WHITEFLAG

主权资产 Re-Optimization Portal

最优配置

Given 8 billion people distributed across the Earth's surface, what arrangement of sovereign structures maximizes aggregate welfare subject to the constraint that every population unit meets a minimum dignity threshold? The Dignity Floor Index (DFI) measures whether populations meet minimum thresholds across seven dimensions — health systems, political voice, housing, education, income, environmental safety, and social connection — not just income alone. Where the World Bank counts 3.5B people below $6.85/day, the DFI captures populations that income measures miss: people above the poverty line who lack adequate governance, health infrastructure, or environmental safety. Baseline figures reflect 2024 data; archetype scenarios model structural outcomes if implemented, not time-bound projections. Climate trajectories (216M internal climate migrants by 2050, per World Bank Groundswell) are modeled separately and suggest the baseline worsens without structural intervention.

正在加载全球尊严评估...

尊严地图

正在加载尊严地图...
0.0
1.0
DFI 分数
排名 国家 DFI 分数 状态 约束条件 置信度 人口

原型配置

每个原型代表一种不同的重组主权结构以最大化尊严底线合规性的方法,范围从最小干扰到完全的结构性转变。

配置浏览器

探索原型配置之间的权衡空间。每个滑块在相互竞争的优先级之间进行混合,从五个结构性原型中产生插值指标。

政策轴
原型权重
混合指标
目标分数
帕累托前沿:可行性 vs. 合规性

结构性上限

在治理传播之后,真正难以解决的问题是什么?此标签显示真正的剩余障碍——哪些国家因结构性障碍而陷入困境,哪些只是因为模型忽略了治理重组而陷入困境。推测性外推已明确标注。

成本上升曲线
约束条件墙
最困难的国家——在完全优化下DFI仍低于0.6
维度失败热图——国家 × 维度

颜色强度 = 完全优化下的差距严重程度。深色 = 无差距。红色 = 剩余差距较大。

干预台

按实体的尊严审计:维度细分、约束条件、排名干预选项和剩余人力资本暴露。

合作架构

前所未有的全球合作实际上需要什么?不是愿望——而是组织结构图、资金、时间表以及每个组成部分的真实可行性。碳税、财富税和补贴重新分配的最大现实年收入达到 $3–4.7T/年 ——相对于 $37.9T/年 的完全优化差距,将30年合规性上限设定在 70–80%。以下每个主张都有引用、计算或陈述的假设。

