基于SPOT5影像多辐射校正水平的植被绿量遥感估算

选用南京市SPOT5图像的灰度值(DN)、星上辐射率(SR)、表观反射率(TOA)和地物反射率(PAC)数据,提取了两种植被指数(VI),即归一化植被指数(NDVI)和比值植被指数(RVI),并与地面实测的绿量(LVV)进行相关分析,建立了165个关系模型.结果表明,LVV与VI呈极显著的相关关系,其相关系数多以相对均质植被高于植被总体,基于灰度值高于常用的地物反射率为主.LVV-VI关系模型的R~2均值以多元线性回归模型最高(0.821),指数模型最低(0.536),而1～3次多项式模型均接近0.7.每种植被样方优选出一个模型,即阔叶林LVV-7.802 RVI_(PAC)-2.455(R~2=0.827,RMSE=0.498);针阔叶混交林LVV=-15.421 RVI_(TOA)+26.971 RVI_(DN)-8.261(R~2=0.918,RMSE=0.356);灌木LVV=-342.591 NDVI_(DN)~3-20.553 NDVI_(DN)~2+14.013 NDVI_(DN)+1.509(R~2=0.764,RMSE=0.689);草地LVV=2.934 RVI_(PAC)+2.147 RVI_(TOA)-3.193(R2=0.903,RMSE=0.464);总体植被LVV=1.789RVI_(PAC)-6.814NDVIs+4.258NDVI_(PAC)+12.854 NDVI_(DN)-0.342(R~2=0.810,RMSE=0.638).这些优选模犁的自变量包括了4种辐射校正水平下提取的两种植被指数,显示基于不同辐射校正水平的植被指数在植被LVV遥感反演中具有一定的应用潜力.

Estimating living vegetation volume from a multiple radiometric correction SPOT 5 Imagery

The images of post atmospheric correction reflectance(PAC), top of atmosphere reflectance(TOA) , satellite radi-ance(SR), and digital number(ZW) of a SPOT5 HRG image of Nanjing were used to derive two vegetation indices(VI), i.e., normalized difference vegetation index(NDVI), and ratio vegetation index(RVI). Between these Vis and living vegetation volume(LW) data which obtained from ground measurement,correlations were analyzed and then 165 relationship models were established. The results showed that LVV was significantly correlated with VI. LVV-VI correlation coefficients of relatively 'pure' vegetation are higher than those of total vegetation, and of digital number(DN) higher than those of post atmospheric correction reflectance(PAC) which is universally used.The average R~2 of multi-variable linear regression LVV-VI models was the highest(0.821),of exponential models the lowest(0.536),and of all polynomial models(linear,quadratic,and cubic) near 0.7.One 'best' model was selected for each of the vegetation quadrats.i.e., broad-leaf forest: LVV=7.802RVI_(PAC) -2.455(R~2= 0.827, RMSE=0.498),broad-conifer leaf mixed forest: LVV=-15.421RVI_(TOA)+26.971RVI_(DN)-8.261(R~2=0.918, RMSE= 0.356),shrub:LVV=-342.591 NDVI_(DN)~3 -20.553NDVI_(DN)~2+ 14.013NDVI_(DN) +1.509(R~2=0.764, RMSE=0.689) .grass:LVV=2.934RVI_(PAC)+2.147RVI_(TOA) -3.193(R~2=0.903, RMSE = 0.464),and total vegetation: LVV=1.789RVI_(PAC)-6.814NDVI_(SR)+ 4.258NDVI_(PAC)+12.854 NDVI_(DN) -0.342(R~2=0.810, RMSE = 0.638) .The independent variables of these selected models include two vegetation indices from 4 radiometric correction lev-els,indicating the potentials of spectral vegetation indices from different radiometric correction levels in LW estimating.