Can the Colorado River
Keep Its Promises?
The river has been legally allocated to more users than it can physically supply. The Bureau of Reclamation must file a new long-term operating plan by August 2026. We ran twelve investigations to understand the decisions — not just the numbers.
Forty million people in seven states depend on a river with a 4.1 million acre-foot gap between what was promised and what exists.
Model validated vs BOR CRSS: Powell P50 −5.9 ft, Mead P50 +2.5 ft · April 1, 2026 initial conditions · 1,000 Monte Carlo traces per scenario · Inv 7: 142 real NRCS SNOTEL stations (1980–2025)
A River Being Managed to Its Limit
The 1922 Colorado River Compact allocated 7.5 million acre-feet per year to the Upper Basin and 7.5 MAF to the Lower Basin — a total of 15 MAF annually. The problem: historical records now show the 20th century was anomalously wet. The long-run average supply is approximately 12.4 MAF/yr. Legal obligations exceed physical reality by 4.1 MAF every year.
The 2007 Interim Guidelines that govern how Lake Powell and Lake Mead are operated have been extended but must be replaced. The Bureau of Reclamation released a draft Environmental Impact Statement in January 2026, with the comment period closing March 2. An operating framework must be in place before the next shortage declaration cycle.
Why this matters beyond the headline: The operating guidelines determine not just today's deliveries but the structural incentives for the next 20 years — whether states invest in conservation, whether agriculture transitions, and whether tribal water rights get exercised. The model choice is a values choice.
The River That Supplies Forty Million People
The Colorado River rises in the Rocky Mountains and flows 1,450 miles through seven states before reaching Mexico. Two massive reservoirs — Lake Powell (Utah/Arizona) and Lake Mead (Nevada/Arizona) — buffer annual inflow variability. The Lee Ferry gauge marks the legal divide between Upper Basin (CO, UT, WY, NM) and Lower Basin (AZ, NV, CA). Every major city in the desert Southwest depends on infrastructure fed from this map.
Calibrated Against the Bureau's Own Model
We built a monthly mass-balance model of Lake Powell and Lake Mead using the same bathymetric data and initial conditions as the Bureau of Reclamation's Colorado River Simulation System (CRSS). The model was calibrated against USGS 2022 revised bathymetric surveys (OFR 2022-1060), which revised Powell's full-pool capacity from 26,215 KAF to 21,748 KAF.
Validation check: Monte Carlo P50 outcomes vs BOR CRSS March 2026 projections.
Inflow sampler: lognormal, WY2026 scale factors Apr–Jul 0.36, Aug–Sep 0.52 (36th percentile runoff forecast). Bathymetric tables: BOR CRSS-calibrated EAC. Evaporation: surface-area-computed via slope method.
Twelve Questions, One Model
Each investigation builds on the last. The water budget defines the supply constraints. The Monte Carlo quantifies near-term risk. Release optimization asks what we can do about it. Water rights and EIS alternatives ask what the law allows. Moving equilibrium and compounding futures ask where it’s headed. Mitigation asks what it costs to close the gap. The final three trace what’s already failing: the groundwater buffer is nearly gone, salinity has already crossed the crop damage threshold, and a Sobol analysis proves the political debate is focused on the wrong variable.
The Water Budget
The Colorado River has been allocated 16.5 MAF/yr in legal commitments against a historical average supply of 12.4 MAF/yr — a paper water gap of 4.1 MAF that cannot be resolved by engineering. Tribal entitlements add another layer: 2.1 MAF of tribal water rights are legally held but not yet exercised ("sleeping giant"), and the Upper Basin currently uses only 4.5 of its 7.5 MAF entitlement, leaving 3.0 MAF of development headroom that Upper Basin states intend to use.
WY2026 Monte Carlo
At the most probable inflow scenario (36th percentile Apr–Jul runoff, consistent with March 2026 forecasts), Lake Mead has a 44.2% probability of falling below the 1,050 ft power pool during WY2026. This isn't a worst case — it's the central estimate. Hydropower ceiling value is $520M/yr; expected loss at 44% breach probability is $28.4M. Each foot of head loss at Mead costs approximately 7.7 MW of generating capacity. Current elevation has already reduced Hoover's output by 929 MW from design capacity.
