世界银行-利用光学遥感和有限训练数据绘制斯里兰卡田间水稻面积和产量图（英）-2025.8.pdfVIP

d

e

z

i

r

o

h

t

u

A

e

rPolicyResearchWorkingPaper11194

u

s

o

l

c

s

i

D

c

i

l

b

u

P

d

e

z

i

r

o

h

t

u

A

e

r

uField-ScaleRiceAreaandYieldMapping

s

o

l

c

s

iinSriLankawithOpticalRemoteSensing

D

c

i

l

b

uandLimitedTrainingData

P

dMutluÖzdoğan

e

z

i

rSherrieWang

o

h

t

uDevakiGhose

A

e

rEduardoFraga

u

s

o

lAnaFernandes

c

s

iGonzaloVarela

D

c

i

l

b

u

P

d

e

z

i

r

o

h

t

u

A

e

r

u

s

o

l

c

s

i

D

c

i

l

b

u

P

DevelopmentResearchGroup

ProsperityVertical

August2025

PolicyResearchWorkingPaper11194

Abstract

Riceisastaplecropforoverhalftheworld’spopulation,andenvironmentalvariablesandsubsequentlyappliedat

andaccurate,timelyinformationonitsplantedareaandthefieldlevel.Theapproachhasenabledthecreationof

productioniscrucialforfoodsecurityandagriculturaltwodecades(2000–2022)ofreliable,field-scalericearea

policy,particularlyindevelopingnationslikeSriLanka.andyieldestimates,achievingmapaccuraciesbetween70%

However,reliablericemonitoringinregionslikeSriLankaandover90%andyieldestimationswithlessthan20%

facessignificantchallengesduetofrequentcloudcoverandRMSE.Thesehighlygranularresults,whichwerepreviously

thefragmentednatureofsmall-holderfarms.Thisresear