Prediction of eucalypt foliage nitrogen content from satellite-derived hyperspectral data

Hyperspectral remote sensing methods are advancing rapidly and offer the promise of estimation of pigment, biochemical, and water content dynamics. The recent Earth Observer 1 (EO-1) Hyperion mission, and associated field campaigns, has allowed a range of biophysical and biochemistry attributes of eucalypt foliage to be analyzed in conjunction with remotely sensed spectra. This paper reports on a study at Tumbarumba (Bago-Maragle State Forest), Australia, which has a wide variety of eucalypt species, ranging in productivity and age. EO-1 Hyperion imagery was obtained in April 2001, and a field program was undertaken involving the establishment of plots, collection of standard forestry inventory data, and green leaf samples. Leaf nitrogen (N) content was measured from leaf samples using wet chemistry techniques and canopy N concentration estimated using leaf mass and proportional species leaf area index data. A number of models were developed from Hyperion reflectance, absorbance, and derivate transformations using partial least squares regression and multiple linear regression. The most significant calibration model predicted N with a correlation coefficient (r)=0.9 (82% variance explained) and a validation r²=0.62 (P<0.01). The standard error of the estimate of foliar N was 0.16% equating to 13% of the mean observed %N at the site. These initial results indicate that predictions of canopy foliar N using Hyperion spectra is possible for native multispecies eucalypt forest. Similar studies worldwide, particular those associated with the flux tower network, will allow these findings to be placed in context with other biomes and functional types.