Analysing Phenological Patterns in Boreal Forests using MODIS Time-Series-Derived and Eddy-Covariance Flux Data
01 RESEARCH DOCUMENT [PART 1].pdf (1014.Kb)
02 SUPPORTING DOCUMENT [PART 2].pdf (2.519Mb)
MetadataShow full item record
Phenological observations are fundamental to our understanding of how vegetation dynamics change with an altered climate. Remote sensing of vegetation activity from satellite sensors such as the Moderate Resolution Imaging Spectroradiometer (MODIS) offer us the opportunity to observe such change on a global scale. However, observations at such latitudes (61.8⁰N), are often negatively affected by various atmospheric conditions and persistence of snow cover. The use of in-situ eddy-covariance (EC) derived ground data such as net ecosystem exchange (NEE), air and soil temperature was used as primary reference and allow the validation of such satellite data as an effective proxy indicator, specifically in the identification of seasonality metrics such as start of season (SOS), end of season (EOS) and growing season length (GSL). The asymmetric Gaussian function within the TIMESAT software was used to fit the QA-filtered satellite data. Satellite seasonality metrics were derived by using the 1st, 2nd and 3rd derivatives (1D, 2D, 3D) of the Normalised Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI). Results from the study based in the southern Finnish boreal forest of Hyytiälä show that relationships between all satellite-derived SOS and all in-situ¬ SOS metrics were poor (R2< 0.5, RMSE> 26 days) . This was similarly so for EOS (R2< 0.25, RMSE> 16 days) metrics. However, GSL evaluations of 2D NDVI and air temperature showed a strong relationship (R2> 0.7), giving confidence that GSL is increasing from GSL trends over the study period. Although relationships of SOS and EOS are poor, results have shown an important trend observation that the length of growing seasons are increasing.