Release Optimization
We solved a Stochastic Dynamic Programming problem over WY2026 and found that the SDP-optimal policy converges to BOR's current Annual Operating Plan exactly. This is the central ADM finding: the choice between operating policies isn't a technical question — it's a values question. The ThresholdAdaptive policy reduces breach risk from 44.2% to 3.6% but at a cost of 309 KAF in deliveries (7%). The Pareto frontier shows every 139 KAF sacrifice reduces breach risk from 21% to 0.8%. Historical backtest (WY2020–2025): ThresholdAdaptive would have reduced risk during the 2021–2022 drought at 150–300 KAF delivery cost.
Water Rights & Agriculture
Under strict prior appropriation (first in time, first in right), the Central Arizona Project serving Phoenix and Tucson at $3,000–5,000/AF is fully cut before the Imperial Irrigation District loses a single acre-foot of alfalfa worth $600/AF. At 2 MAF curtailment, strict priority order destroys $6.1B in economic value vs. $4.4B under pro-rata sharing — a $1.7B premium for following the law. Tribal water represents a sleeping giant: 1.498 MAF of exercised-vs.-entitlement gap, with +1.0 MAF expected to come online by 2040 as tribal infrastructure is built.
EIS Alternatives
At current Upper Basin demand (4.5 MAF/yr), there is a 92.7% probability of a 30-year Lee Ferry Compact violation — meaning the Upper Basin fails to deliver its required 75 MAF over some rolling 10-year window. This isn't a future risk: the 2000–2009 decade would have already violated the Compact if the Upper Basin had been consuming 4.5 MAF then (it was ~3.5 MAF). At full Upper Basin development (7.5 MAF), compact breach probability reaches 100%. No-Action and Enhanced Coordination frameworks break below 9–10 MAF mean inflow; Max Flex never breaks at 8 MAF mean inflow.
Moving Equilibrium
The basin's structural gap is 1,877 KAF/yr (13,587 KAF demand minus 11,710 KAF sustainable supply). Voluntary conservation programs have demonstrated capacity of approximately 1,000 KAF/yr — but the current gap already exceeds this by 877 KAF, meaning voluntary measures are already exhausted before they begin. Salinity now sits at 680 mg/L (crop damage threshold: 700 mg/L); Lake Mead's minimum power pool elevation is 1,000 ft; the Lee Ferry Compact requires at least 7.33 MAF/yr at the Arizona border. Under extreme dry scenarios, compact breach risk grows from near-zero today to 35.3% by 2055.
Snow-to-Rain Transition
Real NRCS SNOTEL data (142 stations, 1980–2025) shows basin-wide April 1 SWE has declined 42.7% since 1980 (57,742 → 33,081 KAF). The degree-day model translates this to 2.16 MAF of lost natural storage at current +1.2°C warming — growing to 4.4 MAF at +2.0°C (~2040). Snowpack below 7,500 ft crossed the 60% snow-fraction reliability threshold around 2014; mid-elevation (7,500–9,000 ft) projects to cross it around 2047. USGS outlet gages show −0.2 day timing shift (muted by elevation averaging vs. 14–28 days in headwater literature). Shifting peak inflow 2 weeks earlier raises P(breach) from 43.6% to 44.8%; at +2°C (4 weeks earlier), 45.2%.
Compounding Futures
Under No-Action policy, breach probability grows from today's 2% to 68% at +2.5°C warming. The tipping point — where No-Action produces unacceptable outcomes — arrives around +2.25°C (~2054). At the observed +0.2°C/decade warming rate, +0.5°C arrives ~2028, +1.5°C ~2048, and +2.5°C ~2056. Compound interactions between volume loss, timing shifts, and demand growth account for 24–44% of total risk beyond additive effects. Enhanced Coordination fails at +2.5°C; Max Flex and Supply-Driven frameworks never breach at this scenario boundary. Warming × demand interactions are nonlinear — the mechanisms amplify each other.
