How Daylight Check works

Daylight Check is a screening tool, not a BRE 209 assessment. It tells you early whether a house extension is likely to take too much daylight from a neighbour's windows, so you can adjust the design, or bring in a daylight consultant, before the application goes in.

This page says what it calculates, the rules it applies, where it stops being reliable, and how its numbers were checked against an independent lighting simulation.

What it is for

BRE 209, Site layout planning for daylight and sunlight (3rd edition, 2022), is the guide planning authorities in the UK usually turn to when an extension might reduce the light a neighbour's window receives. Daylight Check applies its daylight and sunlight tests to the plan you draw. It is not a substitute for a report from a qualified daylight consultant: if your extension is close to a limit, or a planning authority asks for an assessment, you will need one.

What it calculates

Vertical sky component

For each neighbouring window you mark, Daylight Check works out the vertical sky component (VSC): how much of the sky's light reaches the middle of the window's outer face, as a percentage of the light an open, unobstructed sky gives a flat horizontal surface. An open window with nothing in front of it scores 39.6%. A building in front hides some of the sky and lowers the score.

It uses the standard overcast sky BRE 209 specifies, which is brighter overhead than near the horizon, and adds up the sky the window can see in steps of 0.1° across its view. Reflected light is left out, as it is in the guide's VSC. Every house in your plan counts as an obstruction, before and after the extension; the extension is added for the second figure.

The verdict

BRE 209 passes a window if its VSC after the extension is at least 27%, or at least 0.8 times what it was before. That second figure is the ratio: VSC after divided by VSC before. A window that is already below 27% before is not failing for that reason; it fails only if it also loses more than a fifth of its sky. Daylight Check leads with the ratio for this reason.

The fail rule is BRE 209's. The borderline band is ours: it marks a window that passes by less than 10% of a limit, where a small error in what you drew or measured could turn it into a fail. BRE 209 itself only passes or fails a window. A borderline window says which limit it is close to. Numbers are rounded down when shown, so a figure never appears to reach a limit that the window misses. A window with no sky at all before (against a blank wall) cannot lose any light, so it passes.

Where on the window it is measured

Following BRE 209, the point is the centre of the window, on its outer face. For a floor-to-ceiling window such as a patio door (here, one whose sill is no more than 15 cm above the ground) it is a point 1.6 m above the ground on the window's centre line, or the top of the opening if that is lower.

The 25° and 45° tests

BRE 209 also has two quick tests that come before the VSC. The 25° test asks whether any part of a development in front of a window rises more than 25° above the horizontal, measured at the height of the centre of the lowest window on that wall, as BRE 209 asks, so every window on a wall shares its answer. The 45° approach is for an extension beside a window, on the same wall. Daylight Check draws both on your plan. They screen; they do not decide: a window's verdict comes from its VSC alone.

Sunlight to windows

BRE 209 also asks whether an extension takes a neighbour's sunlight (3.2). It applies to windows that face within 90° of due south, which includes due east and due west, and above all to main living rooms and conservatories. A window facing within 90° of due north gets little sun in the UK, and the guide does not ask for it (3.2.3), so Daylight Check gives no sunlight figures for one. Which of the rest matter is your judgement: Daylight Check checks every window you mark that faces within 90° of due south.

For each, it works out the annual probable sunlight hours (APSH): the hours of sunshine a year expected to reach the window, as a percentage of those that reach open ground at the same place. It also works out the winter figure, the part of the same total that arrives between 21 September and 21 March. Both are given before and after the extension, at the same point on the window as the VSC.

Sunlight depends on where the site is, so this needs the site's location. Daylight Check follows the sun through a whole year in 10-minute steps, and takes each month's sunshine from the nearest of 21 Met Office weather stations, as averages for 1991 to 2020. A step counts when the sun is in front of the window's wall and above every building in its way. A site more than 200 km from every station gets no sunlight figures.

The sunlight verdict

BRE 209 (3.2.6 and 3.2.7) treats a window's loss of sunlight as likely to be noticeable if it loses more than 4 points of APSH over the year and, after the extension, either gets less than 25% a year and less than 0.8 times its former annual figure, or less than 5% in winter and less than 0.8 times its former winter figure. So a loss of 4 points or less passes whatever the ratios, and so does a window that keeps 25% a year and 5% in winter.

The fail rule is BRE 209's; the borderline band is ours, and for sunlight it has two parts. Close to a limit is the VSC's: a pass within 10% of every way to pass. Within the estimate's error is new, and it takes in fails as well as passes. A VSC worked out here follows BRE 209's own method, so a narrow fail is a fail. Our sunlight figures use different sunshine data from BRE 209's indicators, so they are an estimate of what the indicators would give, and a narrow fail is as uncertain as a narrow pass. The figures before and after are moved together, because their errors move together: the same window, sunshine and sun, differing only by the extension. A borderline window says why, for example "Passes, but would fail with the figures 3 points (2 in winter) lower" or "Passes on annual 27.0%, close to 25%". Figures are rounded down when shown.

How the sunlight estimate was checked

BRE 209 reads probable sunlight hours off its sunlight availability indicators: charts for London, Manchester and Edinburgh/Glasgow with 100 spots, each worth 1% of the year's probable sunlight hours, laid over the sky a window sees. Daylight Check does not use the charts. It uses the Met Office's records for the station nearest the site, which say how much sunshine each month gets but not at what time of day. How that sunshine is spread over the sky is ours, and it is calibrated to BRE 209's 2022 indicators: fitted so that its figures come close to theirs. It is not a reproduction of them.

