Cadillac Desert,
Forty Years Later

A reservoir makes a promise visible. Water held behind a dam becomes next summer’s irrigation, a town’s drinking supply, a season of boating, or electricity when it is needed. When the reservoir falls, the consequences spread well beyond the shoreline.

A Wall Street Journal report by Christopher Kuo and Carl Churchill prompted this exploration. Their reporting describes how the Rio Grande’s decline is threatening farms and the economy of Elephant Butte, New Mexico. It brings a large environmental problem down to the level at which people experience it: a place, a livelihood, a future.¹

It also brought me back to Marc Reisner’s Cadillac Desert: The American West and Its Disappearing Water. First published in 1986, forty years ago, the book described the institutions and ambitions that transformed the West’s rivers. Reading it beside today’s water accounts, the unsettling question is how much of the present difficulty was already visible then.^{2, 3}

Start with the reserve

Powell and Mead dominate the chart because their inventories are much larger. Their decline after the turn of the century is difficult to miss. Periodic gains interrupt the fall, but the later reserves sit far below the high levels visible in the late 1990s.

Elephant Butte’s blue line is less prominent on this scale. That is a difference in size, not in consequence. A shortfall in a smaller reservoir can still be decisive for the communities and farms it serves. Its own history is easier to read on the separate scale in the appendix.

Shasta offers an essential comparison. Its record has repeated sharp drawdowns and strong recoveries, including the replenishment visible near the end of this window. The four lines do not justify treating the entire West as one reservoir, or declaring that recovery is impossible everywhere. They show why a broad regional problem still demands local water accounts.

Nor does a storage chart isolate causes. The level reflects weather, inflow, operating decisions, releases and other terms. These four case studies are not a statistical attribution of climate change or a census of all Western reservoirs. They are an invitation to examine the reserve and then ask what is passing through it.

What Reisner saw forty years ago

Reisner’s subject was the construction of a water-dependent society: rivers redirected, dams built, land brought into production, and cities able to expand because water could be delivered from somewhere else. His history followed political power and institutional competition as closely as concrete and canals.²

His warning was about the relationship between engineering success and the commitments it encouraged. The ability to store and transport water could make aridity seem less binding, while agricultural, urban and political claims accumulated around the resulting supply. A system could become impressive and vulnerable at the same time. The later scientific assessment published by Sabo and colleagues explicitly examined his concerns about scarcity, ecological stress and competition among users.³

That is the sense in which the book foresaw today’s issue. It identified a structure in which obligations could become harder to adjust than the supply supporting them. It did not need to predict the precise level of Elephant Butte in 2026 to recognize the danger of treating a limited reserve as evidence of unlimited security.

My reading of the present charts is that storage can postpone a reckoning long enough for more people to depend on the old arrangement. Replenishment is what makes that postponement useful. Without it, the buffer is gradually spent.

A dam can move water through time. It cannot make an enduring deficit disappear.

That observation is not an argument against reservoirs. Storage makes seasonally variable water more useful and can carry surplus into dry periods. The mistake would be to confuse that valuable service with an unlimited increase in dependable supply.

A hotter water account

The modern problem also includes a climate mechanism beyond the political history. Research by USGS scientists Christopher Milly and Krista Dunne links Colorado River flow reductions to warming, increased evapotranspiration, and the loss of reflective snow cover. Their model-based estimate was a roughly 9.3 percent reduction in annual mean discharge per degree Celsius of warming. That is a basin-specific research result, not a coefficient we have estimated from these four reservoir charts.⁶

The practical distinction matters. Low precipitation can reduce inflow. Higher temperatures can also change how much of the water falling on a basin reaches its rivers. Planning around a return of rainfall alone can therefore overlook an important part of future supply risk. The operating question is how much usable replenishment the system can reasonably rely on.

Elephant Butte: see the whole year

A falling line shows that the inventory changed. A Sankey chart asks how the year’s water account fits together. The width of each ribbon below is proportional to a calendar-year 2024 volume, measured in acre-feet. One acre-foot is enough water to cover an acre to a depth of one foot.