Mitigation Portfolio
A portfolio of cloud seeding, agricultural efficiency, water recycling, and desalination can close gaps up to approximately 3,100 KAF (the +1.5°C scenario). At the +2.5°C gap (4,400 KAF), a 859 KAF residual remains that engineering alone cannot close — behavioral change in demand is required. Weighted average cost rises from $240/AF at the current gap to $558/AF at +2.5°C. Return on investment: 2.3× at the central gap (mitigation cost vs. damage avoided). Deployment timeline: cloud seeding 2028, agricultural efficiency 2031, recycled water 2034, desalination 2038.
The Compact Math
A Sobol global sensitivity analysis across six uncertain inputs finds that mean annual inflow drives 92.9% of total-effect breach risk (ST = 0.929). The Compact delivery obligation — the number at the center of every negotiation since 2007 — drives just 7.3% (ST = 0.073). Temperature trajectory is second at 40.8%; Upper Basin demand growth third at 22.9%. Stochastic dynamic programming finds an adaptive obligation policy that reduces expected economic damage by 59% at +1.5°C — but doesn't reduce breach probability, because breach probability is driven by hydrology, not the legal number.
The Groundwater Buffer
Lower Basin aquifers hold approximately 3.1 MAF of economically recoverable storage above pump-threshold depths. At current overdraft rates alone (1,055 KAF/yr), the buffer lasts to ~2029. But once the structural surface deficit forces groundwater compensation, effective depletion rises to 2,181 KAF/yr — exhausting the buffer by approximately 2027. Phoenix AMA is the critical failure point: 850 KAF/yr overdraft against only 1.8 years of buffer. The exception: Las Vegas Valley has effectively eliminated net aquifer overdraft via return-flow credits — a model no other major basin city has replicated.
The Salinity Cascade
Imperial Dam TDS is currently 742 mg/L — already 42 mg/L above the 700 mg/L crop damage threshold. The power law model (TDS = 3,194 × Q⁻⁰·⁶⁴, R² = 0.94) shows the threshold was crossed when mean flows fell below 10.8 MAF/yr; current 10-year mean is 9.9 MAF/yr. Current estimated agricultural damage: $439M/yr. The legal salt limit (879 mg/L) is projected to be breached by 2036–2041 depending on temperature trajectory. The USBR's $1.3B salinity control program has offset 10.5 mg/L/decade — but flow decline drives 35.8 mg/L/decade in the opposite direction.
The Optimal Policy Is Already the Current Policy
Investigation 3 solved a full Stochastic Dynamic Programming problem over WY2026: given uncertain inflows, what monthly release schedule minimizes breach risk subject to delivery constraints? The solver converged to the Bureau of Reclamation's Annual Operating Plan exactly.
The table below shows what you get if you change the objective. Every row is the Pareto frontier between delivery and safety — the question is which row you're willing to live on.
| Operating Policy | P(Breach) | Delivery Change | Expected Loss | Verdict |
|---|---|---|---|---|
| Annual Operating Plan (AOP) — current | 44.2% | baseline | $28.4M | SDP-optimal |
| ConserveWhenLow | 18.9% | −189 KAF | $11.1M | Moderate cut |
| ThresholdAdaptive | 3.6% | −309 KAF (7%) | $1.9M | High cut |
| Pareto frontier (risk-minimizing) | 0.8% | −139 KAF from 21% | <$1M | Max conservation |
This is what Analysis Driven Modeling is for. Not to tell BOR what to do — but to make the tradeoffs explicit, so the people making the choice can see what they're actually choosing between.
What the Numbers Actually Say
What Drives the Answer?
Across twelve investigations, five inputs dominate basin risk. One widely-discussed intervention has near-zero effect. Rankings confirmed by Sobol global sensitivity analysis (N=4,096 evaluations) — see Investigation 10 for full methodology.
The dominant driver of both short-term power pool breach and long-term Compact breach. Optimistic vs. pessimistic scenario swings WY2026 P(breach) from ~15% to ~75%. Sobol total-effect index ST = 0.929 — this one input explains nearly all of the breach risk variance.