Each month's sunshine hours are shared among that month's 10-minute steps in proportion to 1 − e^{−(h − 2°)/15°}, where h is the sun's height above the horizon. Two numbers in it were fitted to the indicators: the height below which no sunshine is counted, 2°, and how quickly the share grows above it, set by 15°. A step with the sun 12.4° up counts half as much as one with the sun high in the sky.

We then compared our figures with a count of the indicators' spots in 1,458 cases: a long wall in front of a window, 0, 10, 20, 25, 30, 40° high, with the window facing every 2.5° from 10° north of due east round to 10° north of due west. That is 486 cases for each indicator, each worked out with the sunshine of a station near its city, shown in the table. The gap is the difference between our APSH and the indicator's, either way, in points of APSH.

Nine cases in ten are within 3.3 points over the year and 2.2 in winter, which the label under every window's sunlight figures rounds to: Estimate, calibrated to BRE 209's sunlight indicators: usually within 3 points of them (2 in winter). On average our figures are 0.2 points lower than the indicators' over the year, which is no lean either way, and 0.6 points higher in winter, a slight lean high.

The worst gap is 5.4 to 6.2 points over the year, depending on the indicator, and up to 3.9 in winter. Much of it is in the indicators themselves: their spots are scattered rather than set along the sun's paths, so turning a window 15° can gain or lose several at once, and no smooth spread of sunshine over the sky follows that. The indicators also give the high sun more weight than recent satellite measurements of UK sunshine do. We match the indicators all the same, because BRE 209's 25% and 5% are meant for figures read from them.

Contains Met Office data © Crown copyright 2026, licensed under the Open Government Licence v3.0.

Sun on gardens

BRE 209 also asks how much of a garden gets at least 2 hours of sun on 21 March (3.3.17). A garden should have at least 50% of its area sunlit for that long. If an extension leaves it with less, and the area that gets 2 hours falls below 0.8 times what it was, the loss of sunlight is likely to be noticeable. Daylight Check applies this to each garden you draw, which BRE 209 means to be the main back garden of a house, not a small front garden or a drive. It needs the site's location, because the sun's path depends on it.

It follows the guide's method (3.3.8). The garden is divided into a grid of points no more than 0.3 m apart, and for every minute of 21 March Daylight Check works out the shadow of each house, and of the extension after it is built. A point gets its 2 hours when it is out of every shadow for 2 hours in all. Sun 10° or less above the horizon does not count, as BRE 209 says, because low planting would block it anyway. The ground the extension stands on is not garden after it, so an extension built out into a garden leaves the L-shaped garden the guide measures.

The ratio is the sunlit area after divided by the sunlit area before. As for windows, the fail rule is BRE 209's and the borderline band is ours: within 10% of a limit, a small error in what you drew could change the verdict, and the grid adds its own: where a footprint's edge crosses the grid, the area can be out by up to half a cell along it. Shares are rounded down to a whole percent, so a garden that is 49.9% sunlit never shows as 50%. A garden with no ground getting 2 hours before cannot lose any, so it passes.

The grid sees only what you draw. Walls and fences more than 1.5 m high, which BRE 209 counts as casting shade, cannot be drawn yet, and trees, hedges and slopes are not modelled, so a garden with a high wall or a row of trees along its south side gets less sun than Daylight Check shows. A garden is one rectangle.

What it does not do

• Shapes. Every building is a box, optionally with a gable roof whose ridge runs along its length. There are no overhangs, bays, dormers, hipped roofs, balconies, trees or hedges. A real building that differs from its box is only as good as that approximation.
• The site. Ground is flat and level. A neighbour on higher ground, a hill that hides the low sun, or a window onto a balcony is not represented.
• Which windows. A rooflight cannot be drawn yet, and only the extension is tested by the 25° and 45° rules; buildings behind it are counted in the VSC only. For a room with several windows, BRE 209 takes the best one, or counts the sunlight through all of them without counting any twice (3.2.8); Daylight Check gives each window its own figures and leaves that choice to you.
• Sunlight is an estimate. Its sunshine comes from the nearest Met Office station, not from BRE 209's indicators, and its figures are calibrated to the indicators rather than equal to them: usually within 3 points, and in the worst case 6.2 points out. A site more than 200 km from every station gets no sunlight figures.
• Gardens. A garden is one rectangle, and only the houses and the extension shade it: not walls, fences, trees or hedges.
• What you draw. Results are only as accurate as your measurements. That is part of why there is a borderline band.

How it was checked

We compared the calculation with Radiance, the lighting simulation software from Lawrence Berkeley National Laboratory (version 6.0.2). Radiance models the same buildings as solid shapes and traces light rays through them, a different method from ours. For 18 test layouts we worked out the VSC both ways: a wall, terraces, single blocks, gabled roofs, a window facing along a terrace, a window in a gap, and a scene with several buildings. Both used the same sky formula, so the comparison tests the geometry and the adding-up, not the sky. The sky's own figures are checked separately against the exact answers: 39.6% for an open window, and about 27% for a very long wall whose top is 25° above the window.

Our acceptance bar was agreement within 2 points of VSC. The actual result is much closer: the largest difference in any case is 0.002 points (case extension-beside). Radiance's own error is about 0.002 points, so differences of that size say little. Daylight Check's automated tests repeat the comparison on every change and fail if any case is more than 0.05 points out.

All figures are VSC in percent. The last column is ours minus Radiance's, in percentage points, worked out before the other two are rounded.

Unless a case says otherwise, the window is 1.5 m up, faces north, and the buildings in front of it have flat roofs.

Draw your plan