行動不等於成果：台灣金融與製造如何跨越自然影響宣稱的證據邊界

EIS-024 的全球命題是「行動不等於成果」；台灣市場還要追問：誰的行動、哪個場址的變化、哪家機構可以據此宣稱什麼？歐洲證券及市場管理局（ESMA）2026 年 10 月 12 日的影響宣稱主題說明，以及 10 月 13 日 TNFD、SBTN、Nature Positive Initiative 的協同更新，分別處理金融溝通和自然評估、目標、衡量。它們不是同一規則，更沒有直接創設台灣法定申報義務。 本篇以台灣投信基金、上市櫃企業、製造供應鏈及特定場址為單位，建立「投入／活動 → 產出 → 環境狀態觀察 → 機構貢獻 → 可授權宣稱」的證據階梯。資金已撥、樹已種、供應商已改善或目標已驗證，都應準確記錄；但這些事實不能單獨回答生態是否改善，更不能證明是特定投資人或企業造成。台灣版為 SNN.TW 獨立編輯解讀，保留本地證據與責任。

EIS-024's global proposition is that action is not outcome. For Taiwan the next questions are whose action, which site's observed change, and which institution has authority to claim what. ESMA's 12 October 2026 thematic note on impact claims and the coordinated 13 October updates from TNFD, SBTN and the Nature Positive Initiative address different problems: financial communication on one side, nature assessment, target setting and measurement on the other. They are not a single regulatory programme and do not themselves create a Taiwan statutory filing duty. This independent SNN.TW interpretation follows an evidence ladder: resource or activity, output, observed environmental state, institutional contribution, and authorised public claim. Money disbursed, trees planted, supplier pressure reduced or a target validated are meaningful records of progress. None alone answers whether the ecosystem improved, or whether a particular company or investor caused the improvement. We read the issue through Taiwan fund products, listed and OTC issuers, manufacturing supply chains and place-specific projects, retaining Taiwan evidence and editorial responsibility. The global canonical supplies a thesis; the local analysis tests what it can support in Taiwan's market and institutional setting.

全球制度訊號有兩條。ESMA 的 thematic note 指出投資標的產生的影響，與投資者藉融資或議合所作的額外貢獻不同；結果的衡量和宣稱應說明意圖、途徑、可衡量性與限制。ESMA 自己明言此系列不產生新監理要求。TNFD LEAP 2.0 是依賴、影響、風險與機會的評估路徑；SBTN Nature 2.0 是目標及行動方法；NPI 的 State of Nature Indicators and Metrics 是生態系範圍、狀態、物種滅絕風險與族群的衡量基礎。方法更能銜接，不代表評估、目標、結果及歸因可合併。 台灣制度須另看本地正式文本。金管會投信基金公開說明書準則第 13-1 條要求適用 ESG 主題基金揭露永續投資目標、衡量指標、策略、風險及定期評估等事項；證交所 2026 年 5 月修正的上市公司永續報告書作業辦法、櫃買中心上櫃公司規範與年報準則另有各自對象、年度與欄位。林業保育署《臺灣 TNFD 在地操作指引》支持在地評估，屬自願工具；它與歐洲 thematic note 都不能替代台灣法規、場址許可或企業的揭露責任。

The institutional signals follow two routes. ESMA distinguishes the positive effect generated by an investee from the investor's additional contribution through financing or engagement. An impact claim should explain intention, mechanism, measurement and uncertainty. ESMA expressly says its thematic notes do not create new regulatory requirements. TNFD's LEAP 2.0 helps organisations locate and assess nature-related dependencies, impacts, risks and opportunities. SBTN Nature 2.0 provides methods and resources for targets and implementation. The Nature Positive Initiative's State of Nature Indicators and Metrics provide a measurement foundation for ecosystem extent and condition, species extinction risk and populations. Better alignment permits appropriate input reuse; it cannot collapse assessment, target, intervention, observed result and attribution into one status. Taiwan's legal and policy state must be read from local primary instruments. Article 13-1 of the FSC regulation on securities investment trust fund prospectuses sets required particulars for covered ESG theme funds, including sustainability objectives, evaluation indicators, strategy, risk and periodic assessment. The TWSE's May 2026 rules for listed-company sustainability reports, TPEX requirements for OTC issuers and the public-company annual report regulation have different entities, years and fields. The Forestry and Nature Conservation Agency's Taiwan TNFD operational guide supports local assessment as a voluntary method. Neither that guide nor ESMA's European note replaces Taiwan law, site permits or an issuer's responsibility for its own disclosure.