Reported inflow was approximately 438,501 acre-feet. Releases were about 649,847, and reported evaporation about 78,345. The reservoir’s recorded inventory fell from 466,360 to 186,701 acre-feet, a draw of 279,659. An additional unreconciled balance of about 10,033 acre-feet is needed to close this account.

This makes the central point concrete: the annual release was supported partly by water already in storage. Drawing that reserve is part of what reservoirs are designed to do. Repeating a draw without sufficient subsequent replenishment is what reduces the ability to do it again.

The diagram continues into Caballo delivery accounting because the downstream destinations matter. But the differences between reservoir releases, Caballo releases and listed delivery charges cannot simply be labeled physical losses. They are differences between accounting boundaries. Their causes need separate investigation.

There is even a small disagreement within the annual report: its listed delivery components sum to 530,884 acre-feet, while the printed total is 530,788. We use the components and leave the 96 acre-foot discrepancy visible. The larger point is that an attractive diagram should never conceal the uncertainties needed to make it balance.⁵

Agriculture: ask which share of which water

Agriculture has to be central to a serious discussion of Western water. It is also a subject on which a poorly defined percentage can mislead. Water diverted into a canal, water delivered to a district, and water consumed by a crop are different quantities. Some diverted water returns to the river system; water consumed through evapotranspiration does not remain immediately available for a downstream user.

A separate basin-wide study by Richter and colleagues provides useful context for the Colorado River. For its 2000–2019 accounting period, irrigated agriculture represented 52 percent of overall consumption and 74 percent of direct human consumption. Those percentages answer different questions because the broader denominator also includes natural vegetation and reservoir evaporation.^{7, 8}

Natural vegetation belongs in a physical water budget. That does not make a functioning river ecosystem wasteful. Likewise, identifying the largest human consumptive use does not settle who should bear the costs of adjustment. Farms produce food and support rural communities. Cities, Tribal Nations, downstream communities and river ecosystems also have interests that a volume chart cannot rank for us.

The narrower conclusion is that meaningful savings require evidence about consumptive use. A technology that reduces a farm’s diversion does not automatically save the same amount for the basin if it also reduces return flows. A conservation claim should specify where the saved water remains available, for whom, and for how long.

Our reservoir study does not estimate an agricultural percentage for Elephant Butte, Powell, Mead or Shasta. The downstream delivery branches are a starting point for that research, not its completion. Applying the Colorado study’s percentage to a Rio Grande release would substitute an assumption for a measurement.

The reserve below the surface

A reservoir is visible. Groundwater is harder to see, but it cannot safely be treated as an unrelated backup supply. USGS research explains how pumping can intercept water that would otherwise feed a stream, or induce water to move from a stream into an aquifer. The location and timing of that effect depend on the connected groundwater system.⁹

That connection complicates a simple surface-water success story. Replacing a reduced river delivery with additional pumping may preserve production in the short run while shifting part of the cost into a less visible account. Measuring reservoir recovery without examining connected groundwater would leave an important question unanswered.

The decisions are already here

This is no longer a distant planning exercise. Reclamation’s post-2026 process produced a decision framework for coordinated Powell and Mead operations and the first guidelines for operating years 2027 and 2028, issued on August 21, 2026. Its current account of the process states that consensus on long-term operations was not reached.¹⁰

Better operating rules matter. So do realistic supply assumptions, credible conservation measurement, and arrangements that recognize who bears the costs. The choice is not simply between farms and cities. It concerns which uses remain dependable under a smaller supply, how adjustments are shared, and what is left for the river itself.

There is room for adaptation: changes in crops and irrigation practices, reuse, changes in urban demand, and operating decisions that preserve flexibility. Each proposal should be judged against the same physical test: does it reduce depletion or improve dependable supply at the boundary that matters? A promise to save water is useful only if the water account supports it.

Forty years after Cadillac Desert, the strongest reason to revisit Reisner is not to award him a prediction score. It is to recognize the institutional problem he made visible. A society can become dependent on a water arrangement long before it agrees on how to change that arrangement.

