http://trepo.tuni.fi:80/handle/10024/105882 fi=Pro gradut, Diplomityöt, Syventävät työt (lääketiede), Lisensiaatintyöt|en=Master's theses, Master's theses (medicine), Licentiate theses| Tue, 15 Sep 2026 05:33:48 GMT 2026-09-15T05:33:48Z http://trepo.tuni.fi:80/handle/10024/240134 Bridging the Implementation Gap: Value Sensitive Design Requirements for Digital Sobriety in AI SME Platforms. The DigiDeus Case Schwidrowski, Venus Thu, 01 Jan 2026 00:00:00 GMT http://trepo.tuni.fi:80/handle/10024/240134 2026-01-01T00:00:00Z http://trepo.tuni.fi:80/handle/10024/240024 Scenario-based Assessment of Truck Electrification: Estimating Emissions and Impact on Charging Infrastructure in Helsinki Metropolitan Area A Simulation Based Evaluation Khalid, Muhammad Shuaib Climate change is real and so its repercussions – as visible as the broad day light. Carbon emissions are considered as a major contributor in greenhouse gases, and urban logistics are significantly contributing to these emissions especially in compact and densely populated urban cities where freight vehicles operate intensively. Although these vehicles take a small proportion of the road transport fleet, but they account for a large share of carbon emissions which makes them an important target for decarbonizing and for achieving carbon neutrality. Consequently, the electrification of urban logistics has been widely recognized in the literature as one of the most promising approaches for reducing emissions from freight transport. This thesis explores the pathway towards Zero Emission Urban Logistics (ZEUL) by investigating four electrification scenarios for Helsinki Metropolitan Area in the year 2040 using the SUMO. These scenarios are simulated to evaluate the potential reduction in carbon emissions under different levels of fleet electrification and to examine the resulting impacts on charging infrastructure. In addition, two urban logistics measure i.e., Zero Emission Zones (ZEZs) and Megawatt Charging Systems (MCS) were incorporated into the model to assess their contribution to emission reduction and charging performance. The simulation results showed a clear relationship between the level of fleet electrification and carbon emissions. As it increased, carbon emissions decreased compared with the Base Scenario. At the same time, higher levels of electrification led to an increase in the number of vehicles using charging stations, greater electricity demand, and longer average vehicle charging duration. The analysis also identified locations where additional charging infrastructure would be required to support future electrification. When ZEZs were introduced alongside fleet electrification, the reduction in daily carbon emissions was substantially greater than with electrification alone. It was recorded that, with accelerated electrification alongside with city-wide ZEZ a reduction of 44% of CO2 emissions per day from ICE truck and 55% of CO2 emissions reduction per day from ICE trailer can be achieved. Similarly, upgrading charging stations to Megawatt Charging Systems significantly reduced the average charging time up to 20 mins for heavy-duty electric vehicles. These findings provide practical insights for policymakers and urban planners in the Helsinki Metropolitan Area, supporting evidence-based decisions on charging infrastructure planning, zoning policies, and the transition towards zero emission urban logistics (ZEUL). The study’s limitations are mainly related to the assumptions made regarding vehicle characteristics and their standardization, including vehicle size, vehicle type, emission classification, and battery state of charge. Despite these limitations, the study contributes to the existing scientific knowledge by providing a methodological framework for evaluating metropolitan-scale urban logistics electrification scenarios and assessing their potential in achieving ZEUL. Thu, 01 Jan 2026 00:00:00 GMT http://trepo.tuni.fi:80/handle/10024/240024 2026-01-01T00:00:00Z http://trepo.tuni.fi:80/handle/10024/239994 Työehtosopimuksen rooli palkanmuodostuksessa Suomessa Toivonen, Tomi Työehtosopimuksia neuvotellaan Suomessa yleensä kansallisella toimialatasolla ammattiliiton ja työnantajaliiton välillä. Taloustieteen empiirisen kirjallisuuden mukaan ammattiliitot tyypillisesti nostavat ammattiliittoon järjestäytyneiden työntekijöiden palkkoja suhteessa ammattiliiton ulkopuolella oleviin palkansaajiin. Suomessa tällainen palkkapreemio voi syntyä työehtosopimusneuvotteluiden seurauksena muodostuvan työehtosopimuksen kautta, jossa sovitaan tyypillisesti palkankorotuksista ja palkkatasoista eri ammattiryhmille ja tehtävätasoille. Tässä tutkielmassa arvioidaan ensimmäisenä Suomessa, mikä on työehtosopimuksiin liittyvä keskimääräinen palkkapreemio ja mikä on niiden yhteys palkkarakenteeseen Suomessa. Empiirisessä analyysissä hyödynnetään Tilastokeskuksen harmonisoitua palkkarakennepaneelia yksityiseltä ja julkiselta sektorilta ja yksityiselle sektorille tunnistettua työehtosopimusaineistoa. Analyysi perustuu Abowdin, Kramarzin ja Margoliksen (1999) kaksisuuntaisten kiinteiden vaikutusten menetelmään (AKM-menetelmään). Työehtosopimuksiin liittyvä keskimääräinen palkkapreemio estimoidaan hyödyntämällä työntekijöiden siirtymiä työpaikasta toiseen sekä muutoksia työehtosopimusstatuksessa, joka tunnistetaan työehtosopimusaineistolla sekä työnantajien järjestäytymättömyyttä koskevien tietojen avulla. Tulokset osoittavat, että työehtosopimusten piirissä maksetaan keskimäärin noin 0,9 prosenttia korkeampia palkkapreemioita verrattuna järjestäytymättömien työnantajien palkansaajiin. Arvioitu palkkapreemio on pienempi kuin aiempien maiden tutkimuksissa, mutta pääosin yhdenmukainen sellaisten maiden tutkimusten kanssa, jossa työehtosopimusten kattavuus on korkea. Kun huomioidaan yritysten väliset tuottavuuserot työntekijää kohden lasketulla jalostusarvon logaritmilla, havaitaan työehtosopimusten mahdollistavan erityisesti sellaisten yritysten palkkapreemioiden nostamisen, jossa työntekijäkohtainen tuottavuus on matala. Tutkielma tarjoaa ensimmäisen arvion työehtosopimuksiin liittyvästä keskimääräisestä palkkapreemiosta Suomessa ja luo pohjaa jatkotutkimukselle kattavampien työehtosopimusaineistojen ja AKM-menetelmän avulla. Thu, 01 Jan 2026 00:00:00 GMT http://trepo.tuni.fi:80/handle/10024/239994 2026-01-01T00:00:00Z http://trepo.tuni.fi:80/handle/10024/239671 Machine Learning-Based Detection of Pollution Spikes in Atmospheric Aerosol Data : A Comparative Study Khati, Luv Atmospheric aerosol monitoring at remote background stations such as the Atmospheric Radiation Measurement (ARM) Eastern North Atlantic facility on Graciosa Island, Azores, Portugal, generates large volumes of continuous time series data that require automated methods for identifying pollution spike events. Pollution spikes are sudden and abnormal changes in particle number concentration that deviate significantly from the normal background aerosol levels, and accurately detecting them is important for maintaining the quality and reliability of long term atmospheric datasets. This thesis evaluated and compared three unsupervised machine learning methods, namely Isolation Forest, Long Short-Term Memory (LSTM) networks, and Autoencoders, for the automated detection of pollution spikes in Scanning Mobility Particle Sizer data collected at Graciosa Island from January to December 2025. The three methods were applied to the same preprocessed dataset and evaluated using execution time, Silhouette Score, Spike Persistence, and agreement with the Eastern North Atlantic Aerosol Mask (ENA-AM) rule based baseline method, with these measures selected specifically because no verified ground truth dataset of confirmed pollution spikes exists for this site. The results showed that no single method performed best across every measure considered. The LSTM network achieved the highest Silhouette Score (0.9626) and the highest Spike Persistence (2.0622), indicating the most internally coherent and temporally sustained detections, while Isolation Forest was the fastest method and showed the closest, though still modest, agreement with the ENA-AM baseline among the three methods. Agreement with ENA-AM remained limited overall for all three methods, reflecting both the substantially larger proportion of the dataset flagged by ENA-AM. The auxiliary dataset validation further showed that a number of LSTM detected spikes coincided with flight activity and combustion related trace gases, offering supporting, contextual evidence for these detections rather than confirmed causal explanations. The findings of this study suggest that machine learning based methods represent a viable and flexible alternative to existing rule based approaches for automated pollution spike detection in long term atmospheric aerosol datasets, offering data driven detection without requiring manually defined thresholds. At the same time, the results indicate that the choice of method may depend on which property, internal detection coherence, computational efficiency, or correspondence with an existing rule based baseline, is prioritised for a given application, and that further work is needed to establish a verified ground truth and a more tightly controlled comparison across methods. Thu, 01 Jan 2026 00:00:00 GMT http://trepo.tuni.fi:80/handle/10024/239671 2026-01-01T00:00:00Z