47 research outputs found
Stadtforschung, der schwierige Weg von der Erkenntnis zur Umsetzung
No full textStadtforschung in den atmosphärischen Wissenschaften findet seit über 200 Jahren (Howard 1818-20) statt. Sie umfasst alle Themen und Einflüsse, die die Lebensqualität der Stadtbewohner beeinflussen. Hierzu zählen an prominenter Stelle die städtische Luftqualität, das Strahlungsklima, die Windverhältnisse in Städten und die städtische Wärmeinsel. Laut übereinstimmenden Analysen (z.B. Eliasson 2000; Mills et al. 2010; Parasee etal. 2019) haben die naturwissenschaftlichen Erkenntnisse aus der Stadtklimaforschung bisher nur begrenzt Eingang in die Stadtplanung gefunden. In vier Bereichen, die sich von Mikro- bis zur Meso-Skala erstrecken, lassen sich hier Fortschritte erreichen: (1) Baumaterialien und Gebäudegestaltung, (2) Grün und Blau in der Stadt, (3) Stadtplanung und (4) Einbindung der Städte in regionale und überregionale Infrastrukturen. Der Beitrag wird hier Beispiele für alle vier Bereiche benennen.
Ein wesentliches Mittel um die Auswirkung von Maßnahmen in den vier zuvor genannten Bereichen zu beurteilen und auch unerwünschte Nebenwirkungen einzelner Maßnahmen auf andere Handlungsfelder im Vorfeld zu erkennen, sind holistische Simulationsmodelle wie PALM4U, die von Atmosphärenwissenschaftlern entwickelt wurden (Scherer et al. 2019), aber durch eine entsprechende Oberfläche auch Planern zur Verfügung gestellt werden sollen. Pilotstudien, PALM4U einsatzfähig zu machen laufen derzeit. Der Beitrag wird Fragen zur Simulation der Luftqualität und der Strömung in komplexem Gelände mit PALM4U adressieren. Die Schaffung nachhaltiger Städte mit gesunden Lebensbedingungen für die Bewohner liegt nicht in der Analyse und Bearbeitung von Einzelaspekten, sondern in einer umfassenden Zusammenarbeit von Naturwissenschaftlern, Stadtplanern und Architekten, um vorhandene Städte systematisch zu transformieren und neue städtische Gebiete von Anfang an nachhaltig zu planen. PALM4U ist auf dem Wege, ein wichtiges „Met-Tool“ für diesen Zweck zu werden.
Eliasson, I. 2000. The use of climate knowledge in urban planning. Landscape and Urban Planning 48, 1-2, 31–44.
Fallmann, J., S. Emeis, 2020. How to Bring Urban and Global Climate Studies together with Urban Planning and Architecture? Devel. Built Environ., 4, 100023.
Fallmann, J., S. Emeis, 2021: How to bring urban climate studies to application – A meteorological view from five decades of urban climate research and results from a current study. IAUC Newsletter 80, 12-17.
Howard, L. 1818-20. The climate of London. Deduced from meteorological observations, made at different places in the neighbourhood of the metropolis. London.
Mills, G; H. Cleugh; R. Emmanuel; W. Endlicher; E. Erell; G. McGranahan; E. Ng; A. Nickson; J. Rosenthal; K. Steemer. 2010. Climate Information for Improved Planning and Management of Mega Cities (Needs Perspective). Procedia Environmental Sciences 1, 228–246.
Parsaee, M; M. M. Joybari; P. A. Mirzaei; F. Haghighat. 2019. Urban heat island, urban climate maps and urban development policies and action plans. Environ. Technol. & Innov. 14, 100341.
Scherer, D., Antretter, F., Bender, S., Cortekar, J., Emeis, S., Fehrenbach, U., Gross, G., Halbig, G., Hasse, J., Maronga, B., Raasch, S., Scherber, K., 2019: Urban Climate Under Change [UC]2 – A National Research Programme for Developing a Building-Resolving Atmospheric Model for Entire City Regions. Meteorol.
