How to walk this station如何走完这个站点
Eight stops, in the order a real programme is built: language first, then boundary, then the map, then the maths, then priority, then the benchmark, then the evidence, then the year-on-year loop. Nothing here is a target you must adopt — it is a study reference. Every number marked study range must be replaced by your own measured data before it goes into a report. 八个站点，按真实项目的建设顺序排列：先统一语言，再划定边界，接着绘制地图、算清账、定优先级、对标、备证据，最后进入年度循环。这里没有你必须采纳的目标——它是一份学习参考。凡标注 研究区间 的数字，写进报告前都必须替换为你自己的实测数据。
Define & draw定义与划界
Withdrawal, discharge, consumption, reuse, replenishment. Then the watershed picture that makes the fence line look small.取水、排水、耗水、回用、补充。然后是那张让"厂界"显得很小的流域图。
Count & prioritise计算与排序
Two calculators you click through one step at a time, and a stress × intensity matrix that tells you what must be in scope.两个逐步点开的计算器，以及一个"水压力 × 用水强度"矩阵，告诉你什么必须纳入范围。
Score & prove评分与举证
Score yourself out of 24, then run the yes/no evidence check before an auditor does it for you.先给自己打分（满分 24），再在审核员到来之前，自行完成是/否证据检查。
Define it before you count it先定义，再计量
Most water disputes inside a company are vocabulary disputes. Open each term: the plain meaning, the formula, and the trap that makes numbers non-comparable between two sites.公司内部大多数关于水的争论，其实是词汇之争。逐条展开：通俗含义、计算公式，以及让两个工厂数据无法比较的那个坑。
Term 01Withdrawal · Discharge · Consumption取水 · 排水 · 耗水
Withdrawal is everything you take from any source. Discharge is what you give back to the environment at a stated quality. Consumption is the part the basin never gets back — evaporation, water bound into product, water lost in sludge. 取水是你从任何水源取用的全部水量；排水是你以某一水质归还环境的水量；耗水是流域永远拿不回来的那部分——蒸发、进入产品的水、随污泥带走的水。
The trap常见的坑
A site that recycles internally 5 times can report a huge "water use" number and a small withdrawal number. Only withdrawal and consumption are comparable across sites; internal circulation is a performance indicator, not a footprint. 内部循环 5 次的工厂，可以报出很大的"用水量"和很小的"取水量"。只有取水与耗水在厂际之间可比；内部循环量是绩效指标，不是足迹。
Term 02Reuse vs. recycle — not the same word回用 vs. 再生——不是同一个词
Reuse: water used again with little or no treatment, usually cascaded to a lower-grade duty (final rinse water → first rinse → floor wash → cooling makeup). Recycle: water treated back up to its original grade and returned to the same duty (RO reject → secondary RO → UPW makeup). 回用：几乎不经处理即再次使用，通常按级联降级使用（终洗水→初洗→地面清洗→冷却补水）。再生：处理回到原有水质等级并返回同一用途（RO 浓水→二级 RO→超纯水补水）。
Cascading is almost always cheaper per m³ saved than recycling, because it buys no membranes. Do the cascade audit before you quote a treatment plant.按每节约 1 m³ 计，级联回用几乎总是比再生便宜，因为它不需要买膜。在为处理设施询价之前，先做一次级联梳理。
Term 03Replenishment & "water positive"水资源补充与"水正效益"
Water positive is a ratio, not a slogan: the verified volumetric water benefit you return to a basin, divided by the water you consume in that same basin, over the same year. A benefit in Basin A cannot cancel consumption in Basin B — water does not travel. "水正效益"是一个比值，不是口号：在同一年度、同一流域，你经核证返还该流域的水量效益，除以你在该流域的耗水量。A 流域的效益不能抵消 B 流域的耗水——水不会自己搬家。
Volumetric Water Benefit (VWB) counts only additional, measurable, attributable water — a repaired canal, a restored wetland, an irrigation retrofit — with the counterfactual stated. Awareness campaigns and tree planting without hydrological modelling generally do not count. 水量效益（VWB）只计入额外的、可测量、可归属的水量——修复的渠道、恢复的湿地、改造的灌溉系统——并须说明"若不做会怎样"的反事实情景。没有水文模型支撑的宣传活动与植树，通常不计入。
