UI Revamp — working doc¶

Status: implemented in Phase 17 — this market scan and pattern record drove Phase 17 — Revamp (defined 2026-09-26, build waves B1–B28 landed by 2026-10-01; release, owner walk and mid-review on #1501). The subscription-tier note below stays parked (D20). Started: 2026-07-17 · Market scan checked against public sources and the shipped UI: 2026-09-18 Owner: Giorgos + Claude sessions

Why¶

The application's UI still reads as a descendant of RISK WISE: a fixed parameter rail on the left, a fixed results rail on the right, and a map/chart squeezed between them. With Phase 6 landing real Greek data (EFAS river flood, Storm Daniel, ELSTAT boundaries) and the Εθνική Ασφαλιστική demo as the grading bar, the UI needs to evolve into something visibly not RISK WISE — a modern, interactive, data-dense risk platform — while staying domain-neutral and white-label (product-definition §1 is unchanged by anything in this doc).

High-level goal: replace the static three-column dashboard with a map/chart-first, progressively-disclosed UI that stays responsive at real-data scale (EFAS grids, portfolio exposures) and surfaces information on demand (hover, click, drawers, modals) instead of permanently reserving screen real estate.

Problem areas¶

1. Map presentation¶

Centroid rendering does not scale: raw points clutter at country zoom and the map becomes unresponsive with large datasets. No aggregation strategy tied to zoom level.
Impact centroids sometimes have a fixed radius, visually overlapping unrelated administrative areas — the symbol size carries no meaning.
No multi-layer support (hazard + exposure + impact + boundaries as toggleable layers on one map); each view is a separate static map.
Aggregation/disaggregation (admin 0/1/2) is static and heavy — a button-triggered reload, not a fluid zoom/drill interaction.
The exposure map (choropleth) and impact maps deliver little business value as designed: they show where data is, not what a decision-maker should look at.

2. Chart representation¶

Charts have come a long way from static images, but interactivity is shallow: limited hover detail, no click-through from a chart element to the underlying locations/records, no cross-filtering between chart and map.
Opportunity: on-hover context (contributing drivers, percentiles), on-click drill-down (EP curve point → the events/locations behind it), linked highlighting.

3. Informational & results sections¶

The results rail permanently consumes ~1/3 of a laptop screen and is empty most of the time (see screenshots 02, 04).
Result details are static prose blocks. Information should follow the interactive model of the rest of the app: tooltips for definitions, drawers for drill-down tables, modals for methodology/provenance, inspector panels that appear on selection and get out of the way otherwise.

4. Parameter selection¶

The input rail also permanently consumes ~1/3 of a laptop screen, even after a run when the parameters are no longer being edited.
Direction to explore: collapsible/floating parameter surface — e.g. a setup wizard or command-bar flow that collapses into a compact "active scenario" chip strip after RUN, reclaiming the screen for results.

Reference screenshots (riskwise-v2 — the look we are evolving away from)¶

Stored in assets/ui-revamp/. These document the predecessor's patterns; the revamp is graded on how clearly we depart from them while keeping the workflow legible.

#	File	What it shows
01	`01-home.png`	Home dashboard: card launcher + status/resources column
02	`02-risk-empty.png`	Risk tab, empty state — both rails reserved, center placeholder
03	`03-macro-params-empty-chart.png`	Macro tab: sparse param rail + empty full-height chart
04	`04-adaptation-empty.png`	Adaptation tab, empty state with dead Summary/Details rail
05	`05-settings-country-packs.png`	Settings: custom country pack upload (bundle seam UI)
06	`06-impact-function-detail.png`	Impact-function inspector: chart + raw value table mid-run
07	`07-adaptation-cost-benefit.png`	Cost-benefit bar chart + Active scenario / Summary rail
08	`08-hazard-map-centroids.png`	Hazard map: dense flood-depth centroids along the Nile
09	`09-exposure-choropleth.png`	Exposure/values choropleth at admin level (static Admin 0/1/2 toggle)
10	`10-impact-map-sparse-centroids.png`	Impact map: sparse fixed-radius centroids, little value
11	`11-impact-bar-chart.png`	Impact 2024→2050 bar chart with hover tooltip

Constraints & context¶

Demoable slice first: any revamp work is graded against the Εθνική demo (phase-6 §0 D-S12): probabilistic RF run with EP curve + 1-in-200 PML, Storm Daniel single-event view (EP/PML suppressed, never faked), classic workflow on EFAS data, Greek UI.
White-label core: no client branding or domain assumptions in core UI components; theming/config stays in the bundle seam.
Scale target: EFAS RF grids at 93 m over Greece and portfolio-scale exposures — the map layer must be designed for this, not for the retired toy fixtures.
Render stack: the UI renders with MapLibre GL JS only (maplibre-gl ^4.7.1, plus h3-js and recharts; no deck.gl, no Leaflet). The phase builds on MapLibre. deck.gl is an option, not a dependency: its @deck.gl/maplibre module (9.4.x) supports MapLibre ≥ 4.5.1 in overlaid and interleaved modes. MapLibre itself is at v6 — globe projection needs ≥ v5 — so a major upgrade is a decision for the phase design doc. climate-lama-ui/docs/CLIMATE_LAMA_UI.md still describes react-leaflet and needs correcting.
Engine plurality is in scope. The phase steers functionality as well as appearance: the market scan's section 5 lists the engines that could join climate-lama-engine and CLIMADA behind worker/models/base.py, and pattern 20 names the prerequisite — a shared results vocabulary and a reference scenario every engine must reproduce.
Where the product stands today is stated per pattern at the end of the market scan; the problem areas above describe the starting point the earlier phases have already moved on from (no permanent results rail, a collapsible parameter rail, the Live tab as pattern pilot).

Market scan — products, engines and sources that inform the direction¶

44 entries in 6 groups, checked against public sources on 2026-09-18 (two /reassess fast passes: web verifiers on liveness and on the substance of every statement, plus a completeness critic reading climate-lama-ui at 1c713e9, v0.16.0, and the engine adapter seam in the backbone). A statement that could not be confirmed from public material sits under To confirm hands-on rather than being asserted. Re-check any entry older than 90 days before citing it in a design decision.

1. Commercial catastrophe-modelling and insurance platforms¶

The tools insurers and reinsurers already pay for. They set the expectation of what a professional risk platform looks like and how it is packaged.

Moody's Intelligent Risk Platform (ExposureIQ, Risk Modeler)¶

Paid · Platform · ExposureIQ · Risk Modeler · Developer portal · Platform evolution article
Cloud cat-modelling suite — exposure management, model runs, EP curves/PML at portfolio and treaty level. Offered by Moody's Insurance Solutions (formerly RMS).
Key idea: One shared data layer under several task-specific apps; value mapped as areas and clusters; an in-app assistant.
What it does well:
Role-based applications on one platform (Risk Modeler, ExposureIQ, UnderwriteIQ, TreatyIQ) that share geocoding, financial engines and science models, instead of one screen doing everything.
The ExposureIQ accumulation view combines hazard layers, thematic mapping and evaluation of risk clusters by proximity to hazard zones over very large location sets — aggregated areas rather than raw points.
Event Response footprints and HWind wind fields are available inside the exposure application for immediate portfolio analysis, not as separate datasets to import.
Risk Modeler's Model Composer creates named, maintainable "views of risk" — model settings become an organisational asset rather than temporary form state.
IRP Navigator: an AI assistant integrated into the applications, answering product and workflow questions from support, model and developer documentation.
Risk Data Refinery: AI converts raw broker submissions into structured exposure data in minutes.
Peril Metrics (2026) delivers hazard metrics through an API — analytics sold as data as well as screens.
Moody's has integrated the industry-owned Oasis hosted platform to widen the choice of catastrophe models — a commercial leader embracing multiple engines.
Relevance to our product: Insurer portfolio audience. We already offer named, re-runnable analyses and live event layers. The ideas still open to us are mapping accumulated value as areas (pattern 3), an in-app assistant (pattern 18), and the market validation of our multi-engine approach (pattern 20).
Visuals: Moody's Insurance Solutions YouTube (product walkthroughs), IRP Navigator announcement

Verisk Synergy Studio (successor to Touchstone)¶

Paid · Synergy Studio · Touchstone · Matisse · Launch coverage (Artemis)
Cloud-native cat-modelling platform bringing Touchstone, Touchstone Re and AnalyzeRe into one product. Generally available since June 2026. Touchstone remains supported; new science is released on Synergy Studio first.
Key idea: Configuring a run and exploring its results are separate, purpose-built screens; multi-model analysis in one platform.
What it does well:
Modular analytics (Data Quality, Hazard, Geospatial, Detailed Loss): the workflow is a sequence of named stages, each with its own screen, not one large form.
Reading results is a separate application (Matisse) that traces loss through each contractual term — configuring a run and exploring results are different surfaces.
Sold on streamlined data ingestion and repeatable analysis workflows run with saved preferences, rather than one-off runs.
Multi-model analysis and flexible event catalogues in one platform — the commercial reference point for comparing scenarios across engines.
Verisk acquired McKenzie Intelligence Services (July 2026) to add observed post-event intelligence to the modelling platform.
To confirm hands-on:
Selecting a subset of simulated events on the map; one-click export of maps, charts and tables; dashboards pairing loss curves with loss maps as linked views. These come from earlier Touchstone material and an AWS case study rather than current product pages.
Relevance to our product: Insurer and scientist audiences. Our Assess / Runs split mirrors their configure / explore separation. Linked loss-curve and loss-map views correspond to our open pattern 7.
Visuals: AWS case study with dashboard screenshots (Touchstone era), Synergy Studio FAQ flyer (PDF)

