How the two lakes filled and emptied
One water year, day by day, as the reconstruction runs it — no dam, no canals, both natural lakes in place. Snow and rain arrive at the top. Upper Klamath Lake spreads them over roughly 145 square miles, so it fills slowly and spills over the Link River reef only once it is up. Lower Klamath Lake below is about the same size again, 142 square miles of open water and marsh, and loses water to the air the whole time. When it rises past 4,085 ft it spills in two places at once.
Storage comes from Reclamation’s current curve with Agency Lake and the Barnes units of the refuge reconnected; outflow from the pre-dam Koppen rating over the Link River reef, which reproduces the paired stage and discharge in Reclamation’s own 1915 records to within 1.5% on the annual mean — 1,847 cfs against 1,819 observed — and to between 88% and 116% in every month of that year. The two bars below the lakes are running totals since 1 October: everything that has arrived, and everything that has left past Keno. The gap between them is water still in the lakes or already gone to the air. All elevations on this page are feet on the USBRKB datum — Reclamation’s Klamath Basin datum, the same vertical reference as the USGS lake gage at Upper Klamath Lake and as Reclamation’s stage-storage tables. It is not NAVD88; figures taken from this page should not be compared with elevations on another datum without conversion.
Change in Upper Klamath Lake stage through the water year
Hydrograph — reconstructed natural outflow at Keno
Flow-duration curve — Keno, this water year
What this animation is, and what it is not
It is a counterfactual, and deliberately so. It runs one water year through the two natural lakes as if the Klamath Reclamation Project had never been built — no canals, no Link River Dam, no Keno Dam, the Link River reef uncut and the Keno reef in place. That assumption is false on its face. The dams exist, and the Project exists. The point is not to pretend otherwise but to establish a baseline: what the basin did on its own, so that what it does now can be measured against something.
The water arriving at the top is today’s water, not 1905’s. This is the part that matters most and is easiest to miss. The reconstruction is driven by modern net inflow, which already carries the cumulative effects of a century of diversions — Fourmile Creek sent out of the basin entirely into the Rogue, upper-basin irrigation on the Wood, Williamson and Sprague that did not exist in 1905, and a denser forest after a century of fire suppression. None of those are the Klamath Project. None are within the control of the Project, of Reclamation, of any federal authority, of Klamath Irrigation District, or of any other district. Every one of them shows up here as less water at the top of the animation, and therefore as a lower natural lake and a smaller natural outflow at Keno than the basin would have produced with those diversions absent.
So this is a conservative baseline. A reconstruction driven by depleted inflow understates the natural condition. Whatever gap it shows between the natural lake and the regulated one is the smaller version of that gap, not the larger.
What is deliberately left out. Nothing here concerns the adjudicated storage right, the volume Reclamation may store, or who holds the right to use that water. Those belong to the regulated lake and are held for a second animation.
What this baseline isolates, and what it does not attempt
This is not a “but for” analysis. It shows the effects of the Klamath Reclamation Project in isolation, against a state of nature. That is a different exercise from the one ESA Section 7 consultation and a biological assessment require. Section 7 measures the effects of a federal action against an environmental baseline that already contains past and ongoing non-federal activity. This reconstruction makes no such partition. It strips the basin back to bare ground and asks what the two lakes would have done.
So it leaves out everything that pre-dated the Project or would have arrived without it, including:
· Van Brimmer Ditch Company — adjudicated priority 4 September 1883, 50.0 cfs
· the Ankeny-Henley, or Steele-Ankeny, Canal — adjudicated priority 21 March 1884, begun as the Linkville Water Ditch in 1878
· the Adams ditch from Lost River, later lengthened and renamed the District’s D Canal
· the enlargement of the Ankeny canal and tunnel, already under way when Reclamation arrived
· hydroelectric development on the Link River — the Moore Brothers’ works, and Link River Dam itself, which the power company offered to build
· the growth of forestry in the upper basin
· the growth of the town of Klamath Falls
And that cuts against the District, deliberately. Stripping all of it out means the reconstruction charges the Project with depletions that pre-date it or would have happened without it. A true but-for baseline would leave those diversions standing, and the share attributable to the Project would be smaller than the gap this animation implies. The baseline is drawn against the District’s own interest on purpose — a conclusion that survives a baseline built this way is not resting on a favourable assumption.
