It’s in the pat­terns

Port Douglas & Mossman Gazette - - NEWS -

THE best monthly rain­fall fore­casts come from com­puter pro­grams able to rep­re­sent com­plex re­la­tion­ships be­tween cli­mate data while ac­quir­ing knowl­edge from many ex­am­ples over time for bet­ter pat­tern de­tec­tion.

That’s ac­cord­ing to CQUniversity re­searchers Dr John Ab­bot (pic­tured left) and Jen­nifer Maro­hasy (pic­tured right), who have con­sid­ered re­la­tion­ships be­tween lagged val­ues for tem­per­a­ture, at­mo­spheric pres­sure and rain­fall as well as cli­mate data.

They have pub­lished their find­ings in the At­mo­spheric Re­search jour­nal.

The au­thors have com­pared re­sults from their ar­ti­fi­cial neu­ral net­works (ANNs) anal­y­sis of the In­ter-decadal Pa­cific Os­cil­la­tion in­dex against govern­ment-based sea­sonal rain­fall­fore­cast­ing pro­grams. This in­dex has never be­fore been used for of­fi­cial sea­sonal fore­casts for Queens­land.

‘‘Fore­casts us­ing the ANN for sites in three ge­o­graph­i­cally dis­tinct re­gions within Queens- land are shown to be su­pe­rior ... com­pared to fore­casts from the Pre­dic­tive Ocean At­mos­phere Model for Aus­tralia (POAMA), which is the gen­eral cir­cu­la­tion model used to pro­duce the of­fi­cial sea­son rain­fall fore­casts,’’ say Ab­bot and Maro­hasy.

They say a ma­jor lim­i­ta­tion of govern­ment fore­casts is they pro­vide no in­for­ma­tion about the mag­ni­tude of the ex­pected de­vi­a­tion from the me­dian rain­fall value within the de­fined fore­cast pe­riod.

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.