Deep Listening, the written review

/ Summary

The proposal’s biggest unexamined decision was pointing Deep Listening, a mirror-lined cone, southwest, into both the winter sunset and the prevailing sea breeze. That single choice creates the two most serious issues, glare and wind. The rest are standard engineering and maintenance questions the proposal left open. Each is shown in the blueprint simulations.

One-line reason:
About 550 h/yr of 1–2.2× sun at head height inside and just past the throat, plus a 3.8× line on the pad in Nov and Feb
Confidence:
High: ray-traced from the drawings, full year
Severity:
High
One-line reason:
Open ground on the island already tops 11 mph in 24–36% of daytime hours; the throat raises that to 34–64%, against a 10% comfort yardstick
Confidence:
Medium: measured island wind record plus 2D flow solutions, not a wind tunnel
Severity:
High
One-line reason:
About 4,200 lbf sideways and 2,200 lbf of uplift at design wind, on a slab that was never a foundation
Confidence:
Medium: V from the ASCE Hazard Tool, Exposure D from ASCE 7; force coefficient from a 3D simulation checked against a known shape
Severity:
Medium
One-line reason:
78 in throat top vs 80 in required; rim overhangs more than 4 in above 74 in
Confidence:
High: geometry and ADA text
Severity:
Medium
One-line reason:
Pad on 30–50 ft of loose sand that liquefied in 1989, inside the strip held for future shoreline adaptation
Confidence:
High for the facts; site-specific depth unconfirmed
Severity:
Medium
One-line reason:
About 275 ft from the Bay; the inside is never rinsed by rain; colors fade unevenly
Confidence:
High
Severity:
Medium
One-line reason:
40° inner slope into a 43° ridge up to a 14 ft drop; a steel bowl on a smooth pad
Confidence:
High
Severity:
Medium
One-line reason:
Uplights that conflict with the bird-safe code near parks and water, gull perch, standing water at the base, fingerprints on mirror
Confidence:
Medium
Severity:
Medium

/ Site and climate baseline

The sculpture‘s mouth faces 229° (southwest), straight into the afternoon sun and 18–41° off the prevailing west and west-southwest winds. Every issue below follows from that orientation, the exposed shoreline and the filled ground.

Value:
37.8196° N, 122.3732° W, Cityside Park, Treasure Island
Basis:
Proposal plan, placed on the real seawall
Value:
About 275 ft (84 m) from the west shoreline
Basis:
USGS elevation model¹⁹; OpenStreetMap
Value:
229° (southwest): across the path, onto the lawn, toward downtown SF
Basis:
Proposal drawings
Value:
About 2 mi (3.2 km) of open Bay to the west
Basis:
USGS elevation model¹⁹
Value:
About 12.9 ft (3.9 m) NAVD88 after the park was raised; about 11.3–11.8 ft before
Basis:
USGS 3DEP lidar¹⁹, March 2023 (after) and 2004–2010 (before)
Value:
Sand hydraulically placed from the Bay in 1936–37, over natural shoal sand and Bay mud
Basis:
USGS Professional Paper 1551-B⁹

Sun path at the pad. These are computed with the NOAA solar position algorithm for the exact coordinates.

Sunrise (azimuth):
7:13 AM (89°)
Sun at noon:
52° high
Sunset (azimuth):
7:22 PM (271°)
Sunrise (azimuth):
5:48 AM (59°)
Sun at noon:
76° high
Sunset (azimuth):
8:35 PM (301°)
Sunrise (azimuth):
6:57 AM (89°)
Sun at noon:
52° high
Sunset (azimuth):
7:06 PM (271°)
Sunrise (azimuth):
7:21 AM (120°)
Sun at noon:
29° high
Sunset (azimuth):
4:54 PM (241°)

The sun sits within 45° of the mouth’s axis for about 970 hours a year, all from August to April; it never does from May to July. On the winter solstice it sets only 11° off the axis, which is why the glare peaks in December.

Wind. Daytime wind on the island comes mostly from the west (32% of daytime hours) and west-southwest (23%), strongest on May to July afternoons. The only record on the island is the former Naval Station’s weather station (NCEI ISD 724943⁶, 6,760 readings, 1983–96, with gaps in 1988–91 and 1994). NOAA’s Yerba Buena Island gauge (9414782) has no wind sensor, and NOAA’s harbour stations read 25–43% below the island, most likely because most of them sit at harbour piers sheltered by the city and port. The Wind section has the numbers.