Checking the Slough against Reclamation’s own figure
Reclamation’s Natural Flow Study Part 2 put the average loss to the Lost River Slough at about 6,000 acre-feet a year. This model gives nothing in the dry year, about 700 AF in the median year and about 17,600 AF in the wet one. Set against 6,000 those look wrong in both directions — until you notice that 6,000 is a mean and the median year is not the mean.
A threshold spill produces a badly skewed distribution. The Slough only runs when the lower lake is already above the 4,085 ft sill. In the dry year the lake peaks at 4,084.89 and misses it by an inch and a half, so the Slough never opens at all; in the median year the lake is above the sill on 27% of the model steps, and in the wet year on 53%. Most years therefore contribute nothing and a minority contribute a great deal. Weighting these three years as a quarter dry, half median, a quarter wet gives 4,700 AF; as equal thirds it gives 6,100 AF. Reclamation’s 6,000 sits inside that range. For further scale, 6,000 AF is 0.49% of Part 2’s own 1,233,000 AF average natural year, and the wet year here is 1.0% of its own inflow.
What the answer is actually sensitive to is the sill, not the rating. Doubling or halving the rating coefficient simply doubles or halves the volume. Moving the sill does something far worse:
sill elevation dry median wet weighted mean
4,084.6 ft 65 5,086 45,173 13,852
4,084.8 ft 1 2,126 28,897 8,287
4,085.0 ft 0 687 17,568 4,735 ← as modelled
4,085.2 ft 0 137 9,916 2,548
4,085.4 ft 0 7 5,040 1,264
acre-feet per water year; weighted mean is a quarter dry, half median, a quarter wet
Eight tenths of a foot moves the answer by a factor of eleven. So the figure to defend is the sill elevation. It was left at 4,085.0 ft, which is where the historical descriptions put it — and it happens to be the value that reproduces Reclamation’s mean. The sill was not adjusted to achieve that. Tuning a parameter to hit a single reported average would turn an independent check into a circular one.
How much weight these numbers will bear
Three ratings do the work, and they are not equally well founded. Say so plainly, because the difference matters.
· Upper Klamath Lake, over the Link River reef. Koppen’s pre-dam rating, 392.6 × (stage − 4,138.5)1.712. This one is tested against independent data: Reclamation kept daily records of both lake stage and Link River discharge in 1915, seven years before the dam, and the rating reproduces them to 1.5% on the annual mean and 88–116% monthly. That is a real validation.
· Lower Klamath Lake, over the Keno reef. Q = 371 × (stage − 4,083.1)2.016, fitted to exactly two gauged points — 8,400 cfs at 4,087.8 ft in June 1904, the maximum of record, and 730 cfs at 4,084.2 ft in October 1908, the minimum (Reclamation KBAO, History of Keno Reef, 21 November 2019). Two points fix a two-parameter power law exactly: there is no residual, no degrees of freedom, and no way to test it. It is the best that survives, not a validated curve.
· The Lost River Slough sill. A threshold spill above 4,085 ft. The sill elevation, not the rating, governs the answer — four tenths of a foot changes the annual volume by more than a factor of ten.
The figures are printed at the model’s arithmetic resolution, not at the accuracy of its inputs. Net inflow is itself a residual of a lake water balance and carries error of order tens of cfs on any given day. Read the annual volumes to about three significant figures and the stages to about a tenth of a foot. A total shown as 1,009,939 AF means roughly 1.01 million acre-feet; it does not mean the model resolves single acre-feet.
What “dry”, “median” and “wet” mean here. They are labels for three selected water years, not exceedance percentiles. Against Reclamation’s own average natural year of 1,233,000 AF, the three deliver 578,000 AF (47%), 1,010,000 AF (82%) and 1,734,000 AF (141%). So the median year sits somewhat below the long-run average rather than at it, and the three span roughly a threefold range.
