DWD daily-KL reanalysisProduct v1.0.0 · preregistered frozen v0.1 registry
Is Frankfurt actually warmer?
Frankfurt/Main-Westend, an inner-city DWD climate station, is compared against Frankfurt/Main — DWD's own designated "counterpart" reference station, physically located at the airport — using every day both stations recorded a valid daily mean temperature between 1985 and 2025. The gap and its uncertainty are reported before any "warmer" or "trending" conclusion.
Preregistered protocolReal public DWD dataFrozen result registryUncertainty stated first
01
Sample
14,579 valid paired days
Read this before any "warmer" or "trending" conclusion below
- The reference station is DWD's own designated "counterpart" for Westend — but it is Frankfurt Airport, an open, semi-paved site, not a verified pristine rural baseline.
- This measures the daily mean temperature gap only, not daily minimum (overnight) temperature, which is the variable most often associated with larger reported urban heat island effects.
- Two stations, one pairing: elevation, distance, and instrument history differ between them in ways this comparison cannot separate from any urban-canopy effect.
- This project does not claim to measure "the" Frankfurt urban heat island in general, and does not build an independent land-use model.
Full disclosure: docs/research-protocol.md and docs/research-report.md.
02
Where
Where the two stations actually are
A north-up schematic drawn to real scale from the coordinates in data/provenance.json — not basemap imagery and not a live map service. Hover, click, or tab to a marker for that station's details; the table below carries the same information as accessible text.
Urban station Reference stationSchematic, north-up, drawn to scale from real coordinates — not basemap imagery.
Urban station
Frankfurt/Main-Westend
- DWD station ID
- 01424
- Coordinates
- 50.1269° N, 8.6694° E
- Elevation
- 120.78 m
How the 15.42 km distance and 21.08 m elevation difference were computed
Straight-line (great-circle) distance via the haversine formula, from each station's real coordinates in data/provenance.json:
a = sin²(Δlat/2) + cos(lat₁)·cos(lat₂)·sin²(Δlon/2)
distance = 2·R·atan2(√a, √(1−a)), R = 6371.0088 kmlat₁=50.1269, lon₁=8.6694 (Westend, urban) · lat₂=50.0259, lon₂=8.5213 (Frankfurt/Main, reference) → 15.422 km, initial bearing from the reference station 43.2° (NE).
Elevation difference: 120.78 m − 99.7 m = 21.08 m — Westend sits higher.
| Station | Role | DWD ID | Latitude | Longitude | Elevation, m |
|---|---|---|---|---|---|
| Frankfurt/Main-Westend | Urban station | 01424 | 50.1269° N | 8.6694° E | 120.78 |
| Frankfurt/Main | Reference (“counterpart”) station — physically Frankfurt Airport | 01420 | 50.0259° N | 8.5213° E | 99.70 |
| Straight-line distance between the two stations: 15.42 km (bearing 43.2° / NE from the reference station to the urban station). Elevation difference: 21.08 m (Westend higher). Both computed from the coordinates above via the haversine formula — see the calculation above the table. | |||||
03
Uncertainty first
The measured gap
Mean daily gap (urban minus reference), with a 95% confidence interval from a 30-day block bootstrap (10,000 resamples) — chosen over a naive resample because consecutive days' temperatures are strongly autocorrelated. Read the confidence interval before the point estimate.
Full-period mean gap: +0.455 °C CI excludes zero
95% CI [+0.432, +0.478] °C over 14,579 valid paired days. The interval excludes zero, so Westend is measurably warmer on average than the reference station over this record — but the magnitude is modest, well under one full degree Celsius.
| Period | Valid days | Mean gap, °C | 95% CI | Excludes zero? |
|---|---|---|---|---|
| Full period | 14,579 | +0.455 | [+0.432, +0.478] | Yes |
| Winter (Dec-Feb) | 3,641 | +0.489 | [+0.453, +0.522] | Yes |
| Summer (Jun-Aug) | 3,649 | +0.415 | [+0.361, +0.471] | Yes |
| Spring (Mar-May) | 3,619 | +0.541 | [+0.497, +0.589] | Yes |
| Autumn (Sep-Nov) | 3,670 | +0.374 | [+0.341, +0.406] | Yes |
04
Uncertainty first
Has the gap trended?
