Motivation letter

Please write a brief motivation letter (in English, maximum of 6000 characters with spaces) based on the following points:

1. Why are you interested in this PhD project, explain your choice.
2. What is your previous experience in this field? Explain how your educational and professional background relates to the project you are applying to.
3. What are the analytical/scientific methods you have practiced.
4. Describe briefly the methods and main results of your MSc thesis.
5. Decribe your earlier research activities, including research publications and conference presentations, if available.
Assessment criteria for motivation letter:

- motivation and argumentation of skills and the choice of the project
- relevant study and work experience and other relevant activities (publications, project management etc.) as required to present in the motivation letter.

Interview

The applicant must describe the wider scientific background of the doctoral project and possible applicability of the results, also their motivation to be admitted to PhD studies with particular project. Only applicants whose motivation letter is assessed positively will be invited to the interview (minimum positive result is 35 points out of 50).

The entrance interview is used to assess the following:

- knowledge of the wider scientific background of the project and possible application of the expected results
- applicant’s motivation to pursue doctoral studies in the relevant field of science and to work in this field
- wider analytical and generalization skills regarding the research and study topics.

International applicants who cannot be present at the interview in Tartu, may conduct an online interview. Applicants will be informed of their interview date and time by the respective faculty.

Both the motivation letter and entrance interview are assessed on a scale of 0 to 50 points, minimum positive score is at least 35 points. To be invited to an interview, the applicant must earn at least 35 points for the motivation letter.

Supervisor: Leho Tedersoo

The doctoral student is admitted to a student's place of study, full-time, the nominal study period is four years. An employment contract for a junior researcher is not concluded. The planned start of study and research is November 1, 2023.

This project aims to disentangle the biotic and abiotic factors driving desert microbiome, using state-of-the-art molecular, bioinformatics and statistical methods. The purpose is also to understand how planting of semidesert trees may benefit from the local and inoculated microbiome. Fungi and bacteria are also isolated from rocky substrates, lichen thalli and plant leaves for further collaborative studies of their bioactive properties. In all, the project seeks for ways how to ameliorate climate and land use change impacts on desert ecosystems.

Supervisor: Leho Tedersoo

The doctoral student is admitted to a student's place of study, full-time, the nominal study period is four years. An employment contract for a junior researcher is not concluded. The planned start of study and research is November 1, 2023.

Microorganisms such as fungi and bacteria form diverse communities in grassland ecosystems, with apparent specialisation by vegetation types, soil nutrients and pH. However, at the ecosystem level, there is no holistic understanding to what extent microbial communities overlap within grassland habitats and compared with forest and cropland ecosystems. This project aims to determine habitat specificity in different fungal and bacterial functional groups and the effect of sampling strategies on the efficiency of recovering microbial diversity. The project uses cutting-edge molecular methods for identification, bioinformatics and statistics. The potential importance is very high from the microbial ecology perspective, because knowledge about the relative habitat preferences of fungi and bacteria enable to understand how broad are the microbial niches and hoe this can be used in the conservation perspective.

Supervisor(s): Karl Kruusamäe

In Industry 5.0 robots plays a crucial role in relieving workers from repetitive, unhealthy or dangerous tasks by leveraging human-robot collaboration (HRC). Two high impact areas that seek to benefit from HRC are manufacturing (i.e. lessening the physical demand on workers) and waste management (i.e. keeping people from potentially hazardous materials, including nuclear waste). By addressing these application domains for achieving safe human-robot interaction, one targets the key scenarios in safe HRI: 1) co-located HRC and 2) user-experience during remote telerobotics. In this thesis, the aim is to develop a shared autonomy human-robot system that leverages augmented reality, digital twins, and deep reinforcement learning for safe human-robot interaction with potential use-cases in manufacturing and waste management

Supervisor(s): Chinmaya Kumar Dehury

It is predicted by Gartner that 50% of the enterprise-generated data will be created and processed in the edge infrastructure comprised of ever-increasing billions of edge devices. Instead of implementing the intelligence in the cloud environment giving the full responsibility to manage a large number of edge devices, a minimal and required intelligence is imposed on the edge devices, enabling the edge intelligence that works in a master-worker approach. The major problem with this approach is that an edge device relies mostly on the data collected by the onboard sensors and the built-in or cloud-instructed intelligence, resulting in little scope for cooperation and collaboration among peers and hence no device-to-device knowledge transfer.

Unlike the traditional approach to cluster the IoT devices based on the location and other meta information, a new research direction focuses on clustering the intelligence that the edge infrastructure is equipped with, also known as Clustered Edge Intelligence (CEI). Intelligence discoverability and observability are some of the primary challenges that need to be addressed to realize the full capacity of CEI. This Ph.D. project will focus on designing and developing dynamic intelligence clustering methodology with the discoverability and observability feature for large-scale edge infrastructure.

Supervisor(s): Feras Mahmoud Naji Awaysheh

The advent of Federated learning (FL) unleashed edge intelligence capabilities at full scale without sharing confidential data over large-scale data analytics. Such environments require iteratively sharing model parameters, creating a costly communication bottleneck over the network. In particular, communication overhead induced by the ML updates over resource-limited communication networks, such as wireless networks with limited bandwidth and power. This project aims to address this issue by investigating several communication efficiency techniques, including personalized learning, where the model is tailored based on the client dataset. Correspondingly, the project will examine the advantages of employing automated ML pipelines for the FL framework in an AutoFL architecture. This framework will incorporate best practices and techniques to address privacy and communication latency issues. 

Supervisor(s): Kaido Soosaar, Ülo Mander, Lulie Melling (Sarawak Tropical Peat Research Institute)

Wetlands cover 5-8% of the world's land area and have a tremendous capacity to sequester
carbon (C) from the atmosphere. Natural wetlands effectively accumulate C effectively due to
water-logged conditions promoting highly stable C content. Currently, there is still a great deal
of uncertainty regarding the spatial extent of restored wetlands and the extent of C sinks, as
well as source estimates and sustainable restoration alternatives. In addition, there are
uncertainties related to the impacts of climate change on greenhouse gas fluxes, particularly for
extreme weather events such as droughts and floods. Currently, there is a lack of national
emission factors to account for GHG fluxes from restored wetlands, especially as it relates to
CH4 fluxes. These issues hinder the efficient use of wetlands for GHG mitigation and adaptation
in the context of other LULUCF mitigation options.
The dissertation will add to the current state of knowledge on wetlands, their use and
degradation in Europe. Several new experimental data on ecosystem responses to wetland
management and restoration under different land uses will be collected and summarised in
relation to biodiversity and other ecosystem services.
This work will be based on the data obtained from the ALFAwetlands project joint database
(27 sites in Europe): soil CO2, CH4 and N2O fluxes, automated and periodic environmental
parameters, including precipitation, soil temperature, water table and topsoil moisture, data
collected over a two-year period. To establish the soil carbon balance, C transfer in litter, C
stocks in above- and below-ground biomass, and C turnover in litter decomposition are
quantified by synthesising high-quality research data and data from field studies. GHG flux
data from the database will be linked to detailed information on peat composition, soil and
water biogeochemistry to improve process-based modelling of peatland GHG emissions.

Supervisor(s): Alexander Kmoch

Soil is a complex agglomeration of inorganic and organic particles that is essential to all life. Knowing where and how soil thrives or degrades is important to a wide range of domains. However, despite the rapid growth of readily available remotely sensed data, soil properties are still difficult to map on large scale. Because it is not feasible to study large areas by regularly taking thousands of field samples due to the high time requirements and cost, digital soil mapping (predictive spatial modeling) with machine learning (ML) is becoming an irreplaceable tool to capture the spatial variability of soil properties. There is also a large amount of satellite-derived spectral indices that have not exhaustively been assessed for their use in spatial ML-based soil properties modelling. Another specific challenge with ML approaches in spatial modelling is the phenomenon of spatial dependence, i.e., spatial autocorrelation that is not inherently considered in ML and needs to be integrated properly. Thus, the aim of the project is to develop and interpret spatial machine learning-based models for national (Estonia) and regional (Europe) scale, for the target soil parameters clay, silt, and sand, organic carbon content (SOC), pH (CaCl2, H2O) and to identify the most important features and explain the target variables’ spatial distribution under consideration of spatial autocorrelation and spatial sampling optimization.