Issue 2's verification arc examined the compliance and verification layer of climate finance. The architecture sits on something. The compliance and verification work that platforms like Persefoni and Watershed do, that audit firms apply International Standard on Sustainability Assurance 5000 against, that operators reconcile against the eighteen-month-window Playbook, all of it depends on a deeper layer most practitioners interact with only through their compliance platforms: the data substrate. Earth observation foundation models, atmospheric monitoring satellites, IoT sensor networks, biological monitoring stacks, supply-chain traceability infrastructure, modelled and inferred data. The substrate is the input layer the compliance architecture operates on.

Practitioners reading the compliance layer without understanding the substrate underneath will be surprised when substrate-layer changes cascade into their disclosure data without warning. We name the antagonist the compliance-platform reader, an operator that interacts with climate data only through the compliance platform's interface, treats the platform as the data source, and does not know what the platform's data inputs depend on. We name the protagonist the substrate reader, an operator that maps the data infrastructure beneath the compliance layer and can read how substrate changes propagate upward into disclosure data. This Practitioner's Guide is for substrate readers in training. The five moves below sequence the substrate-mapping discipline.

Move 1: Identify the substrate layers

Six categories matter. Earth observation includes commercial providers (Planet, Maxar, Capella Space, Spire Global) and public-data infrastructure (Copernicus Sentinel-2, Landsat via the United States Geological Survey, AlphaEarth Foundations via Google Earth Engine). Earth observation foundation models sit on top of this layer: IBM and NASA's Prithvi-EO-2.0, the Allen Institute for AI's OlmoEarth, and IBM's TerraMind collaboration with the European Space Agency each produce embeddings of the planet's surface optimised for downstream tasks. Atmospheric monitoring includes methane-specific infrastructure (MethaneSAT, GHGSat, MethaneAIR via Environmental Defense Fund) and broader atmospheric chemistry observation through the Copernicus Atmosphere Monitoring Service.

IoT sensor networks include industrial emissions sensors at production facilities, agricultural sensors across managed land, ocean buoys for surface temperature and acidification, and soil-carbon ground-truthing networks. The density of these networks is growing rapidly as sensor unit costs fall and as wireless connectivity penetration deepens into agricultural and industrial sites that were previously unmonitored. Biological monitoring includes environmental DNA sampling (the gold standard for site-level species inventories), acoustic monitoring through systems like Google's open-source Perch 2.0 bioacoustics workflows, camera-trap networks aggregated through Wildlife Insights, and citizen-science platforms such as eBird and iNaturalist that contribute large volumes of opportunistic species-occurrence data. Supply-chain traceability includes commodity-flow tracking (Trase Earth, Provenance) and sovereign traceability platforms (Brazil's SeloVerde, Ghana's Cocoa Management System, Malaysia's MSNR Trace) that the Foundation Not Forest Case Study catalogued as the operational infrastructure inherited by TNFD and ISSB-nature. Modelled and inferred data includes carbon-accounting factor databases maintained by the GHG Protocol, biodiversity intactness indices, ecosystem-service valuations, and climate scenario outputs from the Intergovernmental Panel on Climate Change and the Network for Greening the Financial System.

For each category, the practitioner needs to know three things: what data the substrate produces, at what frequency and granularity it produces it, and what its known limitations are. The substrate-mapping discipline begins as a six-row inventory with three columns. The inventory is the foundation of every subsequent move.

Move 2: Map the substrate to your compliance use case

For a given disclosure (Scope 1 emissions, Scope 2 emissions, Scope 3 categories, Taskforce on Nature-related Financial Disclosures biodiversity metrics, European Union Deforestation Regulation supply-chain due diligence, Corporate Sustainability Reporting Directive ESRS E4 biodiversity, ISSB IFRS S2 climate-related disclosures), trace which substrate layers feed which data fields. A corporate Scope 3 Category 11 calculation, the use phase of sold products, depends on three substrate inputs: sectoral activity data from agencies like the International Energy Agency or the United States Energy Information Administration, emission factors from the Environmental Protection Agency or the Intergovernmental Panel on Climate Change, and product-use modelling assumptions selected by the operator. Each substrate input has its own update cadence, its own provenance, and its own embedded assumptions. A TNFD biodiversity disclosure pulls from a different stack: site-level biological monitoring data (eDNA, acoustic, camera), geospatial habitat data (Earth observation), and modelled biodiversity intactness layers. An EUDR supply-chain disclosure depends on the geolocation data from sovereign traceability platforms, satellite-derived forest-cover data, and operator-collected commodity-flow data.

