In production with U.S. community banks
v3 Signed certificate and examiner bundle, per portfolio

Climate credit intelligence
for the mortgages you
hold and sell.

CLIMA prices climate exposure into every loan in your portfolio. It surfaces the agency LLPA and aggregator-overlay drag your tape carries today, sorts each loan hold-versus-sell, and matches sale candidates to the buyer most likely to take them.

Independent validation · 4,950-loan community-bank portfolio

20.0bps
Climate pricing spread
Worst decile vs. best decile
$20.4M
Agency LLPA exposure
Surfaced loan-by-loan
1.81s
End-to-end runtime
Full control-layer pass
0.853
Disaster-default AUC
Validated on 14.8M Freddie Mac loans
35
Encoded buyer rules
Each with cited source and confidence
10
Files per examiner bundle
SHA-256 signed, fully replayable

What agency underwriting doesn't tell you

DU and LPA report whether a loan clears the agency box. They don't quantify the climate exposure, agency LLPA drag, or aggregator-overlay conditions already sitting in the tape you hold today. Below are the three figures CLIMA surfaces on a representative community-bank portfolio.

20bps
Climate pricing spread

Spread between the best and worst climate-pricing decile within the same portfolio. Current models do not sort on it; the secondary market does.

$20.4M
Agency LLPA exposure

Price-point adjustments Fannie Mae and Freddie Mac would apply if the tape were sold as-is. CLIMA itemizes each adjustment and flags which are remediable.

338
Aggregator overlay conditions

Conditions added by PennyMac, Wells Fargo, Chase, AmeriHome, and other correspondent buyers above the agency baseline. CLIMA evaluates them per loan.

Five risk layers, all federally sourced

Each property carries a 0–100 reading on every layer plus a calibrated composite. The composite is the input to pricing, hold-versus-sell, and buyer routing. Every component traces to a federal data source documented in the cert manifest.

Flood

FEMA NFHL zones (A, AE, V, VE…), NFIP policy density, residual-market load.

FEMA NFHL · NFIP · FEMA NRI

Wildfire

USFS Wildfire Hazard Potential, fuel-model classification, WUI proximity.

USFS WHP · FEMA NRI · CAL FIRE

Hurricane & Wind

FEMA NRI hurricane & coastal-flood layers, NOAA storm history, residual-market exposure.

FEMA NRI · NOAA SED · TWIA / Citizens

Extreme Heat

NOAA daily-max climatology, heatwave frequency, projected warming through 2050.

NOAA NCEI · NASA SLR · FEMA NRI
New

PFAS Contamination

Proximity to 171k+ EPA-tracked PFAS sites, calibrated XGBoost contamination probability, CERCLA tier flag.

EPA PFAS Analytic Tools · UCMR-5
Why PFAS belongs in a credit risk score

Lender liability for chemical contamination is now a CERCLA secondary-liability concern, and EPA's 2024 PFAS hazardous-substance designation puts mortgaged property within the disclosure perimeter. Proximity to a confirmed PFAS site can trigger property-value discounts, insurance refusals, and material remediation costs — all of which feed default risk. CLIMA flags it before the loan closes.

Inside the control layer

Every loan passes through the same five stations a secondary-marketing desk runs, with climate, LLPA, and aggregator-overlay logic encoded at each stage. Full pass on a 4,950-loan portfolio: 1.81 seconds.

01

Surveil

Continuous monitoring of climate drift, hazard upgrades, PFAS detections, and concentration breaches across the held portfolio. Alerts are deduplicated and ranked by severity.

02

Decide

An AUS unlock simulator evaluates Day-1 Certainty, Value Acceptance, ACE+ PDR, and LCA ≤ 2.5 paths. A hold-versus-sell engine compares retention NPV to net sale proceeds using climate-adjusted MSR multiples.

03

Tailor

A mitigation engine simulates flood coverage, PFAS remediation, paydown to LTV band, rate-and-term refinance, and FICO buy-down scenarios. Bank-funded cost and borrower-funded cost are reported separately.

04

Route

The buyer router enforces agency eligibility gates, then ranks correspondent buyers by expected bid net of climate haircut and overlay conditions. Confidence is recorded on every rule.

05

Reconcile

A signed certificate (v3, SHA-256) and a ten-file examiner bundle are produced on every run. Each decision is replayable, every rule carries provenance, and every figure has a documented source.