Swiss Re CatNet¶

Paid (a basic tier is available to clients) · Product page · Fathom in CatNet
Global natural-hazard atlas: hazard layers combined with the user's own locations, in the browser.
Key idea: Several independent flood models side by side on one map; live event footprints as their own layer type.
What it does well:
Hazard type and time perspective (present day versus multi-decade history) are one dimension picker rather than separate pages.
Near-real-time event footprints form an overlay class distinct from modelled hazard layers.
Fathom's global, US, UK and Japan flood maps are available inside CatNet, so a user can consult several independent flood models on one platform — the closest commercial analogue to an engine-agreement view.
A machine-learning wildfire-probability layer, built with Bellwether, weighs roughly 600 weather, vegetation, terrain, fire-history and human factors.
To confirm hands-on:
The screen layout (map in the centre, layers in a side rail, key figures in a strip above the map) — the product pages cannot be read automatically, and the public brochure screenshots are dated.
Which subscription tier carries the Fathom maps, and the scope of the CatNet API.
Relevance to our product: Decision-maker audience. Our Live section already separates observed from modelled layers. A key-figures strip above the map is worth adopting on Assess and Portfolios.
Visuals: Overview video, Brochure with screenshots (PDF)

Munich Re Location Risk Intelligence¶

Paid (credit-based on-demand tier for single locations) · Product page · On-demand app · Demo request
Location and portfolio scoring for 15 natural hazards plus 13 climate-change hazards under scenarios.
Key idea: One overall score that opens into groups, hazards and detailed sub-scores; features packaged as editions.
What it does well:
A score hierarchy: one overall risk score, three group scores (earthquake, storm, flood), 15 hazard scores and more than 250 climate sub-scores — detail revealed progressively.
Editions: feature bundles on one interface (natural hazards, climate change, financial impact, wildfire, biodiversity, reporting) — close to our own configurable-bundle model.
A Reporting Edition aligned to EU Taxonomy, CSRD, TCFD and ISSB — regulatory output as its own bundle.
The on-demand tier is a simple single-location journey (address, scores, report) alongside the full product; a REST API delivers scores into other systems.
The Wildfire HD edition lists Greece among its four covered countries.
Relevance to our product: Citizen audience. Our Lookup already shows per-hazard cards with a "Why this score?" explanation; the missing piece is the top of the hierarchy — one headline grade and an optional hazard-group level (pattern 9).
Visuals: On-demand app (can be explored without a contract)

Fathom (a Swiss Re company)¶

Paid · How to access · Global Flood Map · Risk Scores · FathomDEM+
Global river, surface-water and coastal flood hazard layers (return periods 5–1000) via portal and API.
Key idea: Return period and scenario are switches on the map, not new runs; simple indices sit on top of expert data.
What it does well:
An uploaded raster or vector file becomes a live interactive map in the browser; layer, region, return period, flood type and climate scenario are switches, not reasons to run again.
One data definition, four delivery routes — API, portal, direct file provision, and embedding in partner platforms (CatNet, CARTO, Climate Engine, EigenRisk).
Two derived indices sit over the depth grids: Relative Risk (usable as an average annual loss for a uniform exposure) and Risk Category (0–100). Depths for experts, indices for everyone else.
Each index is offered per flood type, inland only, or all combined — the aggregation choice is explicit.
The published recipe for the indices (depths at several frequencies, depth-damage relationships, defence standards) is a model for documenting how a single number is derived.
Terrain data is a browsable product in its own right: inputs with clear provenance belong in the catalogue.
Relevance to our product: All audiences. Treating return period as a layer switch is the concrete form of pattern 5 (re-cut a result without re-running it).
Visuals: Fathom and Swiss Re flood data article

JBA Risk Management — JBA Vision¶

Paid · Data products · JBA Vision · Online services · Climate Risk Analytics
High-resolution global flood maps with plain-language flood ratings and a web viewer.
Key idea: Same map, switch the climate scenario, see the difference; plain-language ratings; a pricing-ready output.
What it does well:
Baseline versus climate scenario is the organising interaction: every dataset has a climate-change counterpart, so the same map switches scenario and shows the difference.
Depth and frequency are condensed into a plain-language rating (negligible to significant) per flood type, with a simpler colour-coded floodability scale above it.
Three delivery modes for the same layers: a browser viewer, an API, and WMTS streaming into a customer's own GIS.
Pricing Data offers a rating-ready loss-cost output beside the hazard and score products — the step from hazard to price is a deliverable in itself.
Relevance to our product: Citizen and insurer audiences. Baseline-versus-scenario comparison is our open pattern 10 on Assess and Lookup.
Visuals: JBA Vision product page

ZestyAI¶

Paid · zesty.ai · Model deep dive
AI property-level wildfire, hail, wind, storm and water risk scores for US properties.
Key idea: Scores broken into named components, and regulator-ready model documentation shipped as part of the product.
What it does well:
Scores are broken into named levels rather than one opaque number: wildfire separates exposure risk from structure vulnerability; severe storm reports claim frequency and claim severity.
Property attributes are derived by computer vision from aerial imagery (roof shape and material, vegetation clearance) and combined with permits, terrain and climate records.
Models are trained and validated on real insurance claims, and that provenance is central to the offer.
Regulator-ready documentation ships with the models: methodology, variable-selection rationale, statistical validation and regulatory summaries.
One property record carries many per-hazard scores — one identity resolved, many results attached.
Relevance to our product: Citizen and insurer audiences. Our "Why this score?" explanation exists; still to add are the contributing factors behind each figure (pattern 9) and publishable method documentation (pattern 21).
Visuals: Model deep dive page

Oasis Loss Modelling Framework — platform and hosted service¶

Free / open source (BSD-3); hosted service by subscription · oasislmf.org · Oasis hosted service · GitHub · February 2026 newsletter
The open-source catastrophe-model execution platform, now also an industry-owned hosted service.
Key idea: A vendor-neutral marketplace of models behind one interface — the clearest validation of a multi-engine strategy.
What it does well:
A clear workflow: portfolio, then analysis (portfolio, model and settings), then a run with progress states, then an outputs browser. The analysis is a saved, named, re-runnable and comparable object.
The hosted service is industry-owned, not-for-profit and vendor-neutral, offering several hundred models — positioned to preserve model choice.
A commercial leader consumes it: Moody's integrated the Oasis hosted platform into its own suite to expand model choice.
Release 2.5.0 (February 2026) completed a move to a Python-only codebase — the engine interface stayed stable while the runtime beneath it was replaced.
The REST API, rather than the reference user interface, is the integration surface.
Relevance to our product: All professional audiences. The workflow is already reflected in our Runs section. Oasis is also an engine candidate in its own right — see section 5.
Visuals: Reference UI repository (screenshots in README), Oasis hosted service site

2. Climate risk analytics platforms¶

Design-led products that sell physical climate risk to banks, real-estate owners and corporates. The best source of approachable user experience and of reporting-oriented packaging.

Jupiter Intelligence — ClimateScore Global¶

Paid · Products · Site Intelligence · Story behind the score
Global physical-risk data platform across nine hazards, delivered through a suite of purpose-built applications.
Key idea: One data platform under several small, task-named apps — including adaptation, compliance and model-validation apps.
What it does well:
One data platform under many small purpose-built applications (screening, adaptation, entity modelling, metrics, compliance, model validation, single site) rather than one large configurable interface.
Adaptation Hub is a first-class application that frames resilience options against financial outcomes.
Compliance Hub offers defensible disclosure output as its own application, separating the reporting deliverable from the analysis screens.
MRM Accelerator packages model-validation evidence as a product rather than an appendix.
Scenario coverage is stated plainly: several IPCC-aligned pathways with timeframes to 2100; delivery by web application and enterprise API.
"Story behind the score": each score is documented down to the measurements that produced it.
To confirm hands-on:
A hazard × scenario × year picker that stays visible while results update — inferred from marketing material; no public demonstration exists.
Relevance to our product: Scientist audience. Assess already lets users pick several scenarios and years; still open is letting those pickers update a loaded result in place, summarised as a strip of chips (pattern 5).
Visuals: Products page (application screenshots)

Climate X — Spectra¶

Paid · Spectra · Features
Asset and portfolio physical risk translated into financial loss.
Key idea: Three ways in — search an address, upload a list, draw an area — and a flood-defences on/off switch.
What it does well:
Entry is search-or-upload: address search for one asset, CSV upload for a portfolio, polygon upload for large infrastructure — three ways into one analysis.
Concrete scenario controls: four named climate pathways, years from 2020 to 2100, and one prominent "Toggle Flood Defenses on/off" switch.
Drill-down from portfolio to asset to unit, with annual loss in both money and percentage, and confidence, probability and severity shown beside each score.
Ready-to-use PDF and CSV reports aimed at due diligence — the export is a named deliverable.
A separate adaptation product (Adapt) shares projects with Spectra, so assessment and adaptation planning are distinct screens over one project.
An enterprise API, model-risk-management alignment and ISO certifications are presented as product features.
Relevance to our product: Insurer and citizen audiences. We offer address search on Lookup and upload on Datasets, but the home page has no search and Assess still begins with dataset pickers (pattern 11).
Visuals: Climate X YouTube channel, AWS Marketplace listing (screenshots)

