Dark Sky, NLC Academy Episode 22

Key Takeaways

What you'll walk away with

Four ideas to carry into the next specification conversation you have.

ULOR = Upward Light Output Ratio.

DarkSky does not require zero uplight. A residential luminaire may emit up to 50 lumens above 90 degrees; a commercial one above 1,000 lumens may emit 0.5% of its output, capped at 50 lumens. Zero is required only for the Wildlife Tuned category. And ULOR is the wrong word: CIE S 017:2020 defines it against lamp flux in the reference position, which is meaningless for an LED luminaire. The quantity you mean is ULR, as installed.

Warmer is better outdoors at night.

The certification ceiling is nominal 3000 K, identical across the Residential, Commercial and Pedestrian Comfort categories. 2200 K is not a certification requirement anywhere; it appears as an "ideal characteristic" in the Lodging programme, next to a 2700 K minimum. And CCT is a chromaticity coordinate, not a spectrum: at a fixed 3000 K, measured blue content across real lamps varies by roughly a factor of two.

Full cut-off, not 'fully shielded' marketing.

DarkSky does not use BUG at all. The word does not appear once in its version 3.1 criteria. TM-15 defines the Luminaire Classification System zones and the BUG scale; it sets no thresholds. And TM-15 never contained cut-off classifications: those lived in RP-8 and were deprecated by the IES in 2007, which is precisely why the LCS was written.

Smart curfews multiply the effect.

Curfews are an energy measure, not a sky glow measure. Tucson dimmed about 20,000 luminaires from 90% to 30% of full power after midnight and measured a 5.4% change in zenith sky brightness; street lighting was around 14% of that city's sky glow emissions. DarkSky's actual product requirement is dimmability to 10% or less of full output, which is a capability, not a schedule. No standard sets "50% after midnight".

START WITH THE NAME

IDA is now DarkSky International

DarkSky publishes certification criteria, not consensus standards. Cite the programme and its version: "DarkSky Approved Luminaires, v3.1".

CCT: ONE PRODUCT OR A FAMILY?

A selectable product and a model family are judged differently

So a 3000/4000/5000 K selectable driver cannot be certified. A 3000 K model from a family that also sells 4000 K can be, if it is ordered from that matrix.

METHOD AGAINST THRESHOLD

TM-15 classifies light. It does not limit it.

DarkSky v3.1 uses these zones, then sets its own limits in lumens.

Reference

Dark-Sky compliant spec × typical limit

Thresholds come from certification schemes, ordinances and procurement rules. IES TM-15 supplies the classification method and sets none of them. The Gulf column is what actually binds a project here.

Parameter	Dark-Sky limit	Reference
CCT, DarkSky Approved	≤ 3000 K nominal	DarkSky Approved Luminaires v3.1, sections 2.1.C, 2.2.D, 2.3.D
CCT, AlUla	≤ 2700 K	Royal Commission for AlUla outdoor lighting regulation, with CRI ≥ 80 and ≥ 120 lm/W
Uplight, residential	≤ 50 lm	DarkSky v3.1. Not zero
Uplight, commercial > 1,000 lm	0.5%, cap 50 lm	DarkSky v3.1. Zero only for Wildlife Tuned
Upward tilt adjustment	≤ 5°	DarkSky v3.1 exclusion 1.2.B
Field-selectable CCT	none above 3000 K	DarkSky v3.1 exclusion 1.2.A. One higher option disqualifies the family
Dimming capability	to ≤ 10%	DarkSky v3.1 exclusion 1.2.D. A capability, not a curfew
Upward light, EU procurement	0.0% RULO	EU GPP SWD(2018) 494 TS7, maintained at the installed tilt
Spectral G-index	≥ 1.5	EU GPP TS9, not a DarkSky metric. Higher G means less blue
Light trespass	see LP-11-20 §3.2	ANSI/IES LP-11-20. RP-33 is deprecated

UPLIGHT, IN LUMENS

The uplight limit depends on category and output

"Fully shielded" is not a number. Ask for uplight in lumens, and the category it was judged against.

