简介：ThevariabilityofmainchemicalcompositionofradialChinesefirwasstudied.Analysisofvarianceshowedthatvariancewassignificant,especiallyforholocelluloseandα-cellulose;theholocellulosecontentinheartwoodwaslowerthanthatinsapwood;thelignincontentdecreasedfromPosition1toPosition3,butincreasedgraduallyfromPosition4toPosition8;theα-cellulosecontentinheartwoodwaslowerthanthatinsapwood;andtherelativecrystallinitywas59.97%inPosition1,60.80%inPosition2,andabout42%inothers.

简介：Theradialbasisfunction(RBF)emergedasavariantofartificialneuralnetwork.Generalizedregressionneuralnetwork(GRNN)isonetypeofRBF,anditsprincipaladvantagesarethatitcanquicklylearnandrapidlyconvergetotheoptimalregressionsurfacewithlargenumberofdatasets.Hyperspectralreffectance(350to2500nm)datawererecordedattwodifferentricesitesintwoexperimentfieldswithtwocultivars,threenitrogentreatmentsandoneplantdensity(45plantsm-2).Stepwisemultivariableregressionmodel(SMR)andRBFwereusedtocomparetheirpredictabilityfortheleafareaindex(LAI)andgreenleafchlorophylldensity(GLCD)ofricebasedonreffectance(R)anditsthreedifferenttransformations,thefirstderivativereffectance(D1),thesecondderivativereffectance(D2)andthelog-transformedre?ectance(LOG).GRNNbasedonD1wasthebestmodelforthepredictionofriceLAIandGLCD.TherelationshipsbetweendifferenttransformationsofreffectanceandriceparameterscouldbefurtherimprovedwhenRBFwasemployed.Owingtoitsstrongcapacityfornonlinearmappingandgoodrobustness,GRNNcouldmaximizethesensitivitytochlorophyllcontentusingD1.ItisconcludedthatRBFmayprovideausefulexploratoryandpredictivetoolfortheestimationofricebiophysicalparameters.