Mitiga Solutions — EarthScan¶

Paid (free trial) · EarthScan · Feature article (2022)
Asset and portfolio climate risk with standardised A–F ratings across six hazard families. EarthScan was created by Cervest and has been part of Mitiga Solutions since 2023.
Key idea: A–F letter grades paired with a money figure, and reporting alignment listed framework by framework.
What it does well:
Standardised A–F ratings (A very low to F extremely high) at asset and portfolio level, paired with return periods from 2 to 1,000 years.
Six hazard families, each in acute and chronic form, projected from 1970 to 2100 under three climate scenarios.
Climate Value at Risk converts hazard severity into potential financial loss, so the letter grade and the money figure sit on the same asset.
Reporting alignment is listed per framework — CSRD (ESRS E1), IFRS S2, TCFD, EU Taxonomy, GRESB, CRREM — rather than claimed generically.
API access is part of the enterprise tier, so the delivery mode is part of the packaging.
To confirm hands-on:
A combined physical-risk roll-up and a catalogue of pre-identified assets that can be added to a portfolio in a few clicks — described in the 2022 article; not stated on the current product page.
Relevance to our product: Citizen and insurer audiences. A combined letter grade is the headline our Lookup score cards do not yet have (pattern 9); it should state how many hazards it could not include.
Visuals: 2022 feature article (screenshots), EarthScan comparison page

XDI Climate Risk Hub¶

Paid · Climate Risk Hub · Application · Launch article
On-demand asset and portfolio physical climate risk: 11 hazard models, six time steps to 2100.
Key idea: One engine offered at three scales, engineering-style measures such as downtime, and adaptation options ranked by value.
What it does well:
Tiers by scale, not by feature: the same engine serves a quick screen of up to 30 addresses, a global portfolio, or one large site — a close match to our pin lookup, portfolio and detailed-analysis journeys.
Eleven hazard models over six time steps, condensed into A, B or C ratings for triage, with financial detail behind them.
Results can be read three ways over the same analysis: map view, customisable grids, and APIs.
Engineering-led measures beyond loss: estimated asset downtime, annual probability of asset failure, and change in repair and insurance costs.
Automated Resilience (beta) ranks adaptation options by avoided risk value and break-even investment — the product answers what to invest in first.
Relevance to our product: All audiences. Confirms that audience-named sections (Lookup, Portfolios, Assess) are the right structure. Automated Resilience is the reference for adaptation what-if work (pattern 22).
Visuals: Launch article with screenshots

First Street (an MSCI company)¶

Freemium · firststreet.org · Home-buyer pricing · Interpreting the flood maps · Bulk data and API
Per-property flood, fire, heat, wind and air-quality scores (1–10) for US properties. Formerly published as Risk Factor. MSCI completed its acquisition of First Street in August 2026.
Key idea: An address-first journey with a free score, and a paid tier where owners edit their property to see risk change.
What it does well:
Address-first journey: search, then a property page with 1–10 scores, then detail per hazard. The free tier gives the score, flood history and how risk changes over time.
Scores are calculated to the building outline over an explicit 30-year period under a stated scenario — framed as a period and a threshold, not a raw map value.
Hazard-specific narrative per property rather than raw hazard layers.
A clear line between free and paid: the score is free; interactive maps, damage forecasts and mitigation modelling need a subscription.
Paid users edit their home's characteristics and see risk and damage change — adaptation what-if driven by the owner's own information.
The same models are also licensed as bulk data and developer endpoints; the consumer site is one of several delivery routes.
To confirm hands-on:
A zoom-dependent map that shows shaded areas when zoomed out and individual building outlines coloured by score when zoomed in.
Relevance to our product: Citizen audience — the closest reference for Lookup. Our result map has two zoom levels on one screen and no user control over them (pattern 4).
Visuals: National Risk Assessment report (charts and maps, PDF), Help-centre flood-map walkthrough

Probable Futures¶

Free (non-profit) · Maps · Pro · Map version history · Building with the data
Free global climate maps by warming level, with a public per-location API.
Key idea: Radical simplicity, a slider to compare two scenarios, and a public changelog of every interface change.
What it does well:
Simplicity as a design position: one scenario dimension, one variable at a time, the map carries the meaning.
Two warming scenarios side by side on one map with a slider between them.
Probability in plain language — a warmer / cooler (or wetter / drier) year switch instead of percentiles.
The map toolbar is grouped into View, Download and Screenshot — export and capture are primary map actions.
Keyboard-first search inside the map (Ctrl/⌘ K).
A per-location API accepts an address or coordinates and returns values by map, percentile and warming scenario — a model for a pin lookup service.
Every interface change is dated and described publicly — a simple credibility practice.
Relevance to our product: Citizen and decision-maker audiences. The slider comparison is the missing half of pattern 10 — we offer two synchronised maps for one kind of comparison only.
Visuals: Design case study, Article with embedded live maps

Riskthinking.ai — VELO¶

Paid (free sign-up preview) · VELO · Free sign-up · Insurability reports
Climate exposure across assets, companies and indices with scenario simulation.
Key idea: Pre-loaded content to explore before uploading anything, and named risk states with published definitions.
What it does well:
Explore first, then bring your own: a new user can search millions of physical assets, thousands of indices and listed companies before uploading anything.
Monthly published Climate Insurability Rankings keep the product visible and its method open to scrutiny.
The engine is fully stochastic, running thousands of scenarios per asset rather than a handful of named pathways.
Four named risk states (Low Risk, At Risk, Stressed, Stranded) with a published numerical definition for each.
Relevance to our product: Insurer audience. We deliberately ship no demonstration data, so the equivalent is a browsable catalogue of real ingested datasets — which Datasets and the home page already provide. Named risk states with published definitions are worth adopting (pattern 9).
Visuals: Public insurability report, Feature video (LinkedIn)

S&P Global Sustainable1 — Climanomics¶

Paid · Physical climate risk · Marketplace listing · Methodology (June 2025, PDF) · Climate RiskGauge
Financial climate risk: modelled average annual loss per asset from 270+ asset-type impact functions.
Key idea: One consistent unit everywhere — yearly loss as a share of asset value — that can be broken down by hazard and cost type.
What it does well:
Everything is expressed as modelled average annual loss as a percentage of asset value, for each decade to 2100 under four scenarios — one consistent unit.
Loss can be broken down twice: by hazard, and within each hazard by type of cost.
Vulnerability comes from a library of impact functions for more than 270 asset types, so different assets respond differently to the same hazard.
Two products from one engine: a simple comparable screening score and a detailed explainable loss model.
The methodology is a dated public document, which makes the figures defensible to outside parties.
Relevance to our product: Insurer audience. One consistent unit across Assess, Portfolios and Lookup is a low-effort discipline to adopt.
Visuals: Marketplace listing (screenshots)

3. Public and consumer risk viewers¶

Free tools built for non-experts and for operational forecasters. Strong on plain language, on honesty about uncertainty and on live information.

FEMA Resilience Analysis and Planning Tool (with National Risk Index data)¶

Free · National Risk Index · Resilience Analysis and Planning Tool · Resource centre · Open data
US multi-hazard risk information at county and census-tract level. The standalone National Risk Index application has been retired; its data is now served through this tool and as open data.
Key idea: Let users choose the level of geographic detail — and keep data and documentation available independently of any one app.
What it does well:
Layers are published per geographic level (county, census tract, tribal area), and availability is stated per level — a user-controlled level of detail.
The user guides document a per-community profile that breaks the index down hazard by hazard.
The data, the technical documentation and a bulk download remain available independently of the application — a result should never be reachable only through one screen.
To confirm hands-on:
The county / tract switch and the per-community profile as they behave in the current viewer (documented in the 2025 user guides).
Relevance to our product: Insurer and decision-maker audiences. The level switch is our open pattern 4. Where a level is unavailable for a country, the interface should say so clearly.
Visuals: National Risk Index user guide (PDF, screenshots), Planning tool user guide (PDF)

ThinkHazard!¶

Free, open source (GFDRR / World Bank) · ThinkHazard! · Example: Greece · GitHub · Methods
Hazard level for 11 hazards, with guidance, for any administrative area worldwide.
Key idea: Search a place, get coloured hazard levels and plain guidance — with one permanent web address per area.
What it does well:
The strongest simplification in the field: search for a place, see a grid of coloured hazard levels, open a guidance page per hazard. The map supports the reading rather than leading it.
One permanent web address per administrative area — the best live example of a country or region profile page.
Guidance text per hazard level that recommends further local study where appropriate — a template for honest wording where results are limited.
Relevance to our product: Citizen audience. Our Lookup already shows a clear "not available here" message rather than a zero; guidance text per severity level is the natural next step.
Visuals: Product video, Live report page for Greece

