Leaf Level Chlorophyll Fluorescence Emission Spectra: Narrow Band versus Full 650-800 nm Retrievals
Abstract
Recently, chlorophyll fluorescence (ChlF) retrievals in narrow spectral regions (< 1 nm, between 750-770 nm) of the near infrared (NIR) region of Earth's reflected radiation have been achieved from satellites, including the Japanese GOSAT and the European Space Agency's Sciamachy/Envisat. However, these retrievals sample the total full-spectrum ChlF and are made at non-optimal wavelengths since they are not located at the peak fluorescence emission features. We wish to estimate the total full-spectrum ChlF based on emissions obtained at selected wavelengths. For this, we drew upon leaf emission spectra measured on corn leaves obtained from a USDA experimental cornfield in MD (USA). These emission spectra were determined for the adaxial and abaxial (i.e., top and underside) surfaces of leaves measured throughout the 2008 and 2011 growing seasons (n>400) using a laboratory instrument (Fluorolog-3, Horiba Scientific, USA), recorded in either 1 nm or 5 nm increments with monochromatic excitation wavelengths of either 532 or 420 nm. The total ChlF signal was computed as the area under the continuous spectral emission curves, summing the emission intensities (counts per second) per waveband. The individual narrow (1 or 5 nm) waveband emission intensities were linearly related to full emission values, with variable success across the spectrum. Equations were developed to estimate total ChlF from these individual wavebands. Here, we report the results for the average adaxial/abaxial emissions. Very strong relationships were achieved for the relatively high fluorescence intensities at the red chlorophyll peak, centered at 685 nm (r2= 0.98, RMSE = 5.53 x 107 photons/s) and in the nearby O2-B atmospheric absorption feature centered at 688 nm (r2 = 0.94, RMSE = 4.04 x 107), as well as in the far-red peak centered at 740 nm (r2=0.94, RMSE = 5.98 x107). Very good retrieval success occurred for the O2-A atmospheric absorption feature on the declining NIR shoulder centered at 760 nm (r2 = 0.88, RMSE = 7.54 x 107). When perfect retrievals were assumed (0% noise), retrievals remained good in the low emission regions on either side of the peaks-- those associated with the H alpha line at 655 nm (r2 = 0.83, RMSE =8.87 x 107) and the far-NIR wavelengths recently utilized for satellite retrievals: a K line at 770 nm (r2 = 0.85, RMSE = 8.36 x 107) and the 750-770 nm interval (r2 = 0.88, RMSE = 6.92 x 107). However, the atmosphere and satellite observations are expected to add noise to retrievals. Adding 5% random error to these relationships did not seriously impair the retrieval successes in the red and far-red peaks (r2 ~ 0.85, RMSEs = 6.31 x 107). A greater impact occurred (reducing retrieval success by ~10%) when adding 5% noise for the far-NIR narrow band at 770 nm (r2 ~ 0.70, RMSE ~ 8.5 x 107). When a 10% random error was added, the retrieval successes fell to ~68 ± 7% for all retrieval wavebands, and RMSEs increased by a factor of 10. This laboratory approach will be critical to calibrate space borne retrievals, but additional information across plant species is needed. Furthermore, this experiment indicates that ChlF retrievals from space should include information from the red and far-red peak emission regions, since the true total fluorescence signal is the desired parameter for Earth carbon and energy budgets.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2012
- Bibcode:
- 2012AGUFM.B53G..05M
- Keywords:
-
- 0422 BIOGEOSCIENCES / Bio-optics;
- 0424 BIOGEOSCIENCES / Biosignatures and proxies;
- 0480 BIOGEOSCIENCES / Remote sensing