What the inches mean
They are a change in stage, not rainfall. The inches on this page are the change in Upper Klamath Lake stage since 1 October — what a staff gauge driven into the shore would have read, measured against where it read at the start of the water year. Nothing more complicated than that. A reading of +24.7 in means the water stands a little over two feet higher than it did on 1 October. The figure is signed: it falls back through the summer as the lake draws down, and it would read negative if the lake dropped below its opening stage.
Where the water comes from. Snow and rain fall across the 3,810 square miles that drain to the Link River, and a fraction of it reaches the lake as the inflow shown at the top of the animation. Most of what falls never arrives — it is taken up by forest and range, returned to the air, or recharged to groundwater.
Two rain gauges, and they do not agree. The animation now carries both: Klamath Falls at the top, where the water enters the system, and Keno over the lower lake, at the basin’s outlet. Each shows the running total for the water year, that day’s own rain or snow, and the year’s final figure. Across the wet year Klamath Falls records 15.31" and Keno 27.23"; across the median year, 11.77" against 18.04"; across the dry year, 7.46" against 10.66". Keno takes roughly half again as much, every year, because it sits nearer the foot of the Cascades while Klamath Falls sits in the valley’s rain shadow. Two gauges twenty miles apart, in the same basin, on the same days. That spread is the point: precipitation is not one number for the Klamath, and neither gauge tells you what the lake will get. The water that fills Upper Klamath Lake falls higher still — the Cascade Range and the interior and eastern uplands average more than 30" a year, and Crater Lake about 65".
Where the gauge records come from. Keno is station USC00354403, with complete daily coverage in all three water years — every day present, nothing filled in. Klamath Falls is Klamath Falls International Airport, station USW00094236, for the wet and median years, at 365 of 365 and 366 of 366 days. The airport has no record between WY1972 and WY1997, so the dry year comes from the town co-op gauge, Klamath Falls 2 SSW, at 360 of 366 days. Both totals are almost certainly low. Standard unshielded co-op gauges undercatch snow badly — commonly by 10 to 30% in wind — and in this basin most of the winter precipitation falls as snow. The Keno figure and the Klamath Falls figure are both biased in the same direction, so the gap between them is more reliable than either total, but part of even that gap may be catch efficiency rather than orography.
The inflow figure is Reclamation’s net inflow. It is computed as the residual of the lake’s own water balance, so evaporation off the lake surface has already been subtracted and rain falling directly on the lake has already been added. In a hot dry summer a daily net inflow figure can go negative. None of the three years shown here dips that far; the driest reading on the dry year is 280 cfs.
Why every year starts at the same elevation
Each run opens on 1 October with Upper Klamath Lake at 4,139.46 ft and Lower Klamath Lake at 4,084.01 ft, and it opens there whether you select the dry, median or wet year. There are two reasons, and the second is the important one.
It isolates the hydrology. If each year began at a different stage, you could not tell whether a higher spring peak came from that year’s snow and rain or from where the lake happened to start. Holding the opening fixed means every difference on the screen is the water, not the initial condition.
It does not change the answer. The opening was tested, not assumed. Run the median year from six different starting stages spread across nearly five feet — 4,138.60, 4,139.46, 4,140.50, 4,141.50, 4,142.50 and 4,143.50 ft — and every one of them lands at 4,140.07 ft on 30 September. Carry that ending elevation forward into a second year and a third and nothing moves again: three wet years in a row peak at 4,142.85 ft each time, five dry years in a row at 4,140.63 ft. Running a year warm, opening from the elevation the previous year left behind, changes its peak by a hundredth of a foot.
The reason is the reef. Koppen’s pre-dam rating rises as the 1.712 power of stage above 4,138.5 ft, so a lake standing a foot higher does not release a little more water — it releases a great deal more. A surplus drains off over the winter and a deficit fills, and either way the lake arrives where that year’s inflow and the natural outlet agree it should be. The natural lake carries almost no memory from one year to the next, which is the sharpest difference between it and a reservoir operated to a storage target.
