This thesis deals with the performance measurement of REMAK, a.s. In the theoretical section are explained performance related concepts, there are descriptions of individual systems for performance measurement and analysis used in the desription of the current state of company. The following part focuses on the description of the analyzed company and a description of the current situation using individual analyzes and model START PLUS. Next part is own suggestions for solutions and benefits of the proposals.

Für die Luftqualitätsbewertung in Städten sind Informationen zur raumzeitlichen Variabilität luftgetragener Feinstäube auf kleiner Skala von wichtiger Bedeutung. Standardisierte Messverfahren, zur Bestimmung von Partikelkonzentrationen, sind mit hohem Aufwand verbunden, weshalb dichte Messnetze fehlen. Partikelausbreitungsmodelle sind kompliziert in der Anwendung und/oder benötigen hohe Computerrechenleistung. Infolgedessen gibt es bezüglich örtlicher Partikelkonzentrationen große Informationslücken. Diese Arbeit untersucht die mikroskalige Variabilität von Aerosolen in Raum und Zeit mit unterschiedlichen Methoden. Es wurden Erhebungen mit mobilen Sensoren und eine Passantenbefragung durchgeführt. Weiterhin wurden in dieser Arbeit die physikalischen Partikeltransportmodelle ENVI-met und Austal2000 in ihrer Leistung bewertet und in angewandten Studien eingesetzt. Weiterhin wurde ein neuronales Netzwerk zur Vorhersage von Partikelkonzentrationen entwickelt. Die Untersuchungen erfolgten in den Städten Aachen und Münster. Es konnten unerwartete Verteilungsmuster hinsichtlich der Massekonzentration von Partikeln beobachtet werden. In einem innerstädtischen Park wurden diffuse Partikelquellen identifiziert, mit einem deutlichen Hinweis darauf, dass feuchtgelagerte Wegedecken einen maßgeblichen Anteil an lokalen Partikelimmissionen hatten. Weiterhin wurde Straßenverkehr als wichtiger Beitrag zum städtischen Aerosol identifiziert. Passanten, die verschiedenen Partikelkonzentrationen ausgesetzt waren, konnten diese perzeptiv nicht unterscheiden. Simulationsergebnisse von Austal2000 und ENVI-met wiesen Unterschätzungen im Vergleich zu Messwerten auf. Das entwickelte neuronale Netzwerk prognostizierte Partikelkonzentrationen teilweise mit hoher Genauigkeit. Das große Potenzial von neuronalen Netzen für die Vorhersage von Partikelkonzentrationen in räumlicher und zeitlicher Ausdehnung, auch für den Bereich der Luftqualitätsüberwachung, wurde aufgezeigt. / Knowledge about the micro-scale variability of airborne particles is a crucial criterion for air quality assessment within complex terrains such as urban areas. Due to the significant costs and time consumption related to the work required for standardized measurements of particle concentrations, dense monitoring networks are regularly missing. Models that simulate the transmission of particles are often difficult to use and/or computationally expensive. As a result, information regarding on-site particle concentrations at small scales is still limited. This thesis explores the micro-scale variability of aerosol concentrations in space and time using different methods. Experimental fieldwork, including measurements with mobile sensor equipment alongside a survey, and modeling approaches were conducted. Applied simulation studies, a performance assessment of two popular particle dispersion models, namely Austal2000 and ENVI-met, as well as the development of an ANN model are presented. The cities of Aachen and Münster were chosen as case studies for this research. Unexpected patterns of particle mass concentrations could be observed, including the identification of diffuse particle sources inside a park area with strong evidence that unpaved surfaces contributed to local aerosol concentration. In addition, vehicle traffic was proved to be a major contributor of particles, particularly close to traffic lanes. Results of the survey reveal that people were not able to distinguish between different aerosol concentration levels. Austal2000 and ENVI-met turned out to have room for improvement in terms of the reproduction of observed particle concentration levels, with both models having a tendency toward underestimation. The newly developed ANN model was confirmed to be a fairly accurate tool for predicting aerosol concentrations in both space and time, and demonstrates the principal ability of the approach also in the domain of air quality monitoring.

