I use and analyze satellite observations to study the interactions between land ecosystems, the global carbon cycle, and the Earth system. I use satellite observations with global vegetation models, time series analysis methods, and machine learning approaches to understand and predict changes in vegetation, ecosystem carbon fluxes and stocks, and ecosystem disturbances like fires.

Research questions

  • Phenology: How did the seasonal development of vegetation change in the last decades? What are the impacts on the global carbon cycle?
  • Vegetation dynamics: How did the structure and composition of ecosystems change in the last decades?
  • Fire: What are the controls on recent changes in vegetation fires?
  • Data integration: How can optical and microwave satellite observations be combined to better characterize vegetation changes?
  • Predictive Earth observation: How can satellite observations be used to predict ecosystem states and processes?
  • Model evaluation and development: How can satellite data be used to improve dynamic global vegetation models?

Picture: A typical working screen - analyzing relations between climate and fires from satellite data and global models.

Global Vegetation Change

The greening Earth. The maps shows changes in the greenness of land vegetation between 1982 and 2011 as observed by satellites.

Observations from satellites shown an increasing cover of green vegetation ("greening") during the last ~30 years across wide areas of the global land (Figure). My research focus on mapping such vegetation changes (Forkel et al. 2013) and on identifiying the climatic and socioeconomic drivers behind such changes in global vegetation (Forkel et al. 2015). Furthermore I assess how these changes in vegetation affect the global carbon cycle and hence the climate system (Forkel et al. 2016). Recently, we found that warmer springs result in higher vegetation productivity in spring but in lower productivity in the subsequent summer and autumn (Buermann et al. 2018).

Global Carbon Cycle

Atmospheric CO2 concentration as measured at Point Barrow, Alaska.

Ecosystems take up carbon dioxide from the atmosphere through photosynthesis and release it again through respiration. Based on several datasets, we were able to estimate how long carbon dioxide resides in average in land ecosystems (Carvalhais et al. 2014). Another puzzling question about the global carbon cycle was why the seasonal amplitude of CO2 in the atmosphere is increasing in the last 40 years (Figure). By combining observations with an improved global vegetation model we found that this increase can be explained by an amplification of plant productivity in northern ecosystems (Forkel et al. 2016). Our current work focus on estimating photosynthesis from microwave satellite data (Teubner et al. 2018).

Land Surface Phenology

Seasonal cycle of NDVI with estimated phenology and greenness metrics.

Phenology is the study of the timing of biological changes within a year. In my research, I develop methods to identify phenological metrics like the start and end of the growing season (Figure) from satellite (Forkel et al. 2015) and ground-based vegetation observations (Filippa et al. 2016). Furthermore, I improved the representation of phenology in a global vegetation model (Forkel et al. 2014) which helped to quantify the relative contributions of light, temperature, and water availability on changes in phenology (Forkel et al. 2015). Phenological changes also affect the carbon cycle. Up to know, it was assumed that warmer springs result in higher plant productivity but recently we found that warmer springs results in lower productivity in the subsequent summer and autumn in many northern ecosystems (Buermann et al. 2018).

Wildfires, Extremes, and Mortality

Mean annual burned area in the period 1996-2015 (GFED4 dataset)

Extreme events such as drought or heatwaves can cause a reduced photosynthesis and a release of carbon dioxide from ecosystems to the atmosphere. For example, we found that extreme climate events cause increasing ecosystem productivity in spring but decreasing productivity in summer in Europe (Sippel et al. 2017). Such drought events increase also the mortality in temperate forests while frost events are important in boreal forests (Thurner et al. 2016). Droughts and heatwaves support also the occurrence of wildfires: The Figure shows the average burned fraction per year. Currently, my research focus on the factors that control the occurence and spread of wildfires and how fires can be modelled from satellite data (Forkel et al. 2017).

Model-data integration

Dimensions of global ecosystem models

Ecosystem models or vegetation models are necessary to quantify and predict the carbon and water balances of ecosystems or to estimate how vegetation will change under future climate change. For these purposes different kinds of ecosystem models exist that represent different processes, that are used in diagnostic or prognostic modes, or that are based on empirical or physical descriptions of processes (Figure). I contributed to the development of the LPJmL dynamic global vegetation model (Schaphoff et al. 2018a). The application of ecosystem models requires a careful evaluation against observations (e.g. Forkel et al. 2014, 2015, 2016, Schaphoff et al. 2018b). Moreover, I also use satellite datasets to improve global vegetation models (Forkel et al. 2014) or to develop new model approaches (Forkel et al. 2017).

Time series analysis

Detecting trends and breakpoints in time series

Earth observation data or ecosystem model results cover often several years. This allows to analyze long-term changes or trends. When I started analyzing vegetation trends from satellite data, we had in our research group a discussion about what would be the best approach to do this. This discussion resulted in a publication where we compare and test different methods to detect trend changes (Forkel et al. 2013). Later, I also compared several methods to detect vegetation phenology from satellite data (Forkel et al. 2015) and contributed to the development of methods to analyze phenology from PhenoCams (Filippa et al. 2016).