Design wind and exposure. The ASCE Hazard Tool⁴ gives a basic wind speed of 92 mph at the pad’s coordinates (3-second gust at 33 ft, Risk Category II, ASCE 7-22; also 92 mph under ASCE 7-16). The site is not in a hurricane-prone or special wind region. More than 5,000 ft of open water lies upwind and the pad is within 600 ft of it, so Exposure D applies (ASCE 7-22 Sec. 26.7; the same rule appears in CBC 2022 Sec. 1609.4⁵, which adopts ASCE 7-16).

/ Sun

The polished panels are nowhere near the car-melting kind of hazard, but they turn the sculpture into a glare source. At head height it reaches 1–2.2× direct sun for about 550 hours a year. When the sun lines up with the axis, it also throws a line of about 3.8× sun onto the pad.

How it happens. The inside of the sculpture is a concave cone of mirrors. When the sun is in the mouth, each polished strip bounces light across to the opposite side and toward the axis, and some of it bounces again. The polished panels redirect roughly 10–30% of the sunlight that hits the sculpture, typically about 20%. Where several of those reflections overlap, the light adds up to more than direct sun.

Hours of reflected glare at head height

1× sun or more1.5× sun or more

Hours per month with at least one spot at 1.7 m above the given level. Sampled every 4th day at 10-minute steps.

Brightest reflected spot each month

Head height (1.7 m)Ground / pad

1.0 = the intensity of direct sunlight

When and where.

Head-height glare at 1× or more happens only between about 3:10 PM and 8:30 PM, with the sun under 20° high.
It peaks from mid-December to early January around 4:40 PM, when the sun sets 1–2° up at azimuth 238–240°, almost straight into the mouth. The peak there is 2.1–2.2× over about 8 sq ft.
It is local. Reflected light above 0.3× sun at head height stays within about 11 ft of the pad’s centre: inside the sculpture and just past the throat, exactly where the proposal invites people to stand and walk through.
The 3.8× line on the pad appears in early November and early February around 3:30 PM, when the sun (azimuth 226–229°, 21° up) lines up with the axis. A cone of mirrors focuses along its axis. The grid is 5 in, so the true line is probably narrower and brighter.
“× sun” is relative to the direct sun at that moment. A sun 2° above the horizon is weak, so the December peak is a glare problem, not a heat problem. The 3.8× line comes with a 21° sun and would noticeably warm the pad and anyone standing on it.

Beyond the sculpture. At a winter sunset some of the redirected light leaves at eye level toward Avenue of the Palms. It travels northeast, so it reaches only people facing southwest, who are already looking into the low sun. Around midday, some heads along the path at eye level, so walkers and cyclists see bright glints from the side.

Precedents. This is the same mechanism, at far lower intensity, that made 20 Fenchurch Street in London melt parts of a parked car in 2013 (Physics World²⁵; a Siemens summary²⁴ models the effect). When the same architect’s Vdara hotel in Las Vegas opened in 2009, guests said its pool deck got hot enough to singe hair and melt plastic (Physics World²⁵). Those were buildings; this is a walk-through at face height.

Why it was avoidable. The mouth faces 229° and the winter sun sets at 241°. At this latitude the sun never rises or sets north of 59° or 301°. A mouth facing due north (0°) never has the sun within 45° of its axis. Re-running the same full-year model with the mouth at 0° cuts head-height glare of 1× sun or more from about 550 to about 280 hours a year, lowers the peak from 2.2× to 1.7× and removes the 3.8× line on the pad (its peak drops to 0.4×). What remains comes at sunrise and sunset from April to September, when the low sun swings far enough north to catch the mirrors. A brushed finish on the lower panels would soften that too.

/ Wind

The sculpture is a 175 sq ft scoop turned toward the prevailing wind. It takes about 4,200 lbf at design wind speed. On ordinary days the open lawn around it already breaks San Francisco’s walking-comfort yardstick, and the throat speeds the breeze up by another 20–50%.