Revenue Architecture — Where the Money Comes From
来源机制乐观悲观先例可行性
碳税 对全球排放量的50%(36.8 Gt)征收每吨75美元 $1.38T/年 $0.69T/年 欧盟排放交易体系平均每吨92美元(2023年) 中等
财富税 Graduated 1–3% on wealth above $1M (US+EU+JP+AU+CA+KR) $1.70T/年 $0.85T/年 法国 ISF (1982–2017): capital flight of €200B over 15yr
金融交易税 股票0.1% + 外汇0.01% + 衍生品0.005% $0.35T/年 $0.15T/年 欧盟11国金融交易税于2013年提出,至今未实施
化石燃料补贴重新分配 在10年阶段内重新分配50%的显性补贴($1.3T/年) $0.65T/年 $0.30T/年 印度尼西亚重新分配了156亿美元(2015年);IMF总计:包括隐性补贴在内每年7万亿美元 中等
主权财富转移 1% of GDP from nations w/ GDP/cap >$30K (37 nations, ~$60T GDP) $0.60T/年 $0.30T/年 当前官方发展援助:$211B/年(捐助国国民总收入的0.33%)。只有5个国家达到0.7%的联合国目标
总计 $4.68T/年 $2.29T/年
吸收折扣: $4.7T in revenue becomes $2.3–3.5T in effective spending after governance absorption losses. A dollar transferred to 乍得 (governance effectiveness 0.06) produces $0.06 in DFI improvement. A dollar to 德国 (0.92) produces $0.92. For the worst-governed states, you need to spend $5–17 per dollar of effective outcome. This is not corruption alone — it is absent infrastructure, human capital deficits, and institutional capacity gaps that money alone cannot fill.
Scale Comparison — The Realistic Revenue ($4.7T/yr) Against Reference Points
The 先例 Wall — Why Nothing in History Matches What’s Required
先例成本(占GDP百分比)国家合规性Why It’s Inadequate
蒙特利尔议定书(1987年) 0.003% 198 99% 成本落在20家公司身上,而非80亿人。存在化学替代品。科学滞后时间短。
巴黎协定(2015年) 约0.3%(承诺) 196 30–40% 设计为非约束性。无强制执行。2023年全球排放量上升1.1%。
EU Acquis (1957–present) 约欧盟GDP的1.0% 27 约95% GDP/cap ratio 10:1 (Bulgaria–Luxembourg). Global ratio: 450:1 (Burundi–Luxembourg). 70 years to build.
Marshall Plan (1948–52) 美国GDP的2.5% 16 重建了已经工业化的国家,这些国家拥有现有治理。4年,而非30年。
PEPFAR (2003–present) 美国GDP的0.01% 60 单一疾病干预。2010万人接受治疗。无法扩展到所有7个DFI维度。
所需(尊严底线) 4.5% 150+ 70–80% 10–70× larger than any precedent. Sustained 30年. Requires adversarial great powers to cooperate.
The 时间表 — Three Phases of Realistic Cooperation
PHASE 1: 年份s 1–3
约57%
合规性
Coalition of 18–25 willing nations. Pilot GDF at $50–100B/yr via expanded Green Climate Fund + World Bank IDA. DFI adopted as supplementary metric at UNDP. Carbon border adjustment design begins.
Saves ~300–600K lives/yr. Lifts ~100–200M above floor.
PHASE 2: 年份s 3–10
60–65%
合规性
Carbon tax at $75/ton across coalition. GDF scales to $0.5–1.5T/yr. Physician training pipeline: 50K new doctors by 年份 10 (against 4.2M global shortage). 15–25 “middle-band” countries reform governance.
Below floor: ~3.0–3.20亿. Mortality: ~11.5–12.0M/yr.
PHASE 3: 年份s 10–30
70–80%
合规性
Full revenue architecture: $3–4T/yr. Second-generation health workers enter workforce. 40–60 countries improve institutionally. Climate adaptation partially in place. But: 15–20 hard cases remain. 57-country “democracy wall” partially breached.
Below floor: ~1.6–2.4B. Mortality: ~10.5–11.500万/年. 成本: $50–80T cumulative.
缩小的窗口: Climate tipping points create a countdown. Arctic summer ice loss by ~2035. AMOC tipping estimated 2025–2095 (central ~2057). Greenland ice sheet irreversible above 1.5–2.0°C. Amazon dieback at 20–25% deforestation (currently 17%). Every year of delay in cooperation degrades the environmental baseline against which DFI improvements are measured. The same architecture implemented in 年份 10 faces a meaningfully worse planet than in 年份 1.
The 约束条件墙 — What Money Cannot Buy

在我们的16个原型中,最常阻止国家满足尊严底线的维度是 政治参与 — which depends on electoral democracy, civil liberties, rule of law, and corruption control. These are not purchasable.

民主墙: 57 countries where 政治参与 is the binding constraint under Fiscal Federalism. Combined population: ~3.8B. You can fund health systems, build schools, transfer cash, and train workers. You cannot make an authoritarian regime hold free elections from outside. This is the structural reason the irreducible minimum population below the dignity floor is ~1.5–20亿 even after 30年 and $50–80T in cumulative spending.
真实可行性矩阵
组成部分技术制度政治时间表净评估
碳税(每吨75美元,联盟) 中等 中等 与联盟合作可实现
财富税(协调) 大规模实施可能性极低
金融交易税 中等 中等 适度收入可能
补贴重新分配 中等 中等 部分重新分配可行
全球尊严基金 中等 中等 试点可能;完整不确定
联合国宪章改革 不适用 极低 不可实现
治理改革(57个国家) 中等 极低 约束条件
卫生工作者培训 中等 中等 可实现但缓慢(8年管道)
气候适应基础设施 中等 中等 中等 有资金可实现
真实底线: Maximum realistic cooperation produces 70–80% compliance after 30年 and $50–80T in cumulative spending. This leaves ~1.6–2.4B people below the dignity floor and ~10.5–11.5M excess deaths per year. The irreducible minimum is set by the “democracy wall” — countries where 政治参与 is structurally below the floor and cannot be improved by external actors without violating sovereignty. The problem is not that we do not know what to do. It is that what needs to be done requires a level of cooperation that the structure of sovereign self-interest has never produced and shows no credible sign of producing.
来源: Global Carbon Project 2024 · IMF Fossil Fuel Subsidies (2023) · SIPRI Military Expenditure Database · OECD DAC Aid Statistics · World Bank Annual Report 2023 · Credit Suisse Global Wealth Report 2023 · Saez & Zucman, Brookings Papers (2019) · Barrett, “Environment and Statecraft” (2003) · Collier & Dollar, European Economic Review (2002) · Kentikelenis et al., Lancet (2015) · Ditlevsen & Ditlevsen, Nature Communications (2023) · IPCC AR6 WG2 · WHO Global Health Workforce Statistics · World Bank Governance Indicators · V-Dem Electoral Democracy Index

