|
Forest water–carbon coupling changes under rising atmospheric CO₂: evidence from long-term eddy-covariance measurements.
Study programs: Hydrology, Meteorology, HSE, Flood Risk Management
Begin: immediately
Motivation and aim
Rising atmospheric CO₂ can affect forest water use by increasing photosynthesis and reducing stomatal conductance. Several studies have reported increasing forest water-use efficiency under rising CO₂, but it remains less clear whether this signal translates into detectable changes in ecosystem-scale evapotranspiration (ET), especially across seasons and under different subdaily meteorological conditions. Eddy-covariance observations from ICOS and FLUXNET provide long-term measurements of carbon, water and energy fluxes and are therefore well suited to investigate whether expected physiological CO₂-related signals are visible in forest flux observations.
The study will not aim to isolate a purely causal CO₂ effect. Instead, it will assess whether long-term and subdaily changes in ET, GPP, water-use efficiency and canopy-conductance proxies are consistent with the expected physiological response to rising atmospheric CO₂.
If successful, the thesis could serve as a pilot study for a larger PhD project on CO₂-sensitive forest water cycling and its relevance for future ecohydrological change.
Research questions
-
Do selected long-term forest sites show changes in ET, GPP and water-use efficiency over time, and are observed WUE trends driven mainly by increasing GPP or decreasing ET?
-
Do these changes differ between forest types and hydroclimatic conditions, especially during the growing season, dry periods and high-VPD conditions?
-
Is a subdaily signal detectable at the best-documented sites, i.e. has the daytime coupling between ET, GPP, radiation and VPD changed between earlier and more recent periods under comparable meteorological conditions?
Requirements
-
Strong interest in forest–atmosphere coupling processes
-
Solid understanding of meteorology
-
Beginner to intermediate programming skills
-
Willingness to work toward a publication of results
Literature
-
Keenan, T. F. et al. (2013). Increase in forest water-use efficiency as atmospheric carbon dioxide concentrations rise.
-
Guerrieri, R. et al. (2019). Disentangling the role of photosynthesis and stomatal conductance on rising forest water-use efficiency.
-
Knauer, J. et al. (2018). Towards physiologically meaningful water-use efficiency estimates from eddy covariance data.
From Prompt to Water Balance: Can LLMs Simulate Hydrology?
Study programs: Hydrology, Meteorology, HSE, Flood Risk Management
Begin: immediately
Motivation and aim
Large language models (LLMs) are increasingly used to generate scientific code, explain modelling concepts and support data analysis. However, it remains unclear whether they can go beyond coding assistance and produce physically plausible hydrological simulations from site descriptions and meteorological forcing.
This thesis will benchmark whether selected LLMs can act as hydrological modelling agents for two test cases in Saxony: the DE-Tha eddy-covariance site and the Wernersbach catchment. Two modes will be tested: LLMs as model builders generating executable code for water and energy balance simulations, and LLMs as black-box simulators directly returning hydrological time series from the provided input. The outputs will be compared against observations and/or reference simulations from established hydrological models. The study will focus on subdaily ET, soil-water dynamics and energy fluxes at DE-Tha, and on runoff generation and water-balance closure for Wernersbach. Additional aspect will be transferability: whether an LLM-generated workflow developed for one site can be applied to the other site or to a different period without hidden manual adaptation. This is essential to distinguish general hydrological reasoning from site-specific overfitting or plausible-looking but non-transferable output.
If successful, the thesis could serve as a pilot study for a larger PhD project on safe, explainable and benchmarked AI-assisted hydrological modelling, including its potential and limitations for future ecohydrological applications.
Research questions
-
Can LLMs generate physically interpretable hydrological model code from site and forcing data?
-
Can LLMs directly compute plausible hydrological time series in black-box mode?
-
Are LLM-based simulations accurate, physically consistent and transferable across sites, scales and conditions?