Orientation. The mouth faces 229°. West winds (270°, 32% of daytime hours on the island) arrive 41° off the mouth’s axis, and west-southwest winds (248°, 23%) only 18° off. Two-thirds of daytime hours blow within 45° of the axis, into the concave side, the worst case for drag.

Design wind load. Figures follow ASCE 7-22 Chapter 29 for other structures. The standard has no force coefficient for an open funnel, so I computed one with a 3D simulation (below).

Value:
92 mph, 3-second gust, Risk Category II (ASCE Hazard Tool⁴, ASCE 7-22)
Value:
D / 1.03
Value:
1.0, 1.0 (flat site near sea level)
Value:
22.3 psf
Value:
0.85, 0.85
Value:
1.5 (1.4–1.6), from the 3D simulation
Value:
175 sq ft (16.3 m²)
Value:
8.1 ft (simulation; 7.0 ft assumed before)
Value:
4,000–4,500 lbf (18–20 kN), central 4,200 lbf
Value:
about 2,200 lbf
Value:
32,000–37,000 ft-lbf, central 34,000

3D simulation. I modelled the sculpture at full scale in OpenFOAM²⁷, a standard open-source flow solver: a steady turbulent solution (realizable k-ε) on about 480,000 cells, with the Exposure D wind profile at the inlet and the panels as a thin steel shell, for six wind directions. The same setup gives 1.31 for a hollow hemisphere facing the wind, against a handbook value of 1.42 (Engineers Edge²⁶), so the sculpture‘s values below are scaled up by that 9% (1.42 / 1.31). A coarser mesh gave 4% less, so mesh resolution adds little doubt. Cf is on the fixed 175 sq ft area and the pressure at 7 ft, so it drops straight into the ASCE formula.

Cf:
1.48
Force acts at:
8.0 ft
Note:
Uplift 0.47× the horizontal force
Cf:
1.48
Force acts at:
8.0 ft
Note:
Uplift 0.49×
Cf:
1.49
Force acts at:
8.1 ft
Note:
Worst case; the force swings 38° toward the sculpture‘s axis. Uplift 0.52×
Cf:
0.47
Force acts at:
4.3 ft
Note:
Cf:
0.80
Force acts at:
4.8 ft
Note:
Cf:
1.01
Force acts at:
7.2 ft
Note:
Presses down instead of lifting

The worst case is the ordinary west wind, not one straight into the mouth, and the force acts about 8 ft up, above the 7 ft I had assumed. The top panel also works like a wing, adding uplift of about half the horizontal force. A steady simulation still misses gust peaks and panel flutter, so a stamped design would want a wind-tunnel test.

That is manageable with a proper footing. The 10 × 20 ft pad (about 15,000 lb if it is a typical 6 in slab) outweighs the uplift, but it was never engineered as a foundation. The real risks are connection fatigue and panel flutter: every gust cycle flexes 32 thin panels and their fasteners, thousands of times a day on a windy site.

Speed-up through the throat. The mouth is 5.7 times larger in area than the throat, so it funnels air. Much of that air spills around the rim instead of passing through. Purpose-built wind concentrators report throat speed-ups of about 1 to 2.6× (some studies up to 3×), mostly from simulations rather than measurements (INVELOX review⁸). An open funnel without a closed duct does less. My own 2D flow solutions of this exact shape settle at 1.3–1.4× in the throat, and up to about 2× where air whips around the rim edges. I use 1.2–1.5× for the throat.

How often it matters. Open ground at the pad already passes 11 mph at head height in about a quarter to a third of daytime hours. In the throat that rises to roughly half. Wind builds through the day and peaks on May to July afternoons.

Daytime hours over 11 mph at head height

7 AM–6 PM. Dot = middle, whisker = range across two methods and two ground roughnesses.

Mean wind at head height, mph

Open lawn. Oakland airport 2021–25 mapped onto the island record; head height = 0.78 × 10 m. Numbers mark 11 mph and over.