Z. (Contr. Atm. Sci.), 28, 95-104
Identification of specific interactions between bacteria and heavy metals accumulating plants
No full textCulturable bacteria from Zn- and Cd-accumulating Salix caprea with differential effects on plant growth and heavy metal availability.
No full textAims: To characterize bacteria associated with Zn ⁄ Cd-accumulating Salix
caprea regarding their potential to support heavy metal phytoextraction.
Methods and Results: Three different media allowed the isolation of 44 rhizosphere
strains and 44 endophytes, resistant to Zn ⁄Cd and mostly affiliated with
Proteobacteria, Actinobacteria and Bacteroidetes ⁄ Chlorobi. 1-Aminocyclopropane-
1-carboxylic acid deaminase (ACCD), indole acetic acid and siderophore
production were detected in 41, 23 and 50% of the rhizosphere isolates and in
9, 55 and 2% of the endophytes, respectively. Fifteen rhizosphere bacteria and
five endophytes were further tested for the production of metal-mobilizing
metabolites by extracting contaminated soil with filtrates from liquid cultures.
Four Actinobacteria mobilized Zn and ⁄ or Cd. The other strains immobilized
Cd or both metals. An ACCD- and siderophore-producing, Zn ⁄ Cd-immobilizing
rhizosphere isolate (Burkholderia sp.) and a Zn ⁄ Cd-mobilizing Actinobacterium
endophyte were inoculated onto S. caprea. The rhizosphere isolate
reduced metal uptake in roots, whereas the endophyte enhanced metal accumulation
in leaves. Plant growth was not promoted.
Conclusions: Metal mobilization experiments predicted bacterial effects on
S. caprea more reliably than standard tests for plant growth-promoting activities.
Significance and Impact of the Study: Bacteria, particularly Actinobacteria,
associated with heavy metal-accumulating Salix have the potential to increase
metal uptake, which can be predicted by mobilization experiments and may be
applicable in phytoremediation
Obesity: outcome of standardized life-style change in a rehabilitation clinic. An observational study
No full textHelmuth Haslacher,1 Hannelore Fallmann,2 Claudia Waldhäusl,3 Edith Hartmann,2 Oswald F Wagner,1 Werner K Waldhäusl2,41Department of Laboratory Medicine, Medical University of Vienna, Vienna, Austria; 2Rehabilitation Clinic for Diabetes and Metabolic Diseases, Moorbad Neydhartig, Neydharting, Upper Austria, Austria; 3Department of Radiotherapy, Medical University of Vienna, Vienna, Austria; 4Department of Medicine III, Medical University of Vienna, Vienna, AustriaPurpose: To explore differences in baseline characteristics following three weeks of semi-standardized in-patient care between patients with obesity without and with type 2 diabetes (T2D).Patients and methods: Patients without or with T2D were matched according to age, sex, and BMI. Food intake was restricted to 1,200–1,600 kcal/d to which a 400–600 kcal/d exercise load was added, and data were compared using Student’s t-test, general linear models, and Spearman-rank correlations.Results: At baseline, patients with obesity and T2D displayed, besides elevated blood glucose and HbA1c values, higher serum liver enzymes (p<0.001–0.05), triglycerides, and CRP (p<0.01) and a greater prevalence of treated hyperlipidemia (p<0.001) than those with plain obesity who showed only higher LDL and HDL cholesterol levels (+9.0% and +16.0%). In response to three-weeks of standardized life-style change, both groups improved their vital variables and risk scores (p<0.001). While improvement in cholesterol slightly favored patients with plain obesity, the need for anti-hyperlipidemics (+25%) rose in both groups, albeit that for anti-hypertensives (+50%) increased only in patients with obesity and add-on T2D.Conclusion: Moderate changes in lifestyle improve the clinical condition, including coronary heart disease and premature mortality risk scores (HARD-CHD and ABSI) in patients with obesity both in the absence and presence of T2D, with the latter seemingly increasing the risk of hepatic steatosis and systemic inflammation.Keywords: obesity, standardized life-style change, liver disease, inflammation, rehabilitation clinic
 
Human activity pattern recognition based on continuous data from a body worn sensor placed on the hand wrist using Hidden Markov models