Term 04Water intensity — pick one denominator and defend it用水强度——选定一个分母并守住它
Intensity is only useful if the denominator is physical and stable. For an AI PC chain: m³ per wafer out, m³ per m² of panel glass, m³ per m² of finished PCB, m³ per chassis, litres per kWh of cell capacity, m³ per unit assembled. Revenue-based intensity moves when prices move — never use it as your only metric. 只有当分母是物理量且稳定时，强度指标才有意义。AI PC 链条上：每片晶圆 m³、每 m² 面板玻璃 m³、每 m² 成品 PCB m³、每个机壳 m³、每 kWh 电芯容量的升数、每台整机 m³。以营收为分母的强度会随价格波动——绝不能作为唯一指标。
"An intensity improvement with rising absolute consumption is not progress. The basin only feels absolute volumes.""强度下降而绝对耗水上升，不叫进步。流域只感受得到绝对水量。"
The wall is not the boundary. The basin is.边界不是围墙，是流域
Open zones ① to ④ below the diagram — each one dims the drawing down to its own zone so you can see what you are actually accountable for.展开图下的 ①—④ 区域，每一项都会将图中其余部分淡出，让你看清自己真正需要负责的范围。
Zone ①Upstream catchment — the water you never bought上游集水区——你从未付费的水
Your intake quality, price and permit reliability are all decided upstream: forest cover, farm runoff, upstream abstraction, snowpack. You have no title here, but you carry the risk. This is where most replenishment projects physically live. 你的取水水质、水价与取水许可的稳定性，全部由上游决定：森林覆盖、农业面源、上游取水、积雪量。你在此没有产权，却承担风险。绝大多数补充项目，物理上就发生在这里。
Ask for需要索取
- Basin name and Aqueduct baseline water stress score for every Tier-1 and critical Tier-2 site每个一级及关键二级供应商厂址的流域名称与 Aqueduct 基线水压力评分
- Local abstraction permit volume vs. actual withdrawal (headroom %)当地取水许可量与实际取水量对比（余量百分比）
Zone ②Inside the fence — the only water you fully control厂界之内——你唯一完全掌控的水
Meters, cascades, cooling towers, the wastewater plant. Typically 5–15% of an AI PC brand's total value-chain water sits here if the brand does final assembly only. Small share, total control — so it is where credibility is built, not where the impact is. 水表、级联回用、冷却塔、废水站。若品牌仅承担整机组装，这里通常只占其价值链总用水的 5–15%。份额小但完全可控——所以这里建立的是可信度，而不是影响量本身。
Zone ③Downstream — quality is your footprint too下游——水质同样是你的足迹
A volume returned at poor quality is not returned. Grey-water footprint logic: the volume of freshwater needed to dilute your load back to ambient standard. In electronics the usual named parameters are fluoride, copper, nickel, total nitrogen, COD, and increasingly PFAS. 以劣质水质归还的水，等于没有归还。灰水足迹逻辑：把你的污染负荷稀释回环境标准所需的淡水量。在电子行业，常被点名的参数是氟化物、铜、镍、总氮、COD，以及日益受关注的 PFAS。
Zone ④Aquifer — the slow account地下含水层——那本慢账
Groundwater withdrawal in an over-abstracted basin should be treated as the highest-priority volume you have, even if it is cheap and legal. Recharge takes decades; subsidence is irreversible. If any Tier-1 or Tier-2 site draws groundwater in an extremely-high-stress basin, that site is automatically in scope — see stop 05. 在超采流域抽取地下水，即使便宜且合法，也应视为你最高优先级的水量。补给需要数十年，地面沉降不可逆。若任一一级/二级供应商厂址在极高水压力流域使用地下水，该厂址自动纳入范围——见第 05 站。
The AI PC water flow — follow it, then mark the hotspotsAI PC 的水流——先追踪，再标注热点