技术登月计划

技术突破能否改变轨迹?我们评估了15项技术,针对结构性发现:DFI的约束条件是政治性的,而非技术性的。一项需要功能制度来部署的技术无法帮助生活在失败治理下的人口——无论该技术多么具有变革性。 近期且与DFI相关:太阳能海水淡化(TRL 9)、数字身份(TRL 8,$1/人)、电网储能(铁-空气电池$20/kWh)、CRISPR作物(TRL 5–7,2030–40年)、卫星监测(TRL 9)。 高影响 / 高风险:平流层气溶胶注入——唯一能足够快速改变温度的技术(每降低1°C每年$2–8B),但没有治理框架,且约30个国家可以单方面部署。 太晚或已证明失败:核聚变(2040年代+)、太空太阳能(2050年代+)、大规模直接空气捕获(2045+)、垂直农业(行业崩溃)、北极冰恢复(项目已终止)。

15 Technologies Assessed — TRL, 时间表, What They Solve, DFI影响
技术TRL大规模应用时间解决的问题What It Doesn’t SolveDFI影响
能源
核聚变 5–6 2035–2040 Unlimited clean baseload. CFS SPARC targeting Q>1 in 2027. Helion claims 2028 grid power for Microsoft. Arrives after critical 2025–2035 window. Requires institutional capacity to build and operate. ITER delayed to 2039. 可忽略
小型模块化反应堆(TerraPower, NuScale) 6–7 2031–2040 在OECD国家提供无碳基荷。TerraPower Natrium预计2031年并网。 NuScale costs doubled to $89/MWh. Requires nuclear regulatory capacity. HALEU fuel supply dependent on 俄罗斯. Below-floor countries can’t deploy. 适度
增强型地热(Fervo, Quaise) 7 2028–2035 地球任何地方的基荷。重用石油/天然气钻探专业知识。Fervo Cape Station:到2028年500 MW。钻井时间同比减少70%。 Quaise deep drilling (millimeter wave) still at 100m — commercial needs 10+ km. Induced seismicity risk. Scaling to hundreds of GW takes decades. 如果深钻成功则显著
电网储能(锂离子、钠离子、铁-空气电池) 7–8 现在–2028 实现太阳能+储能作为基荷。锂离子电池成本为70美元/千瓦时。Form Energy铁空气电池成本为20美元/千瓦时,可提供100小时储能。钠离子电池成本约为40美元/千瓦时(CATL)。 Manufacturing/supply chain concentrated in China. Doesn’t exist in most vulnerable nations. Battery materials have their own extraction footprint. 重要推动因素
气候干预
平流层气溶胶注入 3–4 2–5 yrs Only tech that reduces global temp by 1–2°C within 1–2 years. $2–8B/yr. Fleet of ~100 modified tanker aircraft. 海洋酸化未受影响。南亚季风中断。 终止冲击: if stopped, temps snap back at 5–10× rate. No governance framework. ~30 nations could deploy unilaterally. 哈佛SCoPEx项目于2024年取消。 对温度影响高 / 对治理具有破坏性
直接空气捕获(Climeworks, Oxy) 7 十亿吨级:2045年之后 从大气中去除二氧化碳。Climeworks Mammoth:每年3.6万吨。Oxy Stratos:每年50万吨。 $1,000–1,300/ton. Current global capacity is ~0.04 Mt/yr vs. 40 Gt/yr emissions — 1,000,000× too small. 能源-intensive. Gt-scale requires trillions. 在时间框架内可忽略不计
海洋云增亮 3–4 2030年代 局部冷却。可能保护珊瑚礁/北极。比SAI的承诺风险更低。 无法扩展至全球降温。旧金山现场测试取消。区域降水效应了解不足。 极小
Food & 水
太阳能海水淡化 9 现在 Directly addresses water stress for 1.96B people. $0.50–1.50/m³ with solar PV at <$0.03/kWh. 60% cost reduction in 10年. Only coastal/brackish. Brine disposal damages marine ecology. Capital-intensive ($100M–$20亿/plant). Distribution requires governance. 对沿海人口重要
CRISPR耐热/耐旱作物 5–7 2030–2040 直接帮助底层人口。耐热主食作物可在模型预测农业崩溃的地区维持产量。 监管碎片化(美国宽松,欧盟严格)。种子分发需要农业推广服务。脱靶遗传效应。 对食物维度重要
精密发酵 7–8 2030–2035 Protein without land. 90% land reduction. Price parity projected 2027–2029 for some products. $5.8B → $1510亿 market by 2034. Produces protein, not the cheap calories (rice, wheat, maize) that 20亿 people need. 10× cost gap vs. conventional. 能源-intensive bioreactors. 中等(富裕世界)
