Estimating chlorophyll concentration in conifer needles with hyperspectral data: An assessment at the needle and canopy level

This investigation quantitatively links chlorophyll a+b (chl(a+b)) concentration, a physiological marker of forest health condition, to hyperspectral observations of Jack Pine (Pinus banksiana), a dominant Boreal forest species. Compact Airborne Spectrographic Imager (CASI) observations, in the visible-near infrared domain, were acquired over eight selected Jack Pine sites, near Sudbury, Ontario, between June and September of 2001. Supplementing the airborne campaigns was concurrent on-site collection of foliage samples for laboratory spectral and chemical measurements. The study first connected needle-level optical properties with pigment concentration through the inversion of radiative transfer models, LIBERTY and PROSPECT. Next, a chlorophyll sensitive optical index (R750/R710), was scaled-up using SAILH, a turbid medium canopy model, to estimate total pigment content at the canopy-level. Due to the potential confounding effects of open canopy structure and foliage clumping, the analysis accordingly focused on high spatial resolution CASI imagery (1 m) to visually target tree crowns, while accounting for shadowed areas. Chl(a+b) concentration estimation from airborne spectral data using coupled leaf and canopy models was shown to be feasible with a root mean square error of 5.3 mu g/cm(2), for a pigment range of 25.7 to 45.9 mu g/cm(2). Such predictive algorithms using airborne-level data provide the methodology to be potentially scaled-up to satellite-level hyperspectral platforms for large scale monitoring of vegetation productivity and forest stand condition. (C) 2008 Elsevier Inc. All rights reserved.