Skip to content

RA · Foundations

Risk science and methods: what to learn, in what order, and why

A learning guide to the Faculty of Risk Science and Methods at Risk Academy. It sets out the core ideas, a suggested order of study, the standards and references to know, and how learning is recorded.

As of
3 November 2026
읽기
10 minutes
증거
12 sources, cited in order
단계
All levels: a map for new learners and practitioners
학습 시간
15 minutes
NEXUS TWIN · WF-1 · SURFACESEED 31594ba5RA-F04.FACULTYxyzSCHEMATIC · SURFACEST-01 · 0.531 0.556ST-02 · 0.402 0.307
WF-1 · surface Schematic wireframe twin (surface, catastrophe model). An illustration, not data.

What you will be able to do

  • Trace a risk number back through its data, model, forecast and communication.
  • Explain why model validation and data provenance matter, with a public example.
  • Score probability forecasts with a proper score and explain why honest forecasts score best.
  • Name common biases in judgement under uncertainty and how elicitation reduces them.
  • Explain why the shape of a joint tail matters for losses.
단계
All levels: a map for new learners and practitioners
학습 시간
15 minutes
Before you start
None: the guide starts from the basic ideas
리스크 데이터
2013
year the Basel principles on risk data aggregation were published1
Forecasts
15 days
range of the ECMWF medium-range ensemble6
개방 데이터
2016
year the FAIR data principles were published3

01What the faculty of risk science and methods covers

What the faculty of risk science and methods covers

02Why methods decide whether a risk number can be trusted

Why methods decide whether a risk number can be trusted

Every risk number is the end of a chain: data collected by someone, cleaned, combined, fed to a model, tested, and explained. A weak link anywhere makes the number misleading, however precise it looks.

The banking world learned this the hard way. The Basel Committee's principles for risk data aggregation, published in January 2013, start from the finding that during the financial crisis many banks were unable to aggregate risk exposures and identify concentrations fully, quickly and accurately 1.

Public bodies have drawn the same lesson for analysis in government. The UK AQuA Book is guidance on how to produce robust, fit-for-purpose analysis, written for analysts, their managers and the people who commission their work 5.

03How the faculty is organised

How the faculty is organised

The faculty groups its disciplines into departments of the faculty. One group covers measurement and modelling: statistics, catastrophe models, geospatial analysis and the validation of models. A second covers foresight and judgement: forecasts, scenarios, stress tests, expert elicitation and decisions under uncertainty.

A third group covers people and information: how risk is perceived, how it is communicated, and how risk data are governed and shared. A fourth covers risk in organisations: enterprise risk management, business continuity and the culture that makes controls work.

The groups are linked by one question: how do we know, how sure are we, and how do we say so? Every discipline in the faculty answers part of it.

04A suggested learning order in four stages

A suggested learning order in four stages

The Academy describes study at four levels. The introductory level is for anyone new to a field. The intermediate level is for practitioners. The advanced level is for specialists. The integrative level is for people who lead work across disciplines and scales. The four stages below follow the same climb, but they are about subjects, not levels.

Stage one: data and probability. Start with the raw material. Learn basic probability and statistics, and how risk data are described and shared. The Risk Data Library Standard is an open data standard that gives a common description of the data used and produced in risk assessments, from hazard and exposure to vulnerability and loss 2. The FAIR principles, published in 2016, set out how data can be made findable, accessible, interoperable and reusable 3. The course EO/GIS Starter Lab sits here.

Stage two: models and their validation. Next, learn how models turn data into estimates of risk, and how to test them. A model that has never been checked against what happened is an opinion with equations.

Stage three: forecasts, scenarios and judgement. Then study how to speak about the future. Weather centres now run ensembles of forecasts: the ECMWF medium-range system is an ensemble of model forecasts giving information up to 15 days ahead 6. Learn how probabilistic forecasts are scored: Gneiting and Raftery describe scoring rules as assessing the quality of probabilistic forecasts by assigning a numerical score based on the predictive distribution and the event or value that materialises 4. Learn also how human judgement goes wrong under uncertainty 7. The courses Uncertainty and Confidence and Scenario Grammar sit here.