Free wind at head height:
5 mph; 9 mph
In the throat (1.2–1.5×):
6–8 mph; 10–14 mph
Against the SF yardstick:
Mostly comfortable
Free wind at head height:
13–14 mph
In the throat (1.2–1.5×):
15–21 mph
Against the SF yardstick:
Above the 11 mph comfort level
Free wind at head height:
17–19 mph
In the throat (1.2–1.5×):
21–28 mph
Against the SF yardstick:
Well above comfort; can reach the 26 mph hazard level
Free wind at head height:
34–42 mph
In the throat (1.2–1.5×):
41–64 mph
Against the SF yardstick:
Well past the 26 mph hazard level

Head-height values scale the island’s readings (anemometer height assumed 10 m) down by 18–25% (log wind profile, open water to open shoreline). Two methods agree on the shares: the island’s own daytime readings, and Oakland airport’s hourly record from 2021 to August 2025 mapped onto the island’s wind distribution. Oakland is about 6% windier now than in 1983–96, so the older island record does not overstate today’s wind.

The yardstick is the comfort criterion in SF Planning Code Sec. 148¹ until Ordinance 245-25¹ (passed December 2025): 11 mph equivalent wind speed in walking areas, exceeded no more than 10% of the time from 7 AM to 6 PM. That ordinance dropped the comfort criterion and turned the hazard rule (26 mph for a single hour a year) into a Nine-Hour Hazard Criterion: 26 mph equivalent wind speed for nine or more hours a year. Sec. 148 covers only new buildings over 85 ft in downtown districts, not art on Treasure Island, so it is a yardstick here, not a rule. Equivalent wind speed adds a gust allowance to the hourly mean, so the mean speeds above understate it slightly.

Knock-on effects.

The throat is exactly where the proposal invited people to stand and walk through. On a normal July afternoon it would be one of the windiest spots in the park.
Air slows as it leaves the throat and at the dead corners where the sculpture meets the pad. Litter and sand from the adjacent decomposed-granite and dune strips will pile up there.
Sharp rims and open panel seams in a fast flow can whistle or hum. Whether this shape would do so can’t be predicted without a wind-tunnel or field test; it is a known risk for open metal forms.

/ Ground

The pad sits on roughly 30–50 ft of loose, saturated sand (hydraulic fill over natural shoal sand) that liquefied in the 1989 earthquake, inside a shoreline strip the city is holding for sea level rise. A 10 × 20 ft slab laid on the surface was never going to be an acceptable foundation.

Evidence:
Treasure Island was filled from February 1936 to August 1937 with mostly sandy Bay sediment placed hydraulically. In the 1989 Loma Prieta earthquake, 0.06 g on Yerba Buena Island’s rock was amplified to 0.16 g on Treasure Island (east–west). Sand boils formed at 18 places, settlement was generally 2–6 in (locally up to 2 ft) and lateral spreading was generally under 1 ft. On the west side, cracks stayed within about 150 ft of the shore; the pad is about 275 ft in (USGS Professional Paper 1551-B⁹).
What it means for the sculpture:
In a strong quake the pad can tilt or sink unevenly. A 14 ft piece with an 8 ft wind lever and a heavy mouth rim is the kind of object that ends up leaning.
Evidence:
Hydraulic fill and natural shoal sand together are about 30–50 ft thick, over Young Bay Mud 10–120 ft thick (the EIR¹⁰ gives 20–120 ft), with bedrock roughly 100–400 ft down (USGS PP 1551-B⁹; Treasure Island EIR, 2010¹⁰). At the island’s downhole array, 600 m north of the pad, about 12 m (40 ft) of fill and sand sits on about 15 m of Bay Mud, with bedrock at 91 m (Graizer and Shakal, SMIP2000¹¹). No boring log specific to the park was found.
What it means for the sculpture:
Long-term settlement continues; a rigid pad on it can crack or rotate over the years.
Evidence:
The 2010 plan was to densify about 100 acres of development areas by deep dynamic compaction, vibro-compaction and surcharging, about $137 million, leaving about 300 acres of parks untreated (SF Public Press¹³). ENGEO also describes cement deep soil mixing along the shoreline (UCSD seminar¹²).
What it means for the sculpture:
The pad is in a park, the kind of ground the 2010 plan left untreated. A foundation has to be coordinated with whatever the park strip actually received. We never asked.
Evidence:
Building floors go at 12.7 ft NAVD88 or higher: the 9.2 ft 100-year still-water level, plus 36 in of rise, plus 6 in of freeboard. Parks must be at least 9.2 ft (TI Infrastructure Plan, 2011¹⁵; the 2010 EIR¹⁰ and press¹⁴ round it to 9.1 ft). Planning to adapt Phase 1, which includes Cityside Park, to at least 36 in of rise starts once the Bay has risen 30 in (BCDC, 2022¹⁶).
What it means for the sculpture:
The site was raised during construction (SF Chronicle¹⁸ reports about 3 ft; USGS lidar¹⁹ shows about 1.7 ft at the pad, from 11.3 ft in 2010). The pad now sits at about 12.9 ft NAVD88 (USGS lidar¹⁹, March 2023): 3.7 ft above today’s 100-year still water and about 0.7 ft above that level plus 36 in. Wave run-up at the shoreline reaches 10–16 ft NAVD88 (EIR)¹⁰.
Evidence:
The 300 ft width of Cityside Park leaves room for future adaptation, and stormwater gardens run along its south and back edges (CMG¹⁷). The pad sits on the park’s back (east) edge (OpenStreetMap).
What it means for the sculpture:
A permanent foundation there, beside a stormwater garden edge, could clash with future levees or with the garden.

