Design of forecaster workstations

[ For discussion purposes only. Share freely. ]

Design of forecaster workstations

How should forecasting tools be designed?

Modular design should be used. As technology improves, improved modules could be easily installed and quickly implemented.

The two most important forecasting tool modules are, of course, forecasters and tools. In any forecasting tool design, the requirements of these two parts should considered first: the system should be designed around these two parts.

What kinds of tools would best serve the operational forecaster?

Tools should adapt to help forecasters to work more effectively, not vice versa. They should be designed as "intellectual levers", not as "intellectual siphons".

Forecasters need:

Ergonomic graphical user interfaces, designed to help forecasters to maintain their situational awareness, and

Intelligent nowcasting tools, tools that increase forecaster efficiency or increase short-term, local forecast quality, preferably both.

Design for modular decision support systems for weather forecasting

Note in the design how the two smart parts/modules/agents, the two sources of complementary intelligence, the forecaster and the artificial intelligence (AI) module, share top-level positions for controlling data processing.

Other easy-to-update modules would handle communication of data between these two agents -- data is what the agents act upon -- or simply supply improved data to the system. Any improvements in data (observing technology, NWP, and NWP-based products) would be external to the system and would be readily ingested by the system. The system is designed to enable forecasters and forecasting rules to further improve the data -- "post-process", perform "weather watch", "intervene", nowcast, make better weather forecasts -- efficiently.

Click here for PDF version.

The main idea behind the design is to enable National Meteorological Services to take maximal advantage of modernization opportunities based on improving pieces of technology, by building around the three most fundamental "modules" of a forecaster workstation -- forecasters, forecasting rules, and data -- the modules whose requirements must be satisfied:

Forecasters require ergonomic human-computer interaction, e.g.: minimal clicking, maximal clarity in data visualization, and smart timely advice;
Forecasting rules require a standard format, or "language." They should be clear to developers and forecasters, and thus easy to develop, tune, and maintain. They should operate intelligently on heterogeneous data (e.g., observations, model output, and radar and satellite projections) to help forecasting in the 0-to-24 hour period, especially the early part of that period;
Data requires a standard format. Fix and document the format and let would-be developers use it in innovative ways.

I would appreciate any suggestions, comments, or criticisms. Just send me an e-mail at bjarne.hansen@ec.gc.ca.
Bjarne Hansen

Last updated 31 May 2003.
Opinions expressed here are solely those of the author except where indicated as otherwise.