No full textDas Ziel dieser Arbeit ist kontinuierliche und diskrete Daten in einem Algorithmus zu vereinen um komplexe Aktivitäten-Muster zu erkennen,wie Zähne putzen,Zubereitung von Essen oder Hausarbeit.Diese Muster sind oft sehr komplex,da sie aus vielen Untermustern bestehen. Das Zubereiten von Essen zum Beispiel,besteht aus Untermustern wie 'Töpfe aus dem Schrank nehmen',Essen aus dem Kühlschrank nehmen',Schneiden',Kochen'und andere.Der schwierige Teil dieser Aufgabe ist,dass die Essenszubereitung sich nicht nur in der Reihenfolge der Untermuster unterscheidet,sondern auch in den unterschiedlichen Speisen die zubereitet werden.Man stelle sich den Unterschied zwischen der Vorbereitung eines Drei-Gänge-Menüs zu einem Frühstück vor.Diese beiden Aktivitäten unterscheiden sich enorm in der Vorbereitungsdauer und den Untermustern.Das Erkennungssystem welches innerhalb dieser Arbeit konstruiert wurde und menschliche Aktivitäten erkennt,kann mit dieser Art von Unterschieden umgehen. Das InvenSense MotionFit TMSoftware Development Kit(SDK)wird verwendet um die MPU-9150Sensordaten aufzuzeichnen.Der MPU-9150Sensor ist ein neun-achsiges MotionTracking Gerät,welches für diese Art von Anwendungen,wie in dieser Arbeit benötigt,optimiert wurde.Dieser tragbare Sensor kann Beschleunigungssensor-und Gyrometer-Daten senden,welche später herangezogen werden um menschliche Aktivitäten in realen Umgebungen zu erkennen. Die Frequenz der annotierten Daten ist in allen Experimenten auf50Hz gesetzt worden.Die aufgenommenen Aktivitäten sindHaare kämmen',Gesicht waschen',Hände waschen',Zähne putzen',Das Bett machen',Kleidung wechseln',Rollos rauf/runter ziehen',Essen zubereiten',Essen'undFenster schließen/öffnen'.Zwischen diesen Aktivitäten wird eineNULL'-Klasse durchgeführt,welche die Vorbereitung bzw.Nachbereitung der nächsten bzw.vorigen Aktivität beschreibt.Diese rohen Daten werden mithilfe eines verschiebbaren Fensters und verschiedener Features vorverarbeitet.In dieser Arbeit beträgt die Fensterlänge50,genauso wie die Frequenz beim Aufnehmen.Die Verschiebung des Fensters wird mit einer50%-Überlappung durchgeführt.Die verwendeten Features sind Mittelwert,Varianz,Korrelation und die auf schnelle Fourier-Transformation beruhenden Features,spektrale Entropie und Energie. Die Mustererkennung wird mit MATLAB und der von Murphyet al.zur Verfügung gestellten Toolbox PMTK3 [22]bewerkstelligt.Die verwendeten Klassifikations-Algorithmen sind supervised Lernmethoden,dies bedeutet,sie brauchen gelabelte Daten für das Training.Dieser Umstand wird von den Daten in dieser Arbeit erfüllt.Die Klassifikations-Algorithmen die während der Experimente verwendet werden sind einerseits kontinuierliche Hidden Markov Modelle(cHMM)und andererseits k-nächste-Nachbarn(kNN)Klassifikatoren.Die Klassifikationsmethoden werden im Detail beschrieben und ein Vergleich wird durchgeführt um Unterschiede zu erörtern.Die Experimente zeigen schlussendlich,dass das cHMM zu besseren Resultaten führt,im Gegensatz zu dem kNN Klassifikator. Die Erkennung von Alltagsaktivitäten funktioniert gut im Zusammenhang mit Ambient Assisted Living.Es kann gefolgert werden,dass cHMM die geeignetste Methode ist und zu den besten Ergebnissen führt.Verhältnismäßig ist der kNN Klassifikator viel schlechteraufgrund seiner einfachen Annahme.Deshalb ist der kNN Klassifikator nicht der beste Klassifikator,aber trotz seiner Einfachheit können annehmbare Ergebnisse erwartet werden. Nach der Validierung des Models sind verschiedene Features Kombinationen verglichen worden um die geeignetste Kombination zu finden.Andere Experimente konzentrieren sich auf die Verwendung von verschiedenen Training-und Test-Sets,die beste Anzahl von Sensoren,den Einfluss von Filtern,den Einfluss der Teilung von Aktivitäten und die Anwendung von diskreten und kontinuierlichen Daten.Die Experimente zeigen,dass die Bewegungssensor-Daten alleine,die besten Resultate liefert,während Filter und Teilung von Aktivitäten keine qualitative Verbesserung bringen. Die Kombination von diskreten und kontinuierlichen Daten verbessert die Resultate erheblich und führt zu verschiedenen Aktivitäten mit bester Genauigkeit und Trefferquote.Die Genauigkeit fürHände waschen'ist mit100%die beste