hide
Free keywords:
LIGHT-USE EFFICIENCY; PHOTOCHEMICAL REFLECTANCE INDEX; INDUCED CHLOROPHYLL FLUORESCENCE; SPECTRAL VEGETATION INDEXES; RADIATION USE
EFFICIENCY; BOREAL FOREST; PHOTOSYNTHETIC EFFICIENCY;
TEMPERATURE-DEPENDENCE; FLUX MEASUREMENTS; REMOTE ESTIMATIONAgriculture; Forestry; Meteorology & Atmospheric Sciences; Field spectroscopy; Eddy covariance; GEP; PRI; Passive fluorescence; LUE
model;
Abstract:
This study investigates the possibility of monitoring carbon fixation of a terrestrial ecosystem from high spectral resolution field spectroscopy measurements. Canopy radiance spectra were collected under clear sky conditions using high resolution spectrometers which made it possible to estimate sun-induced chlorophyll fluorescence at the oxygen absorption band O(2)-A located at 760 nm (F(760)) as well as spectral vegetation indices. Spectral observations were collected in a rice field monitored with an eddy covariance (EC) flux tower measuring the net ecosystem exchange (NEE) of the crop.
Estimation of gross ecosystem productivity (GEP) from remotely sensed data was based on the widely used light-use efficiency model (LUE), which states that carbon uptake is a function of the photosynthetically active radiation absorbed by vegetation (APAR) and light-use efficiency (epsilon) which represents the conversion efficiency of energy to fixed carbon.
Hyperspectral data were used to derive both the APAR and the e term. Different versions of the LUE model were formalized and tested with EC fluxes during two growing seasons. We started using a LUE model in which epsilon is held constant and remote sensing data are used to estimate APAR. We then investigated the improvements in GEP modelling provided by the partitioning between photosynthetic (PV) and non-photosynthetic (NPV) components of vegetation in APAR estimation, holding epsilon constant. The use of spectral indices related to APAR(PV) instead of APAR resulted in an improvement in midday GEP estimation of about 50% with respect to the basic LUE model, average root mean square error in cross-validation (RMSE(CV)) of the model class from 9.69 to 4.49 mu mol CO(2) m(-2) s(-1). Afterwards, we tested the use of the apparent fluorescence yield (Fy*(760)) and the scaled Photochemical Reflectance Index (sPRI) to derive epsilon. Modelling epsilon further improved the estimation of GEP up to an RMSE(CV) of 3.81 mu mol CO(2) m(-2) s(-1) using Fy*(760) and the MERIS terrestrial chlorophyll index (MTCI) and 3.67 mu mol CO(2) m(-2) s(-1) using sPRI and F760 to estimate E and APAR(PV), respectively. (C) 2010 Elsevier B.V. All rights reserved.