Stage four: communication and management. Finally, connect the numbers to decisions. WHO describes risk communication as the real-time exchange of information, advice and opinions between experts or officials and people who face a hazard 8. Learn how organisations manage risk through frameworks such as COSO's enterprise risk management framework 9 and the UK government's Orange Book 10. The integrative studios of the faculty are in preparation.

The order is a suggestion. A risk manager may begin at stage four and work back. A data scientist may begin at stage one. What matters is that each new idea has the ideas beneath it in place.

Fig. 01Four stages of study in risk science and methods

Four stages of study in risk science and methodsA rising ladder of four steps: data and probability, models and their validation, forecasts, scenarios and judgement, and communication and management, each with the standards and references a learner meets at that step.01Data and probabilityRDLS, FAIR principles02Models and validationTest against what happened03Forecasts and judgementEnsembles, scoring rules, biases04Communication and managementWHO risk communication, COSO ERM,Orange Book
The suggested learning order of this guide, with the standards and references that anchor each stage 2,3,6,4,7,8,9,10. The order is a suggestion, not a required curriculum.2,3,6,4,7,8,9,10Sources: Risk Data Library; Scientific Data; ECMWF; JASA; Science; WHO; COSO; HM Treasury. Figure CC BY 4.0. As of 3 November 2026.
Figure 1 in words

Four steps rise from left to right. Step one, data and probability: the Risk Data Library Standard and the FAIR principles. Step two, models and their validation: testing models against what happened. Step three, forecasts, scenarios and judgement: ensemble forecasts, proper scoring rules and the biases of judgement under uncertainty. Step four, highlighted, communication and management: WHO risk communication, COSO enterprise risk management and the Orange Book, with the integrative studios in preparation.

05The chain of risk evidence

The chain of risk evidence

A useful way into the whole faculty is to follow one risk estimate from start to finish. Each link has its own discipline, its own standards and its own ways of going wrong.

Data come first, and need to be described, licensed and traceable 2 3. Models turn data into estimates, and need validation. Forecasts and scenarios look ahead, and probabilistic forecasts can be assessed with scoring rules 4. Indices combine many measures into one, and need sound statistical method: the European Commission's Competence Centre on Composite Indicators and Scoreboards works on exactly this 11, and the INFORM Risk Index is one open, global example 12.

Communication comes last but is not an afterthought: a correct number that nobody understands changes nothing 8. Figure 2 lays out the chain.

Fig. 02The chain of risk evidence

The chain of risk evidenceFive stacked layers from data to communication: data, models, forecasts and scenarios, indices, and communication, each with the standards and references that govern it.01데이터Described, licensed, traceable: RDLS, FAIR, BCBS 23902모델Validated against outcomes; AQuA Book on analysis quality03Forecasts and scenariosEnsembles, scored with proper scoring rules04지수Sound statistical method: JRC centre, INFORM Risk Index05CommunicationA two-way exchange with the people at risk
A teaching frame by the Academy desk, with a reference for each link 2,3,1,5,4,6,11,12,8.2,3,1,5,4,6,11,12,8Sources: Risk Data Library; Scientific Data; BIS; UK Government Analysis Function; JASA; ECMWF; European Commission JRC; WHO. Figure CC BY 4.0. As of 3 November 2026.
Figure 2 in words

Five layers stacked from top to bottom. One, data: described, licensed and traceable, with the Risk Data Library Standard, the FAIR principles and the Basel principles on risk data aggregation. Two, models: turn data into estimates and are validated against outcomes, with the AQuA Book on analysis quality. Three, forecasts and scenarios: ensembles and probabilistic forecasts scored with proper scoring rules. Four, highlighted, indices: many measures combined with sound statistical method, as at the JRC centre on composite indicators and in the INFORM Risk Index. Five, communication: a two-way exchange with the people at risk, as WHO describes it.

06Six core ideas every learner should hold

Six core ideas every learner should hold

Six ideas carry most of the weight in this faculty. Each is simple to state and takes practice to use.