The fix is ordinary engineering: a geotechnical report, and a mat or drilled-pier foundation sized for liquefaction and the wind moment in the Wind section. Its absence from the proposal is the kind of thing a selection panel notices.

/ Salt air

At about 275 ft from the Bay, with a sea breeze blowing salt at it most afternoons, the sculpture sits in a severe marine exposure, though a sheltered bay is milder than open surf. The mirror finish and the rainbow colors are both the parts that suffer first.

Evidence:
Australia’s stainless steel association says tea staining is most likely within about 1 km of still marine water or 5 km of surf, and much less severe around sheltered bays (ASSDA²⁰). Within 200 m of an open coast with onshore winds, salt deposition can reach the level that defines ISO 9223’s C5 class (Solaracks²⁸). The pad is about 84 m from the Bay and downwind of it most afternoons, though the Bay is not open coast.
Effect on this piece:
Expect a severe marine exposure (likely ISO 9223 C4–C5; a coupon test or survey would confirm it). Type 304 stainless and plain powder-coated steel will not last.
Evidence:
Grade 316 is the minimum within 5 km of surf. Brown “tea staining” forms where chlorides sit on a surface that rain doesn’t rinse. ASSDA rates a mirror polish the most corrosion-resistant mechanical finish and easy to clean, and suggests washing sheltered seaside surfaces 3–12 times a year (for grade 304; higher grades may need less) (ASSDA²⁰).
Effect on this piece:
The sculpture‘s top panel spans its whole depth, so vertical rain never reaches any of the inside. That is exactly the sheltered surface where salt collects; only storm rain blown into the mouth rinses it. Mirror panels need 316 or better plus a washing schedule, or they go brown-streaked.
Evidence:
Aluminium touching stainless in salt air corrodes galvanically. Aluminium sits at about −0.7 to −0.9 V and stainless at about −0.1 to +0.2 V against a calomel reference, a gap of several hundred millivolts (PatSnap²¹).
Effect on this piece:
If color panels were aluminium on a stainless frame, every fastener needs isolation washers and sleeves.
Evidence:
The toughest architectural coating standard, AAMA 2605 (70% PVDF or FEVE resin), requires a 10-year South Florida exposure with fade of 5 ΔE or less (PPG²²).
Effect on this piece:
Pigments fade at different rates, so a rainbow can lose its gradient unevenly. Specify AAMA 2605 or porcelain enamel, and ask for fade data on each color.

None of this rules out the design. It changes the material spec, the budget and the maintenance plan, and the proposal addressed none of them.

/ People

The throat is the proposal’s invitation to walk through, and it is 2 in short of the 80 in headroom the ADA requires on a walking route. Its top edge also curves down over the walkway where a cane can’t detect it.

This design:
78 in (6 ft 6 in) at the centre
Standard:
80 in minimum on circulation paths, or a cane-detectable barrier where lower (ADA 307.4³)
Result:
Fails by 2 in
This design:
The rim curves in over the floor line by 4 in at 74 in, 9 in at 76 in and 22 in at 78 in
Standard:
Objects between 27 and 80 in may protrude 4 in maximum (ADA 307.2³)
Result:
Fails above 74 in: a head-strike edge for anyone over about 6 ft 2 in walking off-centre
This design:
43 in
Standard:
36 in minimum for a route
Result:
Passes
This design:
The wall leans outward from the floor up to 5 ft, then curves back
Standard:
Same 4 in rule
Result:
Passes: no overhang within reach

Whether the ADA applies depends on whether the pass-through is designated as a route. The proposal drew people walking through it, so a reviewer would treat it as one.

