Phenological characterization of Fagus sylvatica L. in Mediterranean populations of the Spanish Central Range with Landsat OLI/ETM+ and Sentinel-2A/B

Abstract

<p>The Spanish Central Range hosts some of the southernmost populations of <em>Fagus sylvatica</em> L. (European beech). Recent cartography indicates that these populations are expanding, going up-streams and gaining ground to oak forests of <em>Quercus pyrenaica </em>Willd., heather-lands, and pine plantations. Understanding the spectral phenology of European beech populations—which leaf flush occurs earlier than other vegetation formations—in this Mediterranean mountain range will provide insights of the species recent dynamics, and will enable modelling its performance under future climate oscillations. Intra-annual series of 211 Landsat OLI/ETM+ images, acquired between April 2013-December 2019, and 217 Sentinel-2A/B images, acquired between April 2017-December 2019, were employed to characterize the spectral phenology of European beech populations and five other vegetation types for comparison in an area of 108000 ha. Vegetation indices (VI) including the Normalized Difference Vegetation Index (NDVI) and Tasseled Cap Angle (TCA) from Landsat, and the NDVI and Enhanced Vegetation Index (EVI) from Sentinel-2 were retrieved from sample pixels. The temporal series of these VI were modelled with Savitzky-Golay and double logistic functions, and assessed with TIMESAT software, enabling the parametric characterization of European beech spectral phenology in the area with the start, length, and end of season, as well as peak time and value. The length of beech phenological season was similar when portrayed by Landsat and Sentinel-2 NDVI time series (214 and 211 days on average for the common period 2017-2019) although start and end differed. Compared with NDVI counterparts the TCA season started and peaked later, and the EVI season was shorter. Sentinel-2 NDVI peaked higher than Landsat NDVI. The European beech had an earlier (21 days on average) start of season than competing oak forests. Joint analysis of data from the virtual constellation Landsat/ Sentinel-2 and calibration with field observations may enable more detailed knowledge of phenological traits at the landscape scale.<em></em></p>