Estimation of absorbed PAR across Scandinavia from satellite measurements. Part II: Modeling and evaluating the fractional absorption

✍ Scribed by Pontus Olofsson; Lars Eklundh

• Publisher: Elsevier Science
• Year: 2007
• Tongue: English
• Weight: 832 KB
• Volume: 110
• Category: Article
• ISSN: 0034-4257
• DOI: 10.1016/j.rse.2007.02.020

✦ Synopsis

The aim of this study is to generate a quality-controlled sub-kilometer dataset of the fraction of absorbed photosynthetically active radiation (FAPAR) across Scandinavia from satellite. FAPAR is required for estimating the amount of PAR absorbed (APAR) by vegetation, which in turn allows for estimation of carbon uptake. In this study, FAPAR was modeled from normalized difference vegetation index (NDVI) which was obtained from the MODIS VI product (MOD13Q1) at 250 m spatial resolution. Modeled FAPAR was evaluated against in-situ measurements of fractional interception of PAR (FIPAR) and FAPAR at nine plots in six forested sites across Sweden and Denmark from 2001 to 2005. High resolution remote sensing data were used to investigate the representativeness of the measurement areas. Furthermore, FAPAR from the MODIS LAI/FPAR product at 1 km spatial resolution (MOD15A2) was investigated and compared the measured and modeled FAPAR. There was good agreement between modeled and measured FAPAR (6.9% average RMSE of the means). A linear relationship between daily values of NDVI and FAPAR was found (R 2 = 0.82), and it is concluded that seasonally adjusted NDVI can be used for accurate FAPAR estimations over forested areas in Scandinavia. However, it was found that the error was correlated with average FAPAR and that it is important to take the understory vegetation into account when measuring FAPAR in open canopies. The observed difference between FIPAR and FAPAR was 2.3 and 1.4 percentage units for coniferous and deciduous stands, respectively. MODIS FAPAR performed well although a few unrealistic values were present, highlighting the necessity to filter out low quality values using the quality-control datasets.