The Modern Flood Estimation Report

FER as a decision record, not a flow table

A Flood Estimation Report should let a competent analyst reconstruct the decision, not merely copy its final numbers. It needs the purpose of the study, the catchment evidence considered, the candidate methods, the checks performed, the selected result, the reasons for selection and the people who approved it. A flow table without that chain cannot show whether the result is appropriate for the decision it supports.

The calculation record still matters. Inputs, intermediate results and final flows should be structured and reproducible. The FER adds the project-specific interpretation that computation alone cannot supply: why this evidence was accepted, why alternatives were rejected and where judgement or uncertainty limits later use.

Study purpose, catchment and flooding mechanisms

The report should begin with the decision being supported, the relevant return periods, climate or policy allowances, the study extent and the required outputs. It should identify the subject catchment and any reaches, tributaries, structures or downstream constraints that change how the estimate will be used.

A short catchment narrative should then explain the credible flooding mechanisms: rainfall response, soil and geology controls, groundwater, reservoirs, lakes, floodplain storage, snow where relevant, urbanisation and known historical events. The current generator records neither the project purpose nor this catchment-specific mechanism account, so this capability is Absent rather than presumed from descriptors.

Hydrometric data, rating quality and conceptual model

Gauged evidence needs an explicit suitability review. The record should identify the flow and level series used, periods of record, gaps, rejected years, datum or station changes, rating revisions and the degree of extrapolation at important peaks. Nearby stations should be considered for consistency without assuming that proximity makes their catchments hydrologically interchangeable.

Those observations inform a hydrological conceptual model: which stores, pathways and controls plausibly generate floods at the site, and how the available gauges represent them. The current generator has no generated data/rating review or conceptual model. A list of station values would not by itself satisfy either requirement.

Boundary and descriptor review

The catchment outlet and boundary should be checked against mapping, drainage paths, structures and the purpose of the estimate. Descriptor values need their dataset edition, extraction location and any manual changes recorded. Material features such as urban extent, attenuation, artificial transfers or nested catchments should be reconciled with the conceptual model rather than accepted as unexamined fields.

Hydrometric can present descriptors and boundary-related evidence, but the review trail is incomplete. The source, checks, edits, reasons and implications are therefore classified Partial. Displaying a descriptor is not evidence that its catchment boundary or project meaning has been reviewed.

Candidate methods and explicit selection

The analyst should record which approaches were eligible for the study, including relevant statistical, rainfall-runoff or local-data routes, and why any method was excluded. Each candidate needs comparable outputs and a stated evidence boundary so differences can be investigated rather than hidden by a single preferred result.

Crucially, the first computed distribution is a candidate, not an adopted answer. Adoption requires recorded analyst selection: the chosen method or combination, the evidence relied upon, the rejected alternatives and the consequences for the design flows. The current generator does not provide that selection record.

Donor, pool and distribution decisions

A statistical-method record should preserve the initial donor search, donor suitability checks, exclusions, adjustments and sensitivity to plausible alternatives. For pooling, it should retain the initial group, record lengths, similarity evidence, additions or removals, the reasons for changes and the final pooled L-moment evidence.

Candidate probability distributions should be compared using fit, stability, physical plausibility and sensitivity over the return periods that matter to the project. Hydrometric produces the statistical calculations, but it does not generate the donor/pool sensitivity narrative or the analyst rationale for an adopted distribution. Both decision records are currently Absent.

Checks, sensitivities, uncertainty and limitations

The adopted estimate should be checked against independent local evidence, nearby gauges, historical floods and credible alternative methods where those sources are available. Sensitivities should target decisions that could move the result: boundary or descriptor changes, donor choices, pooling changes, distribution family, record treatment and other project-specific assumptions.

Uncertainty should remain attached to its evidence layer. Data and rating quality, finite-record sampling, method transfer, model structure and analyst choices do not become one defensible interval merely by listing or combining them. Hydrometric's existing uncertainty presentation is Partial, while generated checks, sensitivity narrative and study-specific limitations are Absent.

Richer assessment summaries, checks, audit records, sign-off, provenance and uncertainty presentation exist only as branch-only prototypes. They are unshipped and do not alter this current-state classification.

Dataset/engine/schema versions and digital file register

Reproduction requires exact dataset editions, engine and calculation versions, schema versions, relevant configuration, run time and stable identifiers for inputs and outputs. Where a dataset or tool is licensed, the record should state the access boundary without reproducing protected content.

A digital file register should enumerate the evidence that belongs to the study: exported descriptors, hydrometric extracts, rating review, calculation files, figures, checks, decision notes and issued outputs. Current dataset/engine/schema provenance is Partial because some identifiers exist but the exact chain is incomplete. The digital evidence manifest is Absent.

Analyst/reviewer approval and revision history

Issue status should be explicit. The FER should name the analyst and reviewer, their roles, the review and approval dates, unresolved qualifications and whether the document is draft, approved or superseded. A revision entry should say what changed, why it changed and which prior issue it replaces.

The current generator does not create analyst/reviewer approval, sign-off or a project revision history. These capabilities remain Absent. A version number on generated calculation output is provenance for that artefact; it is not a substitute for professional approval of the study decision.

Current Hydrometric gap analysis

Hydrometric currently produces a structured statistical calculation report. That component is classified Complete. Descriptor/boundary evidence, existing uncertainty presentation and dataset/engine/schema provenance are Partial. The study purpose and flood mechanisms, hydrometric/rating review, donor/pool sensitivity narrative, adopted-distribution rationale, study-specific limitations, digital evidence manifest and approval/revision/sign-off are Absent.

The taxonomy also permits Manual where an analyst supplies and retains a controlled record outside the generator. None of the Absent rows is relabelled Manual here because no such artefact is part of the generated publication. This prevents undocumented professional work from being inferred from calculation output.

The current release does not yet produce a complete Environment Agency-style Flood Estimation Report. Branch-only prototypes may inform future work, but they are not shipped features, not part of the present audit and not evidence that the missing professional review has occurred.