Read it like an emissions flow: upstream to downstream. Amber nodes are hotspots — high intensity, or high stress, or a discharge with named pollutants. Open a node for its study range, why it is hot, the top actions with expected savings, and exactly what to request from the supplier. 像看碳排放流一样看它：自上游至下游。琥珀色节点为热点——高强度、或高水压力、或含特定污染物的排水。展开节点可见研究区间、成为热点的原因、预期节水幅度的首选措施，以及应向供应商索取的具体资料。
An NPU-bearing SoC moves to a leading-edge node: more mask layers, more litho–etch–clean cycles, and every single clean step ends in an ultrapure water rinse. Add more DRAM die, bigger packages, more advanced packaging steps and a larger vapour chamber, and the water per finished notebook rises even when the box gets thinner. Water intensity per unit is a design decision made in the architecture review, months before procurement sees it. 带 NPU 的 SoC 迁移到先进制程：更多光罩层、更多"光刻—蚀刻—清洗"循环，而每一道清洗步骤都以超纯水漂洗收尾。再加上更多 DRAM 颗粒、更大封装、更多先进封装工序与更大的均温板，即使整机变薄，每台笔电的用水量仍在上升。单机用水强度，是在架构评审会上做出的设计决策——比采购看到它早好几个月。
Tier 2 · Front-end二级 · 前道制造 Wafer fab — SoC / NPU, advanced node晶圆厂——SoC / NPU，先进制程 Hotspot · rank 1热点 · 排序 1
Why it is hot为何是热点
Rinse water must be ultrapure, so every m³ carries an energy and chemical cost before it is even used; and fab clusters concentrate in a handful of basins, so the whole industry's risk is correlated.漂洗水必须达到超纯标准，因此每 m³ 在被使用前已附带能耗与化学品成本；而晶圆厂集群集中于少数几个流域，使全行业风险高度相关。
Top actions & expected effect首选措施与预期效果
- Segregate rinse streams by contaminant (CMP slurry / acid / solvent / clean rinse) → unlocks 30–50% of UPW makeup for reclaim.按污染物分流漂洗水（CMP 研磨液／酸／溶剂／清洁漂洗）→ 可释放 30–50% 超纯水补水用于回收。
- Secondary RO on primary reject → recovers 10–25% of intake.一级浓水加设二级 RO → 回收取水量的 10–25%。
- Cooling tower cycles 3 → 6 → ~20% less makeup on that circuit (maths at stop 04).冷却塔浓缩倍数 3 → 6 → 该回路补水量减少约 20%（计算见第 04 站）。
Request from supplier向供应商索取
Site-level W / D / C for 3 years, reclaim rate methodology, basin name + permit headroom, discharge permit and 12 months of monitoring results for fluoride, copper, TN.近三年厂址级取水／排水／耗水数据、回收率计算方法、流域名称与许可余量、排污许可证及氟化物、铜、总氮 12 个月监测结果。
Tier 2 · Display二级 · 显示 Panel fab — LCD / OLED glass面板厂——LCD / OLED 玻璃 Hotspot · rank 2热点 · 排序 2
Top actions首选措施
- Counter-current cascade rinsing on cleaning lines → 40–70% rinse reduction.清洗线采用逆流级联漂洗 → 漂洗水减少 40–70%。
- Conductivity-triggered rinse valves instead of timer valves → 20–40%.以电导率触发阀取代定时阀 → 减少 20–40%。
- MBR + RO on general effluent back to cooling / scrubbers → up to 60–90% of treated effluent reused.一般废水经 MBR + RO 回用于冷却／洗涤塔 → 处理水回用率可达 60–90%。
Tier 2 · Interconnect二级 · 互连 PCB / HDI substrate — plating & etchPCB / HDI 载板——电镀与蚀刻 Hotspot · quality risk热点 · 水质风险
Top actions首选措施
- Three-stage counter-current rinse + drag-out recovery tank → 50–70% rinse water cut and metal returned to the bath.三级逆流漂洗 + 带出液回收槽 → 漂洗水减少 50–70%，金属返回镀槽。
- Ion exchange on rinse for closed-loop copper recovery → near-zero copper discharge.漂洗水离子交换实现铜闭环回收 → 铜排放接近零。
Tier 2 · Mechanical二级 · 结构件 Chassis — CNC, anodising, painting机壳——CNC、阳极氧化、喷涂 Medium · quick win中等 · 速赢
The cheapest real savings in the whole chain usually sit here: overflow rinse tanks running 24/7 regardless of production. Meter one line for a week before you buy anything.整条链上最便宜的真实节水通常就在这里：溢流漂洗槽不论是否生产都 24 小时长流。在采购任何设备之前，先对一条线连续计量一周。
Tier 2 · Energy storage二级 · 储能 Battery cell & cathode material电芯与正极материал材料 Medium · upstream heavy中等 · 上游偏重