EU Risk Data Hub (Joint Research Centre)¶

Free (European Commission) · Risk Data Hub · Data collection
Pan-European web GIS for hazard, exposure, vulnerability and risk layers plus historical losses.
Key idea: The European public-sector benchmark: modelled risk and historical events presented together.
What it does well:
The European reference for bringing hazard, exposure, vulnerability, risk and historical loss data together at regional level.
Historical events are presented as evidence alongside modelled risk.
To confirm hands-on:
The module structure (exposure, vulnerability, risk, historical events) and the regional granularity — the application cannot be read automatically; capture dated screenshots during a hands-on review.
Relevance to our product: Scientist and decision-maker audiences. We already present observed and modelled information separately; this is the public-sector benchmark our interface should compare well against.
Visuals: Risk Data Hub overview (PDF)

Copernicus EFAS / GloFAS map viewer¶

Free (real-time forecast layers need a registered role) · EFAS map viewer · GloFAS · Reporting points · EFAS v5.6
Operational European and global flood forecasts — also a source of our river-flood hazard data.
Key idea: Markers that appear automatically where a forecast crosses a threshold, coloured by published rules.
What it does well:
Forecast points are symbols coloured by alert level, split into fixed points at known stations and dynamic points created automatically where the forecast crosses a return-period threshold.
Dynamic points follow published rules: where they are placed, how long they persist, and the probability thresholds behind each colour.
Small catchments are drawn differently from large ones, so the symbol itself communicates confidence.
To confirm hands-on:
Layers grouped by theme, user-added external map layers, saved personal map views, and drill-down from a symbol to the full forecast chart — the live layers need a registered account.
Relevance to our product: Decision-maker audience. Creating markers where a result crosses a threshold — not only where a station exists — is a direct idea for the Live section.
Visuals: Viewer documentation (screenshots)

Google Flood Hub¶

Free · Flood Hub · Layer documentation · Flood Forecasting API · Email subscriptions
Free operational AI flood forecasting: river forecasts up to 7 days ahead, flood-extent maps and source confidence.
Key idea: Show how reliable each forecast source is, let users opt in to weaker ones, and offer alerts for a watched place.
What it does well:
Sources are tiered by confidence: gauges where model quality could not be assessed are placed behind an explicit "Additional coverage" switch rather than mixed in.
A flood-probability layer and a flood-history layer sit beside the forecast — forecast, likelihood and precedent on one map.
The coverage of each layer is stated openly, including its geographic limits.
Forecast horizons are explicit: up to 7 days for river floods and up to 24 hours for urban flash floods.
A free API under an open licence, and email subscriptions that deliver alerts for a watched location.
Relevance to our product: Decision-maker audience. Our Live section shows the status of each data feed but not its reliability, and has no subscription alerts (pattern 23).
Visuals: Flood Hub (open to explore)

Windy and Zoom Earth¶

Free (Windy has a premium tier) · Windy · Zoom Earth
Consumer live-weather maps with animated forecast playback.
Key idea: Play and pause a forecast over time with a timeline along the bottom of the map.
What it does well:
A timeline along the bottom of the map with play and pause animates a forecast sequence.
Layers can be switched while staying at the same place and time on the web version.
Both are widely used on phones.
To confirm hands-on:
Neither product publishes feature documentation; record a dated screen capture of the timeline on desktop and on a phone during a hands-on review.
Relevance to our product: Decision-maker audience (automatic playback of the Live timeline, which is manual today) and scientist audience (animating the years 2030, 2050, 2080 on one map).
Visuals: Windy (record a screen capture of the timeline), Zoom Earth

IPCC Interactive Atlas and Copernicus Interactive Climate Atlas¶

Free · IPCC Interactive Atlas · Copernicus Interactive Climate Atlas · Copernicus atlas guide · Implementer's article
The scientific reference for showing model agreement and uncertainty on a map, with accessible colour palettes.
Key idea: Hatching where models disagree, based on a published rule, and colour palettes safe for colour-blind readers.
What it does well:
Two documented hatching methods: a simple one marks low model agreement; an advanced one distinguishes "no robust signal" from "conflicting signals".
The Copernicus atlas formalises three robustness categories using an 80% model-agreement threshold — a published rule that can be cited.
Colour palettes were designed to be safe for colour-blind readers; the tool can be used entirely by keyboard and supports screen readers.
Relevance to our product: Scientist audience. We already calculate where engines disagree; the result should be hatched on the matrix and the map using a stated rule. Our map colours are also fixed, so the high-contrast display setting does not yet reach the map (pattern 15).
Visuals: IPCC Interactive Atlas (open to explore), Implementer's article with annotated images

Climate Central — Coastal Risk Finder¶

Free · Coastal Risk Finder · About the tool · Entry page for citizens
Public sea-level-rise and coastal-flood maps with local statistics.
Key idea: A shareable map — the web address holds the view — with downloadable local charts and audience-specific entry pages.
What it does well:
A location search (state, county, municipality, electoral district) with scenario and settings menus.
The map view is held in the web address, so any view can be linked and shared.
Local maps, statistics, bar charts and high-resolution images can be downloaded for reuse.
Social vulnerability is a primary dimension: exposure is broken down by population group.
A separate entry page per audience (concerned citizen, researcher) frames the same map differently.
To confirm hands-on:
How well the tool works at phone width.
Relevance to our product: Citizen audience. Four of our sections already hold their state in the web address, but none offers a "copy link" button; the home page and Assess are designed for desktop only (pattern 17).
Visuals: The tool itself (open to explore)

4. Geospatial application tooling¶

Not risk products, but the state of the art in map-application ergonomics. Most offer directly adoptable ideas; some offer adoptable open-source code.

kepler.gl¶

Free (MIT, OpenJS Foundation) · kepler.gl · Live demo · Layer documentation · Release notes
Open-source browser tool for large geospatial datasets with configurable layers, filters and aggregations.
Key idea: The reference layer list, and turning a cloud of points into a readable surface with one control.
What it does well:
The reference layer panel: a vertical stack of collapsible layer cards with visibility, type, colour-by-field, scale and opacity; add, duplicate and reorder in place.
Hexagon and grid aggregation layers with a choice of function (count, sum, mean, maximum of any field) turn a point cloud into a readable surface with one control.
Filters sit beside layers in the same panel; brushing a histogram is the filter; a time filter becomes a playback bar.
Configurable tooltips, a split dual-map view, and export of the whole map state as a file.
An AI assistant driven by text, voice or a screenshot, and an in-browser database (DuckDB) for querying data directly.
Raster-tile and WMS layer types — the closest match to hazard rasters and external agency layers.
Supports MapLibre base maps, the same map library our interface uses.
Relevance to our product: All map screens. Our layer panel is a small floating box with three fixed rows (pattern 2). kepler.gl is the design reference; we keep our existing map library.
Visuals: Introduction article (animated interface images), kepler.gl 3.1 announcement

deck.gl¶

Free (MIT) · Layer catalogue · Hexagon demo · H3 hexagon layer · Using with MapLibre · What's new
Open-source GPU rendering library used by kepler.gl, Foursquare and CARTO.
Key idea: An optional rendering layer for very large datasets that works on top of our existing map library.
What it does well:
Hexagon and grid layers aggregate raw points on the GPU (sum, mean, minimum, maximum, count).
A dedicated layer renders pre-aggregated hexagon cells: aggregate on the server per level of detail and choose the level from the zoom.
Hover and click picking feeds tooltips and detail panels.
A dedicated MapLibre integration module (version 9.4, September 2026) supports MapLibre 4.5.1 and later, in overlaid and interleaved modes.
Version 9.4 is announced as the last release before a larger architectural change in version 10.
Relevance to our product: Optional. Our interface renders with MapLibre alone today. Recommendation: deliver aggregation from the server as map tiles first, and adopt deck.gl only if a measured portfolio size requires it (pattern 3).
Visuals: Hexagon layer live demo

Foursquare Spatial Studio and Spatial Desktop¶

Freemium (Desktop: free basic tier through 2026, paid tiers from $25 per month) · Spatial Studio · kepler.gl and Studio compared · Spatial Desktop
A hosted extension of kepler.gl with data connectors, hexagon tiling and analysis modules; plus a desktop edition.
Key idea: Analysis tools beside the layer list, and hexagon pyramids for country-scale data.
What it does well:
Analysis modules (cluster and outlier detection, suitability scoring) offered alongside the layers rather than in a separate application.
Hex Tiles: a tiling system built on H3 hexagons for country-scale point and raster data — the scalable form of area aggregation.
In Spatial Desktop the analytical database sits beside the viewer and reads modern cloud formats (GeoParquet, PMTiles) directly, so filtering needs no server round trip.
A natural-language agent that builds maps and queries datasets, with usage metered per plan.
Relevance to our product: Insurer audience. Hexagon pyramids are the concrete design for pattern 3. Querying data in the browser is worth a short trial before adding more server endpoints (pattern 19).
Visuals: Spatial Desktop page (animated images)