Every number has a history. Know where the data came from, who changed them and under what licence.

All models are tested or they are guesses. Validation against what happened is what separates a model from a story.

Probability needs scoring. Scoring rules assess the quality of probabilistic forecasts by assigning a numerical score based on the predictive distribution and the event or value that materialises 4.

Judgement has known biases. Tversky and Kahneman describe heuristics that are usually effective but lead to systematic and predictable errors 7.

Tails decide losses. Two risks with the same average link can still fail together far more often in extremes, so the shape of the joint tail matters.

Communication is two-way. Risk communication is an exchange with the people at risk, not a broadcast to them 8.

07Worked example: two rivers and a joint flood

Worked example: two rivers and a joint flood

Worked example (teaching values): the numbers in this section are invented for teaching and describe no real place. A city sits where two rivers meet. Each river, on its own, floods the lower town about once in twenty years. The city's analysts know the two rivers tend to run high at the same time.

Worked example, continued (teaching values): one analyst models the link between the rivers with a standard bell-curve assumption. A second uses a model with heavier tails, with the same overall strength of link. Both agree on ordinary years. They disagree on the years that matter: the second model gives a much higher chance that both rivers flood together in the same season.

Neither analyst is wrong to build a model. The lesson is that the choice of model shapes the answer most in the extremes, where data are thinnest. Learning to test that choice, and to report how much the answer depends on it, is the heart of stage two. The twin below lets a learner compare the two.

TwinSame link, different tails: when two risks fail together

Interactive · 3D
Same link, different tails: when two risks fail together
An illustrative statistical model with stated parameters, not real rivers or assets: two risks are joined with the same correlation under a bell-curve (Gaussian) dependence and under heavier-tailed dependence. The model computes exactly how often both pass their 1-in-20 and 1-in-100 levels together under each. It is the worked example of this guide, made explorable. Drag to rotate and scroll to zoom; Explore opens the layers, story steps, time bar and charts.
Method and description

Illustrative model with stated parameters, not data. Exact densities on an 81 × 81 grid; joint exceedance by quadrature; 600 seeded draws for the scatter only.

An illustrative model with stated parameters that depicts no real river, asset or portfolio. Two variables with standard normal margins are joined at the same correlation, first by a Gaussian copula and then by Student-t copulas with several degrees of freedom. The model shows each joint density as a surface, computes tail dependence in closed form, and integrates the chance that both variables exceed their 95th and 99th percentiles together. A sample of points shows how clustered the extremes are under heavier tails. A last panel shows how averages of many draws approach a bell curve while extremes do not. All numbers on screen are model outputs, not observations.

  1. Same correlation. Both copulas join two N(0,1) variables with ρ = 0.6. At ν = 30 the t relief lies almost on the Gaussian wire: peak 0.202 against 0.199, joint 99% exceedance 0.21% against 0.19%, and upper tail dependence already λU = 0.01 instead of 0.
  2. Identical marginals. With the relief hidden, the curtains on the back and left walls are the marginal densities of X and Y under the Gaussian wire. They are the same N(0,1) for every ν: no univariate check can tell the two copulas apart.
  3. Heavier joint tails. At ν = 4 the peak rises to 0.225 and the corners gain mass: P(both above their 99% quantile) is 0.35% against 0.19% for the Gaussian, 1.87 times as likely, and λU = 0.31 versus 0. The coral corner marks that region.
  4. Draws. 600 draws from the ν = 4 copula on the bench, with the 90% footprint in gold: 3 of them land in the coral corner (expected 2.1). Joint 95% exceedance is 2.01% against 1.55% for the Gaussian; at ν = 3 the joint 99% exceedance reaches 0.39%.
  5. The central limit theorem. The lower panel shows why Gaussian marginals are a natural default: the standardised mean of n exponential draws moves from a skewed law (KS distance 0.16 from N(0,1) at n = 1) to 0.024 at n = 30. Joint tails get no such guarantee: at ν = 4 λU stays 0.31.

