Site conditions, honestly: the climatic design basis in one run

Every specification opens with a page of site conditions, and that page is where projects quietly go wrong. Not because the numbers are hard to find, but because two different kinds of number get written in the same column: what the weather did, and what the plant has to be built for. A record maximum is not a design temperature. A gust from a thirty-year series is not a reference wind speed. This guide walks the whole thing end to end in the climatic and environmental design conditions tool: where every figure comes from, which ones it refuses to invent, and how to read the result. The worked example is a real run — every figure below was taken from the tool, not typed in by hand.

What the tool is for

You give it a point on the map. It gives you the design basis: 346 parameters in four blocks, each carrying its own type, source, reference period, height or depth, averaging time and return period.

• A — the site: country, nearest settlements, transport, terrain, distance to the sea
• B — general climatic conditions: temperature, humidity, pressure and altitude, wind climatology, precipitation, solar
• C — natural hazards and environmental actions: design wind, snow, ice, lightning, pollution, dust, corrosivity, seismic
• D — the application block for the plant you picked: photovoltaic, wind onshore or offshore, conventional, substation, with an overhead-line toggle

Nothing has to be filled in. The array tilt is found by transposing the hourly irradiance onto a tilted plane and sweeping it; the wind profile is referred to the upper level the source actually carries; the terrain category, the soil thermal resistivity, the pollution severity class and the IEC 60721 classes are all derived, and each row names the method it used.

The one rule the whole thing is built on

A statistic of the record and a design action are different things, and neither is ever presented as the other.

That is not a stylistic preference. Take a thirty-year daily wind series and read off its maximum: you have the largest value that happened to occur in thirty years, which is neither a 50-year nor a 150-year value, and has no defined probability of exceedance at all. Structures are not designed against it. What a structure is designed against is an extreme-value fit, and the tool does that one properly — Gumbel Type I on annual maxima by the method of IEC 60826:2017, Annex D:

F(x) = exp{ -exp[ -a(x - u) ] }

a = C1 / σ          u = x̄ - C2 / a

x_T = x̄ - (C2/C1)·σ + (σ/C1)·y_T     y_T = -ln(-ln(1 - 1/T))

C1 and C2 come from Table D.1 by sample size. The implementation is checked against the standard's own Table D.2, which tabulates x_T/x̄ for a range of coefficients of variation and return periods — 143 of its 144 cells reproduce exactly, and the one that does not is an internal inconsistency in the table itself, documented in the code rather than smoothed over.

So the report separates them by column. Every row says whether it is Climatological, an Observed extreme, a Design value, Code-based, Derived, or User input. Section 5 collects the observed extremes together on purpose, with a note saying none of them is a design value — because that is exactly the section people write specifications from by mistake.

Where the numbers come from

Everything is free and needs no key. Most sources answer a browser directly; the ones that send no CORS header go through a read-only pass-through of our own — GET only, an exact-host allowlist, no credentials forwarded, a day of caching.

Choosing NASA POWER as the primary source means the whole report can be produced without touching a metered service. That is deliberate: an hourly series over years is the largest request the tool makes, and on the other source it is what exhausts a free budget.

The optimum tilt, determined three times

A single optimum tilt is a number you have to take on trust. So the tool produces three, on three radiation datasets by three methods, and prints them side by side:

• this report — the hourly irradiance of the record, transposed by the isotropic sky model of Liu and Jordan and swept over tilt from 0 to 60°
• PVGIS — optimised by the Joint Research Centre on its own satellite radiation database
• Solargis — the Global Solar Atlas result

On the worked site below, PVGIS and Solargis both return 26°, with in-plane irradiation of 2350 and 2315 kWh/m²·year — 1,5 % apart on two independent models. The report says why they differ rather than leaving a reader to think one is broken: the isotropic model neglects the forward scattering an anisotropic model accounts for, so it reads a little low on a tilted plane and biases the optimum a little low with it. The optimum is also flat, and the sweep is printed in full so you can see that a few degrees either side costs well under a per cent.

A worked site: 25,2000° N, 55,2700° E

A coastal point in Dubai, run against NASA POWER over 2016–2025.

Where it is. United Arab Emirates, Dubai Emirate. Dubai 7,5 km NNE, Deira 8,8 km NNE, Sharjah 22,2 km NE. Nearest classified road D71 at 1,0 km, nearest motorway D68 at 2,5 km, railway 7,2 km, Dubai International Airport 11,6 km. The sea lies 4 km to the SSE. All of it from OpenStreetMap and a reverse geocode, in latin script, because a report whose map a reader cannot read is not a report.

Temperature.

The first line is the one an equipment rating is referred to, and the tool puts it forward as the design ambient. The others sit beside it as what they are. At 39,9 °C the assessment is blunt: ratings referred to a 40 °C ambient are close to the site value, so derating has to be checked rather than assumed adequate, and the enclosure temperature rise verified at the site ambient rather than at the 40 °C of the standard rating.

