Weather Margins Must Be Line Items, Not Hidden in Dry-Bulb or Wet-Bulb

The heat-rejection chapter later had to write a prohibition it had already needed. Margin may not be written into the weather numbers. The Engineer of Record's weather values are the uncorrected base condition, and every margin must appear as an explicit, individually revocable addend. The reason given is forensic rather than about accuracy. Once margin is pre-mixed, nobody downstream can tell whether a given temperature is a weather fact or a fact carrying several layers of conservatism, and the same margin will be counted again in other disciplines that inherit the number.

Four Margin Rows Receive Four Different Treatments

The recommended schedule has four rows, every placeholder value graded as an assumption [TARGET], and they are not treated alike:

• Site heat recirculation adds 1 to 3 K to dry-bulb on the summer and extreme conditions, with a condition attached: the value depends on the final site layout and equipment spacing, and the lower end may not be taken before the layout is fixed.
• Weather-data uncertainty adds 0.5 to 1 K to wet-bulb on the summer condition only, covering station distance, the data period and local site difference.
• Equipment performance degradation, fouling and dust loading is converted into a capacity margin and is expressly not a temperature addend, with the reason written into the row: to avoid double counting against the first two.
• Future climate trend carries no placeholder at all and applies only if the owner requires it.

The third row is where the rule proves itself. It carries no numeric placeholder, and the capacity margin it becomes is nowhere quantified — the row exists to stop a selection margin being spent twice, once as temperature and once as capacity.

Stacking is then a rule rather than an aggregation. The extreme condition is already an extreme percentile, so the data-uncertainty margin is not stacked onto it — extreme-value statistics already contain their own tail uncertainty — and only recirculation and, if adjudicated, the climate trend are added. Repeated stacking pushes the check condition into a range with no physical meaning, oversizes the rejection plant, and then degrades part-load efficiency and annual energy, so the conservative act makes the annual result worse. No penalty is quantified for that, and no annual energy figure exists in the read set that could size it.

Weather band and equipment failure state are independent dimensions to be combined as a matrix, the most onerous joint point being the extreme condition including recirculation under one failure — and whether full capacity must be held there is registered as an owner resilience decision with no answer presumed. If derated operation is accepted, its trigger temperature and depth must still be stated.

Capability Checks Run Hot; Energy Forecasts Stay Unbiased

A later conclusion in the same chapter splits the four-row schedule by purpose. Capability check is biased hot: add temperature, discount capacity. Energy forecast is unbiased: any one-way margin distorts the prediction. One input set cannot serve both. The annual integral therefore applies the recirculation margin and forbids the weather-data-uncertainty margin. Carry the design-point conservatism into the annual integral and the result is an annual ratio nobody can reach and nobody can falsify.

A fifth line item sits outside the original four. A route that is only just feasible at the summer design condition has no temperature left for ride-through, so the thermal clock is zero. That ride-through allowance is an independent margin. It must not be mixed with recirculation, weather-data uncertainty, or fouling.

The Wrong Chain Was Marked Dead; Recomputation Is Unperformed

A controlled erratum then downgraded an entire weather chain in the heat-rejection basis to inapplicable, because it was drawn from a region other than the project's site. Its risk paragraph is the argument for line-item margin stated from the failure side: the rejection capability, approach temperatures, design dry-bulb and wet-bulb points, plant selection margins and every ratio and water figure derived from them all hang on one chain, and they appear in the text in exactly the same form as a legitimate conclusion. The erratum names the most dangerous part as the fact that downstream documents had already cited these figures as established facts. Reading that invalidation as an instance of the margin rule's failure mode is this article's framing; the erratum is about a wrong data source, and the two share a mechanism rather than a cause.

So the first mandated action is not recomputation. It is to mark the chain dead in place so every downstream citation fails visibly, with the original values retained on display, annotated as a wrong chain kept for traceability, and specifically not deleted — because deletion would leave downstream readers unable to identify what they had cited. The heat-rejection revision executed that invalidity declaration. The recomputation remains unperformed.

The erratum is split into two disposal segments, and the split is the honest part of it. The text correction can be closed by the control editor; the recomputation cannot, and is assigned to a licensed Engineer of Record with four prerequisites stated — site coordinates and elevation, a named weather station or climate dataset entry with its version number, the selected design percentiles, and a confirmation scope covering every derived result across four documents. Until those exist no number may be substituted. A textual erratum can only remove a wrong statement and cannot generate a right one. That is also the limit of what this article claims.

Which Kelvin Came From a Weather Station

If the owner declines the future-climate scenario and the recomputed site weather returns a wide summer margin, the four-row schedule becomes bookkeeping and nothing turns on whether it was kept separate. The discipline earns its cost in the opposite case, where the difference between a defensible design condition and an unauditable one is whether anyone can still say which kelvin came from a weather station and which came from a judgement about spacing.

The 1 to 3 K recirculation band and the 0.5 to 1 K wet-bulb band remain placeholder assumptions, not selected addends. [TARGET]

Open: the site weather final values; whether a future climate scenario is stacked; whether full capacity is held at the joint extreme-and-failure point; which weather band the contract ratio is scored against; and the four recomputation prerequisites. Three of those belong to the owner rather than to any engineer. [HOLD]

No corrected design condition exists. No annual energy figure is published. No Engineer of Record has reviewed these quantities. Which of those owner items is being tabled first?

© 2026 K&K Data Service Inc. All rights reserved. Reproduction or republication is permitted only with clear attribution to K&K Data Service Inc. and a working hyperlink to the canonical URL of this article. Excerpts must preserve the technical context, maturity labels, assumptions, and limitations. No excerpt may imply project approval, field validation, certification, or endorsement that the original article does not state.

Source: K&K Data Service Inc., “Weather Margins Must Be Line Items, Not Hidden in Dry-Bulb or Wet-Bulb,” https://www.kkdatasvc.com/lab/heat-rejection-and-pue/weather-margins-are-line-items-not-hidden-inside-db-wb/.