How Accurate Are the Official Forecasts?
How far official forecasts for 12 gauges missed what was observed, by lead time, against assuming no change, from our archive of every issue.
This describes forecasts already issued; it says nothing certain about the next one. Informational only, not for navigation. Read the notice.
Key Facts
- 12 gauges have enough forecast issues to judge (at least 10), from 2026-09-22 to 2026-11-20.
- At 38 of 42 gauge and lead time pairs shown, the official forecast was clearly more accurate than assuming no change; where it was not, the table says so.
- Kaub: 52 forecast issues from 2026-09-22 to 2026-11-20 (BfG forecasts (Germany), via PEGELONLINE). At 0 to 1 days ahead the forecast was off by 2.35 cm on average, against 3.77 cm for assuming the level does not change: better.
- Duisburg-Ruhrort: 52 forecast issues from 2026-09-22 to 2026-11-20 (BfG forecasts (Germany), via PEGELONLINE). At 0 to 1 days ahead the forecast was off by 2.70 cm on average, against 5.19 cm for assuming the level does not change: better.
- Cologne: 52 forecast issues from 2026-09-22 to 2026-11-20 (BfG forecasts (Germany), via PEGELONLINE). At 0 to 1 days ahead the forecast was off by 1.88 cm on average, against 4.63 cm for assuming the level does not change: better.
- Emmerich: 52 forecast issues from 2026-09-22 to 2026-11-20 (BfG forecasts (Germany), via PEGELONLINE). At 0 to 1 days ahead the forecast was off by 2.42 cm on average, against 3.88 cm for assuming the level does not change: better.
BfG forecasts (Germany), via PEGELONLINE
Forecast issues up to 20 November 2026| Gauge | Lead time | Points checked | Average error of the forecast | Average error of assuming no change | Bias | Forecast against no change |
|---|---|---|---|---|---|---|
| Kaub 52 issues since 2026-09-22 | 0 to 1 days ahead | 619 | 2.35 cm | 3.77 cm | 0.37 cm | better |
| 1 to 2 days ahead | 607 | 4.29 cm | 9.02 cm | 1.01 cm | better | |
| 2 to 3 days ahead | 48 | 5.04 cm | 10.98 cm | 0.92 cm | better | |
| Duisburg-Ruhrort 52 issues since 2026-09-22 | 0 to 1 days ahead | 619 | 2.70 cm | 5.19 cm | -0.24 cm | better |
| 1 to 2 days ahead | 607 | 5.67 cm | 13.26 cm | 0.50 cm | better | |
| 2 to 3 days ahead | 48 | 5.98 cm | 15.71 cm | 1.60 cm | better | |
| Cologne 52 issues since 2026-09-22 | 0 to 1 days ahead | 619 | 1.88 cm | 4.63 cm | -0.38 cm | better |
| 1 to 2 days ahead | 607 | 3.98 cm | 12.49 cm | 0.44 cm | better | |
| 2 to 3 days ahead | 48 | 5.62 cm | 15.31 cm | 2.42 cm | better | |
| Emmerich 52 issues since 2026-09-22 | 0 to 1 days ahead | 619 | 2.42 cm | 3.88 cm | 0.26 cm | better |
| 1 to 2 days ahead | 607 | 5.42 cm | 11.24 cm | 0.37 cm | better | |
| 2 to 3 days ahead | 48 | 6.27 cm | 13.50 cm | 0.94 cm | better | |
| Pfelling 50 issues since 2026-09-22 | 0 to 1 days ahead | 596 | 5.27 cm | 6.86 cm | -0.19 cm | better |
| 1 to 2 days ahead | 48 | 7.58 cm | 11.52 cm | -3.25 cm | better | |
| Hofkirchen 50 issues since 2026-09-22 | 0 to 1 days ahead | 596 | 2.80 cm | 4.55 cm | 0.41 cm | better |
| 1 to 2 days ahead | 48 | 4.92 cm | 9.40 cm | 0.29 cm | better | |
| Dresden 38 issues since 2026-09-22 | 0 to 1 days ahead | 452 | 2.56 cm | 3.72 cm | -0.42 cm | better |
| 1 to 2 days ahead | 37 | 3.86 cm | 5.14 cm | -1.32 cm | better | |
| Magdeburg 38 issues since 2026-09-22 | 0 to 1 days ahead | 452 | 0.94 cm | 0.71 cm | 0.19 cm | worse |
| 1 to 2 days ahead | 440 | 2.29 cm | 1.56 cm | 0.06 cm | worse | |