Requirements
-
Strong interest in hydrological and ecohydrological modelling
-
Solid understanding of water and energy balance processes
-
Beginner to intermediate programming skills
-
Willingness to work toward a publication
Literature
-
Kizilkaya, D. et al. (2025). Toward HydroLLM: A benchmark dataset for hydrology-specific knowledge assessment for large language models.
-
Foroumandi, E. et al. (2023). ChatGPT in hydrology and Earth sciences: opportunities, prospects, and concerns.
-
Hou, S. et al. (2024). Can Large Language Models Generate Geospatial Code?
Extending subdaily BROOK90 across vegetation layers and lateral flow paths
Study programs: Hydrology, Meteorology, HSE, Flood Risk Management
Begin: immediately
Motivation and aim
Subdaily BROOK90 is currently strongest as a vertical one-dimensional stand-scale model. However, several applications require more flexible structural representations: multi-layer vegetation, improved soil hydraulic functions and simple lateral redistribution of water. This is especially relevant for heterogeneous forest and catchment settings such as DE-Tha and Wernersbach, where overstory, understory, soil hydraulic properties and lateral runoff generation can influence the simulated water balance.
This thesis will develop a modular roadmap and implement selected structural extensions of subdaily BROOK90. The main developments are: a two-layer canopy concept separating overstory and understory/grass vegetation, replacement or optional extension of Clapp-Hornberger soil hydraulic functions by Mualem-van Genuchten parameterization, and a simple conceptual routing module to connect vertical soil water drainage and runoff production to catchment-scale discharge. The modules will be tested using DE-Tha ICOS and Wernersbach data, including flux observations, soil moisture, discharge and site/catchment descriptors.
Research questions
-
Can a two-layer canopy representation improve simulated partitioning of interception, transpiration and soil evaporation in forest and understory systems?
-
How does replacing Clapp-Hornberger soil hydraulic functions with Mualem-van Genuchten parameterization affect soil moisture dynamics?
-
Can a simple routing concept link subdaily BROOK90 stand-scale water-balance outputs to observed catchment discharge in Wernersbach?
Requirements
-
Strong interest in hydrological and ecohydrological modelling
-
Solid understanding of water and energy balance processes
-
Intermediate to advanced programming skills (R)
-
Willingness to work toward a publication
Literature
-
Federer, C. A. et al. (2003). BROOK90: A simulation model for evaporation, soil water, and streamflow.
-
Kronenberg, R. et al. (2025). An extension of the BROOK90 hydrological model for subdaily water and energy balance simulations.
-
van Genuchten, M. T. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils.
Toward a more physical subdaily BROOK90: soil heat, litter storage and transpiration delay
Study programs: Hydrology, Meteorology, HSE, Flood Risk Management
Begin: immediately
Motivation and aim
The subdaily version of BROOK90 enables coupled water and energy balance simulations at high temporal resolution, but several short-term storage and delay processes are still simplified. This is especially relevant for forest sites, where soil heat exchange, litter interception and the timing between root water uptake and canopy transpiration can strongly affect subdaily water and energy fluxes. In the current setup, soil heat flux is often required as an external input, litter storage is not explicitly represented, and transpiration is assumed to occur without delay after root uptake.
This thesis will develop and test selected vertical process modules for subdaily BROOK90. The main modules are: a physically based soil heat flux module driven by soil temperature dynamics, a conceptual litter or forest-floor storage layer parameterized by litter thickness and organic horizon type, and a simple transpiration delay storage representing the time lag between root water uptake and canopy water loss. The modules will be evaluated using observations from forest ICOS stations, including eddy-covariance fluxes, soil temperature, soil moisture, sap flow and litter measurements where available.
If successful, it could serve as a pilot study for a larger PhD project on observation-constrained subdaily forest ecohydrological modelling.
Research questions
-
Can a soil heat flux module reproduce observed subdaily soil heat flux dynamics and reduce the need for prescribed soil heat flux input?
-
Does an explicit litter/forest-floor storage layer improve simulation of interception, evaporation and percolation?
-
Does a simple transpiration delay storage improve subdaily timing of transpiration-related fluxes compared with the current instantaneous big-leaf assumption?
