PhD project: Data driven models of individual trees to better understand forest water use

In cooperation with the Friedrich Schiller University Jena, the Max Planck Institute for Biogeochemistry houses a unique and flexible research program that grants German and foreign students a broad selection of learning opportunities while still maintaining a research focus.
The International Max Planck Research School for Global Biogeochemical Cycles (IMPRS-gBGC) offers a PhD program specializing in global biogeochemistry and related Earth system sciences.

The PhD program
Successful applicants will be part of the IMPRS-gBGC. Research focuses on the distribution of elements essential to life and the climate among the components of the Earth system and the interaction between biosphere, atmosphere, lithosphere, land and oceans.

Besides doing research for their PhD project, PhD candidates also benefit from a three-month external research visit, specialised courses in e.g. statistics, Earth observation, modelling and analytical techniques, as well as in soft skills. The IMPRS-gBGC is thus an excellent starting platform for a successful career in a field related to global biogeochemical cycles and Earth system science.

Project description

How trees use water is a vital part of the earth system, with links to everything from the amount of carbon forests are capable of storing to regional and global water cycles. Indeed, water taken up via roots and transpired from leaves may precipitate and feed water sheds and rivers thousands of kilometers away. Tree water use is a vitally important process to understand, particularly in regards to how the earth’s forests can cope under a changing climate.

Current modelling approaches today typically use generalizations of forest types (e.g. evergreen needleleaf forest vs broadleaf deciduous forests) and/or structure (e.g. leaf area index) to model the ecosystem as a whole. However, new datasets such as SAPFLUXNET (https://sapfluxnet.creaf.cat ) which contains measurements of tree water use from over 2714 plants of 174 species from around the world, combined with novel machine learning methods such as deep neural networks which incorporate temporal information, gives the opportunity for new types of models which focus on individual trees. Such models have the potential to give a new and data driven perspective on how trees and forests function across different climactic conditions, as well as to be used as a tool for understanding the potential effects of future land use and forest changes on global water and carbon cycles. Furthermore, the project will explore linkages to other datasets, such as forest inventory data and remote sensing products.

Key Objectives

- Combine existing datasets on sap flow, tree information, meteorology, and remote sensing
- Develop a model structure for predicting tree sap flow across species and ecosystems
- Evaluate model performance both on a individual tree level as well as for ecosystems
- Link ensemble of tree models to external datasets such as from forest inventory and/or remote sensing

Working group & planned collaboration

The candidate will be based in Department of Biogeochemical Integration as a collaboration between the Biosphere-Atmosphere Interactions and Experimentation (BAIE) and the Global Diagnostic Modelling (GDM) groups, working in close collaboration with CREAF in Spain and the Terrestrial Ecohydrology group at Friedrich Schiller University Jena, as well as collaborating across international measurement networks such as SAPFLUXNET and FLUXNET. The project will also be associated with the Las Majadas experimental site in Spain (bgc-jena.mpg.de/freiland/index.php/Sites/SpainMajadas?userlang=en)

Requirements for the PhD project are

Applications are open to highly motivated and independent students from any country who:

- a Master's degree in geosciences, atmospheric science, environmental science, bio(geo)chemistry, physics, computer science, mathematics or other disciplines related to environmental sciences.
- experience in data science and machine learning methods with motivation to learn ecophysiology --or-- experience with forest ecophysiology with motivation to learn data science.
- basic programming skills (Python, Julia, Matlab, and/or R)
- excellent oral and written communication skills in English

Further Reading

- SAPFLUXNET data paper: essd.copernicus.org/articles/13/2607/2021/essd-13-2607-2021.html
- Ecosystem scale transpiration estimates: doi.org/10.1111/gcb.15314
- Predicting carbon land-atmosphere fluxes with an LSTM: journals.plos.org/plosone/article?id=10.1371/journal.pone.0211510
- Introduction to recurrent neural networks: colah.github.io/posts/2015-08-Understanding-LSTMs/
- Everything about deep learning: journals.plos.org/plosone/article?id=10.1371/journal.pone.0211510

The Max Planck Society (MPS) strives for gender equality and diversity. The MPS seeks to increase the number of women in those areas where they are underrepresented and therefore explicitly encourages women to apply. The MPS is committed to increasing the number of individuals with disabilities in its workforce and therefore encourages applications from such qualified individuals.

More information about the IMPRS-gBGC + application:
https://www.imprs-gbgc.de/index.php/Application/Main