THE MECHANISM

Why short wavelengths cost more sky

Two caveats: dust and aerosols scatter more evenly across wavelengths, and clouds shift the colour of skyglow. The blue penalty is a clear-sky result.

A CATEGORY ERROR

Melanopic is not a skyglow metric

CIE S 026:2018 defines melanopic quantities for the human ipRGC response. A melanopic metric does not directly measure skyglow.

ONE STUDY, NINE SPECTRA

In Aubé's spectra, unfiltered white LED scored worse per lumen than HPS

Direct light, equal lumens: the 2760 K LED scores 0.359, HPS 0.231, 1.55 times higher. After clear-sky scattering: 1.76 times at the source, 1.60 times at 30 km.

Deep Dive

Why dark-sky matters in Saudi

Five things the Gulf market repeats that the primary documents do not say.

AlUla is certified. NEOM is not.

AlUla holds International Dark Sky Park status across four protected areas totalling 6,146 km², certified in two phases: AlUla Manara and AlGharameel in 2024, Sharaan and Wadi Nakhlah announced 23 December 2025. It is not even the largest in the country: AlNufud, in the Imam Turki bin Abdullah Royal Reserve, covers 13,416 km². On 3 August 2026 Red Sea Global's Desert Rock and Nujuma became the first two properties on earth certified as DarkSky Approved Resorts, a category created because they did not fit the existing ones. NEOM, by contrast, has no certification and no published outdoor lighting requirement in any of its own documents.

Warm helps, but melanopic is the wrong yardstick.

Short wavelengths scatter more: Rayleigh scattering by air molecules goes as the inverse fourth power of wavelength. But two corrections matter here. Melanopic response is a human circadian action spectrum defined by CIE S 026:2018 and has nothing to do with atmospheric scattering, so it is the wrong quantity for sky glow entirely. And in dusty Gulf air the advantage shrinks: aerosols scatter roughly as the inverse first power and coarse desert dust is close to wavelength-neutral, pulling the effective exponent well below four. Clouds reverse it outright, amplifying red sky glow 17.6 times against blue at 7.1.

BUG did not win in the Gulf.

Half right, and the wrong half matters in the Gulf. TM-15 never contained cut-off classes: they lived in RP-8 and the IES deprecated them in 2007, and the LCS was written to replace them. But BUG did not win here. AlUla's 2025 regulation is written entirely in cut-off language and puts numbers on it, semi cut-off at 5% above 90 degrees, cut-off at 2.5%, full cut-off at 0%, and the word BUG appears nowhere across its 66 pages. BUG also cannot be produced for a floodlight at all, because LCS is defined only for Type C photometry and floodlights are tested as Type B.

Three standards, three layers, one confusion.

Three standards bodies own three layers and the market collapses them. ANSI C136.41 owns the seven-pin twist-lock socket. ANSI C137.1-2022 owns the 0-10 V signal on pins 4 and 5, and sets no light level and no schedule. Nobody owns the dimming schedule. There is also no "NEMA standard" for the socket: it is an ANSI standard from committee C136 with NEMA as secretariat. SASO 2927:2019 clause 10.3 already mandates the socket in Saudi Arabia, but "capped (short-circuited)", which hard-wires the lamp permanently on. The socket is a promise, not a function.

Per lumen, LED made sky glow worse.

High pressure sodium measures 1950 to 2200 K depending on family and around 85 lm/W at the lamp; "5 lm/W in dark-sky terms" is not a quantity that exists. And the comparison runs the other way from the marketing. On Aube's Star Light Index, per lumen delivered, a 2700 K LED scores 0.359 against high pressure sodium at 0.231, so it is 1.55 times worse; a 4000 K LED is 2.4 times worse. Chelan County replaced 3,693 HPS units with fully shielded LED, 80% of it at 3000 K, halved lumens per light, and the National Park Service still measured the zenith sky 60% brighter. The LED advantage was never that its light scatters less. It is that you can aim it and dim it, and sodium could do neither.

THREE FIELD STUDIES

What three conversions did to the measured sky

Chelan County: 3,693 HPS to fully shielded LED, 80% at 3000 K. A year later the zenith at one site was 60% brighter.

