将机器学习和人工智能嵌入到天气和气候科学与服务中.docx

A

AframeworkfordatascienceintheMetOffice|2022-2027

Embeddingmachine

learningandartificialintelligenceinweatherandclimatescienceandservices

ThepurposeofthisdocumentistosettheframeworkfortheMetOfficetoachieveits

ThepurposeofthisdocumentistosettheframeworkfortheMetOfficetoachieveitsgoalof‘toharnessthepowerofdatasciencetopushthefrontiersofweatherandclimatescienceandservices’.Indoingso,itbuildsuponthe‘fusingsimulationswithdatascience’themewithinitsResearchandInnovationStrategy(MetOffice,2022).

July2022

ProducedbytheMetOffice.MetOfficeandtheMetOfficelogoareregisteredtrademarks.

?CrownCopyright2022,MetOffice01933

Citethisdocumentas:

MetOffice(2022).Embeddingmachinelearningandartificialintelligenceinweatherandclimatescienceandservices:aframeworkfordatascienceintheMetOffice(2022-2027).Exeter,UK.

2

Contents

Foreword 4

Executivesummary 5

Context 6

Overviewofdatascienceinweatherandclimatescienceandservices 8

Observations 10

NowcastingSimulation 10

Nowcasting

Analysis 12

Productsandservices 12

Challenges 13

Datasciencevisionandgoal 14

Capability1:Discoveryandattribution 16Capability2:Fusingsimulation

Capability1:Discoveryandattribution 16

Capability2:Fusingsimulationwithdatascience 17

EmulationoftheSOCRATESradiationschemeusingmachinelearning

BeyondtheBlackBox

Pillar1–Capabilities 16

BOXA

What

Whatisdatascience?

BOXB

BOXC

Ocean

Oceantemperatureprobeclassificationproject(XBT)

BOXD

BOXE

EthicsCapability3:Uncertaintyandtrust 19

Ethics

Capability4:Datatodecisions 21

Enablingthecapabilities 22

Pillar2–People 23

Attracting,retaininganddevelopingtalent 23

Fosteringadiverseandrepresentativeworkforce 23

Engagingthenextgenerationofdatascientisttocreateaskillspipeline 23

Pillar3–Partners 25

S