Climbing and falls.

The inside of the throat rises from the pad at about 40° at its steepest, and the inside of the mouth at about 60°, curving to vertical. Kids will run up it, and polished stainless is slick when damp with fog.
The top ridge climbs at 43° from the throat’s 6 ft 6 in top to the mouth’s 14 ft top. A fall from 14 ft onto concrete is a serious injury. Playground standards don’t govern art, but a park agency’s risk manager will apply their logic.
The panel and rim edges run down to the pad on both sides of the mouth, passing knee and shin height. They need a rolled or capped edge.

Skateboarding. A concave steel bowl with a 40° lip on a smooth concrete pad is a skate feature. Expect wax, grind marks on the polish and the rims, and a request for skate-stoppers that would change the look.

/ Night lighting, birds, water and upkeep

These are smaller issues, but each is a question a parks agency asks. Together they show whether a team has thought past opening day.

Why:
Two in-grade fixtures aimed straight up break DarkSky’s principles that outdoor light be useful, targeted, low-level, controlled and warm-colored (DarkSky International²³). The mirror panels then throw part of that light back up into the sky and out across the lawn. San Francisco’s bird-safe code says “No uplighting shall be used” for buildings inside an Urban Bird Refuge (open space of 2 acres or more dominated by vegetation, or open water) or less than 300 ft from one in an unobstructed line; the pad is inside Cityside Park, about 275 ft from the Bay (SF Planning Code Sec. 139(c)(1)(B)²). A sculpture may not count as a building, but a reviewer would read it as the standard.
Response:
Drop the in-grade uplights. If the piece must be lit, use shielded fixtures aimed down onto the painted panels, 3000 K or warmer, dimmed late at night and on a timer.
Why:
San Francisco’s bird-safe code treats free-standing glass walls, wind barriers, skywalks, balconies and rooftop greenhouses with unbroken glazed segments of 24 sq ft or more as hazards anywhere in the city (SF Planning Code Sec. 139²). It covers glass, not polished metal, and each mirror strip here is about 12 sq ft and broken up by color.
Response:
Low risk, but a reviewer may still ask. Upward light at night attracts migrating birds more than mirrors do by day.
Why:
The 14 ft rim makes an open perch about 275 ft from the water.
Response:
Expect droppings streaking the mirror and paint. Guano etches coatings, so plan for bird wire on the rim or a regular wash.
Why:
Everything falling into the mouth runs down the inner surface to the line where the sculpture meets the pad. No drainage was drawn.
Response:
Water and salt standing in that joint is where corrosion starts. Add a drainage slot or a slope in the pad.
Why:
A mirror polish shows every hand and tag at touch height, and anti-graffiti coatings dull a mirror.
Response:
Budget for regular cleaning, or use a brushed finish below 7 ft.

/ What we would change

Most of these problems trace back to one decision: facing the mouth southwest, into both the afternoon sun and the prevailing wind. Changing that one decision would have solved more than any single material or detail change.

Design response:
Face the mouth due north (0°). Or keep southwest and give the lower half a brushed (#4) finish.
Effect:
In the same model, north halves head-height glare (about 550 to 280 h/yr), cuts the peak from 2.2× to 1.7× and removes the 3.8× line. Brushed panels spread reflections instead of focusing them.
Design response:
Face the mouth away from 230–300°, or perforate some panels so air passes through.
Effect:
Less scoop, less speed-up, lower load.
Design response:
Engineered footing (mat or drilled piers), stiffened panel frame, fatigue-rated fasteners.
Effect:
Carries a 4,000–4,500 lbf design force, 2,200 lbf of uplift and years of gust cycles.
Design response:
Geotechnical report, and coordination with TIDA on ground treatment and the park’s adaptation zone.
Effect:
Foundation survives the design quake; nothing blocks future shoreline work.
Design response:
Raise the throat‘s top to at least 80 in (6 ft 8 in), with the rim curving in above that height. Or mark it as not a route, with a low curb as a cane-detectable edge.
Effect:
Meets ADA 307.
Design response:
Rolled rims, skate-stop details worked into the pattern, no reachable handholds on the top ridge.
Effect:
Removes the obvious risk-manager objections.
Design response:
316L or duplex stainless, AAMA 2605 or porcelain enamel colors, isolated fasteners, a written cleaning schedule.
Effect:
Keeps the mirror bright and the rainbow intact.
Design response:
No uplights; shielded, dimmable fixtures aimed down at the painted panels; a drain slot at the base; bird wire on the rim.
Effect:
Answers the parks agency’s standard questions.

