A different kind of risk

The six variables before this one all changed a number: how many kilowatt-hours a panel makes in a normal year. Storms change a different number: the probability that the panel, the racking or the inverter is damaged or destroyed. That risk does not belong in a yield estimate, because it is episodic and skewed, a small annual chance of a large one-off loss rather than a steady drag on output. It belongs in two other places instead: the structural specification of the mount, and the insurance and maintenance reserve. This report covers the four perils that govern it, and they fall very unevenly.

Large hail attacks the module itself: a stone of two centimetres or more can crack the glass and the cells beneath it outright. It is also the most concentrated of the four perils:

The 10 of 40 markets with a defensible large-hail figure. Thirty markets are blank, not zero: almost no city publishes a days-per-year count for hail at or above 2 cm, so a number is entered only where a credible city-scale figure exists, and an explicit 0 marks a genuinely hail-exempt site. Bogota and Seoul counts include all hail sizes, not strictly >=2 cm, and are flagged as such. Full sourcing and averaging bases in the public dataset.

Two markets carry a genuinely high hail hazard, and they are the spine of this report. Johannesburg sits on the South African Highveld, with roughly five damaging-hail days a year and, in the wider all-size count, more than sixty. Sydney records about five damaging-hail days a year in the city and far more across its basin, and the April 1999 storm brought nine-centimetre stones; hail-prone days in the Australian hail climatology cited below are rising at roughly ten per cent a decade. In these two markets the mounting and the insurance reserve should assume glass-breaking impacts, and a hail-durability rating on the module earns its price.

Where the danger is, and where it simply is not

Hail is only one of four perils, and across the full 40 markets the thing that governs asset survival varies enormously:

The cyclone coasts are a short list. Tokyo and Shanghai face frequent typhoons, so their structures must be cyclone-rated and, ideally, stow-capable if tracked; Seoul and New York take the occasional storm, with New York's major-hurricane recurrence around one in seventy-four years. Reykjavik has no hail and no cyclone, and still carries Europe's highest design wind in this set, up to 46 metres per second on the exposed coast, so its racking is wind-governed. Five markets sit outside all four perils: Antofagasta on the Atacama, along with Madrid, Athens, Santiago and Casablanca, faces no material extreme-weather peril, which is a signal not to over-specify the mount.

The design-wind numbers carry one essential caveat, and the dataset preserves it: the averaging basis is decisive. Tokyo's code figure of 34 metres per second is a ten-minute mean; converted to the three-second-gust basis Sydney's 45 is already quoted on, it becomes roughly 56, eleven metres per second harsher. The smaller headline number is the tougher structural requirement, and a mount specified by comparing raw code figures ranks these two markets backwards. Eight markets carry no confirmed code wind speed at all and are left blank.

Specify the mount to the real peril

The buying decisions follow from the map. Specify the mount to the real peril: cyclone-rated and stow-capable structures on the typhoon coasts and the high-wind outliers like Reykjavik, a hail-durability rating on the module in Johannesburg and Sydney, and no resilience premium in the five exempt markets. Insurance and the maintenance reserve price to the hazard rather than the yield, because a market can rank high for production and still carry an elevated premium; the number is set by the annual chance of a large loss rather than by output. And two of the perils are moving: the Australian hail climatology cited below has hail-prone days rising in Sydney and Perth, while Li et al. (2016) record total hail declining on the Tibetan Plateau. The storm map a mount is specified against is not static, which is why this series re-reads it annually.

Hazard, not an annual average

This pillar records hazard rather than annualised loss, because storm damage is episodic and averaging it into a percentage would misrepresent it. Hail is assessed at the two-centimetre threshold that damages photovoltaic glass; tropical-cyclone exposure is graded from basin geography and hazard screening; design wind speed is taken from the local building code with its averaging basis (three-second gust or ten-minute mean) preserved; and flood is the higher of riverine and urban screening. Large-hail days are blank for 30 of the 40 markets, because almost no city publishes the figure, and a blank marks the absence of a credible number rather than an absence of hail. Return periods and several wind speeds are approximate bands inferred from hazard ratings and basin climatology rather than single published values, and each such cell is marked. The dominant-peril and insurance-signal fields are a reasoned synthesis across the four hazards, not a single sourced measurement. Every value carries its source and its basis in the public dataset. Coastal storm surge, the dominant flood mechanism for a market like New York, is noted but sits out of scope for a pillar keyed to the panel and its mount.