CARTO Builder¶

Paid (trial) · Builder · Widgets · What's new, Q2 2026 · AI Agents
Cloud-native map dashboards.
Key idea: Charts beside the map that also filter it — the strongest single idea for our results area — and an assistant whose actions can be inspected.
What it does well:
A widget rail as the results panel: formula, category, pie, histogram, range, time-series and table widgets docked beside the map.
Several widget types cross-filter: interacting with them filters the map and the other widgets, within documented limits.
SQL Parameters: dashboard-level dropdowns and sliders that re-parameterise the underlying query — a model for return-period and scenario pickers that update results without a new run.
A selection tool filters by rectangle, polygon, circle or free-hand lasso — choosing part of a portfolio by drawing on the map.
AI Agents with full traceability: the tools an agent ran, with their parameters and outputs, can be inspected; agents can be published as a side panel on shared maps.
Collapsible layer groups, a command-line tool that creates maps from configuration files, and a hosted MCP server for external assistants.
Relevance to our product: Insurer and scientist audiences. Pattern 7 is fully open for us: no chart in our product filters anything today. The traceable assistant is the model for pattern 18.
Visuals: Builder page (screenshots and full video), What's new in CARTO, Q2 2026

Felt¶

Freemium (the free plan cannot upload data files; Professional has a 7-day trial) · felt.com · Interface tour · Dashboards · Pricing
Collaborative web mapping with a consumer-grade feel; a benchmark for map-application ergonomics.
Key idea: A detail panel that appears only on selection, a table drawer under the map, and separate editor and viewer modes.
What it does well:
The detail panel is hidden by default: it appears only when something is selected, shows exactly that item, then gets out of the way.
The left panel leads with the legend for viewers, with an editor-only list tab — one panel, two audiences.
A table drawer at the bottom: a sortable, filterable table linked to the map, for detail without leaving the map.
Editor and viewer modes on one map — suited to an analyst preparing a view that a client then reads.
Dashboards are a separate surface with dropdown and slider filters.
A developer toolkit lets a host application control panels and selection; a mobile field application complements the web product.
Relevance to our product: All map screens. The selection-driven detail panel is already in place on Assess, Live and Datasets. The table drawer exists on Live only and should be extended to Assess and Portfolios (pattern 8).
Visuals: Interface tour (screenshots), Dashboards page

Command palette (keyboard search)¶

Free (open-source component) · cmdk
One keyboard-first search box (Ctrl/⌘ K) over every page, saved item and action.
Key idea: One search box that jumps to any page or saved item — and can run actions.
What it does well:
One palette that searches pages, runs, portfolios, datasets, live watches and places, and jumps straight there.
Actions as well as navigation: the same palette can set the year or scenario, or start a run.
One search service behind both the palette and a visible search box, so each new kind of item is registered once.
A maintained open-source component exists for our technology; on phones it should be paired with a visible search field.
Relevance to our product: All audiences. A low-effort improvement for a product with nine main sections; it shares its search service with a home-page search box (pattern 11).
Visuals: cmdk live demo

5. Risk engines that could run alongside ours¶

Our platform is engine-agnostic by design: today it runs its own engine and CLIMADA side by side and shows where they agree. These are the open and licensable engines that could join them, and what each would add.

Engine	Calculates	Licence	Effort to connect	What it adds
Our own engine	Multi-hazard impact; expected annual loss, return-period loss	Own	In place	Default; fast and lightweight
CLIMADA	Multi-hazard probabilistic impact; cost-benefit	Open source	In place	Scientific reference; also used by the EU insurance supervisor
Oasis LMF	Insured loss with policy terms; market-standard outputs	BSD-3	Moderate effort	Insurance financial structure; opens vendor models
Delft-FIAT	Flood damage from depth maps	MIT	Low effort	Independent third opinion for flood
OS-Climate physrisk	Acute and chronic hazards; impact ranges	Apache-2.0	Low–moderate effort	Heat, drought, water stress; uncertainty
GEM OpenQuake	Earthquake hazard, damage and loss	AGPL-3.0 (review)	Moderate effort	New hazard: earthquake
Deltares SFINCS	Flood maps from forecasts	GPL-3.0 (review)	High effort	Generates hazard for live events
Deltares RA2CE	Road disruption and isolation	GPL-3.0 (review)	Moderate–high effort	New result type for civil protection

Oasis Loss Modelling Framework (as an engine)¶

Free / open source (BSD-3) · Documentation · Open data standards · Model providers
Open-source catastrophe-loss engine with the insurance market's open data standards.
Key idea: Adds insurance financial terms — deductibles, limits, reinsurance — and the market's standard loss outputs.
Calculates: Ground-up and insured loss with policy terms; occurrence and aggregate exceedance curves, average annual loss, in the market's standard results format.
Licence: BSD-3 — suitable for commercial use.
How it would connect: As a separate service beside our platform, in the same way CLIMADA runs today. Our hazard and damage-curve data would be converted to its binned model format.
What it adds: The only candidate that applies insurance financial structure, so insured loss can finally be reported. Speaks the insurance market's own data language (Open Exposure Data / Open Results Data).
Effort: Moderate.
What it does well:
Applies policy terms and reinsurance structures and produces the loss-exceedance outputs insurers report.
Defines open standards for exposure and results data that the market already uses.
Ships a small reference model that any installation can run to prove it works — a model for a conformance test for our own engines.
Relevance to our product: Insurer audience — the most valuable additional engine for portfolio work (pattern 20). Adopting its results vocabulary would make all our engines comparable measure for measure.
Visuals: Documentation site

Commercial models through Oasis (JBA, Fathom, KatRisk, Impact Forecasting and others)¶

Paid, per model · Model providers · Oasis hosted service
Licensed vendor models that run in Oasis format and would become reachable through the same connection.
Key idea: Let an insurer run the vendor model they already license inside our workbench and compare it with ours.
Calculates: Depends on the model — flood, windstorm, earthquake and others, by region.
Licence: Commercial licence per model from each vendor.
How it would connect: Through the Oasis connection above; no separate integration per vendor.
What it adds: Credibility with insurers: a recognised vendor model and our own engines side by side on the same portfolio.
Effort: Low once the Oasis connection exists; commercial agreements needed.
What it does well:
One technical connection opens a catalogue of recognised commercial models.
The industry-owned hosted service offers a vendor-neutral route to the same models.
Relevance to our product: Insurer audience. Depends on the Oasis engine connection (pattern 20).

Deltares Delft-FIAT¶

Free / open source (MIT) · Documentation · Source and licence
Open-source flood impact assessment tool: flood-depth maps, exposure and depth-damage curves in, damage out.
Key idea: The simplest independent third opinion for flood: its inputs are already the ones we use.
Calculates: Flood damage per asset and per map; risk across return periods.
Licence: MIT — suitable for commercial use.
How it would connect: A Python package run as its own small service. Inputs match ours closely: flood-depth raster, exposure points or polygons, depth-damage curves.
What it adds: An independent, deterministic method from a recognised institute — disagreement with our probabilistic engine is scientifically informative.
Effort: Low. Note: recent development activity is concentrated in the FloodAdapt project built on top of it.
What it does well:
Uses the same three inputs our platform already manages, so little conversion is needed.
Widely used in European and international flood-risk studies.
Relevance to our product: Citizen and scientist audiences — a third engine in the agreement view for flood at low cost (pattern 20).
Visuals: Documentation site

Deltares FloodAdapt¶

Free / open source · FloodAdapt
A decision-support application for flood adaptation measures and what-if scenarios, built on Delft-FIAT.
Key idea: A worked example of the adaptation journey: define a measure, compare with and without it, see benefit against cost.
What it does well:
A guided flow: choose an event or scenario, define measures (for example raising properties, flood walls, buy-outs), and compare outcomes with and without them.
Results are shown on the map and as summary figures, designed for planners rather than modellers.
Relevance to our product: Decision-maker audience. Our platform already has a cost-benefit calculation and a panel for it, but no journey for defining measures; FloodAdapt is the reference for that journey (pattern 22).
Visuals: Documentation with screenshots

OS-Climate physrisk¶

Free / open source (Apache-2.0) · Source
Open-source physical climate risk library with vulnerability models, from the OS-Climate / FINOS community.
Key idea: Adds long-term hazards such as heat, drought and water stress, and returns ranges rather than single figures.
Calculates: Asset-level impact distributions for acute and chronic hazards under climate scenarios.
Licence: Apache-2.0 — suitable for commercial use.
How it would connect: A Python package or its companion web service; needs an adapter between its hazard-indicator format and our rasters.
What it adds: Chronic hazards that neither current engine covers, and impact as a distribution — the agreement view gains an uncertainty dimension.
Effort: Low to moderate.
What it does well:
Backed by a finance-industry open-source community, which helps acceptance by banks and investors.
Returns a distribution of impact, not only a single expected value.
Relevance to our product: Decision-maker and scientist audiences (patterns 15 and 20).

GEM OpenQuake Engine¶

Free / open source (AGPL-3.0); other licence terms available from GEM · Source · Licence terms
The leading open-source earthquake hazard and risk engine, from the Global Earthquake Model foundation.
Key idea: The realistic route to earthquake risk — the dominant hazard in Greece and Cyprus.
Calculates: Seismic hazard, scenario and probabilistic damage and loss, with established exposure and vulnerability models.
Licence: AGPL-3.0 — to be run unmodified as a separate service; a licence review is needed before adoption.
How it would connect: It ships its own web service, so the connection is similar to how CLIMADA runs today. Its inputs differ from ours (ground-motion fields rather than depth rasters), so it would be a new hazard line rather than a like-for-like comparison.
What it adds: Earthquake — a hazard neither current engine covers and one every Greek or Cypriot insurer must price.
Effort: Moderate; mostly data preparation.
What it does well:
The international scientific standard for seismic risk, with regional models for Europe.
Mature web service and documentation.
Relevance to our product: Scientist and insurer audiences. Opens a new hazard rather than a third opinion (pattern 20).