Requirements
-
Strong interest in hydrological and ecohydrological modelling
-
Solid understanding of water and energy balance processes
-
Intermediate to advanced programming skills (R)
-
Willingness to work toward a publication
Literature
-
Federer, C. A. et al. (2003). BROOK90: A simulation model for evaporation, soil water, and streamflow.
-
Kronenberg, R. et al. (2025). An extension of the BROOK90 hydrological model for subdaily water and energy balance simulations.
-
Floriancic, M. G. et al. (2022) Potential for significant precipitation cycling by forest-floor litter and deadwood
Effektive Modellierung der Vegetationsstruktur
(engl.: Effective modeling of vegetation structure)
Studiengänge: Forstwissenschaften, RE-NRM, Geographie, Hydrologie
Beginn: sofort
Zielsetzung:
Ziel ist die Erstellung eines Katalogs mit generischen Modellen zur dreidimensionalen geometrischen Repräsentation von Stadtbäumen in numerischen Modellen. Allometrische Funktionen beschreiben die räumliche Struktur der Vegetation als Funktion von Baumart, -höhe und Brusthöhendurchmesser. In einem ersten Schritt sollen in der Literatur vorhandene Funktionen gesammelt und verglichen werden. Auf der Basis von Aufnahmen mit einem terrestrischen Laserscanner von fünf typische Stadtbaumarten sollen allometrische Funktionen entwickelt oder vorhandene angepasst werden. Weiterhin sollen Möglichkeiten der Einbindung der an der TU Dresden aufgebauten und frei verfügbaren Datenbank für urbane Gehölze CITREE (https://citree.de) geprüft werden.
Arbeitsschritte:
-
Literaturstudium zu allometrische Funktionen
-
Vorortbegehung der fünf Straßenzüge in Dresden
-
Datenauswertung und Funktionsanpassung
-
Analyse der Anwendungsmöglichkeiten
Betreuer:
Bestimmung der Pflanzenoberfläche mit optischen Verfahren
(engl.: Assessment of the plant area determined by optical methods)
Studiengänge: Hydrologie, Geographie, Forstwissensch, RE-NRM
Beginn: sofort
Zielsetzung:
Der Leaf Area Index, LAI, hat als die wesentliche Umsatzfläche eine große Bedeutung für den Energie-, Wasser- und Stoffhaushalt der Erdoberfläche. Die exakte Bestimmung des LAI ist dabei nur durch exemplarische Einzelmessung (in Zeit- und Raum) möglich. Häufiger werden zerstörungsfreie optische Verfahren, das heißt Messungen des Strahlungstransfers durch den Pflanzenbestand eingesetzt. Dabei wird genau genommen der Plant Area Index PAI bestimmt (gesamte Pflanzenoberfläche). Das dabei zugrunde gelegte Lambert-Beersche Gesetz gilt streng genommen nur für eine homogene Verteilung der absorbierenden Substanz, vernachlässigbare Variation des Absorptionskoeffizienten, vernachlässigbare Mehrfachstreuung und Eigenemission sowie niedrig konzentrierte „Lösungen“. Diese Bedingungen sind insbesondere in hohen Waldbeständen kaum erfüllt. In der Arbeit sollen verschiedene Methoden zur Bestimmung des PAI verglichen werden (LAI2000, Strahlungsmessungen, Laserscanning). Weiterhin soll die prinzipielle Anwendbarkeit des Lambert-Beersche Gesetzes untersucht werden und anhand eines künstlichen Bestandes mit einem alternativen Modell verglichen werden.
