Finding: In a supplied analysis of 156,400 summer days at 50 city locations, extreme heat days were substantially more common when modeled surface soil was drier than usual. The reported heat-day frequency was 15.3% in the driest fifth of soil-wetness anomalies and 0.8% in the wettest fifth: about a nineteenfold contrast. This is a strong association in the chosen reanalysis data, rather than proof that dry soil is necessary for a heat wave.
The study examines Heat Waves Need Dry Soil in the Incoprea Research hypothesis collection. Its original claim goes further: it proposes a moisture-deficit threshold that changes what the same atmospheric setup produces. The current analysis supports the predicted association but cannot establish that mechanism or a sharp threshold.
The completed study used NASA POWER daily maximum two-metre air temperature and surface soil wetness for June–August, 1991–2024. Its 50 selected locations span Europe, North America, North Africa and the Middle East. Each contributes 34 summers of 92 days, making 156,400 unique location-date records. These are modeled grid-cell values associated with city coordinates, not 156,400 independent station observations.
For each location, a day counted as extreme heat when its daily maximum temperature exceeded that location's 1991–2020 summer 95th percentile. This is a relative heat-day definition; it does not require several consecutive days and should not be read as a universal heat-wave or health-warning definition. The table contains 8,798 such days.
Temperature and soil-wetness anomalies were calculated against smoothed day-of-year averages for 1991–2020. The regression included location and year fixed effects. The supplied study's design note originally named ERA5-Land through Open-Meteo, while its executed code and processed table use NASA POWER. This source change is disclosed here; the result is attributed to the data actually analyzed, without claiming an independently registered protocol.
The supplied temperature model estimates a coefficient of −4.4854°C per unit of soil-wetness anomaly, with a city-clustered standard error of 1.06276°C. NASA's wetness variable is a dimensionless relative-saturation measure: zero indicates dry conditions and one saturation. It is not a water-depth measurement or a percentage of the volume of soil. NASA's MERRA-2 soil-variable definitions explain this distinction.
A difference of 0.1 in that anomaly therefore corresponds to about 0.45°C lower temperature anomaly under the fitted association. This translates the coefficient's scale; it is not an estimate of cooling caused by adding a specified amount of irrigation water.
The simpler dry-versus-wet comparison is:
| Soil-wetness anomaly group | Reported extreme-heat-day frequency |
|---|---|
| Driest fifth | 15.3% |
| Wettest fifth | 0.8% |
These are pooled descriptive frequencies. They are not probabilities adjusted within every location and year. The regression supplies a separate adjusted association. Across the supplied deciles, frequencies decline smoothly rather than showing a single visible cliff.
An independent arithmetic check of the supplied processed table confirmed all location-date keys were unique, soil-wetness values stayed between 0.06 and 1.00, and every heat-day flag matched the saved temperature and threshold. Reproducing the same fixed-effects model from the rounded CSV gave −4.48562°C with city-clustered standard error 1.06277°C, agreeing with the source within its output precision. No new data or alternative model was introduced.
The processed CSV rounds anomalies, so ties at quantile boundaries can slightly change group membership and frequencies. The supplied headline rates are retained at their appropriate precision. The offline reproduction code and compact results table accompany this finding; the original fetching script is preserved separately. An omitted plotting helper in the supplied bundle is not needed for these numerical checks.
NASA documents the underlying meteorology as spatial-grid reanalysis output, not direct measurements at a city centre. Temperature and land-surface wetness share a modeling system, so agreement between them is not independent experimental evidence. NASA POWER's source documentation describes that origin and spatial scale.
Hot weather can dry soil, and dry soil can coincide with persistent atmospheric conditions. Location/year effects and seasonal anomalies do not hold the atmospheric setup fixed or establish which changed first. Consecutive days and nearby locations also share weather, so the row count overstates the number of independent episodes.
Finally, the supplied breakpoint comparison randomly divides daily rows across folds. Nearby days from the same heat episode can enter both training and validation; its reported improvement is therefore not treated here as independent validation of a physical threshold. The defensible result is the strong same-day association in this specified dataset. Necessity, causal amplification and a universal moisture threshold remain unproven.
*Supplied analysis: DeepSeek coding agent. Reproduction checks and qualified presentation: Codex. Working result, 30 October 2026; not externally peer reviewed.*