Wind. Reference wind speed 18,4 m/s at a 50-year return period, from the Gumbel fit above. The hourly record gives 3,72 m/s mean at 10 m and 4,48 m/s at 50 m, a power-law shear exponent of 0,114 fitted between those two levels, and an air density of 1,162 kg/m³ from the ideal-gas law on the hourly pressure and temperature. NASA POWER carries no gust, and the gust rows say so instead of borrowing a number from somewhere else.

Solar. POWER's hourly record gives 2028 kWh/m²·year of global horizontal irradiation over ten years. The Global Solar Atlas gives 2137 — 5,1 % higher. The report reports both and says which is the better estimate: the satellite long-term average, because at a dusty coastal point a half-degree reanalysis grid under-reads the resource. The horizon reaches only 3,8°, so terrain shading is negligible and only near-field shading has to be modelled.

Pollution, and the creepage distance it decides. This is the chain that makes the whole exercise worth doing, because the creepage distance of every outdoor insulator on the site is set by how dirty the air is — a property of the place, not of the equipment:

the air at this site
  → site pollution severity class d (heavy)         IEC TS 60815-1, Table 5
  → reference USCD 43,3 mm/kV                       60815-2 and -3, Figure 1
  → corrections of clause 10 (Ka = 1, Kad = 1)      60815-2, clause 10
  → 900 mm phase to earth at U(m) = 36 kV           60815-1, clause 3.1.6

Class d follows from the sea at 4 km and mapped industry within 1 km. And note the last step, because it is where this calculation is most often wrong by a factor of 1,73: the unified specific creepage distance is referred to the highest operating voltage across the insulator, so a phase-to-earth insulator in a 36 kV system is dimensioned on U(m)/√3 = 20,8 kV, not on 36 kV. The note to clause 3.1.6 exists precisely because the older specific creepage distance of IEC/TR 60815 (1986) used the line-to-line value.

Environmental classification. 4K4 / 4Z10 not applicable / 4B1, or 4B2 where termites are present / not below 4C4 / not below 4S2

Read that carefully, because each part means something specific. 4K4 is a determination against Table 1 of IEC 60721-3-4 using the means of the annual extreme values, which is what clause 5.1 requires — not the absolute record, which would land a class higher. "4Z10 not applicable" is also a determination: the site is at sea level, below the roughly 1400 m that class corresponds to, so the low-pressure severity does not apply. 4B1 is established because mould, fungus and rodents apply at any outdoor site; whether 4B2 applies turns on one question — termites — which is a question about the site and not about its weather. And the two classes written as "not below" are floors: the measured regional background alone already requires them, and a source over the fence can only raise them. Dust reaches 1177 µg/m³ at its worst hour, with 94 days a year above 200 µg/m³.

Seismic. Nothing. There is no open point service for a peak ground acceleration in the UAE — EFEHR/ESHM20 and GEM publish maps but no API — so the row says a national seismic map or a site-specific probabilistic hazard study is required, and the screening level for the emirate is quoted as the screening level it is. Where a national model does answer, the tool uses it: USGS ASCE 7-16 and 7-22 for the United States, the NRCan sixth-generation model for Canada, J-SHIS for Japan. Los Angeles returns PGA 0,844 g and seismic design category D; Vancouver 0,446 g at a 2475-year return period; Tokyo a JMA intensity of 6,3 and 142,3 cm/s — and no PGA, because that model does not publish one and an intensity is not an acceleration.

What it refuses to do

This is the part worth reading twice. The following are not computed, because no free source and no defensible method gives them, and each row says which study or authority would:

measured ESDD and NSDD · snow and ice density · design radial ice thickness · hail · turbulence intensity and the IEC 61400-1 extreme-turbulence parameters · extreme still-water level and surge · salinity · marine growth · a flood level above datum · groundwater level · a peak ground acceleration outside the three countries named above.

The cloud-to-ground flash density is derived from thunder days by IEC 62305-2 Annex A, which the standard states for temperate regions only — so the row carries that caveat rather than pretending otherwise, and a lightning location network remains the proper source.

What comes out

• A Word report — sixteen sections with a title page, a table of contents, the monthly charts, a wind rose, the tilt sweep and the horizon profile as figures, and a closing assessment that says, hazard by hazard, which value governs and what it drives
• A real AutoCAD table — an ACAD_TABLE entity in DXF 2007, not lines with text on top: click it in AutoCAD and the ribbon shows Table Cell
• A ten-parameter summary for a drawing or somebody else's section, as CSV, as an AutoCAD table, or as text for a title block
• CSV, spreadsheet and JSON of the whole dataset

Where a national code governs

ERA5 is a reanalysis — a physical model of the atmosphere fitted to the available observations, about 25 km, or 9 km for the land surface. NASA POWER is a half-degree product. Both are the best free estimate for a point where no station stands, and both are traceable. Neither is the design climatology of a national code.

Where an ASHRAE design-conditions table, a national annex, SP 131.13330 or a figure issued by the authority applies to your site, that document governs. This tool is the check beside it — or the answer where your site falls outside all of them, which in this part of the world is most of the time.

Run it on your own point →