I spent five years on cargo ships, and the one thing you learn about structures at sea is that the ocean doesn't care what the engineer's specs say. The specs tell you what the hull was designed to take. The ocean tells you what it's going to give. When those two numbers diverge, you don't get a memo. You get a noise.
In structural engineering, rubble is honest. Rubble has achieved its final load state. The dangerous building on any block is the one still standing with a fracture in the bearing wall that somebody covered with a fresh coat of paint. People are living in it. They've been living in it for years. They trust it because it hasn't fallen, and they have confused this observation with a structural assessment.
What follows is an inspection of five American systems still standing as of late May 2026. Each one is carrying loads it was never engineered for. Everything described here is operational. People are inside all of them.
I. The U.S. Forest Service Enters Fire Season at 92% of a Number It Says Isn't Enough
The Forest Service was built to manage 193 million acres of national forest through seasonal fire suppression, research, recreation, and stewardship. Fire seasons historically ran June through October in the West. The agency staffed accordingly.
Through April 2026, 1.8 million acres had burned, running 194% above the ten-year average. AccuWeather projects 5.5 to 8 million for the full year. The NIFC seasonal outlook shows above-normal significant fire potential across the West, Plains, South, and Southeast all at once.
The agency lost roughly 5,900 employees through buyouts, layoffs, and early retirements, about 16% of its workforce. Separately, 1,400 personnel holding "red cards," the operational certification required to work active fireline, departed in 2025. Fifty-seven research stations closed. As of May 11, 10,496 wildland firefighters were on board against a stated goal of 11,300.
That's 92% of target. The agency's own workforce page acknowledges that even full staffing "isn't enough capacity."
Grassroots Wildland Firefighters, an advocacy organization representing frontline personnel, reports morale is "terrible right now" and that firefighters are "already tired, and summer's just starting." The USFS's official page, meanwhile, projects organized hiring windows and competitive pay under the 2025 permanent pay reform.
Both statements can be true. The new pay scale has made federal wildland firefighting more financially competitive. California's CAL FIRE operates independently with its own workforce. Congressional members pressed the agency at a January 21 hearing on its hiring plans; the Forest Service committed on February 12 to hiring 2,000 seasonal workers, and the system is not empty. Holding, here, means something specific: those 2,000 seasonals arriving before July, the pay reform translating into sustained application rates rather than a single-year bump, and the red-card pipeline refilling. That pipeline is the fracture line. Red-card holders are the people who walk toward fire. You can post a hiring window for a position. Experience doesn't come on a hiring window. Rebuilding fireline-certified capacity takes years, and the season that's already running double-pace doesn't wait.
II. Residual Insurance Markets, or a Safety Net Doing Floor Duty
State residual insurance markets, the FAIR Plans and Citizens corporations, were designed as insurers of last resort. A safety net for properties the private market wouldn't touch. They were supposed to be small, temporary, and expensive enough to push policyholders back toward private carriers. The whole architecture assumed the net would rarely bear weight.
California's FAIR Plan grew from approximately 140,000 policies in 2018 to over 610,000 by mid-2025. That 43% surge between September 2024 and December 2025 followed the Los Angeles fires. The California Department of Insurance has taken formal action against the FAIR Plan for illegally denying smoke damage claims. A net designed for a few thousand hard-to-place properties is now the primary insurer for more than half a million homes.
Florida, to be fair, tells a different story. Florida Citizens peaked at 1.42 million policies in October 2023. By January 2026, it had fallen to 395,144, the lowest since at least 2012. Seventeen new companies entered the market. Most remaining Citizens policyholders will see an average premium decrease of 8.7%.
This improvement is real. It is also entirely attributable to litigation reform: the elimination of one-way attorney fees and curbs on assignment-of-benefits abuse under Florida's 2022 and 2023 reforms. Average requested rate hikes dropped from 21% in 2023 to 0.2% in 2025. Florida fixed the litigation load. The hurricane load is unchanged. A building inspector who finds one cracked beam repaired on one side doesn't sign off on the whole structure.
California's FAIR Plan is carrying primary-market weight on a last-resort frame. Florida has demonstrated that some of the load was litigation-generated and correctable. Neither state has reduced its physical exposure to climate loss by one square foot.
III. The Colorado River Compact and 104 Years of Spending a River That Was Never There
Here is a fact so absurd that if you put it in a novel, your editor would cut it for being too on-the-nose.
