Vegetation-Climate Interactions in Arid and Hyper-Arid Climates

  • Iyasu Eibedingil

Student thesis: Master's Thesis


Climate is one of the major drivers of ecosystem dynamics. Its control on vegetation is the result of a series of complex dynamical interactions, often non-linear, and is exerted over a wide range of space and temporal scales. In turn, vegetation is expected to affect climate via alteration of albedo, roughness, water conductivity and atmospheric composition. However, a clear understanding of the complete pathway of the climate-vegetation forcing and feedback loop remains still unclear essentially due to the scarcity of data and the intrinsic complexity of the process. Traditionally the analysis of atmosphere-vegetation interactions is based on linear correlation statistics, under the assumption of a dominant linear component and stationary conditions. However, the possible presence of causal dynamical connections, as well as non-linear couplings and non-stationarity can affect the performance of these tools. This thesis explores an alternative approach to the estimation of the coupling between two main climatic variables (precipitation and temperature) and vegetation in arid and transitional (semi-arid) regions. By using monthly globally gridded precipitation and temperature data (NOAA/ESRL/PSD) and remotely sensed normalized difference vegetation index (NDVI, Global Inventory Modeling and Mapping Studies-GIMMS), we analyze both simple linear connections and causal relationships (via pairwise Granger Causality) between large-scale vegetation dynamics and climate. Also, we investigate the role of seasonality in bio-climatic couplings, through a recently developed entropic measure. While classic correlation analysis seems to detect a significant coupling between vegetation and climate with vegetation leading precipitation and temperature, Granger causality mainly detects the forcing of climate on vegetation, in particular over transitional areas like arid and semi-arid regions. The coupling detected by simple linear correlation could be however not causal and mainly due to the interplay of synchronization/de-synchronization effects between the seasonal cycles of vegetation, precipitation and temperature. Furthermore, we discuss a number of different confounding effects that can emerge in the framework of causality estimation, mainly in connection with the spatial and temporal resolution of the used dataset, and instantaneous couplings.
Date of AwardMay 2014
Original languageAmerican English
SupervisorAnnalisa Molini (Supervisor)


  • Vegetation-Climate Interactions; Ecosystem Dynamics; Hyper-Arid Regions; vegetation.

Cite this