Aktivität für kontinuierliche Daten undZähne putzen'100%für den kombinierten Fall.Essen'hat eine der besten Trefferquoten mit97.13%im kontinuierlichen Fall,wohingegen der kombinierte Fall eine Verbesserung nach sich zieht mit100%fürHände waschen',Das Bett machen',Rollos rauf/ runter ziehen',Essen zubereiten',Essen'undFenster öffnen/ schließen'.Im Allgemeinen liefert das System vertretbare Resultate sogar mit einem relativ kleinen Datensatz.The aim of the thesis is to combine discrete and continuous data in an algorithm to detect complex activity patterns such as tooth brushing, food preparation or household work. These patterns are of highly complex nature, meaning they consist of many sub-activities. Food preparation, for instance consists of sub-activities like 'take pans out of the cupboard','take food out of the refrigerator', 'cutting', 'cooking' and others. The complicated parts of this task are, that the food preparation not only differs in the order of the sub-activities, but also in the food which is prepared. Envision the preparation of a three-course menu in comparison to preparing a breakfast. This two activities differ a lot in needed duration and sub-activities. The human activity recognition system built in this thesis can handle these differences. The InvenSense MotionFit TM Software Development Kit (SDK) is used to record data with the MPU-9150 sensor. The MPU-9150 sensor is a nine-axis MotionTracking device, which is optimized for those kind of applications in this thesis and is normally used in mobile devices.[20] The wearable sensor MPU-9150 is able to send accelerometer and gyrometer data, which are later used to detect human activities in real environments. The frequency of the annotated data is 50Hz in all experiments. The processed activities are 'Comb hair','Wash face', 'Wash hands', 'Tooth brushing', 'Make bed', 'Change clothes', 'Put roller blinds up/down', 'Prepare food', 'Eat' and 'Open/close window'. Inbetween this activities a 'NULL'-class is performed, which describes the preparation for the next activity or the closure of the previous activity. This raw data are preprocessed via a shifted window and different features. In this thesis the window length is set to 50, equal the sampling frequency and the shift is accomplished with an 50% overlap. The used features are mean, variance, correlation and fast Fourier transformation based features. The fast Fourier transformation bases features are spectral entropy and energy. The pattern recognition is done in MATLAB using the PMTK3 toolbox from Murphy et al. [22] with real annotated data. The used classification algorithms are from supervised learning structure, meaning that they need labeled data for training. This circumstances are fulfilled for the data used in this thesis. The classification algorithms that are used during the experimentation are continuous Hidden Markov Model (cHMM) and k-nearest-neighbors (kNN) classification. The classification methods are described in detail and a comparison is done to discuss the differences between the results. In the end the cHMM leads to the more accurate outcomes in comparison to the kNN classifier. Daily activity detection works well in the context of ambient assisted living. It can be concluded that, the cHMM is the most proper method and comes to the best solutions. In contrast the kNN classification is much worse, because of its simplifying assumption. Due to that the kNN classifier is not the best classifier to use, but for the simpleness, acceptable results can be expected. After validation of the model the different features combinations are compared to each other to find the most suitable combination. Other experiments focus on different training and test sets, the best number of sensors, the impact of filters, the impact of activity division and the application of discrete and continuous data. The experiments show that the accelerometer data on their own lead to one of the best results, whereas the filters as well as activity division do not lead to a qualitative improvement. The combination of discrete and continuous data improves the results a lot and leads to different activities with highest recall and precision. The precision for 'Wash hands' is with a value of 100% the best in the continuous data case and 'Tooth brushing' in the combined case, also with 100%. 'Eat' has one of the best recall values with 97.13% in the continuous case, whereas the improvement in the combined case can be seen on the recall value 100% for 'Wash hands', 'Make bed', 'Put blinds up/down', 'Prepare food', 'Eat' and 'Open/close window'. Overall the system leads to reasonable outputs even with a relative small dataset