The common thread: each of these can be tested on screen before the submission, as the companion simulation now does.

/ Methods, assumptions and sources

The sun and geometry figures are computed from the proposal drawings, so they are exact for the design as drawn. The wind and ground figures lean on codes and published data and need a site survey, a geotechnical report and a structural engineer to confirm.

What I computed.

Geometry. From the proposal: 18 ft mouth, 14 ft tall; 7 ft throat with its top at 6 ft 6 in; 8 ft deep; 32 panels alternating paint and polished stainless; 10 × 20 ft pad. Both rims are vertical circles, so the surface is an oblique cone.
Sun. NOAA solar position algorithm for 37.81958° N, 122.37321° W, Pacific time with daylight saving.
Glare. 60,000 sun rays per moment across the sculpture, with the 0.53° solar disk and 62% mirror reflectance. Up to 4 bounces; painted panels treated as non-mirror. Binned on a 5 in grid at the ground and at 1.7 m. Run every 4th day of 2026 at 10-minute steps, for 6,636 daylight moments.
Wind flow. A D2Q9 lattice Boltzmann solver on a plan slice at 4 ft and a vertical section along the wind, 10 cm cells, run to steady shedding. The solver’s Reynolds number is far below real wind and each slice is 2D, so it shows the pattern and an approximate speed-up, not exact speeds.
Wind load. ASCE 7-22 Chapter 29 method: V = 92 mph (ASCE Hazard Tool), Exposure D, Kz 1.03, Kzt 1.0, Ke 1.0, Kd 0.85, G 0.85, Cf 1.4–1.6 from the 3D simulation. Kd 0.85 is the ASCE 7-22 value for solid freestanding walls and signs; the row for round tanks and domes (1.0) would raise the load about 18%. G 0.85 is the gust factor for a rigid structure.
3D wind load. OpenFOAM v1912²⁷: snappyHexMesh around a zero-thickness shell of the exact cone, about 480,000 cells (3.9 cm at the surface), simpleFoam with realizable k-ε and wall functions, Exposure D inlet (z0 = 0.005 m), 3,000 iterations per direction, forces averaged over the last 900. Checked against a hollow hemisphere (8% low, corrected) and a coarser mesh (4% low). Scripts: tools/cfd in the project repository.
Wind climate. Treasure Island Naval Station readings (NCEI ISD 724943⁶, 1983–96, with gaps), bridged to 2021–August 2025 by quantile-mapping Oakland airport’s hourly record (ISD 724930)⁶ onto the island by season and hour; six NOAA CO-OPS Bay stations⁷ as a check. Script: tools/site_wind.py in the project repository.

Assumptions to confirm.

Pad position and mouth direction from the proposal plan, placed on the real seawall.
Force coefficient: from a steady 3D simulation corrected against a known shape, good to roughly ±10%. Gust peaks and panel flutter were not modelled; a stamped design would need a wind-tunnel test.
Site wind climate: the island’s own record ends in 1996 and was read every three hours, mostly by day. Head-height factors (0.75–0.82) assume a log profile; an on-site anemometer for a season would confirm them.
Pad elevation: about 12.9 ft NAVD88 from March 2023 lidar; work continued until the park opened in September 2025 and no as-built grade is published. Flood levels are in NAVD88 per the 2011 Infrastructure Plan. Soil depths are island-wide; no boring log for the park was found.

Sources.

Every page above was read in full in September 2026. Where a site blocked automated access, the text came from a copy supplied by hand (SF Planning Code Secs. 148 and 139 as currently codified, including the 2013 amendment to Sec. 139; the Solaracks and MDPI pages; DarkSky’s principles).