Tier 1 · Assembly一级 · 组装 Final assembly & test (ODM)整机组装与测试（ODM） Low volume · high visibility水量低 · 曝光度高
Low volume, but this is the site journalists and auditors visit, and dormitory water touches people directly. Treat it as a social licence issue, not a footprint issue.水量不大，但这是记者与审核员会到访的厂区，宿舍用水直接关系到人。应视为社会许可问题，而非足迹问题。
Adjacent相邻范畴 Model training & cloud — outside the PC, inside the story模型训练与云——在整机之外，在叙事之内 Disclose separately单独披露
On-device inference is the water argument for an AI PC: work done locally is work not evaporated in a data-centre cooling tower. But the foundation model was trained in the cloud. Do not net these against each other in a footprint — state both, with method, and let the reader compare. Mixing them is the fastest route to a greenwashing finding. 端侧推理正是 AI PC 在水议题上的正面论据：在本地完成的算力，就不必在数据中心冷却塔里蒸发掉。但基础模型是在云端训练的。不要在足迹中把两者相互抵消——分别列示、说明方法，让读者自行比较。混算是被判定"漂绿"最快的路径。
Do the maths one click at a time一次点开一步，把账算清
Three machines. Two are step calculators — open step 1, read it, open the next. The third is live: tick the actions and watch the volume move.三台机器。两台是分步计算器——打开第 1 步，读完，再打开下一步。第三台是实时的——勾选措施，看水量变化。
From withdrawal to replenishment ratio, in five steps从取水到补充比，五步走完
Worked with one illustrative site. Replace every figure with your own meter data.以一个示例厂址演算。请将每个数字替换为你自己的计量数据。
Step 1Measure withdrawal by source按水源计量取水
Every source needs its own meter. If a source is estimated, say so in the data sheet — an estimated source caps your data-quality score at "medium".每个水源都需独立水表。若某水源为估算，须在数据表中注明——存在估算水源，数据质量评级最高只能到"中"。
Step 2Measure discharge, then close the balance计量排水，闭合水平衡
Cross-check against physics: cooling tower evaporation + product moisture + sludge moisture should explain the 300,000 within ±5%. If it doesn't, you have an unmetered loss — find it before you set a target.用物理量交叉验证：冷却塔蒸发 + 产品带水 + 污泥含水，应能在 ±5% 内解释这 300,000 m³。若无法解释，说明存在未计量损失——设定目标前先把它找出来。
Step 3Split consumption by basin按流域拆分耗水
Only the stressed-basin volume drives the water-positive obligation. This single split is what turns a corporate number into an actionable one.只有位于高压力流域的水量，才构成水正效益义务。正是这一次拆分，把一个集团数字变成可执行的数字。
Step 4Count only verified volumetric benefit只计入经核证的水量效益
Each project needs: a hydrological method, a counterfactual, a monitoring plan, and third-party verification. Anything without all four goes in a footnote, not in the total.每个项目都需要：水文方法、反事实情景、监测计划与第三方核证。四者缺一，只能写进脚注，不能计入总量。
Step 5Divide — and read the verdict honestly相除——并诚实解读结论
62% is real progress and it is not water positive. The honest headline is "62% of stressed-basin consumption replenished, target 100% by [year]". Note also that this covers own operations only — supply-chain consumption for an AI PC brand is typically an order of magnitude larger, which is why stop 03 exists. 62% 是真实进展，但它不是水正效益。诚实的表述是："高压力流域耗水已补充 62%，目标为 [年份] 达到 100%"。另需注明：这仅覆盖自有运营——对 AI PC 品牌而言，供应链耗水通常要大一个数量级，这正是第 03 站存在的理由。
One dial, 20% of a utility circuit一个旋钮，节约公用回路 20%
Step 1The identity基本恒等式
Evaporation is fixed by the heat you must reject — you cannot negotiate with thermodynamics. Blowdown is the only variable.蒸发量由必须排走的热量决定——你无法与热力学谈判。唯一可变的是排污量。
Step 2Rearrange变形
Step 3Turn the dial转动旋钮
Evaporation held at 100 m³/day蒸发量固定为 100 m³/天