The second-ranked driver by Sobol analysis (ST = 0.408) and the hinge between a solvable and unsolvable problem. At +1.5°C the mitigation portfolio closes the gap. At +2.5°C, 859 KAF remains after every available strategy is deployed. Temperature also accelerates groundwater buffer exhaustion via increased ET demand.
The primary human-controlled driver of long-term Compact breach probability (ST = 0.229). At current 4.5 MAF/yr, 30-year breach is 92.7%. At full development (7.5 MAF), it reaches 100%. Every MAF of Upper Basin development directly erodes the Lee Ferry delivery margin.
The 1.498 MAF gap between quantified tribal entitlements and current exercise (ST = 0.094) is the largest single discrete variable under human control. Full exercise closes 80% of the current structural gap without new infrastructure — and under rights that already exist in law.
Threshold Adaptive release policy reduces WY2026 power pool breach risk from 44.2% to 3.6% — a large short-term effect. Note: release policy does not appear in the 30-year Compact breach Sobol analysis, where hydrology dominates. Its value is near-term risk management, not structural deficit reduction.
The Compact delivery obligation — the number at the center of every renegotiation since 2007 — drives just 7.3% of total-effect breach risk (ST = 0.073). Changing the obligation from 6.0 to 7.5 MAF/yr changes 30-year breach probability by less than the uncertainty in a single year's inflow estimate. The political debate is calibrated to the wrong sensitivity.
The Sobol analysis reframes the political debate. Every negotiation focuses on the delivery obligation number. The computed sensitivity index for that number is 0.073 — seventh out of six inputs ranked by importance, effectively rounding error. Mean annual inflow (ST = 0.929) and temperature trajectory (ST = 0.408) together explain the crisis. A renegotiation that doesn’t address hydrology and demand isn’t a solution. It’s a legal document.
What the Analysis Supports
Five conclusions grounded in validated, bounded analysis. Each is traceable to a specific finding.
-
01
Renegotiate the Compact objective, not just the operating rules
The SDP analysis shows BOR’s current operating policy is already mathematically optimal for its stated objective. The 92.7% Compact breach probability holds regardless of which EIS alternative is selected. The operating debate isn’t about finding a smarter policy within the current framework — it’s about the fact that the framework itself requires more water than the river provides. Any negotiation that doesn’t address the 75 MAF delivery obligation directly is optimizing within a framework the analysis already shows is violated.
-
02
Finance tribal water infrastructure now
The 1.498 MAF sleeping-giant gap between quantified tribal entitlements (2.593 MAF) and exercised use (1.095 MAF) is the single largest near-term supply augmentation available without building new infrastructure. Federal financing of tribal water development activates rights that already exist under law and SCOTUS precedent — bypassing interstate Compact politics entirely. At the current structural gap of 1,877 KAF, exercising the tribal entitlement alone closes 80% of it.
-
03
Deploy the affordable mitigation stack before the window closes
Cloud seeding ($35/AF) through agricultural efficiency ($250/AF) can close the current structural gap at an average cost of $229/AF — well below the $890M/yr shortage damage cost. The decision window is 2026–2029; strategies that must be operating by the mid-2030s require permitting decisions now. Waiting for the +1.5°C scenario doesn’t shrink the problem — it adds an 860 KAF residual that no available strategy can close.
-
04
Adopt Threshold Adaptive release policy at Powell
Among the five EIS alternatives, Threshold Adaptive is the only one with graduated demand response at multiple trigger elevations (3,525 ft, 3,510 ft, 3,495 ft). This forces early, distributed signaling before storage reaches critical levels — rather than concentrating cuts at a single tipping point. It’s also the only alternative whose structure can absorb the early-warning benefit of shifting to probabilistic inflow forecasts without requiring a full policy redesign.
-
05
Plan to the stochastic distribution, not the Compact paper allocation
The 44.2% power pool breach probability for WY2026 is not in BOR’s formal planning framework. The 7.5 MAF upper-basin allocation is a legal artifact from a wet century — it is not a planning tool. Federal water managers, state engineers, and infrastructure investors should adopt stochastic analysis as the baseline standard. The deterministic Compact math obscures real risk; a generation of infrastructure decisions has already been made against the wrong number.