Ordinary least squares trend of each included year's mean gap against year, with a preregistered classical 95% confidence interval on the slope. A post-release serial-correlation-robust sensitivity is reported alongside it. Read both intervals before any "widening" or "narrowing" conclusion.
Trend: -0.0031 °C/year not significant
95% CI [-0.0071, +0.0008] °C/year, p = 0.118, R² = 0.063, n = 40 years. The confidence interval crosses zero and the p-value does not clear this project's preregistered α = 0.05 threshold: no statistically detectable trend in the urban-reference gap over this ~40-year record.
Serial-correlation sensitivity: the annual OLS residuals have lag-1 correlation 0.657. A three-lag Newey–West interval is [-0.0079, +0.0016] °C/year (p = 0.186). It is wider, still crosses zero, and therefore leaves the preregistered conclusion unchanged. Inspect the sensitivity result.
Read the complete daily data table (14,579 rows)
| Date | Westend (urban) TMK, °C | Frankfurt/Main (reference) TMK, °C | Gap, °C | Direction |
|---|---|---|---|---|
| 1985-11-01 | 5.6 | 5.0 | 0.60 | ▲ warmer |
| 1985-11-02 | 5.3 | 4.6 | 0.70 | ▲ warmer |
| 1985-11-03 | 2.0 | 0.2 | 1.80 | ▲ warmer |
| 1985-11-04 | 2.7 | 1.4 | 1.30 | ▲ warmer |
| 1985-11-05 | 11.1 | 11.6 | -0.50 | ▼ cooler |
| 1985-11-06 | 6.1 | 5.4 | 0.70 | ▲ warmer |
| 1985-11-07 | 8.0 | 7.7 | 0.30 | ▲ warmer |
| 1985-11-08 | 8.9 | 8.4 | 0.50 | ▲ warmer |
| 1985-11-09 | 13.8 | 13.6 | 0.20 | ▲ warmer |
| 1985-11-10 | 7.4 | 8.2 | -0.80 | ▼ cooler |
| 1985-11-11 | 3.0 | 1.8 | 1.20 | ▲ warmer |
| 1985-11-12 | 2.4 | -0.1 | 2.50 | ▲ warmer |
| 1985-11-13 | 1.5 | 0.4 | 1.10 | ▲ warmer |
| 1985-11-14 | 0.1 | -0.9 | 1.00 | ▲ warmer |
| 1985-11-15 | -2.7 | -2.4 | -0.30 | ▼ cooler |
| 1985-11-16 | -0.4 | -0.9 | 0.50 | ▲ warmer |
| 1985-11-17 | 0.3 | -0.3 | 0.60 | ▲ warmer |
| 1985-11-18 | -0.7 | -1.5 | 0.80 | ▲ warmer |
| 1985-11-19 | -1.1 | -1.9 | 0.80 | ▲ warmer |
| 1985-11-20 | -1.7 | -2.1 | 0.40 | ▲ warmer |
| 1985-11-21 | -1.1 | -1.9 | 0.80 | ▲ warmer |
| 1985-11-22 | -2.2 | -2.9 | 0.70 | ▲ warmer |
| 1985-11-23 | -0.6 | -1.6 | 1.00 | ▲ warmer |
| 1985-11-24 | -0.4 | -1.7 | 1.30 | ▲ warmer |
| 1985-11-25 | 0.9 | 0.2 | 0.70 | ▲ warmer |
| 1985-11-26 | -2.0 | -3.1 | 1.10 | ▲ warmer |
| 1985-11-27 | -1.6 | -2.2 | 0.60 | ▲ warmer |
| 1985-11-28 | 1.1 | -0.1 | 1.20 | ▲ warmer |
| 1985-11-29 | 0.6 | -0.2 | 0.80 | ▲ warmer |
| 1985-11-30 | 3.2 | 2.5 | 0.70 | ▲ warmer |
| 1985-12-01 | 5.5 | 4.0 | 1.50 | ▲ warmer |
| 1985-12-02 | 6.5 | 5.4 | 1.10 | ▲ warmer |
| 1985-12-03 | 10.6 | 10.1 | 0.50 | ▲ warmer |
| 1985-12-04 | 10.3 | 9.8 | 0.50 | ▲ warmer |
| 1985-12-05 | 10.3 | 10.6 | -0.30 | ▼ cooler |
| 1985-12-06 | 9.0 | 7.4 | 1.60 | ▲ warmer |
| 1985-12-07 | 9.2 | 8.6 | 0.60 | ▲ warmer |
| 1985-12-08 | 8.8 | 8.1 | 0.70 | ▲ warmer |
| 1985-12-09 | 8.3 | 7.6 | 0.70 | ▲ warmer |
