Validating a daylight model before you rely on it

Daylight model validation is the difference between a report that survives cross-examination and one that quietly falls apart when the objector's consultant asks where the neighbouring room depths came from. The software is not the problem. Every mainstream package computes vertical sky component and sunlight hours correctly from the geometry it is given. The errors live in the inputs: a building modelled from a photograph, a reflectance value taken from a default nobody checked, a window schedule that no longer matches the drawings.

Validate in that order. Geometry first, surfaces second, arithmetic last.

What to check in the geometry first

Ask where every surface came from. A defensible model states the provenance of the site, the proposed scheme and each neighbouring building.

The hierarchy of confidence runs roughly: measured topographic survey with building heights, then point cloud or photogrammetry, then mapping data with height attributes, then estimation from photographs. The last of these is sometimes unavoidable for a building you cannot access, and when it is used it has to be declared.

Four checks worth making on any model:

• Do the site levels in the model match the topographic survey, including any fall across the site
• Are neighbouring window positions surveyed, photographed and measured, or assumed from a standard storey height
• Are room depths behind neighbouring windows measured, or taken from the guide's default assumption where the interior is inaccessible
• Does the existing case model the building actually on the site now, rather than a cleared site

That last point produces more disputes than any other. A baseline modelled as an empty plot overstates the existing daylight and therefore overstates the loss, which cuts against the applicant. A baseline that includes a building larger than the one standing understates the loss. Either way the reviewer should see a photograph of the existing site alongside the baseline model.

Reflectance and surface assumptions

BRE 209 sets out default reflectance values for use where measured values are not available. External surfaces are commonly taken at 0.2 and the average internal reflectance of a room at around 0.5, with ground reflectance also low.

Defaults are defensible. Departures from them are only defensible when justified. A model using 0.4 for the reflectance of a proposed facade because the cladding is pale needs the specification to back it, and the reviewer should ask what happens to the results if the default is used instead.

Sensitivity testing is the honest response to uncertainty. Run the model with the default and with the justified value, report both, and explain the difference in a sentence. A report that runs one set of assumptions and presents the output as a fact is making a claim it has not tested.

Internal reflectance matters more for the interior metrics than for vertical sky component, which is measured at the outside face of the window and is unaffected by what is inside the room. Keep that distinction clear when reading a report, because it tells you which assumptions can and cannot have influenced a given number.

Glazing, frames and transmittance

For the metrics that deal with light inside a room, the glazing inputs drive the answer.

Visible light transmittance for the specified glass should come from the manufacturer's data rather than from a generic figure, and it should be the diffuse transmittance where the calculation calls for it. Low-emissivity coatings, solar control glass and triple glazing all reduce transmittance, sometimes substantially, and a model run on clear double glazing for a scheme specified with solar control glass will overstate the daylight.

Frame factor is the other input that gets glossed over. The calculation uses the glazed area rather than the structural opening, so a window with deep frames and transoms delivers less than its opening size suggests. Check that the frame factor used matches the window schedule and not a default.

Maintenance and dirt factors apply too, and they should be stated. Where the report on new interiors follows the approach the 2022 edition of the guide introduced, our note on what the 2022 edition of BRE 209 changed sets out the framework those inputs feed.

Cross-checking against a hand calculation

Never accept a model output you have not sanity checked. Two checks take minutes.

The 25 degree rule is the quickest. Draw a section from the centre of the lowest affected window. If the proposed obstruction sits below a line drawn at 25 degrees above horizontal, daylight to that window is unlikely to be materially affected, and a modelled result showing a large loss deserves investigation. If the obstruction rises well above that line, a modelled result showing no change deserves the same scrutiny.

The second check is a Waldram diagram or an equivalent manual vertical sky component calculation for one representative window with a simple, continuous obstruction. Where the hand figure and the model agree within a percentage point or two, the model geometry in that area is probably sound. Where they do not, something is wrong with the surfaces, the reference point or the orientation.

Pick the window for this check yourself rather than accepting one the author offers. Our explanation of what 27 per cent vertical sky component measures covers the target the check is testing against, and the 0.8 relative test is the comparison that usually decides the case.

Version control between design and assessment

Daylight models are built early and schemes change late. The single most common defect in a submitted report is that it assesses a scheme that is no longer the one on the drawings.

Insist on three things in the report: the drawing numbers and revisions the model was built from, the date the model was run, and a statement that the author has checked those revisions against the drawings submitted with the application.

Where a scheme is amended during determination, the model has to be rerun. A covering letter saying the amendment reduces the massing and therefore improves the results is an assertion, not an assessment, and officers increasingly say so.

Keep the model files with the report. A model that cannot be reopened in three years, when a section 73 application comes forward, has limited value.

Outputs a third party can audit

A Daylight and Sunlight Assessment should be readable by someone without the software.

The minimum is a plan of the surrounding properties with every assessed window numbered, a table giving the existing value, the proposed value and the ratio for each window, the room each window serves and whether that room is habitable, and a clear statement of which results fall outside the guide's criteria. Where a window serves a non-habitable room such as a bathroom or a circulation space, say so, because those are excluded from the assessment and their inclusion inflates a pass rate.

Appendices should include the assumptions schedule, the provenance statement for the geometry, and the sensitivity runs. Photographs of each assessed property from the street help the reader tie the model to reality.

Red flags in a report you are asked to review

• Results presented only for windows that pass, with no explanation of which were excluded and why
• No baseline photograph or survey of the existing site
• Room depths and uses behind neighbouring windows asserted without a source
• Reflectance or transmittance values that differ from the guide's defaults without justification
• A window schedule in the report that does not match the submitted elevations
• A conclusion expressed as overall compliance rather than window by window
• No named author, no qualification, and no date

Reviewers taking this work on through the professional directory should agree at the outset whether they are auditing the model or rebuilding it. Auditing means checking inputs, assumptions and internal consistency. Rebuilding means a new model and a new fee, and the two should never be confused in an engagement letter.

For specialist technical reports, qualified professional input may be required.

PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval.