Job tips, internships & thesis suggestions
Find job tips, thesis and project suggestions and internships, within research or outside academia.
For more information about the intersnhip course, please look on the course page. There you may find inspiration from reports by students from previous internships.
Quick links:
In business, research or organisations outside the department
In research groups within the department
PhD and postdoc positions worldwide
In business, research or organisations outside the department
Write your Master thesis with Resource Solutions
ReSource Solutions Sweden AB is working in environmental consulting and services. The compay has a few proposals for a master thesis to carry out between January and May 2025.
We work with advanced environmental monitoring techniques such as IoT and drone-based mapping and quantification of greenhouse gas emissions, mainly from anthropogenic sources. We can offer our applied knowledge in gas measurements, some field work and assistance in reviewing your thesis work.
We are looking for a student who has interest in technical applications within environmental management and monitoring topics who is acquainted to the following:
- Geospatial and Geostatistical analysis methods
- Gas plume behavior and reconstruction models
- Basic coding.
Depending on the proposal and level of ambition, the thesis might involve some retributed field work. Basic field experience and environmental sampling knowledge is therefore required.
Interested?
Send a one-page short CV to contact [at] resource [dot] se (contact[at]resource[dot]se) detailing which courses you took so far, and 150-200 words abstract on:
• Why you chose those courses
• A few references to the literature you are familiar with
• Your expectations within and motivation for environmental monitoring and management
• What administrative time frames for your thesis are to be expected.
Opportunity in Morocco with financial support
Mohammed VI Polytechnic University (UM6P) is an internationally oriented institution of higher learning, that is committed to an educational system based on the highest standards of teaching and research in fields related to the sustainable economic
development of Morocco and Africa. Research projects within the Centre for Remote Sensing Application focus on water resource management and food security in Africa, addressing some of the most pressing challenges in sustainable agriculture.
We provide a collaborative and supportive environment, with access to field sites, laboratories, and data for compre hensive research. In addition to valuable research experience, we offer free accommodation and meals at the UM6P restaurant, access to campus facilities (e.g., gym), and a salary of 3,500 SEK. This makes it an attractive opportunity for
students interested in gaining international research experience in water resource management, food security in Africa, and the application of crop models and machine learning.
The research is conducted in Ben Guerir near Marrakech (learning. The research is conducted in Ben Guerir near Marrakech (https://um6p.ma/).
Read more and contact:
Job opportunity/Internship as Forest data analyst at Katam Technologies (Autumn 2024)
Katam Technologies AB is a team of ambitious startup entrepreneurs who want to revolutionize the forest
industry through game breaking technology. We are now looking for two internships to strengthen our
forest data analysis team with the possibility for full-time employment if the match is right.
European Union Satellite Centre (SatCen) recruiting students and young graduates
European Union Satellite Centre (SatCen), is an EU Agency working at the intersection of space, security and defense, located close to Madrid, Spain.
Reaching out to students and young graduates in the fields of GEOINT, GIS, Remote Sensing or similar.
- Traineeship programme: SatCen - Traineeship programmes (europa.eu)
- Recruitment conditions: SatCen - Recruitment procedures (europa.eu)
- Current vacancies: SatCen - Vacancies (europa.eu)
SWECO intresseanmälan – Konsult inom Digital Services – Stockholm / Uppsala (ev på distans)
Löpande ansökan, även nyutexaminerade.
Sweco jobbar inom GIS och geodata arbetar SWECO med informationsbearbetning och -samordning, utformar och utför rumsliga analyser eller arbetar med dataflöden och tillgängliggörande av geodata med verktyg som ArcGIS, QGIS och FME. Vidare arbetar de med organisation- och verksamhetsutveckling eller IT-arkitektur kopplat till GIS och geodata.
Internships worldwide
Announced at the Lund University carreer pages
Thesis suggestions from Sustainalink (updates continiously)
Sustainalink is a matchmaker between students and companies, check out their web page sv /en and take the chance to contact them with your cv.
Current projects at Sustainalink
Internships and student jobs in sustainability via Sustainergies
Sustainergies keeps an eye open for student opportunities at many interesting employers in Sweden. They also organise case study workshops and courses.
Sustainabilities student jobs page (in Swedish)
Swedish environmental protection agency openings for students
Internships, summer employments, and thesis suggestions, when available. Many student opportunities in climate, communication, environmental protection, environmental law and more. Note that there are not always vacancies
To the Naturvårdsverket jobs page (in Swedish)
In research groups within the department
Background
Plant traits are typical species-dependent properties of vegetation like leaf size, root depth, or leaf mass to area ratios. Plant traits are responsible for interaction with the environment, and how plant specific processes function in a given situation or environment, for example photosynthetic rates, leaf water exchange or the plant’s carbon uptake.
These plant traits are in general important for modelling purposes, for example to estimate photosynthesis rates of a forest based on given input parameter, or when scaling up measurements from leaf- to canopy scale.
In tall canopies like forests, the traits of leaves are expected to have adopted to the specific conditions in the canopy layer (Konôpka et al. 2016), such as the light availability, temperature and vapor pressure deficit vary within the canopy. Needle properties in the upper, sun-lit part of the canopy might differ from the lowest levels of the canopy where light availability typically is much lower.
Hypothesis and research question
The ICOS ecosystem station Hyltemossa is located in a commercial spruce forest near Perstorp planted in the 1980´s. ICOS is regularly taking shoots from the upper (sun-lit) canopy from a group of trees for analysis, such as nutrient content, leaf area, or leaf mass per area (LMA). The question is if these plant traits vary throughout the canopy layer,that is if average leaf sizes and LMA differ between lower and upper parts of the canopy, or in comparison with a younger spruce stand.
A second research question is how the results from different trees within the target area of the Hyltemossa forest differ, and how they compare with other published data from spruce.
Method:
The project requires some field work in Hyltemossa, where destructive shoot samples will be taken from different heights within the canopy. Analysis of these samples will be done using e.g. a Li-3100 Leaf Area Meter, and taking weights and dimensions of needle samples.
Contact
Background
Leaf Area Index (LAI) is an important variable directly related to the uptake of CO2 via photosynthesis, and release of water from ecosystems through transpiration. LAI is commonly used both within ecosystem modelling and remote sensing for estimates of vegetation productivity and evapotranspiration in different types of ecosystems. LAI can be estimated in a multitude of ways; i) destructively (e.g. collecting plant material and measure leaf area in a lab) or ii) non-destructively (e.g. by estimating radiation transfer through plant canopies, through Digital Hemispherical Photography (DHP) or through remote sensing techniques). Non-destructive methods for LAI estimations are generally preferred but unfortunately, there is no consensus on which method is most accurate for different ecosystems and each method has it strengths and weaknesses. In addition, development of sensors and methods makes it challenging to achieve coherent estimates of LAI over longer time periods.
Aims and Methods
Your task will be to evaluate how a recently purchased state-of-the-art plant canopy analyzer (LAI-2022TC, Li-Cor) performs in comparison with an older plant canopy analyzer (LAI-2000, Li-Cor), a ceptometer (Delta-T SunScan) and with digital photography with a fisheye lens (DHP). It is important to evaluate the LAI-2200TC in ecosystems with high LAI (ICOS Hyltemossa and Rumperöd continuous cover forest), medium LAI (Hyltemossa clear-cut) and low LAI (Alnarp agricultural site). Expect maximum one week of field work during the thesis work.
The results will be of great importance to several ongoing research projects (read more about the projects LUfluxes and PERENNIAL).
Contact
PERENNIAL is the collective name for two research projects funded by FORMAS and ERC. The main purpose of these projects is to investigate how cultivation of perennial crops can have a positive impact on agriculture and the environment and contribute to mitigating increasing carbon dioxide concentrations in the atmosphere through carbon sequestration in the soil. Through these projects, we have established a research infrastructure in Alnarp outside of Lund. Here we run parallel meteorological and carbon flux measurements on two adjacent fields; one planted with the perennial crop Intermediate Wheatgrass (IWG, Kernza) and one planted with an annual crop (currently winter wheat). Read more at
A wide range of possibilities for BSc and MSc thesis work exists and besides the suggestions listed below, you are welcome to suggest other topics of interest to you. Not all suggestions are suitable for a BSc thesis, but some are. Both 50% and 100% pace are possible. More info at https://www.nateko.lu.se/research/ecosystems-functioning-dynamics-and-ecology/perennial
Topics
Contact persons
Jonas Ardö, Patrik Vestin, Torbern Tagesson, Tobias Biermann
During late summer/autumn 2022 the forest was clear-cut around the tower at the research station ICOS – Norunda: https://www.icos-sweden.se/norunda
Aim and tasks
To follow the temporal development of the clear-cutting a time-series of drone images were collected, during the most intense periods daily flights were conducted. These images will be processed to create orthophotos to map the temporal development of the clear-cutting to study how the clear-cut influenced the measurements at the station. The aim of the BSc thesis will be to georeference the orthophotos and map the clear-cut area in the time-series of data.
The thesis can be focusing on (1) the georeferencing part – to what extent can it be automated, or (2) mapping the clear-cut area in the orthophotos – to what extent can the clear-cut area be mapped with image classification methods.