垂直农业 8 城市地区的微型蔬菜和香草。 Industry collapsed. Plenty ($2.3B valuation) bankrupt March 2025. Bowery ($2.3B) shut down fall 2024. Physics makes caloric staples prohibitive — LED photosynthesis vastly less efficient than sunlight.
Governance & Logistics
数字身份(Aadhaar模式) 8–9 现在 绕过腐败中介。印度有13.1亿人注册。节省390亿美元。注册成本约1美元/人。全球仍有8.5亿人缺乏身份。 Requires minimum institutional capacity. Surveillance/weaponization risk under authoritarian regimes. Doesn’t create economic opportunity. 重要推动因素
卫星监测(Planet, Maxar) 8–9 现在 Near-real-time compliance verification. Deforestation >1ha auto-detected. 200+ satellites, 100M+ km²/day. 检测不等于执行。无法观察治理质量或室内条件。30%的不合规是政治意愿问题,而非信息问题。 中等(基础设施层)
人工智能资源分配 6–7 2025–2030 供应链优化。预测性医疗保健。行政效率。 70% of US hospital AI pilots failed (weak endpoints, workflow misalignment, data gaps). If it fails in American hospitals, what about South 苏丹? Requires data infra that doesn’t exist. 中等(仅限OECD国家)
基本不对称: The technical solutions exist or are within reach. 约束条件 is the governance architecture needed to deploy them. A technology that requires functioning institutions to deliver cannot help populations living under failed governance. Solar desalination is mature and affordable — but a country that can’t maintain a water distribution network can’t deploy it. CRISPR crops could preserve food security — but seed distribution depends on agricultural extension services that don’t exist in the Sahel. Digital identity costs $1/person — but requires a state capable of maintaining a database. The people who need the technology most are the people least likely to have access to it, not because the technology doesn’t work, but because the institutions needed to deliver it don’t exist.
平流层气溶胶注入 — The Intervention Nobody Controls

SAI deserves special attention because it is the only technology that operates on the right timescale and cost. At $2–8B/year (roughly the cost of a single aircraft carrier), a fleet of ~100 modified tanker aircraft could reduce global temperature by 1–2°C within 1–2 years. No other technology comes close to this cost-effectiveness ratio on temperature.

其吸引力所在

Buys 20–30年 of time for every other solution to deploy. Directly reduces heat stress, slows Arctic ice loss, preserves some agricultural yields. The atmospheric science is well-understood (volcanic eruptions are natural analogs — Pinatubo 1991 cooled the planet 0.5°C for 2 years). Deployable within 2–5 years of a political decision.

其恐怖之处

终止冲击: If stopped abruptly, temperatures snap back at 5–10× the rate of gradual warming. Once started, it essentially cannot be stopped. 季风中断: Models show reduced rainfall in South Asia and altered Sahel precipitation — potentially harming the very populations most at risk. 无治理: ~30 nations could deploy unilaterally. There is no treaty, no framework, no agreement on who controls the thermostat. Harvard’s SCoPEx field test was cancelled in 2024 under pressure from civil society and Indigenous groups.