Confidence is recorded per rule

The buyer-rule engine encodes 35 rules with citations: 11 classified high-confidence (cited to seller-guide or correspondent term-sheet language), 17 medium (cross-referenced from public market reporting), 7 low (operator-attributed estimates). The full gradient and citation list ship with every certificate and examiner bundle.

Deliverables

Three artifacts are produced for every portfolio: a per-loan record, a portfolio dashboard, and a signed certificate with an examiner-ready evidence bundle. When a federal source is unavailable for a property, the affected component is reported as null and the composite is recomputed accordingly. No values are synthesized.

Per-loan record

  • composite_score0–100
  • flood / fire / wind / heat0–100
  • pfas_score0–100
  • climate_bps_upliftbps
  • llpa_bpsbps
  • aus_unlocksD1C / VA / PDR
  • hold_npv vs sell_proceeds$
  • top_buyer + bid$ / px
  • tailoring_actions[]bps lift
  • borrower_funded_$$

Portfolio dashboard

  • Held-portfolio surveillance ledger with drift alerts
  • Hold-vs-sell decision board (NPV vs. net proceeds)
  • LLPA concentration alerts by FICO & LTV band
  • Aggregator overlay matrix (PennyMac, Wells, Chase…)
  • Buyer routing ranked by expected bid net of haircut
  • Recap mode: greedy sell-down to a target risk profile
  • Mitigation execution tracker (recommend → closed)
  • Confidence gradient on every rule that fires

Cert v3 + 10-CSV examiner bundle

  • Signed cert with SHA-256 + public verify URL
  • Sellability + hold-vs-sell provenance snapshot
  • Mitigation execution history per loan
  • Buyer-rule gate matrix with confidence stamps
  • LLPA matrix application + aggregator overlays
  • LLPA concentration alerts by FICO/LTV band
  • Federal source manifest with versions + dates
  • OCC SR 11-7, CECL, Fair Lending mapping
Validation result · 4,950 loans · 1.81s

On a community-bank portfolio

A Northeast community bank residential portfolio of 4,950 loans. The figures below are the control layer's direct output against this tape, produced through the same code path applied to client engagements.

Climate pricing spread
20.0bps
Best decile vs. worst decile, intra-portfolio
Agency LLPA exposure
$20.4M
Itemized loan-by-loan with remediation flags
Aggregator overlay conditions
338
Across PennyMac, Wells Fargo, Chase, AmeriHome
End-to-end runtime
1.81s
Full control-layer pass over the portfolio
Tailoring actions identified
  1. Force-place flood coverage, 41 loans+12 bps
  2. Paydown to LTV band, 28 loans+18 bps
  3. Rate-and-term refinance, 14 loans+22 bps
  4. Buy-down to FICO band, 7 loans+9 bps
Portfolio actions taken
  • Recapture targeted at the highest-climate 6.4% of UPB
  • Nine loans retained based on superior hold economics
  • Sale candidates routed to ranked correspondent buyers
  • Signed evidence bundle prepared in advance of exam

Figures are the control layer's direct output on the 4,950-loan validation portfolio, generated through the same code path used for client engagements. Buyer rules are sourced from config/buyer_rules.yaml. CLIMA does not pre-compute outputs and does not synthesize values for missing federal data.

Implementation

No system integration required. Four steps from loan tape to signed certificate.

1

Submit the loan tape

Standard CSV: loan ID, property address, UPB, rate, FICO, LTV, occupancy, and purpose. Geocoding and county resolution are performed on ingest. AES-256 encryption in transit and at rest; data is purged on request.

2

Control-layer execution

The five stations — surveillance, decision, tailoring, routing, reconciliation — run against live federal data and the encoded buyer rules in config/buyer_rules.yaml. Reference runtime on a 4,950-loan portfolio: 1.81 seconds.

3

Joint review session

A working session with credit, secondary marketing, and compliance. We walk each flagged loan in the portal, agree on the action list, and assign owners. The review is conducted in-product against live data.

4

Sign certificate, deliver evidence bundle

Certificate v3 is SHA-256 signed with a public verification URL. The accompanying evidence bundle contains ten files covering surveillance, decisions, mitigation actions, certificates, audit logs, buyer-rule provenance, LLPA matrix application, aggregator overlays, and concentration alerts.

Federal data sources

Every score traces to a verified government source. No proprietary blends, no opaque third-party indexes.