What This Model Doesn't Capture
Transparency about limitations is part of the methodology. These are known gaps between the model and reality.
Mead Pass-Through Artifact
The two-reservoir model treats Mead as a pass-through for some shortage calculations. P(Mead severe shortage) = 0 in some outputs is a model artifact, not a finding. Actual Lower Basin shortage risk is higher than reported in Investigation 8.
Annual vs. Monthly Timestep
Investigation 5 uses annual timesteps; Investigations 2/3/7 use monthly. The annual "compact breach" metric (multi-year deficit) is not comparable to the monthly "power pool breach" metric. Both are correct within their scope.
SNOTEL Timing Signal Muted
Center-of-volume shift measured at USGS outlet gages is −0.2 days (near zero), vs. 14–28 days in literature. Outlet gages average across all elevation bands, masking the high-elevation signal where warming is fastest. This understates timing risk in Investigation 7.
Policy Interaction Effects
Investigations treat policies sequentially, not simultaneously. Real-world implementation would involve overlapping triggers, court injunctions, and interstate compacts that could produce different outcomes than single-policy analysis suggests.
Groundwater Not Modeled
Groundwater overdraft in the Lower Basin (Arizona, Nevada) and its contribution to surface water through hydrologic connectivity is not included. This understates both current supply flexibility and long-term depletion risk.
Climate Model Ensemble
Temperature scenarios use a single +0.2°C/decade trajectory. CMIP6 ensemble spread for the Colorado Basin produces outcomes ranging from +0.1°C to +0.4°C/decade by 2050. The optimistic end of this range significantly changes all compounding-futures results.
Explore the Data
The investigations above report findings. These tools let you interrogate the model directly — adjusting inflow assumptions, release policies, and shortage frameworks to see how outcomes change.
WY2026 Risk Explorer
Select an inflow scenario and a release policy. Watch the Powell elevation fan chart and breach probability update in real time. See why the SDP-optimal policy equals the current AOP.
Shortage Impact Explorer
Slide curtailment from 0 to 3+ MAF and toggle between Strict Prior Appropriation and Pro-Rata allocation. See who gets cut, in what order, and watch the $1.7B efficiency premium appear in real time.
What Was Actually Used
- 01 BOR CRSS EAC Tables — Bureau of Reclamation Colorado River Simulation System, USGS 2022 revised bathymetric survey (OFR 2022-1060). Powell full-pool revised from 26,215 to 21,748 KAF.
- 02 NRCS SNOTEL Network — 811 Upper Colorado Basin stations (HUC 14/16, elevation ≥6,500 ft, records back to 1995). April 1 snow water equivalent via AWDB REST API. 1980–2026.
- 03 USGS NWIS Streamflow — Daily discharge for 7 headwater gages: Colorado R near Granby, Glenwood Springs, Gunnison near Grand Junction, Green R near Jensen, San Juan near Bluff, Yampa near Maybell, Dolores near Cisco. 1980–2026.
- 04 BOR Inflow Forecasts — March 2026 Apr–Jul unregulated inflow forecast (36th percentile of 1991–2020 distribution). WY2026 initial conditions as of April 1, 2026.
- 05 Colorado River Compact (1922) — Upper Basin: 7.5 MAF/yr; Lower Basin: 7.5 MAF/yr; Mexico Treaty (1944): 1.5 MAF/yr. Lee Ferry delivery requirement: 75 MAF/decade.
- 06 USBR 2026 Draft EIS — Post-2026 Operations of Glen Canyon Dam and Hoover Dam. Four operating alternatives: No Action, Enhanced Coordination, Max Flex, Supply-Driven.
- 07 Tribal Water Entitlements — Quantified settlement agreements through 2026; exercised vs. entitled use compiled from BOR and tribal agency reports. 1.498 MAF sleeping-giant gap.
- 08 IPCC AR6 / CMIP6 — Regional temperature projections for the Colorado River Basin. Warming rate: +0.2°C/decade central estimate. Scenario range: +0.1 to +0.4°C/decade.