台灣的自然證據尤其受場址左右。同樣是種植或濕地復育，位於重要濕地、河口、農地或工業區，其基線、水文、棲地連通、物種與合法使用條件均不同。2023 年國土生態綠網圖資能協助辨識空間脈絡，卻只是當時發布的圖資；2026 年場址判讀仍應核對最新圖層、現場調查及觀察期間。濕地保育法、野生動物保育法與水污染防治法各自規範特定行為及對象；遵守某項法規既不自動證成「自然正向」，也不由一張圖或植樹張數推定。 成果證據邊界是從「做了什麼」跨到「實際變了什麼」。要事先界定地點、比較基線、指標、採樣頻率、時間尺度、方法變動與不確定性；許多自然反應在專案結束時尚未可見。壓力減少（例如排放下降）、回應完成（例如修復施工）、生態狀態改善（例如棲地條件與物種族群變化）各有意義，不應互相代換。若觀測數據來自不同年份、季節或空間邊界，企業不能挑出有利的單一數值代表整個流域或供應鏈。

Nature evidence in Taiwan is unusually sensitive to place. A tree planting or wetland project at an important wetland, estuary, farmland site or industrial location may have different baseline conditions, hydrology, habitat connectivity, species and permissions. The 2023 National Ecological Network spatial-layer announcement helps identify geographic context, but the historic launch is not a 2026 outcome. A site analysis must check the latest layers and field observations across a stated period. The Wetland Conservation Act, Wildlife Conservation Act and Water Pollution Control Act govern particular places, species or discharges within their own scope. Compliance with one of those instruments is not automatically a nature-positive outcome; a map or planted-tree count cannot prove compliance either. The Outcome Evidence Boundary separates a record of action from an observation of real-world change. The organisation needs an appropriately defined location, baseline, indicator, sampling interval, observation horizon, method version and uncertainty. Ecological responses may lag the project closeout by years. A reduction in pressure, completion of a response and measured ecosystem improvement each answer a different question. Seasonal variation, upstream changes and shifts in the measurement boundary can alter an apparent trend. A favourable isolated figure should not stand for a whole catchment or supply chain. Until observations support a result, a company can accurately describe implementation and a target in progress without claiming that the ecological outcome has occurred.

第二道是「歸因到宣稱」邊界。假設一處流域監測到棲地狀態改善，仍須分析天候、上游治理、其他地主與政府措施等共同因素。投資人可以證明資金投入與契約條款，企業可以證明施工、採購或減壓措施；要稱「我們造成改善」還須與特定貢獻及可能的反事實比較相稱。缺乏可識別增量時，適當敘述可能是「參與」「支持」「觀察到變化」，而不是「實現」「確保」或「使生態恢復」。 台灣投信產品若在銷售文件稱具自然影響，基金經理人應把持倉、資金用途、議合紀錄與被投資事業的成果分開。發行人與工廠也應將供應商填答、場址照片、採樣原始紀錄、法定許可和確信範圍分開。客戶要求的指標可能進入契約或採購評分，但客戶問卷不是台灣法定揭露。NPI 指標可建立共同測量語言，TNFD／SBTN 產出可重用輸入；重用資料不等於重用某一機構的成果或因果結論。

A second, Attribution-to-Claim Boundary applies even after a state-of-nature change is observed. An improving habitat may also reflect rainfall, upstream action, other landowners or public measures. A lender can document disbursement and covenants; an operating company can document procurement and restoration works. Saying that it caused an improvement requires a proportionate basis for contribution, additionality and plausible alternatives. Where the incremental effect is not identifiable, verbs such as participated, supported or observed may be appropriate; achieved, ensured and restored assert more. The stronger public claim needs a stronger chain of evidence and an authorised decision. For a Taiwan fund marketed with a nature impact proposition, the manager should distinguish holdings and use of proceeds from engagement records, investee results and the fund investor's additional contribution. An issuer and plant should distinguish supplier questionnaires, photos, sampling records, statutory permissions and assurance scope. A buyer's metric may enter a contract or procurement score, but a customer questionnaire is not a Taiwan statutory report. NPI metrics can provide a common measurement language; TNFD or SBTN work can provide reusable inputs. Reusing an observation does not automatically transfer a causal conclusion or an assurance opinion across organisations and claim contexts.

