Zambretti Forecast Project

The aim of this project was to develop a web-based version of the Zambretti Forecaster, similar to what appears on many weather station consoles and in some weather station software. The real-time weather data would be supplied in an XML document and then processed and displayed by XSLT.

The original Zambretti Forecaster was developed in 1915 by precision instrument makers Negretti and Zambra of London. It was a hand-held disc calculator, specifically designed for the UK, that provided on-demand short-term forecasts of up to 12 hours. It is reported that the forecasts are most accurate when generated at 9.00 am local solar time.

The calculator uses the mean sea level pressure (MSLP) and pressure trend (steady, rising or falling) as the primary input. The season (Summer or Winter) and the wind regime both play a minor role in determining the forecast.

Fig. 1. The original Zambretti Forecaster.
(left) front: overlaid discs for applying the prevailing weather conditions and reading the forecast letter
(right) rear: forecast text corresponding to the forecast letters displayed in the cut-outs in the top front disc

The Meteor2000D Implementation Guide (hereafter referred to as "the guide") is a valuable resource that describes the mechanics of the Zambretti Forecaster. The guide also outlines the relationship between the Forecast Code (described as "forecast letter" in the guide), Forecast Index ("forecast number" in the guide) and the Forecast Text. These relationships are presented in the following table, together with the modified forecast text and forecast icons used in our implementation of the Zambretti Forecaster.

Top of page

Independent Review

The guide includes details about the Forecast Codes for a range of pressures under varying pressure trends, seasons and wind regimes. We independently reviewed the published results using the methodology described in points 1 - 5 below. Our conclusions are presented further down the page.

1. Tabulate the Forecast Codes and Forecast Index by pressure, wind and season based on whether the pressure is steady (Fig. 2), rising (Fig. 3) or falling (Fig. 4).
2. Compute the seasonal differences in the Forecast Index by pressure and wind regime for a rising pressure (Fig. 5) and a falling pressure (Fig. 6).
3. Determine whether the seasonal difference is influenced by, or independent of the pressure trend.
4. Compute the effect of each wind regime on the Forecast Index for a steady pressure (Fig. 7), a rising pressure (Fig. 8) and a falling pressure (Fig. 9).
5. Determine whether the effect of each wind regime is influenced by, or independent of the pressure trend (Fig. 10).

Fig. 2. The Forecast Code and Forecast Index by MSLP, wind regime and season for a steady pressure.

Fig. 3. The Forecast Code and Forecast Index by MSLP, wind regime and season for a rising pressure.

Fig. 4. The Forecast Code and Forecast Index by MSLP, wind regime and season for a falling pressure.

Fig. 5. Effect of season on the Forecast Index for a rising pressure.

Fig. 6. Effect of season on the Forecast Index for a falling pressure.

Fig. 7. Effect of wind on the Forecast Index for a steady pressure.

Fig. 8. Effect of wind on the Forecast Index for a rising pressure.

Fig. 9. Effect of wind on the Forecast Index for a falling pressure.

Fig. 10. Average effect of the wind on the Forecast Index, from Figs. 7, 8 & 9.

Top of page

Conclusions

The following statements are made on the basis of our review:

• The Forecast Index is specific to a pressure trend.
• A lower Forecast Index corresponds to an improved forecast.
• An adjustment by a negative value (index units) will lower the Forecast Index.
• A rising pressure in Summer (Fig. 5) will produce an average seasonal adjustment of -0.4 index units.
• A falling pressure in Winter (Fig. 6) will produce an average seasonal adjustment of -0.5 index units.
• The impact of the wind (Figs. 7, 8 & 9) varies according to the wind regime.
• The impact of the wind (Fig. 10) is only marginally affected by the pressure trend, and an average wind adjustment can therefore be applied.
• The average wind adjustment (Fig. 10) varies between 0.5 index units for a southerly wind and -1.4 index units for a northerly wind.

Development & Testing

Three different versions of the Zambretti Forecaster set of algorithms were found, including those in the guide. The algorithms in the guide did not produce results that matched the published results, however further testing identified that each algorithm needed a small correction to mitigate the differences. The two other versions required a translation of the programming code from the native python and javascript languages to XSL. The three XSL adaptations of the Zambretti Forecaster that were independently developed comprised: (1) Meteor2000D version (codename ML), (2) Raspberry Pi version (RPi), and (3) Javascript version (JS).

The evaluation of the three XSL adaptations comprised two daily recordings of the Forecast Codes coinciding with the morning and evening manual observations over a thirty day period. The recorded values were assessed on a scale from 0 (very poor) to 3 (very good) against the actual weather during the forecast period to determine the best performing version. Out of interest, the evaluation also included the stand-alone javascript version (BJ).

The results of the evaluation are presented below in the evaluation summary table and detailed evaluation results (Fig. 11). The evening forecast in each case proved to be more accurate than the morning forecast, although the difference was significantly less for the ML and RPi versions. Overall the ML and RPi versions were the best performing versions, with the former rated marginally more accurate.

Project Status

Based on the results of the evaluation, the ML version was selected to provide the Zambretti Forecast on this website.

This 2020 project successfully achieved its aims and is now closed.

Top of page

---

Website developed and maintained by Truganina Weather