Arbeitsschritte:
-
Literaturstudium
-
Durchführung eigener LAI2000 Messungen
-
Vergleich der verschiedenen Messverfahren
-
Vergleich des Lambert-Beerschen Gesetzes mit einem alternativen Modell und Bewertung der Ansätze
Betreuer:
Dr. Ronald Queck [ronald.queck@tu-dresden.de]
Strahlungstransfer durch Baumkronen und Parametrisierung der Verdunstungsmodelle von Waldbeständen
(engl.: Radiation transfer within forest canopies and parameterisation of models evapotranspiration
Studiengänge: Hydrologie, Geographie, Forstwissenschaften, RE-NRM
Beginn: sofort
Zielsetzung:
Die Verdunstung des Unterwuchses von Wäldern mit ausgeprägtem Stammraum ist weitgehend durch die Strahlung gegeben. Diese wird in Abhängigkeit von der Kronenstruktur (Bedeckung, Dimension der Krone, Blattfläche) beschrieben. Die besondere Herausforderung liegt dabei in der korrekten Behandlung der raum-zeitlichen Dynamik des Strahlungsangebots, die z.B. zeitlich durch den Blattaustrieb und räumlich durch Lichtflecken gegeben ist. Für die Abbildung der resultierenden Verdunstungsdynamik soll ein vorhandenes Mehrschichtmodell genutzt werden und mit dem Zweischichtmodell Brook90 verglichen werden. Für die Arbeit stehen Daten an den Forststandorten Ankerstation Tharandt (Fichte, seit 1996) und Buchhübel (Buche, seit 2005) aus unterschiedlichen Messhöhen zur Verfügung, die durch Zusatzmessungen während der Arbeit ergänzt werden können.
Arbeitsschritte:
-
Literaturstudium
-
Überprüfung und Ergänzung der vorhandenen Strahlungsmessungen
-
Auswertung und Analyse der Messungen
-
Erhebung der Modellparameter für den Strahlungstransfer durch Baumkronen
-
Modellvergleich
-
Sensibilitätstest und Modellierung der Verdunstungskomponenten
-
Validierung mit unabhängigen Messungen der Verdunstungskomponenten
Betreuer:
Interception and land use change
(engl.: Effective modeling of vegetation structure)
Studiengänge/Study course: Hydrologie, Geographie, Forstwissenschaften, RE-NRM, GroundwatCh
Beginn/Start: immediately
Zielsetzung/Objectives:
Background: Prospective climate change will cause land use changes with implications on ecosystem water budgets. The current widespread conifer monocultures in Germany/Europe are vulnerable to droughts that are predicted to occur more frequently. Recent droughts in 2018/19 caused direct forest damage (mainly in cities) but showed also effect on forest diseases and pests like the bark beetle (mainly in spruce monocultures). Forest conversion to mixed forests could mitigate the problem but affects the water balance too. A major influence is the change of intercepted rain water.
Problem statement: Interception depends on the plant surface distribution and on evaporation, whereas the conditions for evaporation are also influenced by the plant surface distribution. The question is, how changes the interception and hence the water budget with changing forest structure.
Research Objectives: Determination of the change in water budget by the change of a spruce forest to a mixed forest.
Steps/Methods/Pre-requisites:
- Use of high resolved 3D vegetation model of a spruce forest stand within the Tharander Wald, which is gained on terrestrial laser scans.
- Vegetation models of deciduous forests form literature or measurements
- Application of an interception model with different spatial-temporal resolutions
- Programming skills (R)
Betreuer/Supervisor:
Interpolation von Windmessungen in und über Waldbeständen mittels eines massenkonsistenten Windmodels
Studiengänge: Hydrologie, Geographie, Forstwissensch, RE-NRM
Beginn: sofort
Zielsetzung:
In einem inzwischen weltumspannenden Netz von Messstationen werden über unterschiedlichsten Landnutzungen der Energie- und Stoffaustausch bestimmt. Die dabei verwendeten Methoden setzen weit ausgedehnte homogene Landoberflächen/nutzungen voraus. Diese Bedingung wird gerade in den charakteristischen klein-räumigen Landschaften Mitteleuropas nicht erfüllt. Die Folge davon sind advektive Stoff- und Energieflüsse.