Deltares SFINCS¶

Free / open source (GPL-3.0) · Source
A fast flood-inundation model that turns rainfall, river flow and sea level into flood maps in minutes.
Key idea: An engine that produces the hazard itself — from a forecast to a flood map to an impact.
Calculates: Compound flood extent and depth from rainfall, river discharge and coastal water levels.
Licence: GPL-3.0 — to be run unmodified as a separate program; a licence review is needed before adoption.
How it would connect: As a separate service, through a new "hazard engine" interface distinct from today's impact-engine interface. Needs terrain data and forcing inputs.
What it adds: Today every scenario is limited to hazard maps we have ingested. A hazard engine would let a live forecast become a flood map and then an impact.
Effort: High.
What it does well:
Designed for speed, which suits near-real-time use.
Relevance to our product: Decision-maker audience — a longer-term option for the Live section (pattern 20).

Deltares RA2CE¶

Free / open source (GPL-3.0) · Source
Open-source analysis of how hazards disrupt road networks: which roads are cut, which communities are isolated.
Key idea: Answers the civil-protection question — which roads are cut and who is isolated — rather than a money figure.
Calculates: Road damage, detour times and isolated population from a flood map and a road network.
Licence: GPL-3.0 — to be run unmodified as a separate program; a licence review is needed before adoption.
How it would connect: As a separate service. It uses our flood-depth rasters plus an open road network, and needs a new kind of result (disruption rather than monetary loss).
What it adds: A result type that no current engine can express and that emergency planners ask for first.
Effort: Moderate to high.
What it does well:
Directly addresses accessibility and isolation, which monetary loss does not capture.
Relevance to our product: Decision-maker audience (pattern 20).

6. Regulatory tools, alerting services and data sources to build on¶

Not competitors but reference points and inputs: the supervisor's own tooling, the services civil-protection users already watch, and datasets that would raise the quality of every result.

EIOPA open-source climate-risk tools (CLIMADA application)¶

Free / open source · EIOPA open-source tools
The European insurance supervisor's own open-source tooling for natural-catastrophe and climate risk, built on CLIMADA.
Key idea: The supervisor's own tool runs on an engine we already offer — a natural benchmark and a credibility link.
What it does well:
A supervisor-endorsed front end over the CLIMADA engine for assessing natural-catastrophe risk under climate scenarios.
Designed to support insurers' own risk and solvency assessments.
Relevance to our product: Insurer and decision-maker audiences. We already produce a solvency-assessment report pack and run CLIMADA as one of our engines; the positioning is "the supervisor's model, plus a second opinion, plus a full record of how each figure was produced" (patterns 20, 21 and 25).
Visuals: EIOPA tools page

EU sustainability-reporting requirements (CSRD / ESRS, EU Taxonomy)¶

Public regulation · Revised reporting standards, July 2026
The reporting frameworks that commercial climate-risk products align their outputs to.
Key idea: Target the physical-risk parts of the revised standards and keep the insurer solvency report as the lead regulatory output.
What it does well:
Competitors list alignment framework by framework (CSRD / ESRS E1, IFRS S2, TCFD, EU Taxonomy) and sell reporting as a separate bundle.
The European Commission adopted revised, simplified reporting standards in July 2026; the scope of companies required to report has been narrowed. [To verify against the legal texts before quoting figures.]
Relevance to our product: Insurer and decision-maker audiences. A focused physical-risk output for the revised standards, beside the solvency report pack we already produce, is more valuable than a broad reporting product (pattern 25).

Alerting and rapid-mapping services (GDACS, Copernicus Emergency Management Service, MeteoAlarm, Disaster Ninja)¶

Free · GDACS · Copernicus rapid-mapping activations · MeteoAlarm · Disaster Ninja
The services civil-protection and humanitarian users already watch during an event.
Key idea: Treat each event as a named, time-stamped item with an alert level, a footprint and follow-on products.
What it does well:
Each event is a named item with an alert level, a timeline, an affected area and a list of products that follow.
Alert levels use a small, widely understood colour scale.
Several offer public data feeds that could become additional live sources.
Relevance to our product: Decision-maker audience. Our Live section is built around data feeds; an event-centred view with alert levels, plus these feeds as extra sources, would bring it closer to what this audience already uses (pattern 23).
Visuals: Copernicus activations list, Disaster Ninja

National flood maps for Greece and Cyprus¶

Free (public authorities) · Greece — flood risk management plans · Cyprus — Water Development Department
The official flood hazard and risk maps produced under the EU Floods Directive.
Key idea: Show the official national maps as a labelled reference layer — the first thing a local reader will check.
What it does well:
Authoritative maps for defined return periods, produced by the responsible national authorities.
A free, independent reference against which our own flood layers can be checked.
To confirm hands-on:
Licence and download format of each dataset, and the interactive flood-map viewer of the Cyprus Civil Defence (not reachable during this review).
Relevance to our product: Citizen and decision-maker audiences in our two covered countries.
Visuals: Greek flood maps portal

Destination Earth — Climate Change Adaptation Digital Twin¶

Free for eligible users (EU programme) · Climate digital twin
The EU flagship producing kilometre-scale climate projections and replays of recent extremes under future conditions.
Key idea: "Storyline" scenarios: replay a real recent extreme event under a warmer climate.
What it does well:
Multi-decade projections at a much finer scale than the previous generation of climate data.
Storyline simulations that re-run real recent extreme events under future conditions — an intuitive scenario type for non-experts.
Relevance to our product: Scientist and decision-maker audiences. A future hazard input, and a scenario idea that pairs naturally with our event-replay capability.
Visuals: Overview page

Open building-footprint datasets (Overture Maps, EUBUCCO)¶

Free / open data · Overture Maps · EUBUCCO
Open, regularly updated building outlines with attributes, for Europe and worldwide.
Key idea: Let a map pin resolve to a building, and let a portfolio join to real building outlines.
What it does well:
Building-level outlines with stable identifiers, suitable for joining to customer portfolios.
Height, type and age attributes where available, which improve vulnerability estimates.
To confirm hands-on:
Coverage and attribute completeness for Greece and Cyprus; licence terms of each dataset.
Relevance to our product: Citizen and insurer audiences. Our exposure data today is on a grid; building outlines would make results recognisable at street level and support pattern 4's most detailed level.

Recurring patterns — and where the product stands (UI v0.16.0)¶

3 in place · 15 partly in place · 7 not yet in place. File references are into climate-lama-ui unless prefixed climate-lama:.