| 2 to 3 days ahead | 428 | 2.99 cm | 2.28 cm | -0.61 cm | worse | |
| 3 to 4 days ahead | 33 | 2.97 cm | 2.39 cm | -0.73 cm | worse |
NOAA National Water Prediction Service
Forecast issues up to 20 November 2026| Gauge | Lead time | Points checked | Average error of the forecast | Average error of assuming no change | Bias | Forecast against no change |
|---|---|---|---|---|---|---|
| Memphis 52 issues since 2026-09-22 | 0 to 1 days ahead | 205 | 0.07 ft | 0.40 ft | 0.02 ft | better |
| 1 to 2 days ahead | 201 | 0.21 ft | 1.10 ft | 0.09 ft | better | |
| 2 to 3 days ahead | 197 | 0.33 ft | 1.77 ft | 0.12 ft | better | |
| 3 to 4 days ahead | 193 | 0.46 ft | 2.29 ft | 0.12 ft | better | |
| 4 days or more ahead | 3432 | 2.07 ft | 5.12 ft | -0.91 ft | better | |
| St. Louis 102 issues since 2026-09-22 | 0 to 1 days ahead | 404 | 0.33 ft | 0.65 ft | 0.01 ft | better |
| 1 to 2 days ahead | 396 | 0.54 ft | 1.69 ft | 0.00 ft | better | |
| 2 to 3 days ahead | 388 | 0.64 ft | 2.59 ft | -0.01 ft | better | |
| 3 to 4 days ahead | 380 | 0.78 ft | 3.38 ft | 0.02 ft | better | |
| 4 days or more ahead | 3360 | 2.30 ft | 5.30 ft | -1.88 ft | better | |
| Vicksburg 52 issues since 2026-09-22 | 0 to 1 days ahead | 205 | 0.10 ft | 0.28 ft | -0.01 ft | better |
| 1 to 2 days ahead | 201 | 0.21 ft | 0.81 ft | -0.02 ft | better | |
| 2 to 3 days ahead | 197 | 0.32 ft | 1.28 ft | 0.00 ft | better | |
| 3 to 4 days ahead | 193 | 0.43 ft | 1.71 ft | 0.07 ft | better | |
| 4 days or more ahead | 3432 | 1.76 ft | 4.73 ft | 0.63 ft | better | |
| Baton Rouge 52 issues since 2026-09-22 | 0 to 1 days ahead | 202 | 0.20 ft | 0.25 ft | -0.03 ft | better |
| 1 to 2 days ahead | 198 | 0.35 ft | 0.55 ft | 0.03 ft | better | |
| 2 to 3 days ahead | 193 | 0.47 ft | 0.81 ft | 0.04 ft | better | |
| 3 to 4 days ahead | 190 | 0.56 ft | 1.02 ft | 0.01 ft | better | |
| 4 days or more ahead | 3382 | 1.21 ft | 2.35 ft | -0.74 ft | better |
How This Is Calculated, and What It Cannot Show
The authorities replace their forecast with every new run and keep no public record of the old ones. This site stores every issue. For each stored forecast point whose time has passed, the forecast value is compared with the reading observed closest to that time (within 45 minutes). The average error is the mean of the absolute differences, and the bias is the mean signed difference, so a positive bias means the forecast ran high. The benchmark assumes the level stays where it was when the forecast was issued, using the last reading in the three hours before. The forecast is marked better or worse only when its average error differs from the benchmark's by more than 10 percent.
A lead time is shown only with at least 20 checked points, and a gauge only with at least 10 forecast issues. The archive covers a short period, so it holds one season and a few weeks, and a forecast that did well in a steady period can do worse when the river changes quickly. It is not a statement about how accurate the next forecast will be. The values are of raw readings, and these are our own calculations.
Source: BfG forecast via PEGELONLINE (pegelonline.wsv.de), DL-DE->Zero-2.0. (DL-DE->Zero-2.0 (4 day series only)) Source: NOAA National Water Prediction Service (water.noaa.gov). Gauge data from USACE and USGS. (US government work) Source: WSV and GDWS via PEGELONLINE (pegelonline.wsv.de), DL-DE->Zero-2.0. Raw, unvalidated data. Data as of 2026-11-20 (20 November 2026).