DFI框架: SAI is an option of last resort that becomes more likely as the DFI worsens. The absence of governance for it is itself a major risk factor. It could buy decades of time for cooperation and relocation — or it could trigger exactly the kind of geopolitical conflict that makes both impossible.
来源: CFS SPARC Assembly (Fortune, 2026) · Helion Polaris (TechCrunch, 2026) · ITER Delay (AIP, 2025) · NuScale Cost Escalation (IEEFA) · TerraPower Natrium (Neutron Bytes, 2025) · Fervo 能源 Cape Station Results · Quaise 能源 Milestones (2025) · Harvard Salata Institute SAI Research · Smith & Wagner, SAI Cost 估计值s (2018) · Parker & Irvine, Termination Shock (2018) · WRI DAC Cost Analysis · Climeworks Mammoth Operational Data · Form 能源 Iron-Air Battery · Aadhaar Impact (MIT 技术 Review, 2026) · World Bank ID4D · Planet Labs Forest Carbon Monitoring · BNEF Battery Price Survey (2025) · CATL Sodium-Ion (MIT 技术 Review, 2026) · AeroFarms Post-Bankruptcy (2024) · Plenty Bankruptcy Filing (2025)

计划搬迁

如果前所未有的合作不可能,且所有人类生命具有同等价值,理性的回应是计划大规模迁移——将人们移出DFI将崩溃的地区。 15至21亿 人必须在2050年前迁移。所有可行目的地的最大接收能力为 3–5.5亿。差距—— 10至15亿 人无处可去——是任何技术或条约都无法解决的约束。此标签解除流动性约束并提问:B计划实际上是什么样子?

Departure Zones — Where 人数 Must Leave

Climate risks are multiplicative, not additive. 孟加拉国 does not face flooding OR heat OR agricultural collapse — it faces all three simultaneously. Countries with 3+ overlapping climate risks require departure by 2040.

国家人口当前DFI热浪海平面农业复合得分
孟加拉国 1.74亿 0.498 0.92
巴基斯坦 2.47亿 0.390 0.90
印度(北部) 约5亿 0.547 0.89
也门 3900万 0.199 0.95
苏丹 5000万 0.174 0.88
埃及 1.14亿 0.567 0.85
伊拉克 4500万 0.248 0.82
尼日尔 2600万 0.207 0.80
乍得 1900万 0.222 0.82
索马里 1800万 0.198 0.88
越南 9800万 0.682 0.78

严重   中等   低/无。来源:IPCC AR6 WG2, CMIP6湿球温度预测, ND-GAIN, World Bank Groundswell。

Receiving Capacity — Who Can Absorb Them

排名ed by composite absorption score: physical capacity (land, water, climate stability), governance quality (DFI), infrastructure, and economic integration potential. 约束条件 is never physical — it is political.

排名国家吸收得分最大吸收关键约束
1加拿大0.9240–80M北部基础设施不存在;寒冷气候
2美国0.8880–15000万政治意愿;本土主义反弹
3德国0.8715–25M人口密度;现有住房危机
4法国0.8615–25M政治反弹阈值
5英国0.8515–25M岛屿地理;住房短缺
6澳大利亚0.8415–30M水资源约束;距离
7瑞典0.835–10M经济规模小;寒冷
8挪威0.823–5M容量非常小
9新西兰0.813–5M偏远;规模小
10巴西0.7230–6000万治理缺口;亚马逊约束
11阿根廷0.6815–30M经济不稳定
12俄罗斯0.5550–100M治理;政治障碍
总计286–54500万
算术是毁灭性的: Combined maximum absorption capacity of all capable receiving nations: ~300–55000万. The demand is 1.5–2.10亿. Even under optimistic assumptions, receiving capacity covers only 15–35% of the need. 俄罗斯 has massive physical capacity but its governance makes it politically unsuitable — the single largest potential receiving zone on Earth fails the DFI filter.
The Scale — Nothing in History Approaches This
情景人数持续时间年率成本(中值估计)占全球GDP年百分比
保守 4亿 25年 1600万/年 总计32万亿美元 1.3%
中等 10亿 30年 3300万/年 总计81万亿美元 2.7%
全面 20亿 35年 5700万/年 总计162万亿美元 4.6%

中等情景需要维持迁移率 5–8× the peak of WWII displacement for thirty consecutive years. For context: current global migration stock is ~280M total (including voluntary). The largest single-year refugee crisis was WWII at ~10M/year. Cost per person relocated: ~$81K (transport, housing construction, infrastructure, integration, healthcare transition). At 10亿 people that’s $81T — roughly $2.7T/year, equivalent to global military spending.