The practitioner action is to build a substrate-to-disclosure map for each material disclosure, identifying the substrate-layer dependency for each data point. The map should explicitly note where the operator depends on a single provider for a critical input, because single-provider dependencies are where substrate-layer changes propagate fastest into the operator's compliance architecture.

Move 3: Assess substrate quality and bias

Each substrate layer has known quality issues practitioners should account for. Earth observation faces cloud-cover gaps over equatorial regions, sensor calibration drift over a satellite's operational life, and model-version dependencies. Different foundation models produce different embeddings for the same location: Prithvi-EO-2.0 and AlphaEarth Foundations are not interoperable at the embedding layer, as the Cloud-Native Geospatial Forum's technical-debt critique has documented. Atmospheric monitoring is more fragile than the substrate's narrative implies. MethaneSAT lost contact on 20 June 2025 and Environmental Defense Fund now considers it unlikely the satellite can be recovered. The methane-detection substrate that operators with oil-and-gas Scope 1 disclosures planned to depend on shrank by one critical instrument inside six months, with no warning to the compliance-platform user.

IoT sensor networks face sensor drift, calibration challenges, and network coverage gaps that compromise data continuity at facility level. Biological monitoring faces species-detection bias (eDNA misses species without reference sequences in the relevant database; acoustic monitoring misses species whose vocalisations are not in the training set), spatial sampling gaps, and the reference-database currency problem (new species and new genetic markers update on lags). Supply-chain traceability faces chain-of-custody gaps at handover points and self-reporting bias from intermediate actors. Modelled and inferred data faces assumption sensitivity that audit firms struggle to verify under substantive testing procedures, as the Audit Gap Data Read documented. The Earth observation foundation-model wave creates further fragility through convergence-and-divergence dynamics: providers update their models on different timelines, and the same underlying scene can produce divergent outputs across providers in any given quarter.

The substrate-quality discipline for the practitioner is to maintain a known-issues log for each substrate input, updated quarterly, and to flag in the substrate-to-disclosure map which disclosure fields are most exposed to substrate-quality risk. Practitioners that skip this step will be surprised when the next MethaneSAT-class event removes a substrate input their compliance disclosures depended on.

The systemic exposure is worth naming explicitly. The climate-data substrate is concentrated. A small number of providers operate the load-bearing layers: a handful of commercial Earth observation operators, three or four major Earth observation foundation models, a single provider for several biodiversity-monitoring categories, a small set of supply-chain traceability platforms anchoring most of the operator-visible deforestation tracking. The audit decade's verification architecture assumes substrate inputs are independent of each other. They are not. A provider outage at the foundation-model layer cascades into every downstream platform consuming that model's embeddings. A reference-database update at the biological-monitoring layer cascades into every operator whose TNFD disclosure depends on species presence-absence data. A regulatory change at the supply-chain-traceability layer cascades into every operator with EUDR exposure. The practitioner's substrate-quality discipline must include single-provider-dependency risk, propagation-speed estimates for substrate-layer changes, and an explicit list of substrate-layer events that would force disclosure restatement. Audit firms in the audit-decade architecture will increasingly ask about substrate-layer continuity assumptions as part of reasonable-assurance procedures; operators that cannot answer those questions clearly will receive emphasis-of-matter paragraphs that the Eighteen-Month Window Playbook flagged as the diagnostic signal of substrate-mapping immaturity.

Move 4: Integrate substrate into the compliance platform stack

The compliance platforms catalogued in The Missing Layer Case Study (Persefoni, Watershed, Sweep, Workiva, AuditBoard, Diginex with Plan A) consume substrate data through three integration modes: native application programming interfaces with substrate providers, third-party data-broker integrations, and manual uploads from operator-side teams. Each mode has different update latency, different data-lineage characteristics, and different exposure to substrate-layer change.

The practitioner action is to audit the platform stack's substrate inputs. Which substrate categories are wired through native APIs? Which depend on third-party data brokers (and is the broker's data source publicly identifiable)? Which substrate data comes through manual uploads that bypass automated lineage tracking? When a substrate provider revises methodology, as the Voluntary Carbon Market did with VM0045 v1.2 per the Gate Held Case Study, does the platform propagate the revision into the operator's prior-year baselines or only forward? When a substrate provider goes offline, as MethaneSAT did, does the platform alert the operator or silently substitute an alternative source? The integration architecture between substrate and compliance platform is where data lineage gets tested under reasonable assurance.