| 1985-12-10 | 5.3 | 4.0 | 1.30 | ▲ warmer |
| 1985-12-11 | 4.3 | 3.3 | 1.00 | ▲ warmer |
| 1985-12-12 | 1.4 | 0.4 | 1.00 | ▲ warmer |
| 1985-12-13 | 0.0 | -1.2 | 1.20 | ▲ warmer |
| 1985-12-14 | 3.9 | 3.4 | 0.50 | ▲ warmer |
| 1985-12-15 | 9.0 | 8.8 | 0.20 | ▲ warmer |
| 1985-12-16 | 9.6 | 9.1 | 0.50 | ▲ warmer |
| 1985-12-17 | 8.7 | 7.6 | 1.10 | ▲ warmer |
| 1985-12-18 | 6.7 | 6.0 | 0.70 | ▲ warmer |
| 1985-12-19 | 6.9 | 5.1 | 1.80 | ▲ warmer |
| 1985-12-20 | 7.7 | 6.7 | 1.00 | ▲ warmer |
Page 1 of 292
05
Findings
What this does and does not conclude
Westend runs measurably warmer
The full-period 95% CI on the mean gap excludes zero, and so does every season's CI. This project states that plainly, because the data support it.
No significant trend over ~40 years
The preregistered OLS interval includes zero (p = 0.118), and a serial-correlation- robust sensitivity does too (p = 0.186). Reported as "not detected at this record length," not reworded to imply a trend either way.
Not "the" Frankfurt UHI
This is one urban station vs. DWD's own airport-based reference. A different reference, a different variable (daily minimum), or a different district would very plausibly show a different number.
06
Scientific method
How this was measured
Inner join on calendar date
Only days where both stations report a non-missing TMK (DWD's daily mean temperature column) are included. DWD's missing-value sentinel (-999) is dropped, never imputed.
Block bootstrap, not i.i.d.
30-day contiguous blocks, resampled with replacement, 10,000 resamples, seeded (20260813) for exact reproducibility — because consecutive daily temperatures are strongly autocorrelated.
OLS on annual means, years with ≥300 valid days
A simple, preregistered linear model — not a search over more flexible curves after seeing the result. Classical t-distribution CI on the slope.
Primary quantity
gap = TMK(Westend, 01424) − TMK(Frankfurt/Main, 01420)trend: annual_mean(gap) ~ β₀ + β₁ · yearFull method, exclusion criteria, and station-selection rationale: docs/research-protocol.md.
07
Data & citations
Provenance
Data
DWD CDC Open Data — daily climate summary (Tageswerte KL)
Deutscher Wetterdienst Climate Data Center, historical/ archives for stations 01424 and 01420.
Accessed 2026-08-13. Pinned snapshot, not a live feed — see the protocol.
PairingDWD Urban and Regional Climatology — station pairing table
DWD's own designated urban/rural pair for Frankfurt: 01424 (Westend) & 01420 (Frankfurt/Main).
Verified by fetching raw HTML directly, not only an AI summary.
LicenseCC BY 4.0 — DWD CDC-OpenData Terms of Use
"The Creative Commons BY 4.0 - Licence 'CC BY 4.0' apply." (status: Mai 2024)
Quoted directly from the fetched PDF.
Provenance recorddata/provenance.json
Full station metadata, source URLs, license text, and access date.
Committed to the repository.