Note the curve flattens: 3→6 buys 20%, 6→10 buys only another 7.4%, and above 6 you start paying in scale inhibitor, corrosion risk and side-stream filtration. The engineering answer is "as high as your water chemistry safely allows", not "as high as possible". 注意曲线趋平：3→6 可省 20%，6→10 只再省 7.4%；而超过 6 之后，你开始在阻垢剂、腐蚀风险与旁流过滤上付出代价。工程上的答案是"在水质化学安全允许范围内尽可能高"，而非"越高越好"。
Step 4Evidence an auditor accepts审核员认可的证据
- Makeup and blowdown meter logs, monthly, 12 months补水与排污水表记录，按月，12 个月
- Conductivity trend from the tower controller (proves CoC actually held)塔控制器电导率趋势（证明 CoC 确实维持）
- Water chemistry report justifying the new set point支持新设定值的水质化学报告
- Before/after calculation on this same formula, signed by the utility engineer采用同一公式的改造前后计算书，由公用工程工程师签署
"If we do that, what is the %?""如果我们做那件事，是百分之几？"
Each action targets a different, non-overlapping share of intake, so the effects add. Set your own baseline withdrawal and tick what is credible for your site.每项措施针对取水量中互不重叠的不同份额，因此效果可以相加。填入你自己的基准取水量，勾选对你厂区可行的措施。
withdrawal avoided可避免的取水
m³ saved per yearm³／年 节水量
m³ remaining withdrawalm³ 剩余取水量
What must be in scope — and what can wait什么必须纳入范围——什么可以等
Click any cell: basin stress across the top, process water intensity down the side. The verdict tells you whether that supplier site is mandatory scope, standard scope, or annual monitoring only.点击任一单元格：上方为流域水压力，左侧为工艺用水强度。结论会告诉你，该供应商厂址属于强制范围、常规范围，还是仅需年度监测。
Stress class = WRI Aqueduct baseline water stress for the site's basin水压力等级 = 该厂址所在流域的 WRI Aqueduct 基线水压力
Mandatory scope — Tier 1 priority, act this year强制范围——第一优先，本年度行动
Site-level metered data required, quarterly. Basin context study, discharge quality monitoring, and a site water target with a dated plan. If groundwater is used here, escalate to executive review. This is where a brand's water risk actually lives — a supply interruption here stops your product line, not just your report. 要求厂址级计量数据，按季度提供。需完成流域背景研究、排水水质监测，并制定带时间表的厂址水目标。若此处使用地下水，应上报高管评审。品牌的水风险真正就在这里——此处一旦断供，停的是产线，而不只是报告。
Evidence depth: full — meters, permits, lab reports, third-party verification证据深度：完整——水表、许可证、检测报告、第三方核证
Always include必须纳入Five conditions that override everything五个凌驾一切的条件
- Site basin is classified high or extremely high baseline water stress.厂址流域基线水压力为高或极高。
- Site uses groundwater in an over-abstracted basin, at any volume.厂址在超采流域使用地下水，不论水量多少。
- Discharge contains a named regulated pollutant (Cu, Ni, F⁻, CN⁻, TN, PFAS).排水中含受管制的特定污染物（铜、镍、氟、氰、总氮、PFAS）。
- Supplier is inside the top 80% of spend or is single-sourced for a critical part.供应商属于采购金额前 80%，或为关键零件的独家来源。
- There is an existing community, media or permit conflict over water at that location.该地点已存在与水相关的社区、媒体或许可冲突。
Reasonable to defer可合理延后What is few, and fine to leave for year 2占比很小、可留到第二年的部分
Defer only when all of these hold: low-stress basin, sanitary-only or closed-loop water use, below 1% of chain volume, no regulated discharge, not single-sourced.仅当以下条件全部成立时才可延后：低水压力流域、仅生活用水或闭式循环、占链条水量低于 1%、无受管制排放、非独家来源。
- Carton, label and manual printing (water-based ink, low volume)纸箱、标签与说明书印刷（水性油墨、水量低）
- Plastic injection with closed-loop chillers (meter the chiller makeup, that's all)采用闭式冷水机的塑料注塑（只需计量冷水机补水）
- Accessory and cable assembly, dry processes配件与线材组装等干式工序
