No description available.

When we develop and use species distribution models to predict species' current or potential distributions, we are faced with the trade-offs between model generality, precision, and realism. It is important to know how to improve and validate model generality while maintaining good model precision and realism. However, it is difficult for ecologists to evaluate species distribution models using field-sampled data alone because the true species response function to environmental or ecological factors is unknown. Species distribution models should be able to approximate the true characteristics and distributions of species if ecologists want to use them as reliable tools. Simulated data provide the advantage of being able to know the true species-environment relationships and control the causal factors of interest to obtain insights into the effects of these factors on model performance. I used a case study on Bythotrephes longimanus distributions from several hundred Ontario lakes and a simulation study to explore the effects on model performance caused by several factors: the choice of predictor variables, the model evaluation methods, the quantity and quality of the data used for developing models, and the strengths and weaknesses of different species distribution models. Linear discriminant analysis, multiple logistic regression, random forests, and artificial neural networks were compared in both studies. Results based on field data sampled from lakes indicated that the predictive performance of the four models was more variable when developed on abiotic (physical and chemical) conditions alone, whereas the generality of these models improved when including biotic (relevant species) information. When using simulated data, although the overall performance of random forests and artificial neural networks was better than linear discriminant analysis and multiple logistic regression, linear discriminant analysis and multiple logistic regression had relatively good and stable model sensitivity at different sample size and detection limit levels, which may be useful for predicting species presences when data are limited. Random forests performed consistently well at different sample size levels, but was more sensitive to high detection limit. The performance of artificial neural networks was affected by both sample size and detection limit, and it was more sensitive to small sample size.

When we develop and use species distribution models to predict species' current or potential distributions, we are faced with the trade-offs between model generality, precision, and realism. It is important to know how to improve and validate model generality while maintaining good model precision and realism. However, it is difficult for ecologists to evaluate species distribution models using field-sampled data alone because the true species response function to environmental or ecological factors is unknown. Species distribution models should be able to approximate the true characteristics and distributions of species if ecologists want to use them as reliable tools. Simulated data provide the advantage of being able to know the true species-environment relationships and control the causal factors of interest to obtain insights into the effects of these factors on model performance. I used a case study on Bythotrephes longimanus distributions from several hundred Ontario lakes and a simulation study to explore the effects on model performance caused by several factors: the choice of predictor variables, the model evaluation methods, the quantity and quality of the data used for developing models, and the strengths and weaknesses of different species distribution models. Linear discriminant analysis, multiple logistic regression, random forests, and artificial neural networks were compared in both studies. Results based on field data sampled from lakes indicated that the predictive performance of the four models was more variable when developed on abiotic (physical and chemical) conditions alone, whereas the generality of these models improved when including biotic (relevant species) information. When using simulated data, although the overall performance of random forests and artificial neural networks was better than linear discriminant analysis and multiple logistic regression, linear discriminant analysis and multiple logistic regression had relatively good and stable model sensitivity at different sample size and detection limit levels, which may be useful for predicting species presences when data are limited. Random forests performed consistently well at different sample size levels, but was more sensitive to high detection limit. The performance of artificial neural networks was affected by both sample size and detection limit, and it was more sensitive to small sample size.

Dans le cadre des activités de maintenance des centrales nucléaires, les personnes intervenant sur des chantiers en zone contrôlée doivent effectuer des mesures de contamination surfacique sur leurs outils, gants, sur tenue de travail et sur les parties de leur corps susceptible d’être contaminées (cou et tête par exemple). Or, les mesures de contamination surfacique bêta ne peuvent pas toujours être réalisées au plus près des chantiers. Le bruit de fond gamma présent à ces endroits étant élevé et fluctuant, ce dernier perturbe les appareils utilisés habituellement pour la mesure de contamination surfacique. L’objectif principal de ce travail était de concevoir un appareil de mesure adapté à la détection de contamination surfacique bêta de quelques Bq/cm2 (typiquement < 4 Bq/cm2), dans une ambiance gamma de quelques dizaines de µSv/h (typiquement < 100 µSv/h) et fluctuante (typiquement de l’ordre de 30% en 1 seconde). / In the framework of the nuclear power plants maintenance, the onsite workers incontrolled areas must performed some surface contamination controls of their tools, gloves, workingsuits and over their body parts which may be contaminated (neck and head for instance). However,surface beta contamination measurements can’t always been done as close as possible of the working location. The onsite ambient gamma background, being high and fluctuating, may disturb the radiation protection devices commonly used for surface contamination measurements.The main goal of this work was to conceive a radiation protection device adapted to the betasurface contamination of about a few Bq/cm2 (< 4Bq/cm2 typicaly) in a fluctuating (typicaly 30%variation per second) ambient gamma background of a few μSv/h (typicaly <100 μSv/h).

Surface contamination occurs in nuclear facilities, something that is important to detect easily and efficiently. Using today’s methods to detect nuclear surface contamination may cause certain inconsistencies as the human operator is solely trusted to keep the detector at the correct distance and move it at the correct speed. This thesis project aims to address the problem of inconsistent measurements with respect to the current measurement methods. A system is designed to monitor the measurement process with regards to detector velocity and height. The system will trigger a warning when the minimum detectable activity is too high, as it would lead to inconsistent results. This system consists of a cart-detector setup with a scintillation detector and velocity measurement device(s). Software will utilize the measurement data to implement the aforementioned monitoring. The system aims to be compliant with international standards, such as the ISO 11929 and the ISO 7503 standards, and will thus make use of these standards. The result of the part-analysis for each component of the system showed a large inaccuracy regarding the Intertial Measurement Units (IMUs); hence, the robotic wheels were chosen as the main method of measuring speed for this project. The robotic wheels and the detector were shown to be sufficiently accurate for the desired measurements. The Raspberry Pi 4 model B, the on- board computer, was also shown to be performance-wise and property-wise well suited for the project. This project showed that there is a theoretical way to implement the speed of a moving detector-rig into the Minimum Detectable Activity (MDA) formula. However, the implementation investigated in this project suggests that full compatibility with ISO 7503 was not achievable. / Radioaktiv ytkontaminering förekommer i kärnkraftverk, vilket är viktigt att upptäcka snabbt och effektivt. Dagens metoder för att upptäcka radioaktiv ytkontaminering kan lida av viss osäkerhet eftersom man förlitar sig helt på att operatören kan manövrera detektorn på rätt höjd och hastighet. Detta examensarbete behandlar en lösning till det ovan nämnda problemet. Ett ”proof-of-concept”-system som kan övervaka mätprocessen designas. Genom att mäta hastighet och känna till höjden över marken kan en varning meddelas användaren när den minsta detekterbara aktiviteten (MDA) når ett tröskelvärde. Det färdiga systemet är en plattform på hjul med en scintillator- detektor monterad tillsammans med en eller flesta hastighetsmätningsenheter. Systemet bör vara kompatibelt med internationella standarder, till exempel ISO 11929 och ISO 7503. Resultaten från den utvärdering av varje individuell komponent som gjorts visade på en stor mätosäkerhet i de två utvärderade IMUerna. Detta medförde att robothjulen valdes som enda källa för hastighetsmätning. Robothjulen, samt detektorn påvisade god mätsäkerhet, väl lämpad för detta projekt. Även mikrodatorn, Raspberry Pi 4 model B, visade sig vara lämplig sett till prestanda och egenskaper. Projektet resulterade i att det är trott att det finns en lämpligt sätt att i teorin implementera hastighet som en parameter i formeln för MDA. Det är dock värt att nämna att resultaten tyder på att det i denna implementation inte var möjligt att uppnå fullständig kompabilitet ISO 7503.