Physical risk
Observed hazard history for every African country, admin-1 region and municipality: heatwaves, extreme rain, drought, ENSO phase, WBGT heat stress and livestock heat.
ERA51991–2025SPI
See hazard history
Climate-risk intelligence for African finance
Know what the climate will cost your book, before it does.
ecosphere turns 35 years of observed hazard history, CMIP6 projections and NGFS transition scenarios into decision-grade figures for every African country, region and site, then carries them through to probability of default, expected credit loss and capital.
Read the map through a lens
illustrative valuesValues on this page are seeded stand-ins, not ERA5 output. In the console the same lenses run on ERA5 reanalysis for every country, admin-1 region and municipality.
Highest five
Heatwave days · 2025
ecosphere in numbers
- Countries
- 54African countriesDown to admin-1 region and municipality
- Years of history
- 35yrsObserved historyERA5 reanalysis, 1991–2025
- Hazard families
- 4Hazard familiesWater, fire, geotechnical, extreme temperature
- Physical shock types
- 7Physical shock typesFor stress testing, from drought to seismic
- Scenario pathways
- 4+5Scenario pathwaysNGFS transition and CMIP6 SSP
- Time horizons
- 4Time horizonsToday, 2030, 2050 and 2090
Coverage as published in the data contract srcpublished data contract
The platform
One platform, from the hazard on the ground to the line in your capital plan.
Nine connected capabilities share one data foundation, so a flood depth at a warehouse in Kisumu and the CET1 figure in your board pack trace back to the same versioned source.
Hazard products
Rainfall return periods, 1-in-100 year flood depth, dam-release surge, fire risk, landslide susceptibility, seismic acceleration and heat percentiles, with hazard curves per site.
4 familiesPGA 1:475
Browse hazards
Climate projections
CMIP6 SSP scenarios at near-term, mid-century and end-century horizons, shown as ranges across models rather than a single line.
5 SSPs2030 · 2050 · 2090
View projections
Transition risk
NGFS Net Zero 2050, Delayed Transition, Current Policies and NDCs, translated into carbon cost, revenue and credit impact for each counterparty.
4 NGFSper counterparty
Model transition
Stress testing
Combine physical shocks with transition scenarios and see PD, expected credit loss and CET1 impact, with every result chained back to the hazard that caused it.
7 shocksPD · ECL · CET1
Run a stress test
Site atlas
3D terrain anywhere in Africa with roads, buildings and line of sight, plus counterparty dossiers carrying a risk fingerprint for today, 2030, 2050 and 2090.
3Drisk fingerprint
Open the atlas
Terrain & DTM
A bare-earth terrain pipeline that prefers LiDAR, then our own models, FABDEM and Copernicus. Flood and landslide runs start as soon as a new DTM is published.
LiDAR › own › FABDEM › DSM
How terrain works
Data deliveries & contract
Every delivery validated against a published data contract, reviewed by a second person, versioned, and shipped to the console, as Parquet files or through the API.
ParquetAPIversioned
Read about the contract
Assistant
Ask plain-language questions of your own published data. Bring your own model: Anthropic, any OpenAI-compatible endpoint, Bedrock or Vertex.
your data onlybring your model
Meet the assistant
From hazard to loss
Four steps, one unbroken chain of evidence.
Each number below is shown the way the console shows it: with a range, a source and a version. Figures are illustrative, drawn from a synthetic agricultural lending book.
-
Start with what has actually happened.
Thirty-five years of ERA5 reanalysis give every country, region and municipality an observed baseline, so a “record year” means something specific before anyone models the future.
- Heatwaves, extreme rain and SPI drought, year by year
- ENSO phase alongside, to separate signal from cycle
- WBGT for workers, THI for livestock
Heatwave days, Kano (admin-1), record year61 days · 2024baseline 1991–2020
median 24 dayssrcERA5 · 1991–2025Illustrative -
Bring the hazard down to the site.
Hazard curves and disaggregation per location show how deep the water gets, how often, and why. Flood depth comes from 2D modelling on bare-earth terrain, not a coarse regional average.