FEMA Flood Zones
National Flood Hazard Layer
FEMA NRI
National Risk Index
USFS Wildfire
Wildfire Hazard Potential
NOAA Storms
Storm Events Database
FEMA Declarations
Disaster Declarations
FEMA PA Grants
Public Assistance Awards
NFIP Policies
Flood Insurance in Force
HMDA Lending
CFPB Mortgage Data
Census ACS
Housing Cost Burden
Freddie Mac
14.8M Loan Performance
Residual Markets
FL Citizens, CA FAIR, TWIA
NASA / NOAA
Climate Projections & SLR
New
EPA PFAS Analytic Tools
171k+ contamination sites
New
EPA UCMR-5
Public-water PFAS occurrence
USGS Aquifers
Principal aquifer geology

Who it's for

Four roles inside a community or regional bank. No climate team or data-science department required.

Secondary marketing

A single view of buyer fit and expected net bid per loan. Replaces side-by-side spreadsheets of correspondent rate sheets. Recap mode prioritizes the highest-climate exposure.

Chief risk officers

A unified surveillance ledger covering climate drift, LLPA concentration, and held-portfolio exposure. Alerts are deduplicated, severity-ranked, and assigned to remediation owners.

Credit officers

Loan-level hold-versus-sell analysis, available continuously rather than quarter-end. Retention NPV, climate-adjusted MSR multiple, and recommended tailoring actions are presented per loan.

Compliance and audit

Signed certificate and ten-file evidence bundle covering SR 11-7 model documentation, CECL forward-looking adjustment evidence, Fair Lending checks, and per-rule confidence attribution.

Frequently asked questions

Direct answers to the questions raised by risk, secondary-marketing, and compliance teams in initial calls.

Doesn’t DU/LPA already price climate risk?
DU and LPA tell you whether the loan clears the agency box. They do not tell you the climate spread between your worst and best decile, the LLPA exposure if you sell tomorrow vs. next quarter, which aggregator overlays you can pass with one tailoring action, or which loans price better at retention. The control layer sits on top of DU/LPA, not in front of them.
How is this different from First Street, Munich Re, or Verisk?
Those products give you a hazard score. The control layer gives you a decision: climate-adjusted price, hold-vs-sell NPV, a tailoring playlist with bank vs. borrower-funded cost, and the buyer most likely to take it — plus the cert and bundle that pass an exam. Every component traces back to a federal source, so “where does that number come from” never requires a vendor call.
Do you train models on our loan data?
No. The disaster-default model is trained on Freddie Mac’s public single-family loan-level performance data (14.8M loans). Your tape is scored against that pre-trained model and never enters our training set. Tapes are encrypted in transit and at rest, processed in your tenant, and deleted on request.
Is the score auditable line-by-line?
Yes. Every loan gets the underlying component scores (flood / fire / wind / heat / PFAS), the FEMA flood zone, the WHP fuel class, the disaster history record IDs, the PFAS site distance and source, and the model version that produced the composite. The examiner PDF includes the full federal-source manifest with version dates and a SHA-256 signature.
How long from loan tape to scored portfolio?
Under 60 seconds for a 5,000-loan tape. The longest step is geocoding addresses; once we have lat/lon for a property, we cache it and re-scoring is near-instant. A 50,000-loan portfolio finishes in roughly 8–12 minutes end to end including the dashboard build.
Does CLIMA replace our existing credit risk model?
No. CLIMA produces an additive climate-risk layer that sits beside your existing PD/LGD/EAD models. Most clients consume the score as a feature in their internal scorecard or as an explicit reserve overlay under CECL. We do not touch your origination decisioning unless you wire it that way intentionally.
Will this satisfy OCC SR 11-7 model risk management?
The model documentation, validation backtest, conceptual soundness narrative, ongoing monitoring plan, and use-case scope are all included in the examiner PDF. Several of our deployments are inside SR 11-7 governed environments; we provide the artifacts your validation function needs to onboard a tier-2 vendor model.
What happens when a federal data source is missing for a loan?
We surface the gap. CLIMA never fills missing data with a synthesized value, a national average, or a peer-county proxy — the affected component is set to null and the composite is recomputed from the components that are present. The audit trail records exactly which sources contributed.

Request a portfolio assessment

Pilot engagements are complimentary and include a signed certificate and examiner-ready evidence bundle for your portfolio. Initial review takes approximately forty-five minutes.

Or email us directly at james@clima.solutions