- Warehousing, distribution centres, retail仓储、配送中心与零售
But state the exclusion. "Excluded, <1% of chain water, low-stress basins, reviewed annually" is a defensible sentence. Silence is not.但必须写明排除项。"已排除，占链条用水 <1%，位于低水压力流域，每年复核"——这是站得住脚的一句话。沉默则不是。
Score yourself out of 24给自己打分，满分 24
Six indicators, four levels each: 0 starting · 2 managed · 3 leading · 4 outstanding. Click the level that is true today, not the one you plan. The band at the bottom updates live.六项指标，每项四级：0 起步 · 2 受控 · 3 领先 · 4 卓越。请点击今天真实的水平，而非计划中的水平。底部等级实时更新。
self-score自评分
You can answer questions but not yet direct capital. Next move: get site-level metered data from the top 20 suppliers by spend, and split consumption by basin.你能回答问题，但还不足以指挥资本投向。下一步：向采购金额前 20 名供应商取得厂址级计量数据，并按流域拆分耗水。
Bands — 0–6 starting · 7–12 managed · 13–18 leading · 19–24 outstanding等级——0–6 起步 · 7–12 受控 · 13–18 领先 · 19–24 卓越
"Outstanding is not the highest number. It is the number a stranger can reproduce from your evidence.""卓越，不是最高的那个数字，而是陌生人能凭你的证据复算出来的那个数字。"
Would this survive an audit? Answer yes or no.这能通过审核吗？请回答是或否。
Nine checkpoints. Press YES only if you could put the document on the table within one working day. Open each row to see exactly which document, who owns it, and how long it must be retained.九个检查点。只有当你能在一个工作日内把文件摆上桌，才按"是"。展开每一行，可查看具体是哪份文件、由谁保管、需保存多久。
evidence confirmed已确认证据
Start with the meter register and the water balance — every other document depends on those two.从水表台账与水平衡开始——其余所有文件都依赖这两份。
1 · Meter register — every source and every major user1 · 水表台账——覆盖每个水源与每个主要用水点
Document: meter list with tag ID, location, type, range, install date, calibration date and next due. Owner: facilities. Retain 5 years. Red flag: a meter with no calibration record invalidates every number downstream of it.文件：水表清单，含位号、位置、型号、量程、安装日期、校准日期与下次校准日。责任方：设施部。保存 5 年。红旗：无校准记录的水表，会使其下游全部数据失效。
2 · Annual water balance that closes within ±5%2 · 年度水平衡，闭合误差 ±5% 以内
Document: Sankey or table of W → uses → D → C, with the unaccounted line shown explicitly. Owner: utility engineering. A balance that closes to exactly 100% is more suspicious than one that closes to 96%.文件：取水→各用途→排水→耗水的桑基图或表格，并显式列出未平衡项。责任方：公用工程。恰好闭合到 100% 的水平衡，比闭合到 96% 的更可疑。
3 · Abstraction / supply permits, valid and within volume3 · 取水／供水许可证，有效且未超量
Document: permit scan, permitted volume, actual withdrawal, headroom %. Owner: EHS. Renewal date tracked at least 12 months ahead.文件：许可证扫描件、许可量、实际取水量、余量百分比。责任方：EHS。到期日须至少提前 12 个月跟踪。
4 · Discharge permit + 12 months of lab results4 · 排污许可证 + 12 个月检测报告
Document: accredited lab reports for every permitted parameter, plus any exceedance report and corrective action closure. Owner: EHS. Retain 5 years. Red flag: missing months.文件：所有许可参数的具资质实验室报告，以及任何超标报告与整改关闭记录。责任方：EHS。保存 5 年。红旗：月份缺失。
5 · Reuse / recycle volume calculation with method note5 · 回用／再生水量计算及方法说明
Document: the formula used, the meters feeding it, and a worked example for one month. Without the method note, two sites will define the same rate differently and your group number is meaningless.文件：所用公式、对应水表，以及某一个月的完整算例。缺少方法说明，两个厂区会用不同定义算同一指标，集团数字随之失去意义。
6 · Supplier water data sheet, signed, for top-80% spend6 · 采购金额前 80% 供应商的用水数据表（已签署）