一條可複核的台灣 Evidence Chain 應把每個主張拆成資料物件和支持範圍。第一層保存專案決議、契約、付款與供應商活動；第二層保留場址座標、國土生態綠網圖層版本、適用法規、許可及基線調查；第三層把生態系範圍、狀態、物種風險與族群的觀察方法、季節、抽樣及品質限制連到原始讀值。第四層記錄外部共同因素、反事實假設、投資人與執行者的角色，最後由具權限者確定公開語句。 本篇 18 筆外部來源分別支援全球制度訊號、台灣金融與上市櫃正式規則、台灣自願在地指引、場址法規與歷史空間脈絡；`.net` 正典和本站都不計入門檻。來源的發布日期、驗證日期、支援範圍、state 與限制一同保存。台灣正式法規可支持 binding-state claim；新聞公告只支持公告事實；2023 圖資只支持歷史發布與定位線索。這種證據鏈也讓後續稽核者看見資料缺口，而不把測量框架誤作認證。

A reviewable Taiwan Evidence Chain separates the objects behind each claim. The first layer contains project decisions, contracts, payment and supplier action logs. A second layer retains site coordinates, spatial layer version, applicable statute, permit and baseline survey. A third links ecosystem extent and condition, species risk and population observations to raw data, season, sampling method and quality limitations. A fourth records shared external drivers, counterfactual assumptions and the respective roles of investor, operator and public authority. Only then should an authorised function approve the public wording and its limits. Versioning must preserve earlier methods and corrections instead of silently replacing inconvenient observations. The 18 independent external sources in this edition support distinct propositions: seven global institutional developments, Taiwan's formal finance and issuer rules, voluntary local assessment guidance, site-specific statutory context and historic spatial context. The.net canonical and SNN.TW are excluded from the count. Each record retains institution, document, publication and verification dates, support scope, state and limitation. A formal Taiwan regulation may support a binding-state proposition within its scope; a release supports an announcement; the 2023 map launch supports historical availability. Neither an analytical framework nor an information paper certifies a company outcome. This separation makes missing data visible to an auditor instead of treating a measurement method as a seal of approval.

對台灣資本市場，壓力不是把每個產品都改稱「影響基金」，而是現有基金目標、衡量方式及定期揭露語句是否有相稱資料。金管會基金公開說明書準則對適用商品的義務，與 ESMA 對歐洲市場溝通的示例，應平行核對、分別適用。投信投顧、銷售與法遵若引用標的企業的復育成果，須指出投資者的資金或議合所能支持的較窄結論。 對台灣製造業與上市櫃公司，國際買方與貸款人可能要求自然風險評估、目標、場址觀察及宣稱證據，經採購、融資與永續報告準備流程傳導到供應商。然而是否屬台灣法定欄位，仍依上市／上櫃規則、年報準則、公司及報導年度判斷。開發案、濕地或排放場所的許可與監測，另由場址相關法律決定。供應鏈中的同一筆採樣可以多次引用，但不能因不同買方共用資料便轉移發行人、基金經理人或查驗者的責任。

For Taiwan capital markets, the pressure is to make existing fund objectives, metrics and periodic wording proportionate to the underlying records. The FSC's formal prospectus requirements for covered products and ESMA's examples for European communications need parallel, separate interpretation. A Taiwan manager, distributor and legal reviewer quoting an investee's restoration result should identify the narrower conclusion supported about the investor's own financing or stewardship. The investor's holding of a security and the investee's ecological performance are separate evidence objects. An assurance opinion, if any, applies only to its stated scope. For Taiwan manufacturing and listed or OTC companies, international buyers and lenders may ask for nature-risk assessment, targets, site observations and claims substantiation. Those requests can flow through procurement, financing and sustainability-report preparation to upstream suppliers. Whether a particular item is a Taiwan statutory disclosure depends on the applicable TWSE, TPEX or annual-report instrument, the company and reporting year. Permits and monitoring at a protected wetland or discharge site arise under different, place-specific laws. A single sample can be reused across a supply chain if its provenance travels with it; reuse cannot transfer the issuer's, fund manager's or assurance provider's responsibility for an institution-specific conclusion.

SNN 編輯與揭露前證據基礎設施觀點（SNN Editorial / Pre-Disclosure Evidence Infrastructure Perspective）。事實邊界：ESMA 於 2026 年 10 月 12 日發表影響宣稱主題說明，但不創設新監理義務；TNFD、SBTN、NPI 於 10 月 13 日協同更新自然評估、目標與狀態衡量工具。這些全球方法不等於台灣強制法規，也不證明任何企業已有自然成果或可歸因貢獻。 傳導機制是歐洲金融市場影響宣稱指引與全球自然衡量框架更新，經由投資產品銷售、跨境融資、國際買方採購合約與供應鏈問卷，進入台灣投信基金、上市櫃公司及製造場址的資料要求，影響台灣資本市場的宣稱審查、供應商證據成本及確信準備。應在揭露前把投資活動、場址措施、實際生態觀察、機構增量貢獻及公開宣稱逐層接上，同時分開台灣基金公開說明書與永續報告書的正式規則、自願 TNFD 指引、個別客戶條款。 投信的財務與法遵部門應在基金公開說明書發布前核對投資目標、衡量指標、持倉及投資人貢獻的證據檔案。台灣製造商的採購與永續部門應在回覆國際買方前保存場址座標、供應商活動紀錄、基線和方法版本。場址主管與確信人員應在聲稱濕地或物種改善前抽測原始觀察資料、期間、外部共同因素及確信範圍。董事會與內部稽核應在重大自然正向聲明核准前檢查例外清單、許可文件、歸因假設和簽核權限；外部查驗者依委任範圍負責，不能替發行人作聲明決策。 禁止推論：撥款不等於自然成果；植樹數量不等於棲地恢復；目標驗證不等於成果發生；觀察到改善不等於某企業或投資人造成；TNFD 或 NPI 指標不等於台灣法定義務；客戶問卷不等於台灣法定申報；資料共用不等於因果結論可共用。這是 SNN.TW 的 editorial interpretation，不能取代主管機關、基金經理人、發行人或具資格確信者的判定。