The 1922 Colorado River Compact divided 15 million acre-feet of water per year between Upper and Lower Basin states based on an assumed annual flow of 16.4 million acre-feet. A 1944 treaty gave Mexico another 1.5 million. Total allocations: 16.5 million acre-feet per year.
The 1920s happened to be the wettest period in the river's recorded history. Flows ran as high as 20 million acre-feet. The Compact's framers measured the river during its best years and wrote a permanent contract on the basis of a wet anomaly. As Brad Udall of Colorado State's Water Center has put it:
"The law of the river was written for a river that no longer exists from a hydrologic standpoint."
It arguably never did.
Average annual flows since 2000 have been approximately 12.4 million acre-feet. That's a structural deficit of roughly 4 million acre-feet per year against total allocations. Forty million people depend on this water. Research modeling shows above 80% probability that both Lake Powell and Lake Mead reach dead pool before 2060.
The 2007 Interim Guidelines and 2019 Drought Contingency Plans expire at the end of this year. The Bureau of Reclamation released a draft Environmental Impact Statement in January with no preferred alternative. The seven basin states missed their November 2025 deadline for a consensus plan, then missed a second in February. On May 1, the Lower Basin states submitted a joint proposal offering 1.25 million acre-feet per year in collective reductions, with Mexico contributing 250,000 acre-feet and a new 700,000 acre-foot conservation program, totaling 3.2 million acre-feet in savings through 2028.
That's roughly 1.6 million acre-feet per year against a 4-million-acre-foot annual deficit. Meaningful, but it covers about 40% of the gap and only through 2028. At Metropolitan Water District's May committee meeting, Colorado River policy manager Shanti Rosset said of the Bureau's pursuit of a seven-state deal: "I think they have given up. There's not enough time." The Upper Basin has not agreed. But the Bureau retains legal authority to impose its own operating plan if the states cannot reach consensus, and the Lower Basin proposal at least gives Reclamation something to build a preferred alternative around. Institutional options remain on a shrinking calendar.
The Bureau must announce a decision before October 1. Without one, operations revert to pre-2007 procedures written for a river that had more water and fewer people.
Twenty-nine tribal nations hold senior water rights under Winters v. United States (1908). The 1922 Compact states that nothing in it affects obligations to Indian tribes. It also made no allocations to them, because no tribal representatives were present during negotiations. The obligation is written into the law, without a mechanism for fulfilling it. The compressed post-2026 timeline makes it less likely, not more, that the oldest legal claims to this water will finally be addressed in the new framework. The Bureau's April Drought Response Operations Plan ordered up to one million acre-feet released from Flaming Gorge Reservoir to prop up Lake Powell and cut releases from Powell to Mead from 7.48 to 6 million acre-feet. Emergency measures, all of them, applied to a 104-year-old contract built on a measurement error.
IV. The National Flood Insurance Program at $22.5 Billion in Debt, Expiring During Hurricane Season
The National Flood Insurance Act of 1968 created a program where premiums would cover routine claims and Treasury borrowing would handle catastrophic outliers. It was explicitly not designed to retain funding for extreme events. It was designed for a world where extreme events were rare.
Current status: 4.6 million policies, $1.3 trillion in coverage, $22.525 billion in debt to Treasury, $7.9 billion in remaining borrowing authority before hitting the statutory ceiling. The program accrues $1.7 million in interest daily and pays approximately $619 million per year in debt service. On February 10, 2025, the NFIP borrowed another $2 billion to pay claims.
Since 1968, nearly 45,000 properties have met the criteria for Severe Repetitive Loss, averaging five floods each. Only 24% have had their flood risk addressed through mitigation, down from 27% in 2018. Over the program's history, more of these properties have become uninsured than have received mitigation. The program pays for the same damage to the same buildings with borrowed money, serviced through premiums collected from everyone else in the pool.
The NFIP's authorization expires September 30, 2026. Without reauthorization, new policies stop, borrowing capacity drops from $30.4 billion to $1 billion, and approximately 1,300 property closings per day are affected. Since 2017, Congress has passed 35 short-term reauthorizations. No long-term reform has advanced.
The political incentive to avoid a lapse is strong. Thirty-five consecutive reauthorizations is a pattern, and 1,300 daily property closings concentrate congressional attention. But the $7.9 billion borrowing buffer is the structural question. The debt that created the current hole was generated primarily by the 2005 hurricane season. Congress forgave $16 billion in 2018 and the program still owes $22.5 billion. A single major-metro hurricane could exhaust the remaining buffer in weeks. A system temporarily extended 35 consecutive times, carrying debt it hasn't zeroed in two decades, servicing that debt at $1.7 million per day while insuring properties that average five floods each, has made not-collapsing into its permanent operating condition. Congress will probably reauthorize again. That's a bet on political behavior. Whether the program can absorb what the Atlantic sends this summer is a bet on arithmetic.