Historical 先例s — And Why They All Failed at Scale
事件人数持续时间年率死亡人数教训
印度分治(1947年) 10–20M 约6个月 约4000万/年 1–2M 无计划的大规模迁移导致大规模死亡。火车满载尸体抵达。
二战后欧洲 6000万 12年 500万/年 需要马歇尔计划(占美国GDP的2.5%)。花费数十年。许多人从未返回。
Syrian Crisis (2011–) 1300万 10年 130万/年 约50万 2M refugees triggered Europe’s far-right surge. AfD: 4.7% → 20.8%. Brexit.
孟加拉国内部迁移 40万/年 持续中 40万/年 达卡在重压下崩溃。40%人口生活在贫民窟。目的地成为下一个危机。
美国沙尘暴(1930年代) 250万 10年 25万/年 相同国籍、语言、文化。仍遭遇敌意和歧视。
计划迁移(中等) 10亿 30年 3300万/年 5–8× WWII rate sustained for 3 decades. No historical analog.
政治现实

阻碍因素

无法律框架。 The 1951 Refugee Convention doesn’t cover climate. No nation is obligated to accept climate migrants. The Global Compact on Migration (2018) is non-binding.

本土主义随人数增加而加剧。 Europe’s far-right surge was triggered by ~2M refugees (0.4% of EU population). At 33M/year, receiving countries absorb 5–20% of their population per decade. Every democracy that has faced immigration at 5%+/year has produced authoritarian backlash.

殖民维度。 大多数出发区曾被需要接收移民的国家殖民而贫困。这在道德上是合理的(气候债务+殖民债务),同时在政治上具有爆炸性。

人口学反驳论点

老龄化国家需要人口。 日本每年减少84万人。德国、意大利、韩国、西班牙都在萎缩。合计人口赤字:约每年116万人且加速。

出发区年轻。 Median age in Sahel: ~15. In South Asia: ~28. In Japan: ~49. The matching algorithm writes itself — young workers to aging economies.

But the scale doesn’t match. Aging nations need 2–3M/year. The relocation demand is 33–57M/year. The demographic dividend absorbs <10% of the need.

Three Paths Compared — Cooperation vs. Relocation vs. Inaction
合作
成本: $3–5T/yr
可行性: ~0.13–0.25
30年低于底线: 1.6–2.4B
30年死亡率: ~10.5–11.500万/年
30年死亡人数: ~315–34500万
维护主权。解决根本原因。需要前所未有的政治协调。
计划性迁移
成本: $2.7万亿/年(中等)
可行性: ~0.05–0.15
迁移人口: 300–55000万 max
滞留人口: 1–1.5B
30年死亡人数: ~250–300M
Saves 200–4亿 from worst zones. 10亿+ stranded. Less feasible than cooperation. No one talks about it.
不作为(默认)
成本: $0
可行性: 1.0
30年低于底线: 37亿+(持续恶化)
30年死亡率: ~1270万/年(上升中)
30年死亡人数: ~3.8亿+
政治阻力最小的路径。累计超额死亡人数超过3.21亿。成本完全由生命承担,而非金钱。
此标签页存在的意义: Planned relocation is less feasible than the cooperation model, not more. Its political preconditions — binding receiving quotas, $2.7T/year financing, a new international legal framework — are arguably harder to achieve. It exists to make the arithmetic of inaction visible. When cooperation has a feasibility of ~0.13 and relocation ~0.10, the honest conclusion is that neither will happen at the required scale. The 计划搬迁 tab quantifies what that means: between 250 and 380 million excess deaths over 30年, concentrated in the populations that contributed least to climate change. No one is talking about this seriously because the numbers are politically unbearable. The purpose of this framework is to make them unavoidable.
来源: IPCC AR6 WG2 Ch.4–7 · World Bank Groundswell 2.0 (2021) · ND-GAIN 国家 Index · UNHCR Resettlement Cost Data · Raymond et al., Science Advances (2020) · Mora et al., Nature Climate Change (2017) · CMIP6 Wet-Bulb Projections · BIS Triennial Survey · Credit Suisse Global Wealth Report · OECD DAC Aid Statistics · Ditlevsen & Ditlevsen, Nature Communications (2023) · WHO Global Health Workforce Statistics · FAO AQUASTAT