WHERE THE REAL LEVER IS

Aim beats spectrum, and tilt undoes aim

Inside a city, most skyglow comes from light reflected off the ground. Cutting total lumens is the lever.

THE SAUDI POSITION

Saudi Arabia's street-lighting standard once required 3000 to 6000 K

SASO 2927:2019 clause 10.2 required LED colour temperature between 3000 and 6000 K. Amendment SASO 2927:2019/AMD2:2023 deleted that sentence (Umm Al-Qura, 24 February 2023).

WHERE THE DARK SKY IS

Saudi Arabia's certified dark sky sites

AlUla: four International Dark Sky Parks, 6,146 km², certified in 2024 and December 2025.

Write this instead

The clause to write

Four lines, each naming a document, a quantity and a unit. Not one of them can be answered with an adjective, which is the only test a tender clause has to pass.

a. Luminaires shall hold a current DarkSky Approved listing, or be supplied with an ISO 17025 goniophotometric report stating the flux emitted at or above 90 degrees, in lumens, at the installed tilt and aiming.

b. That flux shall not exceed 0.5% of luminaire output, capped at 50 lm.

c. Every selectable CCT option shall be at or below nominal 3000 K; one higher option renders the family ineligible.

d. Where a C136.41 receptacle is fitted, it shall be wired to the driver's dimming pins, not capped.

Line d is the one nobody writes. SASO 2927:2019 clause 10.3 already mandates the socket in Saudi Arabia and permits it capped, which hard-wires the lamp permanently on.

From the Academy to the Field

Ask for the listing, not the adjective.

"IDA-compliant" is not a status that exists: the body is DarkSky International and the certification is DarkSky Approved, a paid per-product review needing ISO 17025 goniophotometry. Ask any supplier, including us, for what can actually be checked: the DarkSky Approved listing if it is claimed, the measured uplight in lumens above 90 degrees, the CCT of every selectable option, and whether the driver is wired to the C136.41 socket's dimming pins or just fitted with a shorting cap.

See the solution

Sources

Sources & further reading

The certification criteria, standards and regulations behind every number on this page. Where a document defines a method but sets no limit, that is stated.

DarkSky Approved Luminaires guidelines, version 3.1 (October 2025), the current criteria: CCT ceiling 3000 K nominal, the uplight allowances, the 5-degree tilt exclusion and the field-selectable CCT exclusion. darksky.org
DarkSky International, name change announcement, the International Dark-Sky Association became DarkSky in 2023; the organisation calls IDA "the now-defunct acronym". darksky.org
ANSI/IES TM-15-20, the Luminaire Classification System zones and the BUG scale. A classification method that sets no thresholds. store.ies.org
CIE S 017:2020, ILV, terms 17-29-065 (ULOR, against lamp flux in the reference position) and 17-29-066 (ULR, as installed). cie.co.at
EU Green Public Procurement criteria for road lighting, SWD(2018) 494, criterion TS7 requires 0.0% RULO maintained at the installed tilt; TS9 sets the spectral G-index at 1.5, rising to 2.0 near optical observatories. ec.europa.eu
CIE S 026:2018, the melanopic action spectrum, defined for human ipRGC-influenced responses. Not an atmospheric quantity. cie.co.at
SASO 2927:2019, Lighting Products Part III, clause 10.2 set a colour temperature range of 3000 to 6000 K, a floor rather than a ceiling; clause 10.3 mandates the seven-pin NEMA socket, capped. saso.gov.sa
Royal Commission for AlUla, outdoor lighting regulation, maximum 2700 K, CRI at least 80, LED efficacy at least 120 lm/W, written in cut-off language, and excluding municipal street lighting. rcu.gov.sa
DarkSky International Dark Sky Places directory, AlUla's four parks totalling 6,146 km2, AlNufud at 13,416 km2, and the first DarkSky Approved Resorts at Desert Rock and Nujuma. darksky.org
ANSI C136.41 and ANSI C137.1-2022, the seven-pin twist-lock receptacle, and separately the 0-10 V control signal that rides pins 4 and 5. Neither sets a dimming schedule. nema.org

Dark Sky, designing outdoor light that lights the ground, not the stars