The biological-monitoring substrate is particularly worth scrutinising because the integration patterns are still emerging. NatureMetrics serves more than 600 companies across 110 countries with eDNA-derived biodiversity metrics, having raised a USD 25 million Series B in January 2025 and launched an AI-powered nature risk assessment tool in September 2025; the company reports that its eDNA platform has prevented over USD 2.8 billion in potential project delays, regulatory violations and ecosystem restoration costs across its client base. The Wildlife Insights platform, a Google Earth Engine-hosted collaboration with Conservation International, Smithsonian Conservation Biology Institute, the Wildlife Conservation Society, World Wide Fund for Nature and the Zoological Society of London, processes camera-trap data using Google's open-source SpeciesNet model trained on 65 million images across 2,000 species labels. Most compliance platforms do not yet have native application programming interface integrations with either NatureMetrics or Wildlife Insights. Operators reporting TNFD-aligned biodiversity disclosures are running these integrations manually, with the data-lineage exposure that manual integration implies. The substrate-to-platform integration debt is real and unevenly distributed across the climate-disclosure stack: emissions-data integrations are mature, biodiversity-data integrations are nascent, atmospheric-monitoring integrations are uneven, and supply-chain traceability integrations are sector-specific. Practitioners that map the integration mode for each substrate input avoid the failure mode of treating all substrate inputs as equally trustworthy.

Move 5: Where the substrate is heading 2026-2030

Three forces are reshaping the substrate over the launch decade. The first is foundation-model consolidation. The Earth observation foundation-model market is converging around a handful of provider stacks (Prithvi-EO from IBM and NASA, AlphaEarth from Google DeepMind, OlmoEarth from the Allen Institute for AI, TerraMind from IBM and the European Space Agency, Clay and TESSERA from the open community), and consolidation will likely continue through 2027-2028 as the unit economics of training and serving the models reward scale. The second is sensor-density growth. IoT sensor networks are expanding by orders of magnitude as costs fall, with industrial-emissions sensors, agricultural soil sensors, and ocean monitoring buoys all on declining cost curves. The substrate of 2028 will produce materially more raw data than the substrate of 2026, but the methodology layer that turns raw data into assurable disclosures will lag the raw-data growth, replicating the audit gap dynamic at the substrate-to-methodology interface.

The third force is substrate-platform integration tightening. Compliance platforms in 2026 ship with limited native substrate-data feeds; compliance platforms in 2028 will ship with materially more native integrations as the substrate-and-compliance categories converge commercially. The practitioner planning the substrate-mapping discipline for the architecture of 2027-2030 should expect the integration to deepen but not to remove the practitioner's responsibility to know what the substrate produces. The compliance platform that automates substrate integration without surfacing substrate-layer changes to the operator simply moves the data-lineage opacity from the operator's manual uploads to the platform's automated pipelines.

The substrate reader closes Issue 2

The compliance-platform reader interacts with climate data through one interface and treats the platform as the data source. The substrate reader knows what feeds the platform and can read how substrate changes propagate upward. The bifurcation between operators with deep substrate mapping and operators dependent on platform-as-black-box becomes material when substrate-layer changes cascade unexpectedly, and the MethaneSAT loss is the proof that those cascades are no longer hypothetical. The compliance-platform reader, by definition, will not notice that one of the substrate inputs feeding methane-related disclosure has disappeared until either the platform notifies them or the next reporting cycle surfaces the gap; the substrate reader noticed in June 2025.

Issue 2 examined the compliance and verification layer that sits above the substrate. The four-piece verification arc (architecture, operators, limits, action) framed the compliance layer; this Practitioner's Guide pivots downward to the input layer the architecture operates on. The publication's editorial position, closing Issue 2 and looking ahead to Issue 3, is that the substrate is the input layer for every disclosure regime Issue 2 examined and for every Issue 3 topic the launch corpus will cover, including biodiversity finance, transition-finance taxonomies, nature-credit registries, central-bank stress-test cycles and next-generation reporting standards. The compliance-platform reader sees those Issue 3 topics as separate stories. The substrate reader sees them as variations on one structural question: what data feeds the disclosure, and what does that data depend on? The substrate-mapping discipline that this Practitioner's Guide sequences is the same discipline that Issue 3 will require for biodiversity-credit registries, for transition-finance taxonomies, and for central-bank stress-test inputs. Five moves: identify the layers, map to disclosure, assess quality, integrate with platform, anticipate where the substrate is heading. The substrate reader runs them. The compliance-platform reader, by the time they notice the substrate matters, is responding to a cascade they could have anticipated. The audit decade will be navigated more cleanly by the first than by the second.

This Practitioner's Guide closes Issue 2. The next issue moves forward into the topics the keystone Bifurcation Decade Opinion named in its forward editorial pointer: biodiversity finance, transition-finance taxonomies, nature-credit registries, central-bank stress-test cycles, and next-generation reporting standards. Each of those topics rests on a substrate the practitioner now has a discipline for mapping. The substrate reader who finishes Issue 2 is ready for Issue 3.