SNN Editorial / Pre-Disclosure Evidence Infrastructure Perspective. Fact boundary: ESMA published impact-claims thematic guidance on 12 October 2026 without creating new regulatory requirements. TNFD, SBTN and NPI coordinated updates to nature assessment, target setting and state-of-nature measurement on 13 October. These global approaches do not constitute a Taiwan mandatory rule or prove an achieved ecological outcome or an attributable contribution by any company. The transmission mechanism is concrete: European financial-market impact-claims guidance and updated global nature frameworks flow through fund distribution, cross-border finance, international buyer procurement contracts and supplier questionnaires into data requests for Taiwan funds, listed companies and manufacturing sites. This can affect Taiwan capital-market claims review, supplier evidence cost and assurance preparation. Before disclosure, the evidence chain should separately connect investment activity, site intervention, observed ecological condition, incremental contribution and authorised public wording. The FSC prospectus rule, TWSE or TPEX issuer rules, voluntary Taiwan TNFD guidance and customer contract terms each retain their own state and scope. Before publishing a fund prospectus, the finance and legal leader should reconcile investment objectives, indicators, holdings and an evidence file for investor contribution. Before replying to an international buyer, a Taiwan manufacturer's procurement and sustainability teams should preserve site coordinates, supplier action records, baseline and method version. Before asserting wetland or species improvement, the site operator and assurance practitioner should sample raw observation data, observation periods, shared drivers and the assurance scope. Before approving a material nature-positive statement, the board and internal auditor should examine the exception log, permit documents, attribution assumptions and delegated approval. An outside verifier remains responsible within the engagement scope and does not assume an issuer's claims authority. Prohibited inference: disbursement is not a nature outcome; a tree count is not habitat recovery; target validation is not achieved impact; observed improvement does not prove that a named company or investor caused it; TNFD or NPI metrics are not Taiwan statutory duties; a buyer questionnaire is not a Taiwan filing; shared data are not a shared causal conclusion. This is SNN.TW editorial interpretation and does not replace decisions by a regulator, fund manager, issuer or qualified assurance practitioner.

後續應觀察四個不同時鐘：自然基線建立的時間、介入措施完成時間、生態指標可見變化的時間，以及基金或發行人發表宣稱的時間。若最後一個時鐘跑在前面，應把語句降到當時證據能支持的範圍，並列明仍需追蹤的場址、季節與反事實。NPI 指標與 GBF 監測對接，可改善共用觀察語言；其資訊論文並非國家批准個別企業宣稱。 台灣責任編輯在來源更新或本地規則變動時，應逐筆更新 state、日期與限制，再由人類審核是否改寫結論。SNN.TW 保留本地標題、publisher、本站 canonical、雙語同頁與台灣證據鏈；全球正典提供論點起點，不取代這份獨立解讀。

Four clocks require continued attention: when the baseline was established, when an intervention ended, when a state-of-nature indicator may show a reliable change, and when a fund or issuer proposes to make a claim. If the public-claim clock runs ahead of observation, the wording should be narrowed to the evidence available at that date, with the site, season and counterfactual still under review identified. NPI's alignment with GBF monitoring may improve reuse of observations, but its information paper is no national approval of an organisation's claim. A cross-company portfolio number cannot erase place-specific ecological variation. When an official source changes or a Taiwan rule is amended, the responsible SNN.TW editor should update the evidence state, date and limitation before any conclusion is revised, with human review of the change. The page retains its local title, Taiwan publisher, SNN.TW canonical, bilingual same-page content and Taiwan evidence chain. Its relationship to the.net global canonical is independent regional interpretation. Neither machine summaries nor a reusable data object can escalate an observed result into a verified outcome or a legal duty without authorised review.