Zur Bestimmung dieser advektiven Stoffflüsse ist die Kenntnis der horizontalen Windgeschwindigkeitsverteilung notwendig. Die Aufgabe der besteht darin, Messungen der Windgeschwindigkeit in und über einem Waldbestand mittels des massenkonsistenten Windmodells MCF zu interpolieren und mit Messungen zu vergleichen.
Arbeitsschritte:
- Literaturstudium und Einarbeitung in das MCF
- Anwendung des MCF auf die weitere Umgebung der Ankerstation Tharandter Wald ASTW (2-5 km²) und Vergleich mit bereits vorhandenen Modellierungen
- Mitarbeit bei der Entwicklung einer Parametrisierung des MCF für Strömungen in Waldbeständen
- Anwendung des MCF auf die Strömungen in Waldbeständen in unmittelbare Umgebung der ASTW
- Validierung der Modellergebnisse auf der Basis von vorhandenen Windmessungen
Betreuer:
Dr. Ronald Queck [ronald.queck@tu-dresden.de]
Comparison of different measurement and model-based methods for determining the atmospheric boundary layer height
(DE: Vergleich verschiedener mess- und modellbasierter Methoden zur Bestimmung der atmosphärischen Grenzschichthöhe)
Study programs: HSE, Hydrologie, Geographie, Forstwissensch, RE-NRM, GroundwatCh, FRM
Start: immediately
Goal:
The boundary layer height is an important parameter for various applications in micrometeorology and its accurate determination is important to understand processes related to the ecosystem-atmosphere exchange within in the atmospheric boundary layer. For example, it is needed in important corrections for measurements of atmospheric fluxes such as ecosystem evapotranspiration. There are various methods to determine the boundary layer height. It can be measured directly with ceilometers, but these devices are often not available and only measure the boundary layer height locally. A long-term operation at all atmospheric flux stations would be very costly. Another possibility is to model the boundary layer height, for example with a thermodynamic encroachment model or with numerical models. Due to the models’ limitations, it is questionable how well the modeling approaches can predict the boundary layer height, especially over complex terrain such as (low) mountain ranges and urban areas. The aim of this master thesis is to investigate the two modeling approaches with ceilometer measurements at our site in the Tharandt forest.
Steps:
- Literature research
- Modeling of the boundary layer height with the thermodynamic encroachment model
- Evaluation of the ceilometer measurements in the Tharandt forest
- Collection of ERA5 reanalysis data
- Comparison of the different approaches, statistical evaluation
- Discussion of the results, derivation of conclusions on the suitability of the model-based methods
Language: English/German
Requirements: some experience with R (or python/other programming language of choice)
Supervisor: Dr. Luise Wanner [luise.wanner@tu-dresden.de]
Unraveling Decadal Trends and Climate Sensitivity of Soil CO2 Efflux in a Norway Spruce Fores
Study programs: Geographie, Forstwissensch, RE-NRM
Start: immediately
Goal:
Global forests represent one of Earth’s most important carbon reservoirs, yet they are increasingly impacted by repeated and long-term extreme weather events. After photosynthesis, soil CO2 efflux is the second-largest carbon exchange in terrestrial ecosystems, strongly influencing future climate trajectories. However, long-term observations of soil CO2 efflux remain scarce, creating critical gaps in our understanding of how forest soils will respond to continued climate variability. This master’s thesis addresses that gap by leveraging a decade (2012–2022) of soil CO2 efflux measurements in a mature Norway spruce stand, integrated with high-resolution meteorological data from the ICOS station in Tharandt.
You will apply statistical time-series analyses to assess whether and how temperature and moisture controls on soil CO2 efflux have shifted over time, testing how forest soils are adapting or becoming more vulnerable to warming trends and drought even.
Steps:
- Literature study on the topic of soil CO2 efflux under varying environmental conditions
- Analyses of soil CO2 efflux data collected during the growing seasons 2012-2022 along with soil and atmospheric data
Language: English/German
Requirements: Ability and willingness to work with large datasets, interest in statistical data analysis
Supervisor: Dr. Kyohsuke Hikino, Dr. Luise Wanner [luise.wanner@tu-dresden.de]
|