The four reservoir histories are an effective place to begin. The ribbons add the annual flows. The next step is to bring those physical accounts into the decisions about consumption, livelihoods and growth. The water stored behind a dam is a reserve. Its replenishment is a condition of the promises built around it.

Appendix: charts, water accounts and replication

The material below supports the article without treating every reservoir as the same system. Figures are SVG for clear enlargement. Each Sankey has its own width scale; compare the numeric volumes, not ribbon thickness across separate figures.

A. Individual storage histories

B. Cumulative inflow and releases

C. The remaining annual Sankeys

D. Calendar-year 2024 water accounts

*Residual = inflow − releases − evaporation − change in storage. A negative number is an unexplained input requirement; a positive number is an unexplained output. Inflow is operationally reported and may be computed. Mead’s gross inflow, evaporation and residual are unavailable in the selected record, rather than zero.

The full physical balance can include precipitation on the reservoir, bank storage, groundwater exchange, diversions and other terms as appropriate. The displayed residual captures what is not reconciled by the selected series; it does not identify a particular omitted process.

E. Item IDs, dates and transformations

Annual flows cover January 1–December 31, 2024, with 366 daily observations in every selected flow series. Storage change uses December 31, 2023 and December 31, 2024 endpoints. Daily cfs values are multiplied by 86,400/43,560, approximately 1.98347, before summing. RISE’s Shasta evaporation export labels aggregation as sum despite cfs units; the daily equivalent-rate interpretation follows the operations tables. Negative reported computed flows are retained.

The historical charts cover January 1, 1986–December 31, 2024 and reproduce the plotting points returned by RainbowStats. The storage CSV files contain those chart points, not a claim to include every daily historical record. The archived annual exports contain the daily observations used for the 2024 accounts. Recent and historical source values may be revised. No interpolation fills missing dates in the annual checks.

The four reservoirs are case studies, not independent observations of a common river. Powell releases can contribute to Mead inflow; summing them would double-count water passing through the connected system. Acre-foot inventories are not standardized percentages of usable capacity.

F. Download the data and reproduce the analysis

Open the working RainbowStats project. The saved script and chart output in this package preserve the version used to prepare these figures; some text slides retain development notes from that saved project. The charts and calculations were successfully rerun for this article.

• Annual water accounts — CSV
• Working RainbowStats script — TXT
• Returned chart specifications and plotted values — JSON

Source exports, storage chart points and Sankey links:

Keep the assets and data folders alongside index.html when publishing so the download links resolve. All figures are also embedded directly in the article.

Sources and further reading

1. Christopher Kuo and Carl Churchill, “A Disappearing Rio Grande Is Reshaping Life Along the River,” The Wall Street Journal, October 21, 2026.
2. Marc Reisner, Cadillac Desert: The American West and Its Disappearing Water. First published 1986; publisher page describes the revised edition.
3. Sabo and colleagues, “Reclaiming freshwater sustainability in the Cadillac Desert,” PNAS; USGS publication record. A scientific assessment of Reisner’s themes.
4. Bureau of Reclamation, RISE: daily operational series and metadata for the four reservoirs. Analysis window ends December 31, 2024.
5. Reclamation, Calendar Year 2024 Report to the Rio Grande Compact Commission, Table 14, report page 59: Caballo releases and delivery charges.
6. Milly and Dunne (2020), “Colorado River flow dwindles as warming-driven loss of reflective snow energizes evaporation,” Science; USGS publication record.
7. Richter and colleagues (2024), “New water accounting reveals why the Colorado River no longer reaches the sea,” Communications Earth & Environment. Consumption accounting for 2000–2019.
8. NASA Landsat, “Meat of the Matter: Colorado River Over-Consumed,” April 30, 2024. Explains and illustrates the Richter study.
9. Barlow and Leake (2012), Streamflow Depletion by Wells—Understanding and Managing the Effects of Groundwater Pumping on Streamflow, USGS Circular 1376.
10. Reclamation, Colorado River Post-2026 Operations: decision framework and 2027–2028 operating guidelines issued August 21, 2026.
11. Reclamation, Shasta Lake Daily Operations, January 2025: daily rate units and definition of computed inflow.