- Rainfall return periods and 1-in-100 year flood depth
- Dam-release surge, landslide and seismic acceleration
- Projected under each SSP at 2030, 2050 and 2090
Flood depth at 1-in-100 years, grain store1.1 mrange 0.8–1.4 m
SSP2-4.5 · 2050src2D flood · DTM · FABDEMIllustrative -
Translate it into credit impact.
Physical shocks and NGFS transition pathways move each counterparty’s probability of default and expected credit loss. Physical and transition risk stay as two separate dimensions, so you can see which one is doing the damage.
- Seven physical shock types, combinable with transition scenarios
- Carbon cost and revenue impact per counterparty
- Results in your reporting currency
Portfolio PD uplift, drought + Delayed Transition+38 bpsrange +22 to +51 bps
horizon 2030Counterparty Physical Transition Band Maize cooperative +64 +9 Stressed Cement producer +12 +41 At risk Logistics park +7 +4 Low PD Δbps · NGFSIllustrative -
Land on capital, and defend the number.
Expected credit loss and CET1 impact arrive with the full chain behind them. Official runs are separated from exploratory ones, and a scenario report exports ready for committee.
- Official and exploratory runs, clearly labelled
- Scenario report export for risk committee and supervisors
- Every figure traceable to hazard, version and reviewer
CET1 ratio impact, combined stress, 2030−42 bpsrange −31 to −58 bps
official runrunofficial · reviewed · ZARIllustrative
Who it serves
Built for the institutions that carry Africa’s climate exposure.
The same evidence, framed for the decision each institution actually has to make: lending, underwriting, allocating, supervising or adapting. See every sector.
-
Banks
Climate-adjusted PD and ECL by counterparty, and CET1 impact for supervisory stress tests.
PD · ECL · CET1 -
Insurers & reinsurers
Hazard curves and return periods per insured site for flood, fire, seismic and landslide.
Hazard curves · 1:100 · 1:475 -
Asset managers
Physical and transition exposure across holdings, compared side by side under each scenario.
NGFS · SSP · per holding -
Development finance
Screen projects and sovereign programmes against long-horizon hazard to 2090.
Site atlas · 2050 · 2090 -
Corporates & sustainability
Site-level hazard for facilities and supply chains, ready for climate disclosure.
Sites · disclosure · Parquet -
Regulators & central banks
A consistent hazard baseline for system-wide exercises, with every figure versioned.
Official runs · versioned -
Agriculture lenders
SPI drought, ENSO phase and livestock heat stress mapped against seasonal loan books.
SPI · ENSO · THI -
Infrastructure & utilities
Flood depth on bare-earth terrain, dam-release surge and seismic loading for linear assets.
2D flood · DTM · PGA
How it works
From your loan tape to a committee-ready view.
No new data science team required. You bring the book; we bring the hazard, the scenarios and the pipeline that connects them.
-
Connect your book
Upload counterparties and sites, or connect through the API. Each organisation and its companies are isolated, and nothing is shared unless you choose to share it.
- Inputs
- sites · exposures · sector
- Isolation
- per organisation
-
Run scenarios
Combine physical shocks with NGFS pathways and compare them side by side. Keep exploratory runs apart from the official ones you will stand behind.
- Scenarios
- 4 NGFS + 5 SSP
- Horizons
- today · 2030 · 2050 · 2090
-
Report & decide
Export a scenario report, pull Parquet files into your own models, or ask the assistant about your results, all in your reporting currency.
- Outputs
- report · Parquet · API
- Currency
- ZAR · KES · NGN · …
Data & methods
Data you can put in front of a supervisor.
Climate figures are only useful if they can be defended. Every ecosphere number carries its source, version and reviewer.
How we source, review and label data- Published data contractEach delivery is validated against a contract you can read before you rely on it.
- Second-person reviewNo data is published until someone other than its author has reviewed it.
- Versioned, with provenanceEvery figure shows where it came from. Synthetic or stand-in data is always labelled as such.
- Your data stays yoursOrganisations and their companies are isolated; sharing is a deliberate act.
See it on your own book
Bring a sample portfolio. Leave with a stress result.
In a working session we run a sample of your counterparties through a physical shock and a transition scenario, and walk you through every step of the chain from hazard to capital.