Contact
Background/topic
The ICOS-site at Hyltemossa (HTM) is located within a ca 40 year old commercially used spruce forest. The main task of the station is to measure the carbon balance of this forest, to quantify the forest’s growing conditions and to assess processes related to the carbon uptake of the forest ecosystem. Ongoing measurements related to the forest canopy are PAR above and under the canopy at high temporal resolution. Additionally, LAI (Leaf Area Index, or Green Area Index) is estimated by hemispherical photos on a regular basis. Since 2019, litterfall is measured with 5 litter traps in all 4 CP and reported in 2-4 week intervals.
Tasks
The task is to use continuous PAR and discontinuous LAI data to try to estimate the temporal behavior of LAI at HTM following Holst et al. (2004) where the extinction of PAR in the canopy layer was calculated for different situations and applied in Beer’s law to assess LAI. Furthermore, litter data should be analyzed to assess annual patterns and to relate these to calculated LAI patterns.
Method
Data analysis of existing data
100% pace (50% possible), flexible starting time; both as BSc or MSc thesis; possible extension to other sites.
Contact
100% pace (50% possible), flexible starting time; both as BSc or MSc thesis; possible extension to other sites.
Background/topic
The ICOS-site at Hyltemossa is located within a ca 40 year old commercially used spruce forest. The main task of the station is to measure the carbon balance of this forest, to quantify the forest’s growing conditions and to assess processes related to the carbon uptake of the forest ecosystem. Specifically, the eddy-covariance (EC) method is used to continuously measure carbon exchange and evapotranspiration of the forest. However, the EC method is not able to distinguish transpiration and evaporation. Both photosynthesis and evapotranspiration are dependent on the water supply for the trees, and 2018 was a year with very limited water availability.
Tasks
The task is to parameterize the hydrological model BROOK90 for the Hyltemossa forest site, and to run the model based on the available input data (from 2015 onwards) from Hyltemossa, and to compare model results with measured variables (evapotranspiration, soil moisture, possibly transpiration). Additionally, climate station data from SMHI for pre-ICOS time (before 2015) can be used to increase the timeseries. (BROOK90 model: http://www.ecoshift.net/brook/brook90.htm).
Method
Data analysis of existing data, application of an existing hydrologic model.
Contact
100% pace (50% possible), flexible starting time; MSc or BSc thesis; possible extension to other sites.
Background/topic
The ICOS-site at Hyltemossa is located within a ca 40 year old commercially used spruce forest. The station is both an ecosystem station and an atmospheric station within the ICOS network, and an ACTRIS site. Measurements of for example CO, CH4, NOx or particles at the station are influenced by source areas far away from the forest where long-range transport is important to understand the patterns observed. One example is the CH4 peak overserved at Hyltemossa caused by the NordStream pipeline blowup. The HYSPLIT model is a combined Eulerian and Lagrangian transport model which can be used to calculate trajectories both in forward- and backward mode to identify source areas.
Tasks
Analysis of timeseries from the ICOS (eventually ACTRIS) network (e.g. CO, CH4, NOx) to identify time periods when non-local processes dominate measured concentrations, and to do a first source appointment based on HYSPLIT results. Potential medium-range source areas would be Greater Copenhagen, long-range transport could originate from Germany or Poland.
Method
Data analysis, application of an existing trajectory model (HYSPLIT)
Contact
100% pace (50% possible), flexible starting time; both as BSc or MSc thesis
Background/topic
Measurements in a sub-arctic birch forest (Stordalen, near Abisko) were done in 2022 to try to track and quantify the impact of an Autumnal moth outbreak in the forest, when larvae were at consuming birch leaves in larger patches within the forest. Besides other measurements tailoring at the BVOC and CO2 fluxes from this herbivore attack, we also collected data on LAI and PAR interception to estimate the impact of the Autumnal moth outbreak on the forest canopy.
Tasks
The task is to use continuous PAR and discontinuous LAI data to try to estimate the temporal behavior of LAI at Stordalen birch forest following Holst et al. (2004) where the extinction of PAR in the canopy layer was calculated for different situations and applied in Beer’s law to assess LAI.
Method
Data analysis of existing data of PAR interception, with a possibility to combine this with analysis of remote-sensing data (satellite, evtl. drone images, automatic cameras).
Contact
100% pace (50% possible), flexible starting time; preferably MSc thesis but BSc possible as well.
Background/topic
Measurements of particle size distribution (sub-µm size) were done in a sub-arctic birch forest (Stordalen, near Abisko) in summer 2022 with the aim to detect and quantify new particle formation (NPF) and particle growth rates. Particles may be formed based on atmospheric precursors, such as BVOC. In 2022, there was a partial Autumnal moth outbreak in the forest, when larvae were at consuming birch leaves in larger patches within the forest, with a suspected large emission of BVOC.
Tasks
Analysis of field data from two different instruments (SMPS and AIS) with continuous data on particle size distribution (<1nm to ca 400nm size bins) to detect growth events and NPF. There is also an older dataset (2006-2008) for comparison.
Method: Data analysis, see Svenningsson et al. 2008
Contact
We have available a range of possibilities for projects related to understanding the form and function of the world’s forests, from local to global scale. Possible techniques include:
- Process-based ecosystem modelling using our state-of-the-art global vegetation model, LPJ-GUESS
- Empirical analyses of observations of trees from forest research and inventory plots
- Inference of ecosystem properties and behaviour from remote-sensing products
Contact Thomas to discuss your interests and match to possibilities.
Contact
Background
Preliminary analysis suggests that a large share of forest harvest (clear cuts) in Sweden is harvest of old forest stands that have never been clear cut before. There is however no official data on what types of forests are being cut, but we do know the location of past cuts for the last 10 years and planned cuts for each coming year. We have made the first steps towards a homepage that indicate when panned cuts are cuts of old forests, but our estimates are surrounded by large and unidentified uncertainties.
Our ambition is to create a high-quality webpage that tracks planned cuts of old forests that store much carbon and may have high biodiversity and recreational values. Our current estimate is based on a forest age map from year 2010, but it has not been evaluated and the uncertainties increase with forest age (it is difficult to distinguish an 80 year old forest from a forest that is 200 years old).
Objectives
There are multiple projects that may contribute to this goal, ranging from technical and visual developments of the webpage to scientific predictions of forest age and diversity. Multiple datasets are available on forest type, species composition, tree height from airborne laser scanning, and around 100 thousand ground plots. There are also a multitude of other maps available.
To summarize, there are multiple projects available for driven and independent students that want to contribute to the project. These include:
- Technical implementations that may involve automatic updates of the map or adaptations for users to find potentially old forests that are reported to be cut.
- Developments of the visual appearance of the homepage.
- Evaluation and corrections of the existing forest age map using plot data.
- Creation of a new forest age map and/or an index indicating the age and values of forests marked for harvest.
We are in the beginning of this work, and other ideas are welcome.
Contact
Do you want to help us understand old-growth forests in Sweden (gammelskogar)? This project is suitable on for both MSc and BSc level. It can start at any time.
Background
We have recently produced a digital (GIS) map of old-growth forests in Sweden. In a larger long term project we aim to investigate if old-growth forests store more carbon than production forests do. In the suggested project the student would perform carbon inventory in a single or a few old-growth forests and in the surrounding managed ecosystems (forest plantations, pastures or croplands). The work would be performed in collaboration with a PhD student. Several old-growth forests are located in Skåne and in Southern Sweden, but there are sites all over the country.
Two or more students in collaboration
Because there are multiple forests there are multiple projects, and two or more students could collaborate. Because the project involves field work the ideal candidate(s) would be willing to organize and perform field trips and measurements. The project also involves analysis of collected data and existing inventory data.
We also welcome other ideas that are related to this topic.
Contact
This project is suitable on for both MSc and BSc level. The project also works for E-learning program theses. Any semester, there are multiple projects that can start at any time.
Background
There are still large uncertainties on how land use, such as forestry and agriculture affects the carbon cycle and biodiversity. A larger long term project at the department aims to further our knowledge on this by contrasting pristine ecosystems (ecosystems that are more or less natural, such as old growth forests) to surrounding ecosystems under land use. Right now we have produced a digital map (GIS map) of old growth forests in Sweden, but we are also interested in expanding the map to more countries. Data also exists for European Russia, but we have not analyzed it.
Objectives
This project can take many shapes, for example.;
- Produce a map for an unmapped region. This can be done by collecting information from known sources, digitizing older maps or using remote sensing information such as aerial or satellite images to digitize known forests. Our long term aim is to build a global map, you would contribute to this.
- Analyze our already mapped old-growth forests in Sweden. There are several questions here;
(a) if old-growth forests differ in terms of their topography or soil characteristics from surrounding ecosystems. This is important to know when performing paired analysis, if the old growth forest has been left unused because it is a mountain or a wetland, then the paired analysis of that forest will need to be modified. This mainly involves GIS analysis.
(b) Use forest inventory data or remote sensing data to compare carbon storage between the old growth forests and surrounding ecosystems under land use.
- Perform the first analyses on old growth forests in European Russia, all of the above types analysis can be performed.
- A topic of your own that uses our new old growth forest map.
Possible work aspects
The student will work with our project on these questions. There are many things to do, the projects can be varied to focus on carbon cycle, ecology or technical GIS / remote sensing aspects. The scope of the project and analysis can be adjusted to the type of thesis (bachelor or master) and it is also possible to divide the work between multiple students. The project is not limited to the allocated time allocated for the thesis; it could start earlier and continue after. Multiple students are also welcome, working in groups or focusing on different regions.
Contact
Terrestrial ecosystem models such as the dynamic global vegetation model LPJ-GUESS include a large amount of process parameters for which the exact values are often not known or are derived from lab measurements. These parameter values are then employed at larger (global) scales lacking robust upscaling methods and this uncertainty in parameter values is one of the contributors leading to the overall model uncertainty. However, for improving our understanding of terrestrial carbon cycling as well as future predictions, it is important to quantify and reduce this uncertainty as much as possible.
Objective
This project aims to calibrate some of the key process parameters embedded in LPJ-GUESS by applying the Land Variational Ensemble Data Assimilation Framework (LAVENDAR, Pinnington et al., 2020 ) together with observations of the terrestrial carbon cycle. LAVENDAR has been developed by colleagues at the University of Reading. Its main feature is that it separates the data assimilation analysis step from the model integration. It requires a pre-constructed ensemble of model runs to calculate necessary statistics.
The main research work will be based on performing ensemble simulations with LPJ-GUESS and integrate these into the LAVENDAR framework. As a first step so-called “twin” experiments (using model generated observations for the assimilation) will be performed to determine the systems performance. In a second step the framework will be used together with relevant observations of terrestrial carbon cycling such as CO2 fluxes provided by ICOS. Initial work will be based on single site simulations, which, depending on progress, can be extended to multi-site/regional/global simulations.
Requirements
This project requires the candidate to have good programming skills (mainly Python but also C/C++) and an interest in terrestrial carbon cycle modelling and in statistical analysis. Previous experience with LPJ-GUESS is advantageous but not necessarily required. The project offers the possibility to collaborate with colleagues from the University of Reading, UK.
Contact
Marko Scholze
Referencess
Pinnington, E., Quaife, T., Lawless, A., Williams, K., Arkebauer, T., and Scoby, D.: The Land Variational Ensemble Data Assimilation Framework: LAVENDAR v1.0.0, Geosci. Model Dev., 13, 55–69, https://doi.org/10.5194/gmd-13-55-2020, 2020.
Master thesis: Can the new merged gridded rainfall product replace limited rain gauge measurements for hydrological modelling in poorly gauged basins? a case study using the HYPE hydrological model in Ethiopia. Read more
Background
Renewable solar energy needs to be promoted around the world to substitute for the fossil fuels that currently still dominate worldwide energy consumption. Ideally, implementations of the large-scale solar photovoltaic panels over global drylands can utilize substantial solar power which is the most abundant renewable energy source. However, due to the radiative forcing and associated with complex atmosphere-land(vegetation) feedbacks, the profound impacts on global climate induced by these hypothetical solar farms are poorly understood.
Tasks
The student will analyze existing model output from a state-of-art Earth System model EC-Earth for different solar farm scenarios, and focus on the quantification of global climate response to these large-scale solar farms, such as the characteristics of El Nino-Southern Oscillation or climate extremes; then, try to advance our understanding of the mechanisms for the interaction processes in the Earth system that drive these impacts regarding such land-use changes.
Requirements
Basic knowledge of data analysis software/language such as Matlab and Python.
Contact
Zhengyao Lu, zhengyao [dot] lu [at] nateko [dot] lu [dot] se
Anna Schultze, anna [dot] schultze [at] nateko [dot] lu [dot] se
When: Spring 2025
What: Soil moisture is a critical component in the global water cycle. Reliable estimation of soil moisture is essential for various applications, including drought monitoring, water resource management, precision irrigation, and forecasting crop yields. While surface soil moisture (SSM) receives considerable attention, root zone soil moisture (RZSM) is particularly significant as it directly influences plant health, photosynthesis, and transpiration rates. As a result, RZSM is a crucial indicator for assessing vegetation stress, crop water needs, and optimizing agricultural water use.
Different optical satellite sensors, such as MODIS onboard the Terra and Aqua satellites and OLCI onboard Sentinel-3, typically have different central wavelengths, bandwidths for the same red, NIR bands etc. This leads to inconsistencies when building a long time series of vegetation indices (VI) from multiple sensors, causing problems when analyzing vegetation trends using VI. Therefore, to generate a single product from multiple satellite sensors—known as a virtual constellation (Claverie, 2023)—bandpass adjustments are necessary. This involves correcting the bandpass using the spectral response function of each sensor.
In this master’s thesis project, you will investigate the inconsistency in BRDF reflectances between the MODIS and Sentinel-3 OLCI sensors.
The results of your work are expected to contribute to our ongoing project of Copernicus Land Monitoring Service (CLMS) MR-VPP product continuation using Sentinel-3.
Requirements: Experience in satellite data analysis and programming is essential for this project.
This topic requires expertise in Geographic Information Systems (GIS) and machine learning. The project involves training and fine-tuning large language models, as well as handling and organizing geospatial datasets and managing their metadata. The primary programming language for implementation is Python, with TensorFlow serving as the main framework for model development and testing.
The integration of Large Language Models (LLMs) into geospatial workflows presents a promising avenue for enhancing data retrieval and interpretation within complex Geographic Information Systems (GIS) environments. This thesis aims to explore how LLMs can be effectively trained or fine-tuned to understand and process GIS-specific queries that involve multi-layered datasets. The research will investigate the capabilities of LLMs to navigate intricate directory structures and retrieve accurate information from multiple layers stored in diverse formats. It will focus on improving LLMs' comprehension of geospatial data representations and enhancing their ability to providecontextually relevant responses by understanding the relationships between different layers in a GIS.
Key challenges addressed in this thesis include developing methods for LLMs to accurately interpret metadata and file structures and enable them to identify and retrieve the appropriate geospatial data. This research will also explore effective strategies for indexing and describing GIS layers to facilitate efficient data access by LLMs. The objective is to create a framework that allows LLMs to seamlessly interact with complex GIS datasets, providing precise and meaningful responses to geospatial queries. By advancing LLM capabilities in this domain, the study aims to contribute to more intuitive, automatic and efficient use of GIS technology in various applications, ranging from environmental monitoring to urban planning, particularly for users without extensive technical knowledge.
Contact:
Ali Mansourian: ali [dot] mansourian [at] nateko [dot] lu [dot] se
Rachid Oucheikh: rachid.oucheikh@nateko.lu.se
Improved knowledge of trees in semi-arid ecosystems is essential to understand how these ecosystems function, and for improving our understanding of the role of the trees within the climate system, the global carbon cycle, and local livelihoods. These trees are important for survival and biodiversity of flora and fauna, and for ecosystem services such as carbon storage, food resources, fuelwood, and shelter. Currently, dryland-restoration development projects are given enourmous amount of funding and attention, and trees play a central role in these campaigns to mitigate climate change, and to stop threats of deforestation and desertification. Since the trees of semi-arid environments are becoming increasingly important within environmental initiatives across Africa, but we still do not know much about their climate mitigation effect, it is increasingly important to understand their role within both these ecosystems, and within the full Earth system.
Tasks
The Environmental Mapping and Analysis Program (EnMAP) is a hyperspectral satellite mission that monitors and characterizes Earth’s environment on a global scale. The satellite provides a high resolution hyperspectral image with 230 spectral bands from 420 to 2450 nm with a ground resolution of 30 m x 30 m. These hyperspectral data may help in estimating the species of the sparse trees in the semi-arid savannah landscape. The main aim with this thesis is to study the spectra of various tree species of a semi-arid savannah in Senegal. Secondly, we will aim at mapping the species distribution across a semi-arid savannah landscape using EnMAP data.
Contact
Torbern Tagesson torbern [dot] tagesson [at] nateko [dot] lu [dot] se
This topic requires experience in GIS and knowledge in machine learning. The project involves training and fine-tuning large language models, and potentially the use of Retrieval-
Augmented Generation (RAG) method. Python will be the primary programming language used for implementation, with TensorFlow as the core framework for developing and
evaluating the models.
In recent years, the integration of artificial intelligence with geospatial data has opened newavenues for automating complex data analysis and reporting tasks. Geospatial data, generated through various sources such as satellite imagery, remote sensing, and Geographic Information Systems (GIS), is inherently complex and multidimensional, often requiring expert knowledge to interpret and present. Traditional methods for creating geospatial reports are Ntime-consuming and labor-intensive which limits the ability to rapidly extract actionable insights. The emergence of Large Language Models (LLMs) provides an opportunity to revolutionize this process by enabling the automated generation of customized reports that can communicate complex geospatial information effectively to diverse stakeholders. Thispotential shift could democratize access to geospatial insights, making it easier for decision- makers in fields such as urban planning, environmental management, and disaster response to understand and act on geospatial data.
This thesis aims to explore methods for leveraging LLMs to automate the generation of customized reports from standardized geospatial data. The research will focus on training LLMs to generate tailored reports based on user-defined inputs, such as specific regions, features, or time periods. It will investigate techniques for integrating LLMs with geospatial data workflows to ensure the generated reports are accurate and align with the specific needs of user queries. The study will address key challenges in processing complex GIS outputs, including the identification and presentation of trends, outliers, and actionable insights. Acritical component of this research will be the development of validation frameworks for assessing the quality and reliability of LLM-generated geospatial reports. The thesis will also examine how to enhance the clarity and coherence of the reports through natural language generation improvements. By combining qualitative and quantitative evaluation methods, the research aims to establish criteria for verifying the accuracy and interpretability of automated geospatial reporting systems. Methods for optimizing the customization of report content through user feedback loops and adaptive learning strategies may be also considered focus.The outcomes of this study are expected to contribute to the broader field of geospatial intelligence and decision support and provide a foundation for more intuitive and effective use of geospatial data for non-expert GIS users.
Contact
Rachid Oucheikh: rachid [dot] oucheikh [at] nateko [dot] lu [dot] se
Ali Mansourian: ali [dot] mansourian [at] nateko [dot] lu [dot] se
We are seeking one MSc student with a background in ecosystem analysis and remote sensing. We are suggesting an MSc project that builds on data from the ICOS and SITES infrastructures. ICOS Sweden collects CO2 flux and meteorological data, and SITES collects spectral data from drones, fixed sensors, and satellites. Combined, these data provide good opportunities to investigate how to accurately monitor carbon fluxes for Sweden. Read more here:
https://www.icos-sweden.se/
https://www.fieldsites.se/en-GB/sites-thematic-programs/sites-spectral-32634403
Tasks
This task can be modified depending on the interest and background knowledge of the students. Here are suggested components:
- Explore the synergy between ICOS (Integrated Carbon Observation System) and SITES (Swedish Infrastructure for Ecosystem Science) datasets.
- Investigate how remote sensing can enhance the understanding of vegetation dynamics and carbon fluxes in Swedish forests.
- Develop methods to integrate and analyse these datasets to provide comprehensive insights into carbon cycling.
- Focus on the temporal aspects of vegetation dynamics by utilizing time-series remote sensing data.
- Examine how changes in climate and land use influence carbon fluxes in Swedish ecosystems.
- Implement advanced time-series analysis techniques to capture the nuances of carbon dynamics over different seasons and years.
- Explore machine learning for training a model of carbon fluxes with inputs from satellite (for student who has taken machine learning course)
Contact
Jose Beltran
Department of Physical Geography and Ecosystem Science
jose [dot] beltran [at] nateko [dot] lu [dot] se
Suserup and Biskopstorp have had at least two prior inventories where the entire forests have been measured (DBH and height). The project would be to do measurements with a TLS (Terrestrial Laser Scanner) to extend these prior measurements with a new data point and also to make an estiamte of the total volume. From these measurements we would also get more detailed data to update allometric relationships used in e.g. LPJ-GUESS. This project could also include using these measurements to setup LPJ-GUESS for the site(s).
Contact
Stefan Olin, stefan [dot] olin [at] nateko [dot] lu [dot] se
The future of tropical forests under climate change is uncertain. One of the biggest questions is how these forests will respond to increasing levels of water stress likely to result from a warmer climate. To make projections of how the forests are likely to change we need computer models. But to have confidence in those models we have to evaluate them against suitable observations.
At INES we develop one of the most advanced models for making such projections of forest futures – LPJ-GUESS. However, suitable observations to test the model against are very rare in the tropical forests. In a new collaboration with partners in China we have the opportunity to make one of these rare tests. In the Xishuangbanna forest in southern China, a rainfall exclusion experiment has been installed and now has several years of data.
Tasks
In this project you would carry out simulations for the forest with the LPJ-GUESS model and evaluate them against the various observation available. The challenge is to pick apart when and why the model and observations differ and what this says about the underlying processes.
The project requires some skills in coding to run and analyse the model output, but should be within the reach of anyone who has enjoyed the coding parts of their degree course so far. You will develop not only your research skills and understanding of tropical forests, but also transferrable skills in code development.
Logistics allowing, there is potentially a chance to visit Xishuangbanna to work with the local scientists during the project.
Spring 2024 or 2025
Contact
I have collected old PhD theses where stands have been measured sometime in the 60's or 70's, the idea is to go back to these locations and measure them again with the aim of setting up LPJ-GUESS to simulate these forests. Examples are:
- Andersby in Uppland
- Dalby Söderskog in Skåne
- Skirö in Småland
- Biskopstorp in Halland
- But there can be more potential places.
LPJ-GUESS have recently been developed with the ability to be initialised with a known forest structure, the idea is to evaluate this for some forests where we have some good knowledge of how the forest have evolved.
Contact
Stefan Olin stefan [dot] olin [at] nateko [dot] lu [dot] se (stefan[dot]olin[at]nateko[dot]lu[dot]se)
Background
Marine fisheries play a vital role in food security and social and economic development. Affected by fishing technology, market demand and high-intensity fishing, marine fisheries are facing dilemmas such as resource exhaustion and declining benefits.
Aim
The aim of this Master project is to explore spatiotemporal patterns of fishing activities from satellite-based Automatic Identification System (AIS) data using GIS and machine learning.
The project is a cooperation with Marine center of Simrishamn municipality, Sweden. In particular, we are interested in understanding how the spatial fishing pattern for Swedish pelagic fishing developed in Hanö Bay, in the Blekinge archipelago, and east of Öland when individual transferable quotas (ITQ) were introduced.
For example, how quickly did the large industrials vessels take over in the fishery? Can we observe that the presence/activity of the large vessels forced the smaller ones to seek other areas?
The student should be familiar with spatial data analysis and machine learning. Programming skills (Python or R) are beneficial. A more specific introduction to the topic will be given by the supervisors upon request.
Supervisors
Pengxiang Zhao and Josefine Larsson
Background and aim
Shared e-scooters, as one of the environment-friendly and sustainable transport modes, have shown great potential to help mitigate urban mobility issues. They also provide a flexible transport mode for connecting to public transport to solve the first-mile-last-mile problem. Shared e-scooters can be both complementary and competitive with public transport, but less attention has been paid to the spatiotemporal variations of these complementary and competitive relationships.
Motivated by this, the goal of the thesis is to explore and analyze the integrated usage of shared e-scooters and public transport across space and time. Three major cities in Sweden are chosen as the case studies, including Stockholm, Gothenburg, and Malmo, for comparison.
Tasks
Tasks include
- to collect and process shared e-scooter trip data and public transport data,
- examine the usage relationships at the trip level with machine learning,
- conduct spatial and temporal analysis of the integrated usage to uncover how the complementary and competitive relationships vary in space and time.
- A comparison study will be implemented in the three Swedish cities.
The student should be familiar with spatial data processing, analysis, and mining. Programming skills (Python or R) are beneficial. A more specific introduction and data for the topic will be given by the supervisor upon request.
Contact
Background
Mitigating the urban heat island effect is essential for creating more sustainable and livable cities, especially considering the ongoing global trends of urbanization and climate change. This effect appears in almost every urban area, no matter large or small city.
Previous studies have indicated that green spaces can reduce urban heat islands. Research on relationships between urban greenery and urban heat island at a fine scale is still scarce. One important task is to measure urban greenery at the street level.Compared with remote sensing image data, street view image data can better capture the street-level, profile view of urban greenery (e.g., a green wall).
Aim
This topic aims to understand the role of urban greenery in reducing the land surface temperature at the street level using street view image data and deep learning.
Method
The tasks include to collect land surface temperature data and prepare the training data for the deep learning model development, develop a deep learning model to measure urban greenery at the street leve, and examine the relationships between urban greenery and land surface temperature at the street level.
The student should be familiar with spatial data processing, analysis, and mining. Programming skills (Python or R) are beneficial. A more specific introduction and data for the topic will be given by the supervisor upon request.
Contact
Background
Timely and accurately monitoring the spatiotemporal dynamics of urban expansion and shrinkage plays a crucial role in urban planning, sustainable development, and social well-being. Traditional methods of monitoring urban shrinkage and expansion mainly rely on population data and statistical yearbooks. However, such traditional data have difficulties in collection, especially on a large scale.
With the advances of big data technology, night light satellite image data have been collected and used to study urban expansion and socioeconomic activities (population, GDP, electricity consumption). It is important to develop a method for identifying urban areas from night light image data. Deep learning opens up opportunities for extracting insights from these images.
Although Sweden has undergone a significant and long-term urbanization process over the years, it is still significant to understand its dynamics of urban expansion and shrinkage in recent years. In particular, Sweden has been actively engaged in smart city initiatives and planning to enhance urban living in recent decades. This study will be helpful for planners in designing urban development strategies.
Aim and tasks
The aim of this topic is to uncover the spatiotemporal dynamics of urban expansion and shrinkage in Sweden using night light data and deep learning.
The tasks includes:
- to collect the night light image data in Sweden, which is openly available
- to develop a deep learning model to identify urban areas from night light images for each year
- to conduct a spatiotemporal analysis to uncover the dynamics of urban expansion and shrinkage in Sweden
The student should be familiar with spatial data processing, analysis, and mining. Programming skills (Python or R) are beneficial. A more specific introduction and data for the topic will be given by the supervisor upon request.
Contact
Aim
To address the following two intriguing questions:
(1) A prevailing hypothesis suggests that a longer growing season length (GSL) leads to higher gross primary productivity (GPP). However, how does time use efficiency (defined as the GPP/GSL) changes with prolonged growing season length?
(2) How do these three resources use efficiencies (carbon, water, and time) interact with each other? Do they exhibit trade-off or synergistic effects?
The Tibetan Plateau - the Third Pole of the world
The Tibetan Plateau, known as the Third Pole of the world, has an average elevation of 4,000 m and along the altitudinal gradient. plant functional types change from subtropical forests to desert grasslands. The warming trend on the Tibet Plateau is three times greater than the global average, making it an ideal nature laboratory to study how plants respond and adapt to the warming climate.
The central concern in ecological research is how plants use different resources (i.e., carbon, water, and time) in response to climate change. Especially for the time, there is still little knowledge about how plants use time throughout the entire growing season (Meng and Yan et al., unpublished). Specifically, there have been extensive studies about the carbon use efficiency and water use efficiency for plants based on which ecologists try to understand how plants optimize the carbon allocation and water usage. For instance, different plants would take different strategies to adapt to drought: some species would allocate more carbon in aboveground biomass so that it can grow taller to get more light while others prefer more belowground biomass in order to grow deeper roots to access deep soil water or groundwater.
In drier climate, plants may change the stomatal size and density to reduce transpiration and improve the water use efficiency. However, it remains largely unknown that how plants would allocate time during its growing season. For example, how do plants allocate time in canopy development and senescence and is the time ratio changing given the longer growing season that is widely observed on the Tibet Plateau under warmer climate?
Materials and methods
To answer these two questions, we will apply multiple sources of dataset that are available: field observations, remote sensing (e.g., NDVI), and simulations from terrestrial ecosystem model (e.g., TRENDY). To explain the mechanisms and interrelationships governing three resources use efficiency. Statistical analysis, and/or machine learning techniques, and/or terrestrial ecosystem modeling might be employed, subjecting to the work development.
Contact
We have large amounts of UAV and field data over agricultural fields and forests collected within different project that can be used for MSc theses. There are several potential research questions to answer and different methodologies can be applied depending on interest. Even if a thesis is individual work, it possible to collaborate with other student(s) that are working with similar methods and data but with different research questions.
Forest
Potential research questions: How early can bark beetle attacked trees be detected? How accurately can machine learning models for segmentation (e.g. https://segment-anything.com/) delineate individual trees?
Agriculture
Potential research questions:
- Can UAV data be useful for monitoring crop growth?
- How can UAV data be used to estimate crop biophysical variables such as LAI during the growing season and for different Crop?
- Can machine learning models be developed to accurately predict crop yields based on UAV data, ground measurements?
- How does the accuracy of these predictions vary during the growing season and across different crop types?
- When is the best time to forecast yield for each crop?
Read more (pdf, open in new window)
Contact
Per-Ola Olsson
Background
Semantic 3D city models have the potential to help cities grow sustainably. From an urban densification planning perspective, they can enable urban planners to evaluate a higher number of scenarios. This is achieved as they support access to simulation input/output data along with related metadata either through information stored locally or through linking their 3D geometries to semantic information stored in external databases. Having access to standardized, high-quality geodata & semantic information decreases pre-processing time & makes simulations easier to execute.
Task
In this master thesis you will investigate which the requirements of 3D city models are, to act as input data to simulations (either noise or daylight) and/or possibilities to store simulation result as part of a 3D city model.
Requirements
The work is conducted at the LU campus. The student must be either a geomatics student or an LTH student in geographic information technology.
Contact
Spring or autumn 2024
Background
The decay of dissolved organic carbon (DOC) in freshwater causes significant global CO2 emissions and is a main contributor to water oxygen depletion. However, DOC decay is conventionally measured in the laboratory. Such measurements give an idea of the intrinsic degradation potential of the DOC, but there is little knowledge of how this ‘lab decay’ translates into actual rates of CO2 production in nature. This project will use a new method to measure DOC decay directly in the field.
The thesis work will evaluate the method as such, but also test ideas about the patterns of DOC decay in time or space, e.g. in relation to land use/land cover or hydrological aspects.
Contact
Background
Natural forests are a complex patchwork of trees of different types and sizes. If we are to understand how the state of these forests is likely to change under a changing climate, or by implementing different management practices, then we need to be able to make accurate computer simulations of how they work. Only then can we answer questions around how to best make our forests sustainable, best mitigate climate change or maximise biodiversity.
The complex patchwork of trees would seem to make such simulations a big challenge, but maybe forests are simpler than they appear? In the LPJ-GUESS vegetation model, we have implemented the capability to test a range of different hypotheses of what needs to be considered in order to accurately simulate the structure and dynamics (growth, mortality) of forests. Now there is an opportunity to use this new model capability to do some science! If you like the idea of using a cutting edge modelling tool to figure out if there are simple rules that underlie forest complexity, then get in contact with us!
Requirements
The LPJ-GUESS vegetation model developed at INES is regularly used to make assessments of forest carbon uptake across the world (Pugh et al., 2019; Friedlingstein et al., 2020). The project will contribute towards a wider development agenda working to improve our ability to simulate tree mortality, linking with researchers across the world. Advanced programming skills are not a pre-requisite, as the LPJ-GUESS modelling can be done without changing the underlying C++ code (only requiring some basic commands in Linux). But ability to analyse datasets using a programming-based software such as Matlab or R will be fundamental to the project, as well as a basic understanding of the concepts of ecosystem modelling.
Contact
Background
With a warmer Arctic, woody plant species are moving north. The increased plant biomass can increase carbon (C) storage in parts of the Arctic and provide a possible negative feedback to Arctic warming. However, some areas have experienced vegetation damage, which offsets the higher C uptake. An increase in extreme winter warming (WW) events in the Arctic—periods of 5–10 days with above-freezing temperatures that cause the snow to melt and the upper soil to thaw, can cause damage to tundra vegetation, particularly to evergreen shrubs.
Aims and methods
Your task will be to figure out which plant functional trait that, as consequence of WW events, impedes plant N uptake and C storage the most. Thereby, which plant functional types (evergreen shrub, deciduous shrub, and moss?) that is the most sensitive to WW events in terms of its C storage.
In order to do that, you will model an Arctic tundra ecosystem in West Greenland. You will be in charge of adapting a process-oriented ecosystem model (CoupModel) to the West Greenland tundra heath using measurements of released N, where in the ecosystem WW-released N moves, soil properties, vegetation C and N content, and meteorological variables.
Contact
Wenxin Zhang (INES) and Laura Rasmussen (University of Gothenburg).
Modelling peatland carbon dynamics at different scales
Peatlands comprise 30% of the present-day soil organic carbon pool and are one of the biggest carbon reserves in terrestrial ecosystems. They play an important role in the global carbon cycle, as a persistent long-term CO2 sink and a moderate source of methane (CH4). Only a few large-scale dynamic global vegetation models (DGVM) and land surface models (LSM) have incorporated peatland dynamics to hindcast Holocene peat carbon accumulation and used them to predict how these ecosystems will respond in future climate conditions. In this project, we will perform hindcast and future experiments with the established state-of-the-art peatland-vegetation model (LPJ-GUESS) at three peat-dominated regions under different climate-warming scenarios.
Aim
The main aim of this project is to determine the fate of peatland carbon stocks in a warming world. To be specific, the project will (1) evaluate the model results using observed datasets from northern latitudes; (2) identify temporal and spatial patterns of peatland carbon sink in these regions; and (3) quantify the importance of climatic drivers that are responsible for modulating carbon fluxes. The overall purpose of this study is to minimize the uncertainty surrounding peatland carbon balance at different spatial and temporal scales.
Managing peatlands to mitigate climate warming
Currently, the majority of global efforts are focused on keeping global warming to 1.5-2°C. To achieve this goal, early and large-scale carbon emission reductions are required to keep the global temperature rise within safe limits and in line with the Paris Climate Agreement. Peatlands are important carbon sinks on land that have been playing a significant role in buffering the effects of climate change, and they will play an important role in supporting climate mitigation in the coming decades. If peatlands continue to sequester carbon in the future, their conservation will be a simple, inexpensive, and reliable mitigation option.
However, currently, damaged and disturbed peatlands are emitting at least two billion tons of carbon annually. In addition, unplanned management practices, together with ongoing warming, have turned many of the intact peatlands into major carbon sources due to higher soil respiration. Curbing emissions from both disturbed and intact peatlands requires planned and informed peatland management strategies. Any unplanned activities related to drainage, rewetting, and land cover change will alter the prevailing balance of CO2‒CH4 fluxes and enhance carbon and methane emissions, which will potentially trigger positive peatland-generated climate feedback and thereby accelerate global warming.
Aim
This project will investigate the role of peatlands in mitigating climate change and assess the wider impacts of climate change on peatland carbon stocks. Different modelling experiments will be performed to identify ideal sets of management practices for moderating CO2‒CH4 fluxes during the restoration of degraded peatlands and the conservation of intact peatlands. Initially, the focus will be on peatlands in Sweden, and later, the scope of the study will be expanded to include the entire Nordic region.
Representation of tropical peatlands in vegetation models
Tropical peatlands cover 90 to 170 million ha and store approximately 152 to 288 Pg C. They make up only 6% of global peatlands but contain 11-14% of the global peat carbon pool. Studies focusing on tropical peatlands have just started emerging in the last few decades. Although peat accumulation in all types of peatlands largely follows the same underlying principles, there are processes and mechanisms that are strikingly different between tropical and other types of peatlands.
One unique characteristic of tropical peatlands is that they are primarily dominated by trees and are commonly known as peat swamp forests. These peatland types contain a high proportion of woody material in their litter pool, which is mainly derived from dead roots and branches. In addition, the microbial communities that thrive on these unique peat materials feature inherently different characteristics than typical northern peatland microbial communities.
Aim
In this project, we will incorporate important tropical peat processes and parametrizations in the LPJ-GUESS Peatland. The main purpose of this project is to understand the current and future rates of carbon exchange in tropical peatland ecosystems at the regional level.
Contact
Atmospheric inverse modeling is an approach that consists in estimating the emissions (or removal) of an atmospheric species (e.g. CO2 or CH4) into the atmosphere, given the observed impact of these emissions on the atmospheric composition (changes in the atmospheric concentration of the modelled species). The approach relies on an atmospheric transport model, which establishes the link between the emissions and the concentrations, and on a fitting algorithm, that adjusts the emissions in order to optimize the fit to observations.
The capacity of an inversion system to estimate emissions depends largely on the accuracy of the transport model, and on the quantity of observations that can be used to constrain the emissions.
Objective
The objective of this master thesis project will be to implement the capacity to model satellite retrievals of CH4 and/or CO2 into the LUMIA inversion system. LUMIA is our in-house inversion system, and can currently use in-situ observations such as those from the ICOS network, but lacks the capacity to assimilate satellite retrievals.
Satellite retrievals are column-measurement (they measure the CO2/CH4 content integrated over the whole atmospheric column, and not at a specific point). As such, their representation in transport models is not totally straightforward. The first task in the project will therefore be the validation of the implementation (comparison with in-situ observations and with other models, diagnostic of potential biases, etc.). Depending on the student interests, the work can be extended by analyzing at specific features of the data (e.g. estimation of CO2 emissions from a large city), or by using the data in actual atmospheric inversions.
The project requires a basic understanding of meteorology, as well as good programming skills (python, at least, and ideally also Fortran).
Contact
The CH4 emissions from the NordStream leakage have been observed at several ground-based ICOS atmospheric observation sites and it has been shown that the peaks in the CH4 observations are linked to the NordStream pipeline damage (see more: https://www.icos-cp.eu/event/1221).
Instead of using atmospheric tracer transport models in a forward mode (relating emissions to concentrations) one can also use atmospheric modelling in an inverse mode that relates observed greenhouse gas (GhG) concentrations to emissions (or removals) of GhGs.
Objective
In this project the student will use pre-calculated atmospheric backward footprints from an atmospheric transport model over the leakage period to estimate an emissions profile the best fits the observed concentrations at the ICOS sites. These so-called footprints are essentially providing an estimate of how the emissions from regions upstream of the observation sites infuence the greenhouse gas concentrations at the sites (one can also think of them as the sensitivity of the concentration at the observation site to previous emissions from a pre-defined domain). Using assumptions on the possible amount and time profiles of the leaked gas the student will create an ensemble of prior emissions profiles that, together with a simple optimisation algorithm, will be used to estimate an emissions profile that best fits the observed concentrations.
Requirements
The project requires a basic understanding of meteorology, a good mathematical understanding, as well as good programming skills (python, at least, and ideally also Fortran).
Contact
Fall 2023 / Spring 2024
Background
Nuclear bomb tests in the 1950s and 1960s introduced a huge amount of radiocarbon (14C)-enriched carbon to the atmosphere. The ocean and the biosphere rapidly absorb this carbon. However, the biosphere slowly releases it back into the atmosphere through heterotrophic respiration. The isotopic disequilibrium is the difference between the radiocarbon content entering and leaving a carbon pool. It can be used to diagnose the turnover time of carbon in terrestrial carbon pools. It is also an important quantity for the estimation of fossil fuel emission from atmospheric inverse models.
Objective
In this thesis, you will estimate the global biospheric isotopic disequilibrium from 1901 to 2020 using the isotope-enabled LPJ model version developed by Scholze et al. (2003). You will also perform atmospheric transport simulations with the lagrangian model FLEXPART to compare the modeled concentrations with atmospheric ∆14C observations from the ICOS network and evaluate the feasibility of using the isotopic disequilibrium fluxes in atmospheric inverse modeling applications. You should have some programming skills in Python and FORTRAN.
Contact
Background
Terrestrial ecosystems play a key role in the global carbon cycle and climate. However, due to the complicated vegetation-climate interactions, both vegetation dynamics and climate are highly uncertain in the tropical regions and they have become one of the main challenges for further understanding the vegetation and climate dynamics in the future Anthropocene, in particular under climate extremes. Global warming along with the continuous changes in land use and land cover further complicate the overall picture.
Objective
Will climate become more extreme if we consider the role of vegetation in the future climate? For addressing this question, we need a sophisticated research strategy to further understand the vegetation-climate relationships in the Earth system.
This degree project will allow you to explore the vegetation dynamics and vegetation feedback (e.g., how the ecosystems could affect climate) under tropical regions, which include but are not limited to Africa, South America (with a focus on the Amazonian region) and Australia. Special attention will be paid to vegetation responses and feedback under extreme climate events (e.g., droughts and heat waves).
Examples of research questions
- Vegetation responses (e.g., LAI, GPP and ET) to the intensity and duration of climate extremes
- Will vegetation feedback amplify or buffer climate extremes (e.g., heatwaves and droughts)?
Methods
You’ll have access to the interesting modelling data from the coupled vegetation-climate Earth system models over then tropical regions, with the assistance of multiple observational datasets for exploring the research questions. The use of ecosystem model LPJ-GUESS and the coupled model RCA-GUESS is possible in a later stage.
We are a supportive supervision team looking for independent students who are interested in modelling vegetation dynamics and terrestrial carbon cycles from an Earth science perspective. We expect you have good analytical skills using programming-based software such as Matlab or Python, which will be fundamental to the project tasks. You will have the opportunity to promote your research findings in a wide scientific community.
Contact
Dr. Minchao Wu and Prof. Ben Smith for more information.
Do you want to help us understand primary forests in Sweden (pristine forests)?
We have recently produced a digital (GIS) map of old-growth forests in Sweden and performed a nationwide inventory of vegetation and soils. One of the motivations behind this extensive data collection was to understand how human activity affects the forests, in terms of biomass and carbon storage, in terms of structure and biodiversity, etc.
There are several opportunities to dig into our dataset for a thesis project. Two obvious topics would be measures of diversity, species or structural (i.e. tree sizes) and the influence of the largest trees for carbon storage. Students interested in these topics, or other related topics are welcome to contact me for potential thesis work.
Contact
Note: This thesis includes a field trip to Öland.
Background
The Alvar on Öland is a world natural heritage with a very high biodiversity. However it is also a cultural site since the current high biodiversity vegetation can only sustain when managed in a proper way. Over the last decades shrub encroachment has been a major threat due to a decrease in grazing intensity.
The public administration is trying to optimize their effort and to clean the areas which are mot in need. In this thesis the student was using an automatic classification scheme, trying to use RGBI images to generate a map of shrub encroachment. However the map proved to have a rather low accurracy as well as computational issues only allowed to generate classifcations for some rather small areas. Two reasons were identified for not having a better classification.
One was that for an RGB(I) classification more validation sites would be needed and the second was that not all available data was used. The final goal of the work on Öland is not only to generate a shrub map for the Alvar now but also to be able to use earlier aerial photos to identify areas with high increase in shrub encroachment over time. Therefore in this thesis you are not only using RGBI data for the classification but also vegetation height data based on stereometric analyses of RGBI data provided by Lantmäteriet.
Methods
The first step is to generate a method to classify shrubs using RGBI+Digital surface model + ground height model. Once this is done this should be used as training and validation data to generate a classification method based only on RGBI for historical orthophotos.
Contact
Background
Medium to large scale sports events in the Swedish Mountains are becoming more and more common and the number of participants is rising steeply. These events are causing a new type of damage by trampling or cycling of the vegetation.
Methods
In this thesis you are evaluating how well the effects of previous sports events can be monitored using remote sensing. You are using drone images (the flights can either be done by the student or can be arranged) and aerial photographs to evaluate the effect of the event on the vegetation.
Field trips
While this thesis contains a number of field visits and vegetation evaluation (mainly estimation of bare ground), the majority of the work lies on the processing of remotely sensed data.
Contact
The use of plastic in agricultural applications is a large problem since the plastic is often staying in the field or in the surrounding, or at least parts of it and it may take a very long time until it decomposes. Currently companies are developing alternatives for the plastic (sheets of organic origin which decompose easily) used to increase temperatures for early potatoes, however one of the hinders of market introduction is that there is no knowledge about how large the market for these products are.
In this study you will use Sentinel data to make an assessment of how much plastic is used for the production of early potatoes in Sweden. You will develop an automatic classification method to perform the analysis at a country scale.
Dryas octopetala is a common dwarf shrub in the alpine areas of both Sweden and Switzerland. While it typically forms a complete cover of the area or cushions, in rare cases it grows in the shape of rings.
As of now 3 sites are known where these rings were found, one on Svalbard and two in the Swiss national park. In all sites the special growth forms are found on an area with very small slopes, well above the tree line, and on a ground which is mainly dominated by pebbles. Also the total vegetation cover of the area is rather low (<20%).
Objective
In this thesis you are supposed to generate maps of potential occurrences of this special growth form both for the Swiss Alps as well as for the Swedish fjäll area. The study is mainly focusing on the remote sensing and analysis of spatial data by using GIS, but if there is an interest, a field work part can be added in which some of the sites can be visited to evaluate the model performance.
Contact
Background
Lake surface water temperature (LSWT) is an important physical property of lakes. LSWT is a critical parameter for evaluating the water quality and biodiversity in lakes. It is also an indicator of climate change. Therefore, it is crucial to monitor LSWT to improve our understanding of the spatiotemporal dynamics of LSWT for many applications.
Conventionally and ideally, we can install in-situ gauge stations or monitoring sites with devices to measure surface water temperature in lakes, and these in-situ measurements are generally the most accurate. However, in-situ measurements in lakes are often sparse and limited in terms of spatial coverage (many lakes have no measurements) and temporal length (long-term continuous measurements are lacking). Thus, in-situ measurements cannot sufficiently capture the spatiotemporal dynamics of LSWT in lakes, particularly large lakes. Satellite remote sensing and lake models are two main methods to estimate LSWT, which can complement the limited in-situ measurements for monitoring spatiotemporal dynamics of LSWT.
Aim
This Master thesis aims to estimate/model lake surface water temperature in Lake Vänern, Sweden (other lakes could also be considered). Satellite products and classic lake models would be explored. This thesis will involve processing large-size satellite data, implementing models and evaluating data and model results. This thesis is suitable for students with a good knowledge of programming (ideally in Python) and good skills in spatial analysis using GIS tools.
Contact
This master thesis topic is in line with a project related to monitoring and risk mapping of Bark Beetle attacks in Southern part of Sweden.
Aim
The aim of this master thesis is to model the association between Bark Beetle attacks with the environmental factors using Machine Learning. The contribution of the study is to study how spatial auto-correlation and spatial errors can be integrated into the ML modeling and if the new developed model increases the accuracy of Bark Beetle attack prediction.
The topic is suitable for students who like to research and learn more about Geo-data Science and Machin Learning. A good knowledge of Python programming is required for the implementation of the model, while available ML libraries will be used.
Contact
Ali Mansourian
The hot drought in 2018 triggered a large spruce bark beetle outbreak in southern Sweden. The outbreak has killed several million m2 of spruce trees and is still ongoing. We have recently started a project to develop methods to monitor spruce bark beetle attacks with Sentinel-2 data.
In the project we have work packages focusing on:
- studying risk factors for bark beetle outbreaks,
- detection of bark beetle attacked trees in Sentinel-2 data
- modelling of bark beetle phenology
We welcome master students to work in the project and have different suggestions for thesis topics. The thesis can be focusing on remote sensing, how can we detect bark beetle outbreaks in remote sensing data? It is also possible to write a thesis directed towards modelling e.g. bark beetle phenology or with a focus on management strategies. A thesis can also be more related to GIS with a focus on risk factors for outbreaks.
Contact
The number of 3D city models in the Worlds is growing and the most common standard for 3D city models is CityGML. A new version, CityGML 3.0 (Conceptual model) was recently approved as an official OGC Standard with several changes compared to CityGML 2.0 that opens up for interesting studies.
Some changes are that the core-, building- and transportation modules are updated and that a module for versioning of city models is introduced. One major difference compared to CityGML 2.0 is the space concept that is used to "build" the objects in the city model. Spaces can be either physical spaces or logical spaces. For example the physical parts of a building (building, building parts) consists of physical spaces but there is also a possibility to divide the building into logical spaces based on e.g. ownership or function. One thing to study is if the logical spaces can be used to divide a building into different (logical) parts based on the function. This is something that will be tested in 2022 by Lantmäteriet for a new national specifications for buildings.
There are also other aspects of CityGML 3.0 that can be studied, related to transportation and versioning of city models.
Contact
Background
Pristine forests are rare in boreal Eurasia and it is debated if they sequester carbon or if they are in equilibrium with no net growth. One way to investigate that is to use remote sensing to investigate trends in vegetation indices from the 1980s until present. We have done this for Swedish pristine forests and found strong trends, in this project a similar study would be done for western Russia.
Objective
We have a copy of the original GIS files used for the map and the report (http://old.forest.ru/eng/publications/intact/). It was first produced around the turn of the century but it isn’t fully clear what is included and what definitions were used. The first part of this project would be to read up on the production of the map and compare to other available datasets on intact forests. Most information is probably given in the publications for which the map was used. When that is done, trends can be extracted from e.g. Google Earth engine to investigate if these forests has become greener and if productivity has increased.
The project is suited for a driven student with very good skills in GIS and some programming.
Contact
Background
Spectral remote sensing can be used to discriminate between different types of forest. The optical characteristics of trees are determined by properties such as the canopy arrangement, and the sizes, shapes and internal structures of their leaves. Many of these properties vary between tree species, which makes it possible to map different tree species with remotely sensed spectral data. However, because many tree species have relatively similar optical characteristics, there is a high risk of confusing species, particularly if only one remotely sensed image is used.
Tree species also differ in their phenological characteristics, such as the dates of bud burst in the spring, and the change in coloration and the dropping of leaves during autumn. These phenological characteristics of trees influence their physical appearance, and can be monitored with the help of earth observation data, if the data has short revisiting intervals such as those acquired by the Copernicus Sentinel-2 satellite mission.
Objective
In order to compare the spectral reflectance of different tree species, field-collected validation data are needed. A common limitation in remote sensing studies is a lack of high quality validation data. Luckily, some cities regularly make inventories of trees in parks, along streets, and other urban areas.
In Helsingborg, the municipality have stored data at the level of individual trees, including information on the species identity of several thousands of trees, and in some cases also information on the tree height and crown width. Can these data be used to increase our knowledge about the spectral characteristics of different tree species, or to monitor changes in the status of individual trees over time?
Contact
Background
Gross primary production, the uptake of CO2 by the vegetation, is an essential climate variable for monitoring of the terrestrial biosphere and its response to climate change.
We recently produced a novel satellite based product on global-scale gross primary production of the vegetation 1982-2015. However, the dynamics of this data set have not yet been fully analyzed. Recent advances in time-series and breakpoint analysis open new possibilities, as they allow for the detection of nonlinear trends and turning points in these data.
Objective
The overarching objective of this thesis is to apply novel time-series analysis to study dynamics and possible reasons for trends and breakpoints in the vegetation productivity 1982-2015.
Contact
Semantic web techniques and linked data (knowledge graphs) are increasingly used in the geospatial domain and are interesting for e.g. the next generation of spatial data infrastructure. At the department, we have conducted some research in linked geospatial data and are open for master thesis in the topic. The thesis could either be devoted technical issues (ontologies, RDF stores, etc.) or have a more applied focus.
Contact
Background
The ice phase of storms is the cause of many aspects of severe weather. Freezing rain, hail and lightning from ice in clouds together cause many billions of euros of damage per year in Europe and USA. Yet the processes of ice initiation in storm clouds are still uncertain. There is evidence from aircraft observations that fragmentation of ice somehow can generate most of the ice particles in such clouds. Thus, lab experiments are required to quantify some mechanisms of fragmentation.
Objective
The prospective student will perform an experiment in our laboratory at INES to characterise one such mechanism. The student will create apparatus to observe collisions between a graupel particle grown by riming and a snow particle from by accretion of crystals using a cold box in an open-top freezer. Video imagery will be analysed computationally so as to count the numbers of fragments from snow-graupel collisions in the cold-box at various temperatures. An existing theoretical formulation will be adapted to fit the results (e.g. for application in a numerical model of clouds).
The student should have an interest in the physics of materials and have experience in programming and in practical experimentation.
Contact
Background
Given the importance of gross primary production (GPP) for monitoring of the terrestrial biosphere and its response to climate change, a consortium earth observation-based GPP products is of high societal relevance.
Today, the publicly available satellite based products derives from just a few satellite sensors. In order for spatially explicit GPP to be properly assessed, it is extremely important that monitoring products are independent and based on different input data. Sentinel is a series of newly launched satellites having global coverage of reflectance in high spatial and spectral resolution, but there are still no GPP products generated from the Sentinel data.
Objective
The overarching objective of this thesis is to combine information from the Sentinel-3 satellite with light use efficiency estimates from a dynamic global vegetation model to generate a state-of-the-art satellite product of GPP. The product will thereafter be evaluated against ground observations.
Contact
Background
The Dahra field site in Senegal, West Africa, was established as an in-situ research site to improve our knowledge regarding properties of semi-arid savanna ecosystems and their responses to climatic and environmental changes. A strong focus of the instrumental setup is to gain insight into the relationships between ground surface reflectance and savanna ecosystem properties for earth observation upscaling purposes using satellite data.
Objective
This master thesis will make use of a unique in-situ data set of reflectance between 350-1800 nm and investigate how ecosystem properties of semi-arid savanna ecosystems influence the reflectance from the ground surface.
Contact
Background
The ongoing rapid warming over the Arctic has caused dramatic changes in environmental conditions (e.g. snow cover thickness and duration, active layer depth and soil moisture and temperature) and ecosystem dynamics. The carbon cycle in dry and wet ecosystems may respond to climatic warming in contrasting ways.
Objective
This project aims to elucidate the long-term dynamics of CO2 fluxes for the neighboring High-Arctic dry and wet ecosystems respectively and identify their biotic and abiotic drivers. The results will improve our understanding or prediction on how dry and wet Arctic ecosystems contribute to greenhouse gas emissions due to the present or future climate change.
The main research work will be based on an integration of a process-oriented model and fluxes measurement data sets (the method is similar to Zhang et al., 2018).
This project should not necessarily require the candidate to have advanced programming skills, but the candidate should know about ecosystem modelling and be able to make statistical analysis and plotting using Matlab or R or other similar tools.
Contact
There is an increasing amount of 3D building models (BIM) and 3D city models. There are currently several research and innovation projects of the usage of these models in the digitalization of planning and building processes in the society, concentrating on issues such as automation of building permits, 3D cadaster issues and environmental modeling.
The department is involved in several of those projects (many as part of the Smart Built Environment program – see http://www.smartbuilt.se/).
We are open for M.SC. projects in several topics concerning the usage of BIM and city models connected to our projects.
