BAMOS Vol 31 Special Issue October 2018 Bulletin Vol 31 Special Issue 01 2018 | Page 10

10
BAMOS
Special Issue
Reflections on 30 years of AMOS
John McBride
The scientific revolution of advances in
forecasting current 5–7 day forecasts have the same level of skill as the
1-day forecasts from 30 years ago.
The changes in our ability, as scientists, to forecast the weather
a week or more ahead are profound. To my mind they represent
the greatest achievement in our science. In 1987 when our
Society began, weather forecasts were considered to give useful
guidance on the time scale of one day. Now, the agricultural
industry, building industry and emergency managers make
practical decisions for the coming week based on the operational
forecasts and outlooks. The general public also makes regular
use of the week-ahead forecast in planning picnics, barbecues
and other social functions. The science behind this revolution
involves, of course, the continuous developments in numerical
weather prediction, data assimilation, and data availability
derived primarily from satellite-based observing systems.
Figure 1 (below) documents the change in skill of maximum
temperature forecasts for Melbourne. It can be seen that the Seasonal forecasting
1987 was a few years before the first seasonal outlooks were
issued by the Bureau of Meteorology. In the early years, seasonal
forecasts were derived from statistical relationships with sea-
surface temperature patterns or ENSO indices as predictors. The
science has advanced such that the operational forecasts here
and over the rest of the globe are now derived from coupled
ocean-atmosphere dynamical models. Essentially seasonal
forecasting is a new science, developed during the 30 years
AMOS has been in existence.
Ensemble prediction systems
To some extent ensemble predictions are the backbone of
modern forecasting. The first operational ensemble forecasts
were issued by ECMWF in 1992.
Figure 1. Evidence of increasing skill in the Melbourne maximum temperature predictions, expressed as a departure from
the seasonal daily average. Skill is defined in terms of the PVOE (Percentage Variance of the Observations Explained)
by the forecasts. The PVOE is calculated over the preceding 12 months, that is, the lines show 12-month running means.
Source: Stern and Davidson, Quarterly Journal of the Royal Meteorological Society, 2015.