Global land surface vegetation phenology using 13 years of SPOT VEGETATION daily observations

Vegetation is a main component of the terrestrial biosphere and plays a key role in the global water, energy and carbon cycles. The need to monitor the impact of recent climate changes on vegetation brought back to light an old scientific discipline studying the timing of recurring biological events, Phenology. Phenology is experiencing a renewal over the last 15 years also by extending its traditional local or regional scope to the global scale. Satellite time series of coarse resolution sensors, recording multispectral reflectance on a daily basis for many years, stand today for being the adequate approach to study phenology on the whole land surface of the Earth. While various studies already tackled different questions in this field of land surface phenology, the ambition of this thesis is to define and implement a framework describing the “Climatology” or baseline behavior of the vegetation cycle for any squared kilometer on Earth. The strategy is to capitalize on a daily reflectance values archive of 13 years (1999-2011) acquired by the SPOT VEGETATION sensor to improve the description of the vegetation cycle, both in the spatial and the temporal dimensions. SPOT VEGETATION is selected for its very good geometric performances and two vegetation indices, EVI and NDVI, are chosen as proxies to describe the foliage development. First, the seasonal and inter-annual variations of the phenology are characterized by computing the mean, the median and the standard deviation profiles of the thirteen EVI and NDVI annual time series. At the exception of some tropical areas, a sufficient number of observations were available to capture accurately the vegetation cycle with a 7-day time step along the year. Second, these reference profiles proved to be consistent enough to estimate precisely specific moments of the vegetation cycle. The method used to compute these phenological reference metrics demonstrates its adaptability to deal with the global diversity of vegetation behaviors and its ability to handle multiple seasons patterns. This is successfully demonstrated for two very different environments, Europe and the Congo Basin in Central Africa. Third, the 13-year time series are tested for long term trend. However, the very low magnitude of the EVI and NDVI increase observed consolidated the choice of no trend removal before averaging the data. Last but not least, a new vegetation map consistent for the whole Congo Basin is realized. This map improve the description of the spatial distribution of the vegetation types in this region heavily affected by cloud coverage by taking advantage of several years of MERIS and SPOT VEGETATION data. The different sections of the thesis fit together to form a coherent framework describing the land surface reference phenology of all terrestrial biomes between 75° N and 55° S as observed by the SPOT VEGETATION instrument. Moreover, the thesis delivered two operational products for the international community, a Congo basin vegetation types map and a comprehensive EVI-NDVI phenological reference dataset. While the first one is already adopted by the “Observatoire des Forêts d’Afrique centrale” as reference information, the land surface phenology reference dataset opens very exciting perspectives for further phenological studies and to detect future vegetation cycle anomalies on a near-real time basis.