Mentor – a performance support system for forecasters Chris Ryan 1, Ankur Jha 2 and Swarup Joshi 2 1 Commonwealth Bureau of Meteorology, Melbourne, Australia 2 RMIT University, Melbourne, Australia
Goals Record some of the knowledge of experienced forecasters Provide data and statistics on subjective facets of the forecasting process Provide real-time assistance for forecasters Assist training of forecasters
Technical Details Modelled on the “Coach” system at Tulsa WFO Web forms using Java Server Pages (JSP) and Javascript Oracle database Thin client architecture (application runs on central server with distributed output to clients) Each day’s entry is represented by one row in the database relation.
Using Mentor – Initial Data Entry Entered once per forecasting shift (2-3/day) Entered by shift representative or supervisor Automatically determines current “Synoptic Type” and calculates relevant statistics Entries are requested on: Expected weather Mesoscale features affecting the forecast Possible hazardous weather Problems of the day Confidence in the NWP guidance Degree of model dispersion Level of difficulty of the forecast
Using Mentor - Follow-up Data Entry Entered once per shift (2-3/day) Entered by shift representative or supervisor Entries record the performance of the forecasts made the previous day Entries are requested on the same topics as for “Initial Data Entry”, plus: Perceived overall forecasting performance Level of surprise at the observed weather
Using Mentor - Search The Mentor database can be searched for situations similar to the current one by: Examining the most recent entries Specifying a date Specifying values of any recorded parameters such as NWP confidence, Synoptic Type, Problems of the Day, etc.
Plans Operational trials in first half of 2003 Improve interface Routine operational use in 7 RFCs from mid Analysis of database when sufficiently populated (1 year)
Issues Fitting another job into the forecaster’s schedule Efficient user interface Perceived value of subjective data