08Canonical references to know

Canonical references to know

A learner in this faculty should know a short list of references, what kind of evidence each holds, and how often it is renewed. Primary documents come first; summaries are useful only after them.

For data: the RDLS 2, the FAIR principles 3 and BCBS 239 1. For analysis quality: the AQuA Book 5. For forecasts: the ECMWF ensemble documentation 6 and the theory of proper scoring rules 4. For judgement: the classic paper on heuristics and biases 7. For indices: the JRC composite indicators centre 11 and the INFORM Risk Index 12. For management: COSO ERM 9 and the Orange Book 10. For communication: WHO risk communication 8.

09Skills that make the knowledge usable

Skills that make the knowledge usable

Knowledge becomes useful through a small set of skills. The faculty treats these as the practical core of every discipline.

Tracing data. Record the source, date, licence and changes of every dataset used. Validating a model. Compare its outputs with outcomes it was not fitted to. Scoring a forecast. Use a proper score and look at reliability, not only at single hits and misses. Eliciting judgement. Ask experts in ways that reduce anchoring and overconfidence. Saying how sure you are. State a range and what drives it, in words a decision maker can use.

10How expertise is shown at the Academy

How expertise is shown at the Academy

Expertise in this field is shown by work that others can check. The Academy records learning, not credentials.

Finishing a course gives a completion record. It is not a certificate, an accreditation or a qualification, and it does not state that a person is competent in a profession. Achievements on the Academy recognise assessed learning only.

Three courses on NXS Learning are linked to this faculty: EO/GIS Starter Lab at the introductory level, and Uncertainty and Confidence and Scenario Grammar at the intermediate level. On 3 November 2026 their pages say they are in preparation, as are the foundations and frontier courses for the other disciplines and the integrative studios. Nothing is open for enrolment until it is published on NXS Learning.

Outside the Academy, the strongest evidence of expertise is a body of work: a documented analysis that cites its sources and dates, a reproducible calculation, or a clear brief that a reviewer has checked. Learners are encouraged to keep such a record from their first course.

11Emerging areas to watch

Emerging areas to watch

Several areas are changing what a risk scientist needs to know. Each is a place to read on once the core is in place.

Open risk data. Standards such as the RDLS aim to make hazard, exposure, vulnerability and loss data easier to find and reuse 2. Ensemble forecasting. The ECMWF medium-range system is an ensemble of model forecasts giving information up to 15 days ahead 6. Analysis quality in government. Guidance such as the AQuA Book, updated in July 2025, sets out what robust analysis looks like 5.

What this guide does not claim: it is suggested learning, not a curriculum that anyone must follow, and it is not professional, legal or technical advice. It names methods, standards and frameworks for learning and endorses no product, model vendor or tool. It reports figures as their issuers published them and draws no new estimates from them; the only invented numbers are the labelled teaching values. Risk Academy does not certify competence, and a completion record is not a professional qualification. The Academy records no partners, cohorts or learner outcomes for this faculty, and this guide implies none.

질문

질문과 답변

Where should a complete beginner start in the Faculty of Risk Science and Methods?

Start with stage one of this guide, data and probability, then look at the course EO/GIS Starter Lab on NXS Learning, which is in preparation.

What is model validation?

Checking a model against evidence it was not built from, such as past outcomes, to find out where it can be trusted and where it cannot.

What is a proper scoring rule?

A way of scoring probability forecasts that rewards a forecaster for stating what they believe, so honest forecasts score best on average.

Why does the shape of a joint tail matter?

Because two risks can be linked with the same overall strength but fail together much more often in extremes, which is where large losses come from.

Do courses in this faculty lead to a qualification?

No. A completion record shows that you finished a course. It is not a certificate or a professional qualification.

이용약관

용어집

출처 이력
The record of where data came from and every change made to them.
검증
Testing a model against evidence it was not fitted to.
앙상블 예측
A set of forecasts run from slightly different starting conditions or models, giving a range of outcomes.
적정 점수 규칙
A score for probability forecasts under which stating one's true belief gives the best expected score.
Elicitation
A structured way of asking experts for judgements, designed to reduce bias.
Composite index
A single measure built by combining several indicators with stated weights and methods.