The lower lake behaves the same way, and its opening was computed rather than chosen. Run through WY2025 on Reclamation’s daily record, every trial opening between 4,083.1 and 4,085.0 ft converges to 4,084.01 ft by 30 September. That is what Jacobs (1906) and Voorhees (1913) both reported from observation — the Keno reef holding the lower lake near 4,084 ft even at low water.
The water arriving at the top has been falling, and not because of the Project. Fourmile Creek has been diverted out of the Klamath basin entirely since the 1920s, through the Cascade Canal into the Rogue — USGS station 365 records it discharging into a lava bed above Fish Lake. A century of fire suppression has left the upper-basin forest far denser, so more precipitation is caught in the canopy and transpired before it ever reaches a stream. And the Wood, Williamson and Sprague now carry irrigation that did not exist in 1905. None of these are the Klamath Project, and none are within the Project’s power to change — but every one of them shows up here as less water at the top of the animation.
An observation, offered as a question rather than a finding. The natural lake and marsh system returned an enormous volume to the air — roughly 528,000 AF a year across both lakes and their marshes. Traditional knowledge of this basin holds that water lifted from the marshes came back to the mountains, and that draining them dried the country above as well as below. The arithmetic sets a ceiling on how much: spread over the 3,810 square mile upper watershed, all of that moisture would be about 2.6 inches of rain a year, and a realistic local recycling fraction perhaps half an inch. Small against 25 to 40 inches in the mountains — but arriving in the growing season, when the lakes evaporated hardest and the country is driest. It is not something this model can test, and it is not counted anywhere in these figures.
The red rule is what the lake used to reach. In April 1904, before any dam, high-water marks put Upper Klamath Lake at 4,144.98 ft — a month when 64.9% of Klamath County stood in the most extreme wet class on the national drought record. The highest reading in fifty-one years of the modern regulated gage is 4,143.44 ft, in Water Year 2000, also a wet year and 1.54 ft lower. The grey rule marks 4,143.3 ft, about as high as Link River Dam has held the lake since 1922, shown here only so the two eras can be compared. It is a limit on the regulated lake, and the reconstruction is not held to it: the capacity curve driving this model runs to 4,145.0 ft, so the natural lake here is free to rise past it if the water arrives to do it. In these three years it does not — the wet year peaks at 4,142.68 ft, just under. The space between the grey rule and the red one is not a limit of the model but a real gap: stage the lake reached in 1904 that no year in the gaged record has matched.
Every ribbon is drawn to the same scale — a ribbon twice as wide is carrying twice the water, whether it is the inflow at the top, the spill over the Link River reef, the Lost River Slough or the outflow at Keno. The Slough is barely a thread even at its peak of 186 cfs, against 6,309 for inflow in a wet spring; that is the correct proportion, not a drawing choice.
What the Slough actually costs. The readout now carries the running volume out the Lost River Slough as well as the rate. Over the whole water year it comes to nothing at all in the dry year, about 700 AF in the median year and about 17,600 AF in the wet one — about 1.0% of that year’s inflow at its largest, and 0.07% in the median year. The Slough is a spill, not a loss of consequence: it opens only when the lower lake is already above 4,085 ft, and in a dry year the lake never gets there. Volumes are integrated by trapezoid on the model’s own step from the slough discharge, so they carry the same uncertainty as the stage that drives them.
Watch the Slough, and watch the gap. The Slough runs for months in a wet year, opens briefly in the spring in a median one, and in a dry year never opens at all — the lower lake never reaches 4,085 ft and everything that leaves goes over the Keno reef. Meanwhile the two running totals tell the other half of the story: they separate through winter and spring as the lakes fill, and close again through summer as the water is given back or evaporates. In the median year about 1,010,000 AF arrives and about 777,000 AF reaches Keno — the difference is what the natural lakes consumed. By September, in every year, the lower lake has fallen back to spilling at Keno alone.