The NFIP expires September 30. The Colorado River decision is due October 1. Two load-bearing systems hitting their limits in the same week, during hurricane season.
V. Climate Risk Disclosure, or the Load-Bearing Wall Someone Chose to Remove
The first four systems in this inspection are overstressed. They're carrying more than they were designed for. The fifth was functioning under load and was deliberately weakened.
Property-level climate risk scores, developed by First Street and integrated into real estate listings, were designed to make physical risk visible to buyers before purchase. Flood, wildfire, wind, heat, and air quality, scored 1 through 10.
The system worked. Zillow's own data showed homes listed in June 2024 with high flood risk were less likely to sell by March 2025: 52% sold versus 71% of low-risk homes. An NBER study found buyers who see flood risk data search for lower-risk properties. During the January 2026 LA fires, First Street's maps identified over 90% of homes that burned in the Eaton Fire as severe or extreme wildfire risk. California's own maps classified just 21% of those homes as very high risk.
In November 2025, Zillow removed the scores from its listings after pressure from the California Regional Multiple Listing Service. The CRMLS CEO explained the concern:
"Displaying the probability of a specific home flooding can have a significant impact on the perceived desirability of that property."
Yes. That is what disclosure does.
The scores remain on Redfin, Realtor.com, and Homes.com. Zillow still links to First Street's site, but the data requires extra clicks. First Street, it should be noted, has restructured from nonprofit to for-profit and now serves institutional clients in insurance and finance, a dual role that complicates its incentives. And a 2024 study that contacted nine climate risk modeling companies found only two willing to participate in an independent evaluation; First Street was not among them. Those methodology questions are legitimate.
But methodology had nothing to do with it. A disclosure tool was pulled from the largest real estate platform in the country because it was doing what disclosure tools do, under exactly the conditions that made disclosure most necessary. An inspector has a word for when someone takes out a load-bearing wall during a storm. The risk stays, and moves from a pre-purchase decision into a post-purchase liability, from the buyer who could have known to the owner who finds out the hard way.
Inspection Summary
Five systems. All operational. The Forest Service is staffed at 92% of a number its own leadership calls insufficient, heading into a season running double the historical pace. California's insurer of last resort carries four times its intended load. The Colorado River's legal framework expires in four months with no replacement and a 4-million-acre-foot annual deficit baked into its original arithmetic. The NFIP owes $22.5 billion on a program designed never to carry such debt, authorized in temporary increments 35 consecutive times, expiring during hurricane season. The one disclosure tool that actually changed buyer behavior got pulled from the biggest platform because it changed buyer behavior.
Yesterday was Memorial Day. We laid wreaths and said the right words about people who were sent into harm's way. We have always been good at that part. Inspecting the structures we send the living into, less so. We are not good at looking at a building that still has its lights on and asking whether the lights are evidence of integrity or just the last thing you see before they go out.
None of these systems has collapsed. People are inside all of them. An honest inspector would note the loads, note the fractures, note the remaining capacity where it exists, and tell you plainly: the building is occupied, the damage is not visible from the street, and nobody has ordered an evacuation.
The lights are on. That is not the same as safe.
- USFS experimental forests in limbo: The Forest Service restructuring plan closed 57 research stations but does not specify what will happen to the 82 experimental forests that house decades-long climate and timber studies, some of which cannot be restarted once interrupted.
- Risk Rating 2.0 premium shock: FEMA's overhaul of NFIP pricing, the biggest change since 1968, is projected to increase premiums for roughly 66% of policyholders, with increases exceeding 100% on average in some analyses, layering affordability pressure onto a program already carrying $22.5 billion in debt.
- Colorado's insurance transparency law: Colorado's HB25-1182, which requires insurers to disclose their risk models and credit household mitigation in pricing, has inspired similar legislation in at least 18 states in 2026, making it the closest thing to a national template emerging from the state level.
- Oregon's vanished snowpack: Oregon's 2026 winter produced the smallest snowpack ever recorded, with state climatologist Larry O'Neill saying there is no historical equivalent in records dating to the 1890s, a baseline condition that underlies much of the West's fire and water risk this summer.