Document: one row per site — legal entity, address, basin, W by source, D, C, reuse rate, permits, treatment type, data quality flag (metered / estimated), signature of the site EHS manager. Owner: procurement. This is the single highest-leverage document in the whole programme.文件：每个厂址一行——法人主体、地址、流域、分水源取水量、排水、耗水、回用率、许可证、处理方式、数据质量标识（计量／估算）、厂区 EHS 经理签字。责任方：采购。这是整个项目中杠杆最高的一份文件。
7 · Basin risk screening output, dated, with tool version7 · 流域风险筛查结果，注明日期与工具版本
Document: site coordinates, basin name, stress class, tool and version used, screening date. Re-run annually — basin classifications change.文件：厂址坐标、流域名称、水压力等级、所用工具与版本、筛查日期。每年重跑——流域等级会变化。
8 · Replenishment project files with third-party verification8 · 补充项目档案及第三方核证
Document per project: location and basin, hydrological method, counterfactual, monitoring data, verification statement, contract, and the volume claimed for the reporting year. No verification = footnote only.每个项目须有：位置与流域、水文方法、反事实情景、监测数据、核证声明、合同，以及本报告年度声明的水量。无核证 = 只能写脚注。
9 · Restatement log — every change to a published number9 · 重述记录——已披露数字的每一次变更
Document: old value, new value, reason, date, approver. Auditors trust a company with a restatement log far more than one whose history is perfectly smooth.文件：原值、新值、原因、日期、批准人。相比历史曲线完美平滑的公司，审核员更信任备有重述记录的公司。
Year on year — the loop that makes it a system年复一年——让它成为系统的那个闭环
A real sequence, not a decoration. Each year unlocks the next: you cannot set a credible basin target before you know which basins you are in.这是一条真实的时序，不是装饰。每一年都为下一年解锁前提：在不知道自己身处哪些流域之前，无法设定可信的流域目标。
Year 1 · Measure第一年 · 测量Meters, map, baseline水表、地图、基线
- Meter register + water balance for all own sites完成所有自有厂区的水表台账与水平衡
- Geolocate top-80%-spend Tier-1 sites, run basin screening定位采购金额前 80% 的一级供应商厂址，完成流域筛查
- Issue the supplier water data sheet; accept estimates, but flag them下发供应商用水数据表；接受估算值，但须标注
- Publish baseline W, D, C — absolute and by basin. No targets yet.披露基线取水、排水、耗水——绝对值并按流域拆分。暂不设目标。
Year 2 · Reduce第二年 · 减量Cascades, cycles, reuse loops级联、浓缩倍数、回用环路
- Execute the no-capex list first: rinse valves, tower set points, leak survey先执行零投资清单：漂洗阀、冷却塔设定值、管网查漏
- Set absolute basin-specific reduction targets on the Year-1 baseline以第一年基线为准，设定分流域的绝对减量目标
- Move 20% of Tier-1 data from estimated to metered将 20% 一级供应商数据由估算升级为实测
- Add water criteria to supplier scorecards and new-part qualification将水指标纳入供应商记分卡与新料号认证
Year 3 · Replenish & assure第三年 · 补充与鉴证Projects in the basins you actually use在你真正取水的流域里做项目
- Fund replenishment only in your own priority basins, with verification built into the contract只在自身优先流域投入补充项目，并在合同中内置核证条款
- Take own operations and top suppliers through external assurance自有运营与主要供应商接受外部鉴证
- Publish the replenishment ratio per basin — including the ones below 100%按流域披露补充比——包括低于 100% 的那些
- Restate Year 1 if the method improved, and log why若方法改进，重述第一年数据并记录原因
Sources to work from可依据的资料来源
Frameworks, tools and standards behind the logic on this page. Always check the current version before you cite one — methods here move fast.本页逻辑所依据的框架、工具与标准。引用前请务必确认最新版本——这一领域的方法更新很快。
All quantitative ranges on this page are study values for orientation only, compiled for benchmark and reference practice. Replace them with measured, site-specific data before any external claim.本页所有量化区间均为仅供定位参考的研究值，为对标与参考实践而整理。在对外做出任何声明前，请替换为实测的、厂址特定的数据。