我们的模型如何比较

Our excess mortality estimates sit within a landscape of institutional projections. The differences are instructive — they reveal what each model counts and what it ignores.
来源 估计值 年份 时间范围 计算内容
WHO 25万/年 2014 2030–2050 仅4种直接途径:高温、疟疾、腹泻、营养不良
Lancet Countdown ~70万/年 2025 观测值(2024) 热浪 deaths + wildfire PM2.5 — already 3× WHO’s projection from just two pathways
GBD / IHME 810万/年 2024 基线(2021) Air pollution alone — now the 2nd leading risk factor for death globally
Zhao et al. / MCC 500万/年 2021 Baseline (2000–19) 所有非适宜温度:460万寒冷 + 48.9万高温 = 占所有死亡的9.4%
本模型(基线) 2025 预测值 4个与死亡率相关的维度:物质安全、医疗可及性、环境安全、住房充足性
WEF / Oliver Wyman 总计1450万 2024 到2050年 6 climate event categories at 2.5–2.9°C trajectory
Bressler (Columbia) 总计8300万 2021 2020–2100 热浪-related mortality only, business-as-usual (4.1°C)
Climate Impact Lab +73 每10万人 2022 到2100年 Temperature-mortality only — equal to current rate from all infectious disease
1,000-Ton Rule Synthesis ~10亿总计 2023 Next 100–200 yrs 所有途径。来自经济学、哲学、气候科学的趋同估计
本模型(崩溃) 2025 预测值 Full state failure scenario — our worst-case archetype
本模型(下限) 2025 预测值 Structural minimum — even perfect global cooperation cannot eliminate

Why WHO’s Number Is Already Wrong

WHO’s widely-cited 250,000 deaths/year projection covers only four direct pathways: heat stress, malaria, diarrhoea, and child undernutrition. It excludes air pollution (810万/年 at baseline), flooding, drought, displacement, conflict, cardiovascular disease exacerbation, ecosystem collapse, and all cascade effects. WHO itself describes the figure as 保守的.

The Lancet Countdown’s 2024年观测数据 — not a projection — already shows ~700,000 annual deaths from heat and wildfire smoke alone, nearly 3× what WHO projected for all four pathways combined. Twelve of twenty health-threat indicators reached record levels in 2025. There is no observed inflection point.

我们模型的位置

我们的死亡率模型使用凸曲线(赤字1.5) across four dimensions that kill people: material security, health access, environmental safety, and housing adequacy. Three DFI dimensions — 政治参与, education, and social connection — matter for dignity but have weak direct mortality links.

我们的基线原型产生 ~1270万超额死亡/年. This is consistent with the GBD’s 8.1M from air pollution alone plus additional mortality from malnutrition, preventable disease, and inadequate sanitation that our material, health, and housing dimensions capture. It is more pessimistic than WHO (which undercounts) but less extreme than the broadest estimates (which project over longer horizons and include indirect cascades our model doesn’t attempt).

我们的结构性下限 exists because meeting DFI minimum-dignity thresholds (0.35–0.40 per dimension) does not eliminate excess mortality — the “safe level” above which a dimension stops contributing to death is higher (0.60–0.70). Minimum dignity and minimum safety are not the same thing.

