Sun Hours Map — A Whole Day of Shadow in One Plate
The 3D model on the Sun & Shadow sheet shows the shadow at one instant. It is good for dragging the clock and watching shadows sweep, and useless for "how many hours of sun does this spot get in a day?" The sun-hours map answers that: shadows are solved every ten minutes from sunrise to sunset and added up.
The same computation appears in the diagram set as the SUN HOURS panel. This page explains what is calculated, what the numbers on the sheet mean, and why the plate is not a shadow study in the regulatory sense.
The numbers on the sheet
| Value | Meaning |
|---|---|
| Daylight | Hours from sunrise to sunset on the chosen date at the site |
| Mean sun hours | Average total daily sunlight on the ground inside the plate |
| Under 1 h | Share of ground that gets less than one hour of sun all day |
The Korean edition also reports two Korean sunlight criteria — two continuous hours between 9 and 15 on the winter solstice (Building Act Enforcement Decree, art. 86) and four hours in total between 8 and 16 (case law). The US edition leaves them out: there is no federal sunlight standard, and a percentage next to a legal-sounding window reads like a pass/fail test that no US jurisdiction sets. Where shadow analysis is regulated in the US it is local and framed differently. New York's CEQR Technical Manual, for instance, assesses shadows on sunlight-sensitive resources over a window from 1.5 hours after sunrise to 1.5 hours before sunset on four representative dates, and asks about the duration and extent of new shadow. The plate will not compute such a test — but the map of sunlit hours is the same picture that test starts from.
Solar noon comes from the site longitude and the time-zone value in the EPW file header. Estimating the zone from longitude alone is a trap — it puts Korea at UTC+8.5 and rotates the sun's azimuth by 7° — so the header value is read instead.
How it is computed
The building height map used by the wind and thermal-comfort sheets is reused. No CFD is involved; shadow is pure geometry, so there is no reason to wait for a three-minute GPU job. A 1,000 ft radius takes about 4 s, 1,640 ft about 14 s, ½ mi about 28 s.
For the chosen date (solstices, equinoxes or any day) the sun's position is found at ten-minute steps from sunrise to sunset; at each step every ground cell is tested for sun or shade and the results summed. Ten minutes differs from six by 0.01 h.
A design we threw away — the optimistic bias
The first version worked differently. Bake, once per cell, a horizon map by azimuth: how high the skyline is in each direction. Then any date and any hour is free — just test whether the sun is above the horizon in its azimuth. Four to nine seconds to bake, then zero. Attractive.
Integrated over a day, though, it returned 0.15–0.29 more hours of sun than the truth. Each instant was off by only 1–5%, but the errors were all in one direction and accumulated. Doubling the azimuth resolution did not help — the cause was the distance sampling along each ray, which stepped over thin buildings.
People read this map to decide where outdoor space and ground floors will get sun. A bias in the direction of "it's sunny" is not acceptable. We discarded it.
Instead the shadow test was folded into a recurrence: walking away from the sun, carry a running maximum of "highest thing seen so far minus the drop over distance". Nothing can be stepped over, and it is 30% faster. Cost was cut not by lowering precision but by restricting the computation to the plate (dropping the buffer ring outside it, a factor of 5.5).
The thermal-comfort sheet's shadow function had the same flaw and now shares this one.
Measured — a dense downtown, 1,000 ft radius
| Winter solstice | Summer solstice | |
|---|---|---|
| Mean sun hours | 1.8 h | 6.9 h |
Most side streets in a dense core get under an hour of winter sun — the dark slate-blue on the plate. Midtown Manhattan reads the same way: side streets nearly sunless in December, the avenues bright.
Reading the plate
Color breaks at 0·1·2·3·4·5·6·8·10·12 hours. There are deliberately no hard edges in the ramp — a hard edge invites reading pass/fail off the color, which this plate cannot deliver (below).
Buildings are filled gray. In the first version buildings and the outside of the plate were both white and a downtown plate looked empty — the same fix as graying the right-of-way on the plot size panel.
A "plate only" save button drops the title, color bar and caption. The default save keeps them; if you take the bare plate, add the caption yourself.
This is not a shadow study
- Buildings without a height are missing. Where neither the city or county data, our USGS LiDAR nor Overture has a height or a floor count, the building cannot be extruded and casts no shadow. Those spots read brighter than reality.
- It measures the ground. Regulatory tests are usually about windows, open spaces or specific resources. A window is vertical, receives the sun at a different angle, and differs by floor; a sunlit share of the ground says nothing about the apartments above it.
- No terrain, no trees. South-facing slopes are brighter in reality; trees are omitted here on purpose (for a winter test they are leafless anyway).
What the plate is for: reading, at concept stage, where winter sun reaches around the site and where it never does — where outdoor space will be usable in January, which frontage a ground-floor shop will be dark on, which part of your lot the neighbor's tower shades all day. A regulatory shadow study needs a survey and a validated tool, and should be done as such.