Climate-Based Daylight Modelling

The defining move of CBDM is the sky. Instead of the daylight factor's single, sunless CIE overcast sky, CBDM drives the simulation with a time series of real skies - one for every daylight hour of a typical year - reconstructed from the location's EPW file, which carries hourly direct-normal and diffuse-horizontal irradiance measured or modelled for that place.

Each hour's sky is generated with a model such as the Perez all-weather sky, which turns those two irradiance numbers into a realistic luminance distribution: where the sun sits, how bright the circumsolar region is, how the rest of the dome grades. The two inputs deserve a moment: direct-normal irradiance is the beam straight from the sun's disc (what a tracker pointed at the sun would see), and diffuse-horizontal irradiance is the scattered light from the rest of the sky dome. A clear day is beam-dominated and produces a bright circumsolar hot spot - the source of most glare and direct-sun ASE hours - while an overcast monsoon hour is almost all diffuse, giving a soft, sourceless sky. The Perez model reads that ratio hour by hour and shapes the dome accordingly, which is precisely why CBDM can tell a harsh clear afternoon from a gentle grey morning.

Sum or sequence 8760 of these and you have captured what the daylight factor cannot - orientation (a south room and a north room now differ), season (December's low sun versus June's high sun), local climate (Bangalore's mildness versus Delhi's clear harsh sun), and the sun itself, present in most hours. This is why every metric in Lesson 5.1 that carries a percentage-of-hours (sDA, ASE, UDI) is by definition climate-based: the hours only exist because CBDM simulated them. A daylight-factor world simply has no clock to count against.

Running Radiance from scratch for every one of 8760 hours would be brutally slow. The breakthrough that made annual daylighting practical is the daylight-coefficient (DC) method. The insight: the room's optical response - how a given patch of sky ends up contributing light to a given sensor - is a fixed property of the geometry and materials. It does not change hour to hour; only the sky's brightness does.

So the method splits the sky dome into a manageable number of patches (a common Tregenza subdivision uses 145 patches, plus the sun). It then runs the expensive ray tracing once to find each sensor's coefficient for each patch - essentially 'if this patch alone glowed at unit brightness, how much light reaches this point?'. That produces a daylight-coefficient matrix. From then on, computing any hour is just fast arithmetic: multiply the fixed coefficients by that hour's patch brightnesses (from the Perez sky) and sum. The heavy physics is paid once; the 8760 hours are cheap multiplications. This 'simulate the room once, apply every sky' factoring is the conceptual heart of CBDM - and the reason a full annual illuminance run finishes in minutes, not days.

The basic DC method treats the whole light path in one step, which is fine for a fixed room but awkward when the interesting part - the glazing or shading - is exactly what you want to change or operate. The three-phase method solves this by splitting the light's journey into separable stages, each its own matrix: D (a daylight matrix, sky patches to the outside of the window), T (a transmission matrix describing the glazing or blind system itself, via its bidirectional transmittance), and V (a view matrix, from the inside of the window to the sensors). The annual result is these multiplied together against the sky series: roughly V x T x D x sky.

Why bother? Because the T matrix isolates the fenestration. You can swap in a different blind, a fritted glass or a complex daylight-redirecting system - or operate the blind hour by hour - by changing only the small T matrix, without re-running the expensive interior and exterior calculations. This makes it the standard for studying switchable glazing, venetian blinds and complex fenestration across a year. You will not build these matrices by hand; Honeybee and ClimateStudio assemble them for you. But knowing the path is split at the glazing explains why annual shading studies are feasible and how tools model operable blinds - the exact controls Lesson 5.4 leans on for glare.

The payoff of all this machinery is judgement you can trust. A single daylight-factor snapshot can be actively misleading: a north-lit studio and a south-lit one can post the same DF, yet the south room delivers far more usable annual daylight (and more glare risk) - a difference only an annual model reveals. Because CBDM counts real hours, it directly answers the questions clients and codes ask: for how much of the working year is this desk usefully lit? How many hours does that west wall get blasted? That is why LEED, IES LM-83, the Well standard and green-rating daylight credits worldwide have shifted decisively from DF to CBDM-based sDA/ASE.

Be honest about the limits, though. CBDM inherits the uncertainty of its weather file - a 'typical' year, not next year, and often from the nearest station, not your exact site (a real issue for Indian locations with sparse EPW coverage). Its realism depends on the sky model and on honest material and geometry inputs, just like any simulation. And it remains decision-support: superb for comparing options and demonstrating a target, but the absolute hours carry error bars, and statutory daylight compliance still defers to the relevant code and authority. Used well - comparatively, with sensible weather data - CBDM is the most defensible daylight prediction available, and the foundation of everything else in this module.

It is worth walking one full loop, because CBDM is only valuable if it changes a drawing. Suppose you are testing a 10 m-deep open studio in composite-climate Delhi with a south facade, and the question is window height and overhang depth. You download the Delhi EPW, build the room in Honeybee, assign honest reflectances (ceiling 0.8, walls 0.5, floor 0.2) and a clear glazing VT of 0.65, drop a 0.5 m sensor grid, and run the annual daylight recipe. The daylight-coefficient method traces the room once and evaluates all 8760 hours in a couple of minutes, returning an sDA of, say, 64% and an ASE of 22% - daylit, but glary and over-sunned, well past the 10% ASE guard.

Now you iterate, which is the entire point of a fast annual method. Add a 0.6 m overhang: ASE drops toward 12%, sDA barely moves because the overhang mostly clips high summer sun. Add interior venetian blinds operated on a schedule - here the three-phase method earns its keep, because only the small T matrix changes, so the re-run is nearly free - and ASE falls under 10% while sDA holds near 60%. Raise the window head by 300 mm and the back third of the room lifts, nudging sDA to 68%. In four cheap runs you have converged on a section that satisfies sDA >= 55% and ASE <= 10% - a defensible, evidence-backed decision no daylight-factor snapshot could have produced. That loop - model once, vary the option, read the annual metric, decide - is the working rhythm of climate-based daylighting, and it is exactly why the coefficient and three-phase machinery was worth understanding.