Contact
Detection and analysis of large scale ecosystem functional relationships
Functional relationships are powerful tools to analyze large datasets, from dynamic models and from upscaled observations. The aim of the project would be to find new regions where a functional relationship between environmental drivers and ecosystem response can be found. Suitable regions vary mainly along a single environmental variable, such as water availability or temperature, but not both. That way the ecosystem response to a single environmental driver can be analyzed.
Objectives
The proposed workflow involves scientific programming to identify potential regions, regions that mainly vary in one variable across space. The second step would be to compare the functional relationships of empirical datasets on ecosystem carbon cycle (GPP, NPP, biomass, soil carbon etc) as well as outputs from climate models and ecosystem models. The project therefore involves handling of large datasets, something that is common in global analysis and climate change research.
The ideal candidate has some knowledge in scientific programming and spatial data.
For more information see our previous publication on a functional relationship in Amazonia
Contact
Background
Lightning in storms is caused by electric fields arising from charge separation in re-bounding ice-ice collisions. The project aims at evaluating laboratory observations of charge separation in ice-ice collisions in the absence of liquid. This is a weaker form of charging than that usually studied but may be significant owing to the prevalence of ice-only cloud in storms.
Objectives
The prospective student will simulate the ice crystal growth and collisions in a specific experiment observing charging that was performed years ago and described in a published paper in 2016. The student will create a simple model to replicate the experiment and then do variational analysis to deduce a formulation for the charge separation that can be applied in cloud models. The student should have an interest in computer modelling and programming (Fortran or C) and should be familiar with a LINUX-based operating system.
Contact
Background
Map labeling, the process of placing names and icons on maps, is a critical and time-consuming task of cartography. It significantly affects the usability and effectiveness of maps for navigation, communication, and decision-making. Evaluation of map labeling is essential to ensure accurate, legible, and user-friendly maps. Traditionally, this evaluation has been conducted manually, which is time-consuming and subject to human bias. Automation of the map evaluation step can be valuable, particularly in high-density labeled maps. It can detect automatically the bad labels or labeled regions and it can help selecting the best solution from a multiple generated solution, thus saving time compared to manual inspection. Therefore, deep learning techniques offer a promising avenue for automating map labeling evaluation.
Aim of the Thesis
The primary goal of this master thesis is to develop a framework for the automated evaluation of map labeling using deep learning. This framework will help to learn the patterns of good labeled maps and assess the quality of label placement.
Method
The supervisor will provide the London map data with different quality categories (bad and good labeled maps). The student will develop a deep learning model to identify the bad labels that should be improved. This involves the preprocessing of the raw provided data to make it suitable for the input of the model. Additionally, the student will analyze the patterns and insights gained from the model's learning process.
The student should be familiar with spatial data processing, analysis, and mining. The student should have intermediate programming skills in Python. A more specific introduction and data for the topic will be given by the supervisor.
Contact
Background
Predicting energy consumption is crucial for sustainable urban planning, resource management, and environmental conservation. Traditional energy consumption prediction models often rely on historical data, but these models may not capture the dynamic nature of energy use. Nighttime light data, captured by satellites, is a valuable source of information, as it reflects human activity patterns and urban development. However, utilizing night light data for energy consumption prediction remains a challenging task due to data complexity and spatial-temporal dynamics. Deep learning techniques have shown promise in handling complex data and making accurate predictions, making them an ideal approach to address this challenge.
Aim of the Thesis
This master's thesis aims to develop a novel framework for predicting energy consumption in urban areas using night light data and deep learning. The primary objectives are to create a model that can accurately forecast energy consumption patterns and understand the impact of human activity on energy usage. Additionally, the thesis seeks to explore the potential applications of this predictive model in urban planning, energy resource management, and environmental sustainability.
Method
The workflow starts with the collection of the night light image data in Sweden, which is openly available. Then, a deep learning model should be developed so to estimate the energy consumption in urban areas from night light images on annual basis. The last step is to conduct a spatiotemporal analysis to unveil the evolving patterns of energy consumption in Sweden.
The student should be familiar with spatial data processing, analysis, and mining. The student should have intermediate programming skills (Python or R). A more specific introduction and data for the topic will be given by the supervisor upon request.
Contact
Rachid Oucheikh and Pengxiang Zhao
Background
Estimating solar radiation using Google Street View data has diverse applications, including urban planning for optimized natural light, environmental impact assessment, solar energy planning, health and well-being promotion, microclimate studies, energy-efficient building design, public health campaigns, agriculture optimization, climate change mitigation, educational initiatives, and environmental conservation. It is a versatile tool for fostering sustainability and comfort in cities while addressing energy efficiency, public health, and climate change concerns.
Estimating solar radiation in urban street canyons is complex due to challenges such as shading from buildings, the intricate geometry of canyons, microclimatic variations, temporal changes in solar patterns, limited data availability, modeling intricacies, and calibration needs. These challenges demand advanced modeling, comprehensive data strategies, and interdisciplinary cooperation to achieve accurate solar radiation estimates in city environments.
Aim
The aim of this topic is to to utilize Google Street View data and deep learning techniques to estimate solar radiation within urban street canyons.
Methods
Some specific required tasks are: The first step is to collect the street view image data and DEM. The data may undergo some preprocessing steps, including hemispherical projection and segmentation. The second step is to develop a deep learning model enabling the analysis of solar irradiation patterns over time and space and allowing to estimate the solar radiation. The third step is to analyze the results obtained by the model.
The student should be familiar with spatial data processing, analysis, and mining. The student should have intermediate programming skills (Python or R). A more specific introduction and data for the topic will be given by the supervisor upon request.
Contact
Topic
Ongoing climate change and land use affect ecosystem properties and the provisioning of ecosystem services. There is a growing societal need of developing adaptation and mitigation strategies in relation to urban development, forest management, agricultural practice and nature conservation.
The thesis work can focus on one sector and cover one or several of the following aspects: 1) identify key stakeholders and evaluate their perspectives and preferences for different ecosystem services, 2) perform scenario analyses and present possible future development directions, and 3) review goal conflicts and uncertainties in a climate adaptation process.
Contact
Different optical satellite sensors, such as MODIS onboard the Terra and Aqua satellites and OLCI onboard Sentinel-3, typically have different central wavelengths, bandwidths for the same red, NIR bands etc. This leads to inconsistencies when building a long time series of vegetation indices (VI) from multiple sensors, causing problems when analyzing vegetation trends using VI. Therefore, to generate a single product from multiple satellite sensors—known as a virtual constellation (Claverie, 2023)—bandpass adjustments are necessary. This involves correcting the bandpass using the spectral response function of each sensor.
In this master’s thesis project, you will investigate the inconsistency in BRDF reflectances between the MODIS and Sentinel-3 OLCI sensors.
The results of your work are expected to contribute to our ongoing project of Copernicus Land Monitoring Service (CLMS) MR-VPP product continuation using Sentinel-3.
Requirements: Experience in satellite data analysis and programming is essential for this project.
Read more contact researchers (project 2)
The current Bidirectional Reflectance Distribution Function (BRDF) correction for high-resolution vegetation remote sensing products (e.g., HLS, Claverie et al. 2018) is primarily based on the simplified method developed by Roy et al. (2016, 2017), which assumes a fixed set of BRDF parameters globally. However, these BRDF parameters are not static, and such an approach often leads to suboptimal correction performance.
In this master’s thesis project, you will explore several alternative methods to improve BRDF correction for vegetation estimation using Sentinel-2, including:
- Land cover-specific parameters
- Temporally varied parameters based on vegetation indices (VI)
- A simple VJB inversion method (Vermote et al., 2009)
Read more and contact:
Background
Estimating solar radiation using Google Street View data has diverse applications, including urban planning for optimized natural light, environmental impact assessment, solar energy planning, health and well-being promotion, microclimate studies, energy-efficient building design, public health campaigns, agriculture optimization, climate change mitigation, educational initiatives, and environmental conservation. It is a versatile tool for fostering sustainability and comfort in cities while addressing energy efficiency, public health, and climate change concerns.
Estimating solar radiation in urban street canyons is complex due to challenges such as shading from buildings, the intricate geometry of canyons, microclimatic variations, temporal changes in solar patterns, limited data availability, modeling intricacies, and calibration needs. These challenges demand advanced modeling, comprehensive data strategies, and interdisciplinary cooperation to achieve accurate solar radiation estimates in city environments.
Aim
The aim of this topic is to to utilize Google Street View data and deep learning techniques to estimate solar radiation within urban street canyons.
Methods
Some specific required tasks are: The first step is to collect the street view image data and DEM. The data may undergo some preprocessing steps, including hemispherical projection and segmentation. The second step is to develop a deep learning model enabling the analysis of solar irradiation patterns over time and space and allowing to estimate the solar radiation. The third step is to analyze the results obtained by the model.
The student should be familiar with spatial data processing, analysis, and mining. The student should have intermediate programming skills (Python or R). A more specific introduction and data for the topic will be given by the supervisor upon request.
Contact
Topic
Ongoing climate change and land use affect ecosystem properties and the provisioning of ecosystem services. There is a growing societal need of developing adaptation and mitigation strategies in relation to urban development, forest management, agricultural practice and nature conservation.
The thesis work can focus on one sector and cover one or several of the following aspects: 1) identify key stakeholders and evaluate their perspectives and preferences for different ecosystem services, 2) perform scenario analyses and present possible future development directions, and 3) review goal conflicts and uncertainties in a climate adaptation process.
Contact
Want to add or remove suggestions?
Contact Susanna Olsson