In the era of data-driven decision-making and innovation, synthetic data serves as a promising tool that bridges the need for vast datasets in machine learning (ML) and the imperative necessity of data privacy. By simulating real-world data while preserving privacy, synthetic data generators have become more prevalent instruments in AI and ML development. A key challenge with synthetic data lies in accurately estimating its utility. For such purpose, Population Fidelity (PF) measures have shown to be good candidates, a category of metrics that evaluates how well the synthetic data mimics the general distribution of the original data. With this setting, we aim to answer: "How well are different population fidelity measures able to indicate the utility of synthetic data for machine learning based classification models?" We designed a reusable six-step experiment framework to examine the correlation between nine PF measures and the performance of four ML for training classification models over five datasets. The six-step approach includes data preparation, training, testing on original and synthetic datasets, and PF measures computation. The study reveals non-linear relationships between the PF measures and synthetic data utility. The general analysis, meaning the monotonic relationship between the PF measure and performance over all models, yielded at most moderate correlations, where the Cluster measure showed the strongest correlation. In the more granular model-specific analysis, Random Forest showed strong correlations with three PF measures. The findings show that no PF measure shows a consistently high correlation over all models to be considered a universal estimator for model performance.This highlights the importance of context-aware application of PF measures and sets the stage for future research to expand the scope, including support for a wider range of types of data and integrating privacy evaluations in synthetic data assessment. Ultimately, this study contributes to the effective and reliable use of synthetic data, particularly in sensitive fields where data quality is vital. / I eran av datadriven beslutsfattning och innovation, fungerar syntetiska data som ett lovande verktyg som bryggar behovet av omfattande dataset inom maskininlärning (ML) och nödvändigheten för dataintegritet. Genom att simulera verklig data samtidigt som man bevarar integriteten, har generatorer av syntetiska data blivit allt vanligare verktyg inom AI och ML-utveckling. En viktig utmaning med syntetiska data är att noggrant uppskatta dess användbarhet. För detta ändamål har mått under kategorin Populations Fidelity (PF) visat sig vara goda kandidater, det är mätvärden som utvärderar hur väl syntetiska datan efterliknar den generella distributionen av den ursprungliga datan. Med detta i åtanke strävar vi att svara på följande: Hur väl kan olika population fidelity mätvärden indikera användbarheten av syntetisk data för maskininlärnings baserade klassifikationsmodeller? För att besvara frågan har vi designat ett återanvändbart sex-stegs experiment ramverk, för att undersöka korrelationen mellan nio PF-mått och prestandan hos fyra ML klassificeringsmodeller, på fem dataset. Sex-stegs strategin inkluderar datatillredning, träning, testning på både ursprungliga och syntetiska dataset samt beräkning av PF-mått. Studien avslöjar förekommandet av icke-linjära relationer mellan PF-måtten och användbarheten av syntetiska data. Den generella analysen, det vill säga den monotona relationen mellan PF-måttet och prestanda över alla modeller, visade som mest medelmåttiga korrelationer, där Cluster-måttet visade den starkaste korrelationen. I den mer detaljerade, modell-specifika analysen visade Random Forest starka korrelationer med tre PF-mått. Resultaten visar att inget PF-mått visar konsekvent hög korrelation över alla modeller för att betraktas som en universell indikator för modellprestanda. Detta understryker vikten av kontextmedveten tillämpning av PF-mått och banar väg för framtida forskning för att utöka omfånget, inklusive stöd för ett bredare utbud för data av olika typer och integrering av integritetsutvärderingar i bedömningen av syntetiska data. Därav, så bidrar denna studie till effektiv och tillförlitlig användning av syntetiska data, särskilt inom känsliga områden där datakvalitet är avgörande.

The Middle Fork John Day Basin in Northeastern Oregon is prime habitat for spring Chinook salmon and Steelhead trout. In 2008, a major tributary supporting rearing habitat, Big Boulder Creek, was restored to its historic mid-valley channel along a 1 km stretch of stream 800 m upstream of the mouth. Reduction of peak summer stream temperatures was among the goals of the restoration. Using Distributed Temperature Sensing (DTS) Fiber Optic Technology, stream temperature was monitored prior to restoration in June 2008, and after restoration in September 2008, July 2009, and August 2009. Data gathered was used to determine locations of groundwater and hyporheic inflow and to form a stream temperature model of the system. The model was used both to develop an evaluation method to interpret components of model performance, and to better understand the physical processes important to the study reach. A very clear decreasing trend in surface temperature was seen throughout each of the DTS stream temperature datasets in the downstream 500 m of the study reach. Observed reduction in temperature was 0.5°C (±0.10) in June 2008, 0.3°C (±0.37) in September 2008, 0.6°C (±0.25) in July 2009, and 0.2°C (±0.08) in August 2009. Groundwater inflow was calculated to be 3% of the streamflow for July 2009 and 1% during the August 2009 installation. Statistically significant locations of groundwater and hyporheic inflow were also determined. July 2009 data was used to model stream temperature of the 1 km (RMSE 0.28°C). The developed model performance evaluation method measures timelag, offset, and amplitude at a downstream observed or simulated point compared with the boundary condition, rather than evaluating the model based on error. These measures are particularly relevant to small scale models in which error may not be a true reflection of the ability of a model to correctly predict temperature. Breaking down model performance into these three predictive measures was a simple and graphic method to show the model's predictive capability without sorting through large amounts of data. To better understand the model and the stream system, a sensitivity analysis was conducted showing high sensitivity to streamflow, air temperature, groundwater inflow, and relative humidity. Somewhat surprisingly, solar radiation was among the lowest sensitivity. Furthermore, three model scenarios were run: a 25% reduction in water velocity, a 5°C increase in air temperature, and no groundwater inflow. Simulations of removal of groundwater inflows resulted in a 0.5°C increase in average temperature over the modeled time period at the downstream end, further illustrating the importance of groundwater in this stream system to reduce temperatures. / Graduation date: 2012