Check your understanding

  1. Why did the Basel Committee publish principles on risk data aggregation?

    Because in the crisis that began in 2007 many banks could not aggregate their risk exposures and spot concentrations fully, quickly and accurately.

  2. What makes a scoring rule proper?

    Under a proper rule, a forecaster gets the best expected score by stating their true belief, so the rule cannot be gamed.

  3. Two models agree on ordinary years but not on extremes. Which result should be reported?

    Both, with the reason they differ: the extremes are where data are thinnest and the model choice matters most.

Continue with a course

출처

참고문헌

12 sources, numbered in the order they are first cited
  1. Bank for International Settlements, Basel Committee on Banking Supervision (2013-01-09). Principles for effective risk data aggregation and risk reporting (BCBS 239). www.bis.org. Accessed 2026-11-03.
  2. Risk Data Library (2026). Risk Data Library Standard (RDLS). riskdatalibrary.org. Accessed 2026-11-03.
  3. Scientific Data (Wilkinson, Dumontier, Aalbersberg, Appleton and others) (2016-03-15). The FAIR Guiding Principles for scientific data management and stewardship. doi.org. Accessed 2026-11-03.
  4. Journal of the American Statistical Association (Gneiting and Raftery) (2007-03). Strictly Proper Scoring Rules, Prediction, and Estimation. doi.org. Accessed 2026-11-03.
  5. UK Government Analysis Function (2025-07-30). The AQuA Book. www.gov.uk. Accessed 2026-11-03.
  6. European Centre for Medium-Range Weather Forecasts (2026). Medium-range forecasts. www.ecmwf.int. Accessed 2026-11-03.
  7. Science (Tversky and Kahneman) (1974-09-27). Judgment under Uncertainty: Heuristics and Biases. doi.org. Accessed 2026-11-03.
  8. World Health Organization (2026). Risk communications. www.who.int. Accessed 2026-11-03.
  9. Committee of Sponsoring Organizations of the Treadway Commission (2026). Enterprise Risk Management: Integrating with Strategy and Performance. www.coso.org. Accessed 2026-11-03.
  10. HM Treasury and the Government Finance Function (2026-07-29). Orange Book. www.gov.uk. Accessed 2026-11-03.
  11. European Commission, Joint Research Centre (2026). Competence Centre on Composite Indicators and Scoreboards. knowledge4policy.ec.europa.eu. Accessed 2026-11-03.
  12. European Commission, Joint Research Centre, Disaster Risk Management Knowledge Centre (2026). INFORM Risk Index. drmkc.jrc.ec.europa.eu. Accessed 2026-11-03.

기록

Editorial record

Facts and course states are given as of 3 November 2026. This guide is reviewed at least every 270 days and whenever a cited report publishes a new edition or a course of the faculty opens.

Published by the Academy desk of Risk Academy. To report an error or request a correction, write through the Risk Academy contact page; corrections are dated and listed with the page. The Academy's statement on AI use explains how AI assistance is used and reviewed. Text, visuals and data are licensed CC BY 4.0, except where a cited source states other terms.

Cite this article

Academy desk, Risk Academy (2026, November 3). Faculty of Risk Science and Methods: a learning guide. Risk Academy. https://risksacademy.com/risk-science/learning-risk-science-and-methods/

Republish

Republish or translate this article free of charge, in print or online, under CC BY 4.0: keep the byline and Risk Academy, link to the original and note any change. Figures, maps and twins made by the institutions carry the same licence; third-party data keeps its own credit line.

Attribution to paste
<p>Risk science and methods: what to learn, in what order, and why — Academy desk, Risk Academy, Risk Academy. First published 3 November 2026 at <a href="https://risksacademy.com/risk-science/learning-risk-science-and-methods/">https://risksacademy.com/risk-science/learning-risk-science-and-methods/</a>. Republished under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).</p>

비중

EmailLinkedInBlueskyMastodon