Towards operational remote sensing of forest carbon balance across Northern Europe
- 1Department of Physical Geography and Ecosystems Analysis, Lund University, Sölvegatan 12, SE-223 62, Lund, Sweden
- 2Mathematical Statistics, Centre for Mathematical Sciences, Lund University, Sölvegatan 18, SE-221 00, Lund, Sweden
- 3Department of Botany, Göteborg University, Box 461, SE-405 30, Gothenburg, Sweden
- 4Max-Planck-Institute for Biogeochemistry, Box 10 01 64, 07701, Jena, Germany
- *now at: Dept. of Geography and Environment, Boston University, 675 Commonwealth Avenue, Boston, MA 02215, USA
Abstract. Monthly averages of ecosystem respiration (ER), gross primary production (GPP) and net ecosystem exchange (NEE) over Scandinavian forest sites were estimated using regression models driven by air temperature (AT), absorbed photosynthetically active radiation (APAR) and vegetation indices. The models were constructed and evaluated using satellite data from Terra/MODIS and measured data collected at seven flux tower sites in northern Europe. Data used for model construction was excluded from the evaluation. Relationships between ground measured variables and the independent variables were investigated.
It was found that the enhanced vegetation index (EVI) at 250 m resolution was highly noisy for the coniferous sites, and hence, 1 km EVI was used for the analysis. Linear relationships between EVI and the biophysical variables were found: correlation coefficients between EVI and GPP, NEE, and AT ranged from 0.90 to 0.79 for the deciduous data, and from 0.85 to 0.67 for the coniferous data. Due to saturation, there were no linear relationships between normalized difference vegetation index (NDVI) and the ground measured parameters found at any site. APAR correlated better with the parameters in question than the vegetation indices. Modeled GPP and ER were in good agreement with measured values, with more than 90% of the variation in measured GPP and ER being explained by the coniferous models. The site-specific respiration rate at 10°C (R10) was needed for describing the ER variation between sites. Even though monthly NEE was modeled with less accuracy than GPP, 61% and 75% (dec. and con., respectively) of the variation in the measured time series was explained by the model. These results are important for moving towards operational remote sensing of forest carbon balance across Northern Europe.