Remote sensing-based estimation of gross primary production in a subalpine grassland

M Rossini,M Migliavacca,C Siniscalco,L Busetto,E Cremonese,S Cogliati,T Julitta,R Colombo,M Meroni,U Morra Di Cella,M Galvagno

doi:10.5194/bg-9-2565-2012

Abstract

Abstract. This study investigates the performances in a terrestrial ecosystem of gross primary production (GPP) estimation of a suite of spectral vegetation indexes (VIs) that can be computed from currently orbiting platforms. Vegetation indexes were computed from near-surface field spectroscopy measurements collected using an automatic system designed for high temporal frequency acquisition of spectral measurements in the visible near-infrared region. Spectral observations were collected for two consecutive years in Italy in a subalpine grassland equipped with an eddy covariance (EC) flux tower that provides continuous measurements of net ecosystem carbon dioxide (CO2) exchange (NEE) and the derived GPP. Different VIs were calculated based on ESA-MERIS and NASA-MODIS spectral bands and correlated with biophysical (Leaf area index, LAI; fraction of photosynthetically active radiation intercepted by green vegetation, fIPARg), biochemical (chlorophyll concentration) and ecophysiological (green light-use efficiency, LUEg) canopy variables. In this study, the normalized difference vegetation index (NDVI) was the index best correlated with LAI and fIPARg (r = 0.90 and 0.95, respectively), the MERIS terrestrial chlorophyll index (MTCI) with leaf chlorophyll content (r = 0.91) and the photochemical reflectance index (PRI551), computed as (R531-R551)/(R531+R551) with LUEg (r = 0.64). Subsequently, these VIs were used to estimate GPP using different modelling solutions based on Monteith's light-use efficiency model describing the GPP as driven by the photosynthetically active radiation absorbed by green vegetation (APARg) and by the efficiency (ε) with which plants use the absorbed radiation to fix carbon via photosynthesis. Results show that GPP can be successfully modelled with a combination of VIs and meteorological data or VIs only. Vegetation indexes designed to be more sensitive to chlorophyll content explained most of the variability in GPP in the ecosystem investigated, characterised by a strong seasonal dynamic of GPP. Accuracy in GPP estimation slightly improves when taking into account high frequency modulations of GPP driven by incident PAR or modelling LUEg with the PRI in model formulation. Similar results were obtained for both measured daily VIs and VIs obtained as 16-day composite time series and then downscaled from the compositing period to daily scale (resampled data). However, the use of resampled data rather than measured daily input data decreases the accuracy of the total GPP estimation on an annual basis.

Highlights

The correlation analysis between vegetation indexes (VIs) and different canopy variables suggested the possibility of using normalized difference vegetation index (NDVI) as an indicator for Leaf area index (LAI) and f IPARg (r = 0.90 and 0.95, respectively), the Medium Resolution Imaging Spectrometer (MERIS) terrestrial chlorophyll index (MTCI) for leaf chlorophyll content (r = 0.91) and the PRI551 for LUEg (r = 0.64);
– the spectral vegetation index MTCI, designed to be more sensitive to chlorophyll content, explained most of the variability in gross primary production (GPP) in the ecosystem investigated, which was characterised by a strong seasonal dynamic of green-up and senescence;
– accuracy in GPP estimation improved when taking into account high frequency modulations of GPP driven by incident photosynthetically active radiation (PAR) (in the form of ln(PAR)) or modelling LUEg with the PRI in model formulation; the model formulation that gave the best results in GPP estimation was based on f APARg and ε;

Summary

Introduction

The availability of simultaneous acquisition of near-surface spectral observations and gas flux measurements quantified with the eddy covariance (EC) technique (Baldocchi et al, 1996) has notably increased in recent years (Sims et al, 2006a; Nakaji et al, 2007, 2008; Hilker et al, 2008a; Cheng et al, 2009; Middleton et al, 2009; Rossini et al, 2010) due to its potential to identify effective links between optical signals and photosynthesis at canopy level (Gamon et al, 2006, 2010). Several research groups have developed different automatic devices to collect canopy spectral properties (Leuning et al, 2006; Hilker et al, 2007; Nakaji et al, 2007, 2008; Daumard et al, 2010; Hilker et al, 2010; Ide et al, 2010; Balzarolo et al, 2011; Meroni et al, 2011) for the purpose of gaining new insights into the quantification and monitoring of plant photosynthesis on a temporal scale Such devices are generally operated automatically for long periods in the sampling area of flux towers. One of the most widely applied approaches to modelling gross primary production (GPP) based on remote sensing (RS) data is the light-use efficiency (LUE) model proposed by Monteith (1972, 1977), in which GPP is modelled as a function of the incident photosynthetically active radiation absorbed by vegetation (APAR), determined as the product of the fraction of photosynthetically active radiation absorbed by vegetation (f APAR) and the incident photosynthetically active radiation (PAR), and the conversion efficiency of absorbed energy to fixed carbon (light-use efficiency, ε).

Methods

Discussion

Conclusion