A detail panel that appears on selection, instead of a permanent side panel — In place. Seen in: Felt, XDI, First Street.
Today: Assess, Live and Datasets open a detail panel only when something is selected; Assess is a collapsible side rail plus a full-width map.
Remains: Keep the behaviour consistent on every new screen.
Evidence: src/components/AssetDetailPanel.tsx:82; src/components/Dashboard.tsx:2593,2762-2778; src/components/live/Inspector.tsx; src/components/datasets/DatasetInspector.tsx
A layer list where hazard, exposure and results are equal, stackable layers — Partly in place. Seen in: kepler.gl, Felt, CARTO, CatNet, ExposureIQ.
Today: A small floating panel with three fixed rows (hazard, exposure, impact), each with a visibility switch, a legend and an opacity slider.
Remains: A full layer list: reorderable cards, colour and scale per layer, collapsible groups, an "Add layer" control that opens the dataset catalogues so several hazards can be stacked, with the layer set held in the web address. Screens: Assess, Lookup, Live, Portfolios.
Evidence: src/components/LayerPanel.tsx:222-331; src/components/MapView.tsx:234
Show aggregated areas instead of individual points, with a choice of what to aggregate — Partly in place. Seen in: deck.gl, kepler.gl, Foursquare Hex Tiles, ExposureIQ.
Today: Points are grouped by count only. A hexagon library is installed but used for a single outline on Lookup.
Remains: Hexagon aggregates calculated in the database per level of detail and served as map tiles (level chosen from zoom), with a picker for count, total value, average loss or maximum loss that only changes colouring. Set a measured performance target. Adopt deck.gl only if measurements require it. Screens: Assess results and Portfolios.
Evidence: src/components/PortfolioMap.tsx:417-429; src/map/styles/results.ts:49-56; src/components/lookup/LookupMap.tsx:5. Tech check 2026-09-18: PostGIS ST_AsMVT tiles first, PMTiles second, @deck.gl/maplibre 9.4.x third; GeoJSON fill is fine to ~10^4 cells only
Detail that follows the zoom, with a visible switch for the level — Partly in place. Seen in: First Street, FEMA, XDI.
Today: Two automatic levels (regions when zoomed out, points when zoomed in) on the Assess result map only; the user cannot choose a level.
Remains: A visible switch — Automatic, Region, Municipality, Hexagons, Assets — held in the web address; more automatic steps; a clear message where a level is unavailable for a country. Screens: Assess first, then Portfolios.
Evidence: src/components/ResultMap.tsx:81-147
Parameters that stay visible as a compact summary after a run — Partly in place. Seen in: Jupiter, Spectra, CARTO, Fathom, Probable Futures.
Today: The Assess side rail collapses and resizes, but when collapsed it shows only icons, and it still behaves as a form to submit.
Remains: After a run, show one chip per chosen parameter (hazard, exposure, scenario, year, engine, damage function); click a chip to change it. Settings that need no recalculation apply immediately; the rest offer a re-run. Screens: Assess and Portfolios.
Evidence: src/components/Dashboard.tsx:1042-1053,2571; src/components/ConfigPanel.tsx:536
Analyses saved as named items that can be re-opened, re-run and compared — In place. Seen in: Oasis, Risk Modeler, Spectra, Synergy Studio.
Today: The Runs section lists every saved scenario and batch as a named item with progress status, and items can be compared.
Remains: Nothing structural.
Evidence: src/pages/RunsPage.tsx:19-39,215-222; src/hooks/useJobPoller.ts; src/components/assess/RunStatusPanel.tsx
Charts that also filter the map and each other — Not yet in place. Seen in: CARTO Builder, kepler.gl, Verisk.
Today: Every chart is display-only; nothing can be brushed, and selecting on the map does not highlight a chart.
Remains: Make the loss-frequency curve, loss-by-region bars and per-asset histogram clickable so they filter the map and each other. The most economical route is to drive this from the table drawer's filter (pattern 8).
Evidence: No onClick/Brush in src/components/FreqCurvePanel.tsx, RiskSeriesChart.tsx, lookup/RiskOverTimeChart.tsx, live/LossSparkline.tsx; src/components/MapView.tsx:257
A table drawer under the map, linked to it in both directions — Partly in place. Seen in: Felt, CARTO.
Today: Available in the Live section only. Portfolios has a separate table; Assess has none.
Remains: Make it a shared component and add it to Assess (loss per asset) and Portfolios (portfolio members). Selecting a row selects the map feature and opens its detail; selecting on the map brings the row into view. Collapsed by default.
Evidence: src/components/live/AssetsDrawer.tsx; imported only by src/pages/LiveEventPage.tsx
One headline grade that opens into hazards and the factors behind them — Partly in place. Seen in: Munich Re, EarthScan, First Street, ZestyAI, Fathom, Riskthinking.
Today: Lookup shows a score card per hazard with a "Why this score?" explanation and clear messages where a hazard is not covered.
Remains: One headline grade per location (stating how many hazards it could not include), an optional hazard-group level, named risk states with published definitions, and the contributing factors in each explanation: hazard intensity, damage-curve point and exposed value. Reuse on the portfolio page.
Evidence: src/components/lookup/ScoreCardPanel.tsx:14-57,280; src/components/lookup/HazardDetail.tsx:226
Comparison built in: a slider across one map, or two synchronised maps — Partly in place. Seen in: Probable Futures, JBA, kepler.gl.
Today: Two synchronised maps exist for comparing saved scenarios, reached from Runs. There is no slider, and no comparison on Assess, Lookup or Live.
Remains: A shared comparison component with side-by-side and slider modes (built in-house; the available open-source slider is no longer maintained), used for year A vs. B, engine A vs. B and return period A vs. B. Show the difference on the slider handle.
Evidence: src/pages/ScenarioComparePage.tsx:75-131. Tech check: @maplibre/maplibre-gl-compare last released 2021
Start from a search, with a browsable catalogue; upload second — Partly in place. Seen in: First Street, ThinkHazard!, Spectra, VELO, Climate Central.
Today: Lookup has address and coordinate search; Datasets has browsable hazard and exposure catalogues with upload for advanced users. The home page has no search, and Assess begins with dataset pickers.
Remains: One search field on the home page that accepts a place, an administrative area, coordinates, an asset or a dataset and opens the right screen; a "start from a place" entry point on Assess that pre-selects datasets from coverage.
Evidence: src/pages/HomePage.tsx:70-82; src/components/home/StartHereCard.tsx:141-162; src/components/lookup/AddressSearch.tsx
Observed events as their own kind of layer, labelled honestly — In place. Seen in: ExposureIQ, CatNet, EU Risk Data Hub, ThinkHazard!.
Today: Observed layers appear beside modelled ones with their own legends; replayed and near-real-time events are labelled differently; a hazard that is not covered is shown as such, never as zero.
Remains: Add a reliability indication per data source, following Google Flood Hub.
Evidence: src/components/LayerPanel.tsx:7-9,105-113; src/components/live/EventKindChip.tsx:21-24; src/components/lookup/ScoreCardPanel.tsx:14-20
Reports and exports from every screen — Partly in place. Seen in: Verisk, Spectra, Munich Re, Climate Central.
Today: CSV and PDF export on Assess and Portfolios, and a composed solvency-assessment report pack per batch. None on Lookup, Live or the scenario comparison.
Remains: A location risk profile from Lookup, an event summary from Live, and a side-by-side comparison report (which needs new server work). Every export carries the same coverage notes and caveats as the screen. Make download and screenshot primary map actions.
Evidence: src/components/ExportButton.tsx:13-16; src/components/assess/ComposedReportButton.tsx:22-38
Animated playback over time — Not yet in place. Seen in: Windy, Zoom Earth, kepler.gl.
Today: The Live timeline is moved by hand; no other screen has a time control.
Remains: Play and pause with automatic advance on Live; animate the year dimension on Assess.
Evidence: src/components/live/SnapshotScrubber.tsx
Show disagreement and uncertainty on the map; accessible colours that follow the display setting — Not yet in place. Seen in: IPCC and Copernicus atlases, OS-Climate physrisk.
Today: Disagreement between engines is calculated but drawn as a confident colour. Map colours are fixed, so the high-contrast display setting does not reach the map.
Remains: Hatch matrix cells and map areas where engines disagree, using a stated, published-style rule; move risk bands and hazard colours onto the display theme with a colour-blind-safe scale.
Evidence: src/lib/riskLevels.ts:2-6; src/components/LayerPanel.tsx:132; src/components/live/EventLossHeadline.tsx:25-26
A keyboard command palette and universal search — Not yet in place. Seen in: Probable Futures, Linear-style palettes.
Today: Nine main sections and separate lists of items; no keyboard navigation.
Remains: One Ctrl/⌘ K palette over pages, runs, portfolios, datasets, watches, places and actions — sharing its search service with the home-page search (pattern 11).
Evidence: No cmdk/metaKey usage in climate-lama-ui/src
Phone-friendly screens and a "copy link to this view" button — Not yet in place. Seen in: Climate Central, Windy, Felt.
Today: Four screens hold their state in the web address but none offers the link. The home page and Assess are designed for desktop only.
Remains: Home and Lookup usable at phone width (390 px: full-width map, score cards as a bottom sheet) and a share button on every screen that holds its state in the address.
Evidence: useSearchParams in Dashboard.tsx:266, RiskLookupPage.tsx:239, LiveEventPage.tsx:194, ScenarioComparePage.tsx:453; responsive prefixes: HomePage.tsx 0, Dashboard.tsx 1
One set of named actions that both the controls and a future assistant use — Not yet in place. Seen in: CARTO AI Agents, kepler.gl, Foursquare, Moody's IRP Navigator.
Today: No assistant; each screen wires its own controls.
Remains: Define a small set of named actions (set region, set return period, set scenario, open run) used by the interface controls. A limited assistant that only combines existing actions, shows which ones it ran, and cites the result it read — never producing a figure itself — can follow later. Decide explicitly whether it belongs in this phase.
Evidence: No assistant/chat component in climate-lama-ui/src
Querying result data directly in the browser (trial only) — Not yet in place. Seen in: kepler.gl, Foursquare Spatial Desktop.
Today: Every new way of cutting a result needs a new server endpoint.
Remains: A short, time-limited trial of an in-browser database over result files, measured against adding server endpoints. No commitment.
Several engines over the same exposure, compared in place — Partly in place. Seen in: Oasis hosted service, Moody's, CatNet, Verisk Synergy Studio.
Today: Two engines (our own and CLIMADA) can be selected and compared, with an agreement matrix across scenarios.
Remains: Add engines in order of value — Oasis (insured loss and vendor models), Delft-FIAT (independent flood method), OS-Climate physrisk (long-term hazards, ranges), OpenQuake (earthquake). Adopt the insurance market's results vocabulary (occurrence and aggregate exceedance, probable maximum loss, an optional range) and a small reference scenario every engine must reproduce before it is offered.
Evidence: climate-lama: src/climate_lama/worker/models/base.py:13-31 (ImpactResult has no OEP/AEP/PML/uncertainty; insured_loss unfilled), registry.py, climada_adapter.py; UI: assess/EngineComparisonPanel, AgreementView
Open methods: a fact sheet per engine, a record of how each figure was produced, and a bundle to reproduce it — Partly in place. Seen in: ZestyAI, Jupiter, S&P Climanomics, Fathom, Probable Futures.
Today: Results carry provenance information and Lookup shows the data source per score; reproducibility exists in the platform but has no user-facing form.
Remains: A fact sheet per engine (version, hazards, curves, known limits) reachable from every comparison cell; a downloadable bundle that reproduces a result; a public, dated method document and changelog.
Evidence: UI: src/components/lookup/ProvenanceChip.tsx; backbone: src/climate_lama/core/engine_comparison_service.py
Adaptation what-if: change an asset or add a measure and see risk and benefit change — Partly in place. Seen in: First Street, XDI, Climate X Adapt, Jupiter Adaptation Hub, FloodAdapt.
Today: A cost-benefit calculation and a results panel exist, but there is no journey for defining measures, and the future-climate side of the calculation needs correcting before benefit figures are shown.
Remains: A guided flow following FloodAdapt: choose a measure, compare with and without it on the map, show avoided loss and break-even. Let users edit asset characteristics on Lookup and Portfolios.
Evidence: UI: src/components/CostBenefitPanel.tsx; backbone: worker/models/base.py:108-148, engine_adapter.py:314-317 (present hazard passed as future hazard)
Alerts and subscriptions for watched places, with events as named items — Partly in place. Seen in: Google Flood Hub, GDACS, Copernicus rapid mapping, EFAS.
Today: Live has watches and the home page has a notifications card inside the application; there are no email or push alerts, and events are organised by data feed rather than as named events with an alert level.
Remains: Subscriptions for a watched place or portfolio; an event-centred view with alert level, timeline and affected area; additional public alert feeds as sources.
Evidence: src/components/home/NotificationsCard.tsx; no subscription/email code in climate-lama-ui/src
The same results through several routes: screens, API, files and embeddable maps — Partly in place. Seen in: Fathom, JBA, First Street, Google Flood Hub, Probable Futures.
Today: A documented API and a software development kit exist alongside the interface.
Remains: A simple public per-location query, map-tile streaming of result layers into a customer's own GIS, and embeddable map views — each with a clear licence statement.
Evidence: climate-lama: /v1 REST API + generated SDK (sdk-smoke.yml)
Regulator-ready outputs — Partly in place. Seen in: Munich Re Reporting Edition, EarthScan, Jupiter Compliance Hub, EIOPA tools.
Today: A composed solvency-assessment report pack is produced per batch of scenarios.
Remains: A focused physical-risk output for the revised EU sustainability-reporting standards and the EU Taxonomy climate-risk assessment, with alignment stated framework by framework.
Evidence: climate-lama: src/climate_lama/core/reports/templates/orsa_report.html; UI: assess/ComposedReportButton.tsx