Impact of urban imperviousness on boundary layer meteorology and air chemistry on a regional scale
No full textIt has been long understood that land cover change from natural to impervious modifies the surface energy balance and hence the dynamical properties of the overlying atmosphere. The urban heat island is manifested in the formation of an urban boundary layer with distinct thermodynamic features that in turn govern transport processes of air pollutants. While many studies already demonstrated the benefits of urban canopy models (UCM) for atmospheric modelling, work on the impact on urban air chemistry is scarce. This study uses the state-of-the-art coupled chemistry-climate modelling system MECO(n) to assess the impact of the COSMO UCM TERRA_URB on the dynamics and gas phase chemistry in the boundary layer of the urban agglomeration Rhine-Main in Germany. Comparing the model results to ground observations and satellite and ground based remote sensing data, we found that the UCM experiment reduces the bias in temperature at the surface and throughout the boundary layer. This is true for ground level NO2 and ozone distribution as well. The application of MECO(n) for urban planning purposes is discussed by designing case studies representing two projected scenarios in future urban planning – densification of central urban areas and urban sprawl. Averaged over the core urban region and 10‑days during a heat wave period in July 2018, model results indicate a warming of 0.7 K in surface temperature and 0.2 K in air temperature per 10 % increase in impervious surface area fraction. Within this period, a 50 % total increase of imperviousness accounts for a 3 K and 1 K spatially averaged warming respectively. This change in thermodynamic features results in a decrease of surface NO2 concentration by 10–20 % through increased turbulent mixing in areas with highest impervious fraction and highest emissions. In the evening and nighttime however, increased densification in the urban centre results in amplified canyon blocking, which in turn results in average increase in near surface NO2 concentrations of about 10 %, compared to the status quo. This work intends to analyse regional scale features of surface-atmosphere interactions in an urban boundary layer and can be seen as preparatory work for higher resolution street scale models
Secondary effects of urban heat island mitigation measures on air quality
No full textAbstractThis study presents numerical simulations analysing the effect of urban heat island (UHI) mitigation measures on the chemical composition of the urban atmosphere. The mesoscale chemical transport model WRF-Chem is used to investigate the impact of urban greening and highly reflective surfaces on the concentrations of primary (CO, NO) as well as secondary pollutants (O3) inside the urban canopy. In order to account for the sub-grid scale heterogeneity of urban areas, a multi-layer urban canopy model is coupled to WRF-Chem. Using this canopy model at its full extend requires the introduction of several urban land use classes in WRF-Chem. The urban area of Stuttgart serves as a test bed for the modelling of a case scenario of the 2003 European Heat Wave. The selected mitigation measures are able to reduce the urban temperature by about 1 K and the mean ozone concentration by 5–8%. Model results however document also negative secondary effects on urban air quality, which are closely related to a decrease of vertical mixing in the urban boundary layer. An increase of primary pollutants NO and CO by 5–25% can be observed. In addition, highly reflective surfaces can increase peak ozone concentration by up to 12% due to a high intensity of reflected shortwave radiation accelerating photochemical reactions
