Add site massing model for the Mueang Chi candidate parcel

Models the plot and everything on it in FreeCAD, and writes up the
decisions the modelling forced.

Parcel geometry is traced from the DOL LandsMaps cadastral polygon and
reconciled against that portal's own measured edge lengths (40.86 + 26.61
/ 166.81 / 42.48 + 14.94 / 152.46 m), which the trace reproduces to within
about 1% with 1.67 m of closure error over a 444 m perimeter. Note the
mapped extent comes out near 9,570 sqm against a deed area of 8,266 sqm,
a ~16% gap that needs a surveyor before any offer.

On the site: 60x25 m production building on a 10 degree monopitch, a
45x38 m phase-2 building, roof and carport PV, fire and raw water tanks,
a detached boiler/chiller utility yard, perimeter wall and gate, guard
house, roofed staff parking, and the loading and fire-access roads.

Docs:
- scale.md: expansion headroom for a second building, and the detached
  utility yard, which frees ~126 sqm inside the main shed
- sustainability.md: roof orientation study (rotating the building for
  solar gains under 1 point and costs 4 m of clearance, so it stays
  aligned to the plot), and a new water supply section
- facility-security.md: fire water supply behind the existing
  suppression design

Figures marked as estimates are planning-grade, not engineering
calculations; the PV module spec is a generic placeholder, and zoning for
factory use on this plot is still unverified.

docs/sustainability.md

@@ -42,6 +42,62 @@ at 10 bar / 180 °C) and the 250 kW cooling system.
 - System sizing (kWp, storage, grid-connection model) is open until the site
   is chosen.
 
+### Roof design for PV (candidate Mueang Chi site)
+
+Sized against the candidate 5-0-66.5 rai parcel in Mueang Chi (18.4708 N,
+98.9642 E), whose narrow skewed shape fixes how the building can sit. Two
+findings drove the roof form:
+
+- **Orientation barely matters at this latitude.** An annual geometric
+  clear-sky model puts the optimum fixed tilt near 18° facing due south, but
+  the response is very flat — at 10° tilt, swinging the azimuth from due
+  south to ESE costs under 2%. Rotating the building toward due south would
+  gain **under one percentage point** of yield while cutting boundary
+  clearance from 16 m to 12 m. The building therefore stays aligned with the
+  plot, which also keeps truck circulation on the site's long axis.
+- **Roof form matters more than orientation.** The plot puts the long axis on
+  a 212° bearing, so a symmetric gable would strand half the array on a
+  WNW-facing slope worth only ~94% of ideal. A **monopitch roof at 10°,
+  falling toward ESE**, puts the entire roof on one plane at ~98% of ideal
+  and yields a single uninterrupted rectangular array field — simpler
+  racking, stringing, and roof penetrations.
+
+10° was chosen over a shallower pitch deliberately: the yield difference
+between 6° and 10° is under 0.2%, but northern Thailand's dry-season dust and
+burning haze make self-cleaning in the rains worth more than the ~1.8 m of
+extra building height the steeper pitch costs.
+
+| Item | Figure |
+|---|---|
+| Sloped roof plane | ~1,520 m² |
+| Module assumed | generic 580 W, 2,278 × 1,134 mm (~225 W/m²) |
+| Modules (1 m perimeter strip, walkway every 4 rows) | 425 |
+| Installed capacity | **~247 kWp** |
+| Array coverage of the roof plane | 72% |
+| Expected yield | **~335 to 360 MWh/yr** |
+| Share of mature two-shift use (1,000 to 1,500 MWh/yr) | ~25 to 35% |
+
+**No module has been selected.** The 580 W / 2,278 × 1,134 mm figure is a
+generic placeholder for the current large-format n-type TOPCon class (the
+size and power band shared by Trina Vertex, JA Solar DeepBlue, Longi Hi-MO
+and Jinko Tiger Neo in their 570 to 600 W ratings), chosen so the layout
+count is realistic. Capacity scales almost linearly with whatever module is
+actually procured: the roof fits ~425 modules of that footprint, so a 550 W
+module gives ~234 kWp and a 620 W module ~264 kWp. String sizing must then be
+matched to the Deye inverters' MPPT voltage window before anything is fixed.
+
+Caveat on method: the tilt/azimuth comparison is a geometric clear-sky model,
+so it is reliable for **ranking** orientations but not for absolute energy.
+The kWh/kWp figure behind the yield column is the regional 1,400 to 1,500
+kWh/kWp/yr assumption, derated to ~98% for the ESE azimuth. Confirm against
+PVGIS or a TMY dataset before committing capital.
+
+This supersedes the older "150 to 200 kWp on a ~1,500 m² roof" figure in
+[scale](scale.md), which assumed lower module power density; current
+large-format modules reach roughly 225 W/m². Roughly 70 m of plot depth
+remains undeveloped behind the building for ground-mount, which matters if
+the electric-boiler route is chosen.
+
 ### Energy efficiency architecture: recover before generating
 
 The deepest efficiency in this plant comes from its own thermodynamics: the
@@ -89,6 +145,47 @@ diesel means a fuel tank with price exposure. The electric and biomass paths
 avoid a fossil tank on the property entirely, which matters for both the
 sustainability commitments and the BOI story.
 
+## Water supply
+
+The zero-discharge commitment below governs water *leaving* the plant. It
+says nothing about water arriving, and that side is still open: **no source
+has been confirmed for the candidate site** — PWA mains, a licensed deep
+well, or both. That choice drives storage, treatment, and part of the
+utilities budget.
+
+Estimated demand at mature two-shift operation (16 h/day):
+
+| Draw | m³/day |
+|---|---|
+| Cooling tower evaporation | 6.0 |
+| Cooling tower blowdown (4 cycles of concentration) | 2.0 |
+| Boiler makeup + blowdown (80% condensate return) | 1.6 |
+| Galvanics rinse, after the recycling below | 3.0 |
+| Domestic, canteen, showers (25 staff) | 2.0 |
+| **Total** | **~15, call it ~20 with margin** |
+
+The cooling tower is over half of it — worth remembering when the heat
+recovery above is sized, because every kW the feedwater heat pump lifts off
+the chiller is also water the tower does not evaporate.
+
+Storage is **two tanks, not one**:
+
+- **Fire reserve ~200 m³** — dedicated, never drawn on by process or
+  domestic, per [facility security](facility-security.md).
+- **Raw/process ~50 m³ usable** (5 m dia × 3 m) — roughly 2.5 days at peak,
+  feeding a softening/RO skid for boiler feedwater and dosing for the tower.
+
+A **combined ~250 m³ tank**, with the fire volume protected by taking the
+process suction from a higher level, is cheaper and has a real side benefit:
+process turnover stops the fire reserve going stagnant. Worth pricing — but
+only the fire design and the insurer can approve it, so the split above
+stands as the safe default until they do.
+
+Open items: confirm the source, its yield and its quality (a well needs a
+licence and a pump test; hardness and iron set the treatment spec), and
+settle whether racking height pushes the fire reserve past 200 m³ before
+either tank is ordered.
+
 ## Zero contaminated water
 
 The reference process schematic shows a conventional open "dirty water"