这向你提出的问题: No credible institution — ours or anyone else’s — projects a pathway to substantially lower excess mortality without assuming either unprecedented global cooperation, technology that does not yet work at scale, or both. The IPCC deliberately avoids giving a single headline number because every honest projection is grim. The choice space the archetypes above map out is between bad and catastrophic. The value of making that visible with numbers is that it replaces rhetoric with structure — and structure is what you need if you intend to act rather than look away.
来源: WHO Climate Change and Health Fact Sheet (2023) · IPCC AR6 WGII Chapter 7 (2022) · Lancet Countdown 2025 Report · GBD 2023, IHME · Zhao et al., Lancet Planetary Health (2021) · WEF/Oliver Wyman, Quantifying the Impact of Climate Change on Human Health (2024) · Bressler, “The Mortality Cost of Carbon,” Nature Communications (2021) · Carleton et al., Quarterly Journal of Economics (2022) · Pearce, “Updating the 1000-Ton Rule,” Energies (2023) · NPR, “The Undercount” (2024) · Carlson et al., Nature Climate Change (2025)

What We Don’t Know

像Palantir、Swiss Re、Citadel这样的组织以及国家情报机构拥有我们无法获取的数据。本节承认这一差距,并询问这是否改变了局面。

资源更丰富的参与者拥有什么

专有卫星 + 移动数据 — Cell phone location data for 5B+ people, tracking migration patterns in real time at district level. We model departure zones at country level; they see it at village level, month by month.

再保险巨灾模型 — Swiss Re, Munich Re, and Lloyd’s have the largest proprietary catastrophe databases on Earth, combining satellite data, machine learning, and decades of claims history. Insured losses reached $107B in 2025. These firms are actively declaring certain regions “uninsurable” — a stronger statement than any public climate model makes. When an insurer says a region is uninsurable, they are saying their proprietary models show expected losses exceed any premium the market will bear.

对冲基金气候情报 — Citadel hired PhD meteorologists and built in-house weather forecasting that earned $16B in 2022. Bridgewater has built “a top-down understanding of the net zero transition.” These firms treat climate risk as a tradable information asymmetry. They are not building models that show a more optimistic picture — they are building models that show where the damage hits first and positioning accordingly.

军事威胁评估 — The US DoD’s Climate Risk Analysis labels climate change a “threat multiplier.” The classified versions almost certainly contain specific timeline estimates for state failure in several of the 57 countries our model identifies, with higher-resolution migration flow predictions combining satellite and cell phone data for billions of people.

他们的数据讲述了一个不同的故事吗?

几乎肯定没有。 专有数据增加了精确度(哪个地区、哪个月、哪个供应链节点),但每个能访问这些数据的实体 的行为 表明情况至少和我们模型显示的一样糟糕,甚至更糟。再保险公司正在撤出承保。对冲基金正在构建私人气候情报。军事学说正在转变。富人在建造掩体。信息最灵通参与者的显示性偏好与我们的模型一致。

There is one documented case where proprietary data led public models: the insurance industry’s catastrophe models began signaling the severity of secondary perils (severe convective storms, wildfire, flood) 5–10年 before public climate models adequately captured them. The 2017–2023 period of “unexpected” insurance losses was not unexpected to the reinsurers. This suggests our model’s mortality estimates may be 保守的

专有数据可能产生分歧的一个领域: 富裕人群的适应有效性。 公司和富裕国家可能拥有数据显示,适应措施的效果比公共模型所显示的 对于他们服务的人群更好。这使富裕世界的体验更好,但对低于尊严底线的38亿人毫无说明。如果有的话,这反而扩大了差距。

我们的立场: We build with the data that is public. We acknowledge that organizations with proprietary access to mobility data, actuarial models, supply chain intelligence, and classified threat assessments have higher resolution than we do. We do not believe this higher resolution reveals a fundamentally different picture. If it did, the best-informed actors would not be retreating from risk — they would be investing in it. They are not.
来源: Swiss Re Natural Catastrophe Review (2025) · SIPRI Military Expenditure Database · Citadel 2022 Returns (public reporting) · US DoD Climate Risk Analysis (2021, unclassified) · UK 中等 Strategic Trends Programme · IMD, “Climate Change: The Emergence of Uninsurable Areas” (2024) · Planet Labs Forest Carbon Monitoring · Palantir/Tomorrow.io Climate Resilience Partnership (2025)
方法论: The Dignity Floor Index (DFI) is a 7-dimension conjunctive measure. Each dimension must independently exceed its threshold for a population unit to meet the floor. DFI models preconditions for dignity, not dignity itself. Sub-national data is incomplete where most needed. The optimizer cannot force political change — transition feasibility reflects this constraint. Redistribution capacity depends on political will. The Pareto frontier shows structural possibility, not political likelihood.
CODEX