioned to consider carefully the applicability of this rating before using it for decisions that concern personal or public safety or operational consequences."

How Long a Number Keeps Its Flag

Every value arrives with a letter beside it. The daily-values files define them in their own header: "A Approved for publication -- Processing and review completed." and "e Value has been estimated." and "P Provisional data subject to revision."

Approval is not a formality that clears in days. Circular 1123 puts it on an annual cycle: "At the end of the year, all measurements are available for review. The entire set of measurements are used to reevaluate the rating conditions for the year. Final decisions are made about the stage-discharge relation that were in effect during the year and the record is refined or recomputed as necessary. This record is then passed through a rigorous review process and, once approved, the data are considered final and are placed in the archives and published."

The boundary is visible in the record. Queried on 2026-09-03, the daily discharge series for 01646500 carried A through 2026-04-06 and P from 2026-04-07 onward: roughly five months of published numbers still marked as subject to revision. The estimated flag shows up too. In that same series the values from 25 January through 18 February 2026 carried A:e, and 19 February returned to plain A. Every one of the 119 Baton Rouge pairs above carried P.

The Threshold Sits on the Measured Column

Flood thresholds are set on the column that is recorded, not the one that is computed. The National Weather Service forecast point for the Mississippi River at Baton Rouge, identifier BTRL1, publishes four categories as stages in feet: action at 30, minor flood at 35, moderate at 38, major at 40. The record crest in the same record is 47.28 feet, dated 15 May 1927. No flow value is attached to any of the four.

So the quantity that decides whether a place is under a flood category is the one a sensor reads. The quantity that gets quoted in coverage, cubic feet per second, is the one derived from it. Both are useful and USGS says the derived one travels better: "Streamflow values are better indicators than gage height of conditions along the whole river." That is a statement about which number describes the river, not about which one was observed.

Where This Doesn't Apply

None of the above says the published discharge is wrong. It says the discharge column is a computed column, and that its errors are the errors of a fitted relation rather than of a bad sensor.

• The field measurements themselves are direct. A crew wading, or working from a bridge or cableway, is measuring the flow. The gap is the interval between visits, not the visits.
• A rating that was just checked is not a stale rating. On a channel that is stable, a check every 6 to 8 weeks is what Circular 1123 describes as sufficient to verify the relation has not changed unrecognized.
• Gage height can be wrong too. It is a recorded value, not an exempt one; the A:e flag exists precisely because some days have no usable record and the figure is estimated instead.
• Revision is the system working. A value that shifts when it moves from P to A has not failed; the flag is there to record which stage of that process the value is in.
• The forecast point and the streamgage are separate records. In the BTRL1 record the vertical datum field is empty and no USGS site identifier is filled in, so a stage in one system should not be assumed to be the gage height in the other without checking.
• Ratings differ by station. The near-straight line at 07374000 over 6.64 to 25.79 feet is one reach in one window and says nothing about the shape of the relation anywhere else.

What Would Change These Numbers

Four inputs govern everything above. The gage datum is fixed per site and published with the station; at 01646500 it is 37.04 feet on the 1988 vertical datum. The rating is rebuilt when the channel makes it wrong, and its identifier and remarks change when it is. The shift moves whenever a measurement says the recorded height needs correcting, and its three breakpoints move with it. The flag moves when the record is published, and USGS puts that on the water year: "Each station record is considered PROVISIONAL until the data are published" and "The data are usually published within 6 months of the end of the water year."

Only the first is stable. Across the streamgaging network, the second number on a station page is answering a narrower question than it looks like it is answering: not how much water went past today, but how much water the relation in force that day assigns to the height the recorder wrote down.