Most worth studying hands-on: CARTO Builder (charts that filter the map; traceable assistant), kepler.gl (layer list, aggregation picker), Felt (table drawer, editor/viewer modes), First Street (address-first journey, owner-editable what-if), Probable Futures (slider comparison), the IPCC / Copernicus atlases (disagreement hatching, accessible colours), Google Flood Hub (source reliability, alerts), FloodAdapt (adaptation journey), and Oasis LMF + Delft-FIAT as the first additional engines.

Open questions¶

Answered 2026-09-26 in phase-17-revamp.md (design record D1–D22 and the Direction board). The list below is kept as the record of what was open.

Settled by the earlier phases: the parameter surface is persistent-but-collapsible, not a wizard; the map is one surface with layer rows, not a view per step; this work is its own phase (17).

Open for the Phase 17 design doc:

Aggregation strategy (pattern 3). Recommended order, from the 2026-09-18 technical check: (1) H3 aggregates per resolution in PostGIS served as vector tiles (ST_AsMVT), resolution from zoom, styling at style level; (2) PMTiles archives of the same aggregates; (3) deck.gl H3HexagonLayer via @deck.gl/maplibre 9.4.x only if a measured portfolio defeats both. A GeoJSON fill layer from h3-js is fine to ~10⁴ cells only. Acceptance = a measured frame and parse time at a named cell count.
MapLibre major. Stay on v4 (≥ 4.5.1) for the phase, or upgrade to v5/v6 first?
Which engines, in what order (pattern 20). Recommended: results vocabulary (OEP/AEP, PML, optional range) + conformance scenario first; then Oasis LMF (BSD-3; insured loss, opens vendor models) and Delft-FIAT (MIT; independent flood method); OS-Climate physrisk (Apache-2.0) next; OpenQuake (AGPL-3.0), SFINCS and RA2CE (GPL-3.0) only after a licence review and only as unmodified separate services.
Assistant scope (pattern 18). Build the named action surface and leave the assistant out, or ship a bounded, traceable one?
Phone-width scope (pattern 17). Lookup + Home only (the citizen moment), or every surface?
Headline grade (pattern 9). A–F letters vs. the existing EAI bands, and how a grade states the hazards it could not include.
Swipe comparison (pattern 10). Build in-house — @maplibre/maplibre-gl-compare has had no release since 2021.
Regulatory output (pattern 25). ORSA pack stays the lead; add a focused ESRS E1 / EU-Taxonomy physical-risk output? [Scope figures of the 2026 CSRD simplification are [ASSUMPTION] until read from the legal texts.]
Client-side query spike (pattern 19). In or out of the phase.

Next steps¶

~~Market scan: products, engines, sources; patterns graded against the UI~~ — done 2026-09-18 (two /reassess fast passes).
Collect screenshots / video per entry into the stakeholder Word document (Market scan.docx, seeded from the same scripts/market_scan/data.js) and close the "To confirm hands-on" items with trial accounts. Capture first: First Street, the 2022 EarthScan article, Verisk Touchstone, FEMA's NRI user guides.
Take the open questions above to stakeholders, then write docs/plan/phase-17-ui-revamp.md and run /phase-split.
Fix the react-leaflet description in climate-lama-ui/docs/CLIMATE_LAMA_UI.md.

Subscription tiers — design note (parked, Phase 17 D20)¶

plan.md keeps the business model out of focus, so nothing is built for tiers or billing. This note only records which capabilities a tier would gate, from the owner's comments C24, C25 and C97 and the CatNet Basic / Premium / API split the owner liked:

Capability	Basic	Premium	API
Place lookups and the place report	limited per month	unlimited	per call
Locations per portfolio	capped	uncapped	per call
Interactive map depth (hexagons, assets, 3D)	regions only	full	tiles by licence
Scenario × horizon grid, compare, engines	today + one scenario	full	full
Adaptation and reports studio	view only	full	—
Assistant	—	included	—

Revisit only if the owner changes plan.md's "not in focus" line.

Phase 17 charter stub — inputs from Phase 14 (2026-09-05; renumbered 2026-09-08 and 2026-09-11)¶

Renumbered 2026-09-11. The owner defined Phase 16 — Consistency — a small no-epic phase that fixes the inconsistencies a prod walkthrough surfaced — in the slot this stub held; the revamp becomes Phase 17. Phase 16 hands it a surface set that already behaves consistently (grouped pickers, an honest map tier, popups and dialogs that behave, a visible nav state), so the revamp starts from presentation, not from repair.

Renumbered 2026-09-08. The owner defined Phase 15 — Live Risk in the earlier slot; the revamp then became Phase 16. Nothing below is dropped. Phase 15 builds its Live tab as the pilot of this document's recurring patterns (map-first, selection-driven inspector, layer cards, bottom drawer, timeline scrubber — patterns 1, 2, 8, 10, 12), so the revamp inherits proven components.

This document is the seed of Phase 17 — UI revamp, which opens after Phase 15 — Live Risk and Phase 16 — Consistency close. Phase 14 hands it a smaller, consistent surface set (Home · Assess · Lookup · Portfolios · Datasets · Runs · Settings · Admin) and the following deliberate deferrals, each recorded with the reason it was not solidification:

Deferral	Why not Phase 14	Precondition recorded
Visual redesign of every surface (this doc's four problem areas: map, charts, results, parameters)	design work, not a fix	Phase 14 leaves each surface answering its question with a working contract underneath
Adaptation measures: create/edit UI, seeded catalog, cost-benefit on Assess and Portfolios	new user capability	the engine adapter passes the present hazard as the future hazard (`worker/models/engine_adapter.py:314-317`) — fix before any BCR is shown
Stochastic event-set hazards (CLIMADA `.h5` with a dated multi-event catalogue, tropical-cyclone track sets)	needs an event-set store; the raster-per-event model and `hazard_events`' two writable shapes cannot hold one	Phase 14's `.h5` reader refuses these with a message naming this track
GeoTIFF exposure upload through the API	adapter exists operator-only (ADR-055); completing it is a new ingress	`core/ingest/raster_exposure.py`
Engine selector beyond `climate_lama` + the CLIMADA sidecar; a third engine	engine plurality product commitment, no candidate yet	Assess carries the Engine option + Compare toggle from Phase 14
Per-asset impact-function binding (`impf_*` columns from CLIMADA exposures)	the impact-function model has no per-asset binding	Phase 14 keeps `impf_*` in `properties` untouched
Country onboarding beyond Greece and Cyprus	#447	—

Problem-area 4 (parameters) is partly consumed by Phase 14: Assess gains a collapsible rail, a horizon-drives-dataset control and a scenario × horizon multi-select; the "chip strip after RUN" idea was not built in Phase 15 and remains open for Phase 17 (pattern 5).