Abstract. New particle formation (NPF) in the atmosphere is globally an important source of climate relevant aerosol particles. Occurrence of NPF events is typically analyzed by researchers manually from particle size distribution data day by day, which is time consuming and the classification of event types may be inconsistent. To get more reliable and consistent results, the NPF event analysis should be automatized. We have developed an automatic analysis method based on deep learning, a subarea of machine learning, for NPF event identification. To our knowledge, this is the first time that a deep learning method, i.e., transfer learning of a convolutional neural network (CNN), has successfully been used to automatically classify NPF events into different classes directly from particle size distribution images, similarly to how the researchers carry out the manual classification. The developed method is based on image analysis of particle size distributions using a pretrained deep CNN, named AlexNet, which was transfer learned to recognize NPF event classes (six different types). In transfer learning, a partial set of particle size distribution images was used in the training stage of the CNN and the rest of the images for testing the success of the training. The method was utilized for a 15-yearlong dataset measured at San Pietro Capofiume (SPC) in Italy. We studied the performance of the training with different training and testing of image number ratios as well as with different regions of interest in the images. The results show that clear event (i.e., classes 1 and 2) and nonevent days can be identified with an accuracy of ca. 80 %, when the CNN classification is compared with that of an expert, which is a good first result for automatic NPF event analysis. In the event classification, the choice between different event classes is not an easy task even for trained researchers, and thus overlapping or confusion between different classes occurs. Hence, we cross-validated the learning results of CNN with the expert-made classification. The results show that the overlapping occurs, typically between the adjacent or similar type of classes, e.g., a manually classified Class 1 is categorized mainly into classes 1 and 2 by CNN, indicating that the manual and CNN classifications are very consistent for most of the days. The classification would be more consistent, by both human and CNN, if only two different classes are used for event days instead of three classes. Thus, we recommend that in the future analysis, event days should be categorized into classes of "quantifiable" (i.e., clear events, classes 1 and 2) and "nonquantifiable" (i.e., weak events, Class 3). This would better describe the difference of those classes: both formation and growth rates can be determined for quantifiable days but not both for nonquantifiable days. Furthermore, we investigated more deeply the days that are classified as clear events by experts and recognized as nonevents by the CNN and vice versa. Clear misclassifications seem to ...

Abstract. The uncertainty in the radiative forcing caused by aerosols and its effect on the climate change calls for research to improve knowledge of the aerosol particle formation and growth processes. While the experimental research has provided large amount of high quality data on aerosols in the last two decades, the inference of the process rates is still inadequate, mainly due to limitations in the analysis of data. This paper focuses on developing computational methods to infer aerosol process rates from size distribution measurements. In the proposed approach, the temporal evolution of aerosol size distributions is modeled with the general dynamic equation equipped with stochastic terms that account for the uncertainties of the process rates. The time-dependent particle size distribution and the rates of the underlying formation and growth processes are reconstructed based on time series of particle analyzer data using Bayesian state estimation – which not only provides (point) estimates for the process rates but also enables quantifying their uncertainties. The feasibility of the proposed computational framework is demonstrated by a set of numerical simulation studies.

No abstract

Abstract. Bayesian state estimation in the form of Kalman smoothing was applied to differential mobility analyser train (DMA-train) measurements of aerosol size distribution dynamics. Four experiments were analysed in order to estimate the aerosol size distribution, formation rate, and size-dependent growth rate, as functions of time. The first analysed case was a synthetic one, generated by a detailed aerosol dynamics model and the other three chamber experiments performed at the CERN CLOUD facility. The estimated formation and growth rates were compared with other methods used earlier for the CLOUD data and with the true values for the computer-generated synthetic experiment. The agreement in the growth rates was very good for all studied cases: estimations with an earlier method fell within the uncertainty limits of the Kalman smoother results. The formation rates also matched well, within roughly a factor of 2.5 in all cases, which can be considered very good considering the fact that they were estimated from data given by two different instruments, the other being the particle size magnifier (PSM), which is known to have large uncertainties close to its detection limit. The presented fixed interval Kalman smoother (FIKS) method has clear advantages compared with earlier methods that have been applied to this kind of data. First, FIKS can reconstruct the size distribution between possible size gaps in the measurement in such a way that it is consistent with aerosol size distribution dynamics theory, and second, the method gives rise to direct and reliable estimation of size distribution and process rate uncertainties if the uncertainties in the kernel functions and numerical models are known.

The interpretation of X-ray photoelectron spectroscopy (XPS) data relies on measurement models that depend on several parameters, including the photoelectron attenuation length and X-ray photon flux. However, some of these parameters are not known, because they are not or cannot be measured. The unknown geometrical parameters can be lumped together in a multiplicative factor, the alignment parameter. This parameter characterizes the ability of the exciting light to interact with the sample. Unfortunately, the absolute value of the alignment parameter cannot be measured directly, in part because it depends on the measurement model. Instead, a proxy for the experimental alignment is often estimated, which is closely related to the alignment parameter. Here, a method for estimating the absolute value of the alignment parameter based on the raw XPS spectra (i.e. non-processed photoelectron counts), the geometry of the sample and the photoelectron attenuation length is presented. The proposed parameter estimation method enables the quantitative analysis of XPS spectra using a simplified measurement model. All computations can be executed within the open and free Julia language framework PROPHESY. To demonstrate feasibility, the alignment parameter estimation method is first tested on simulated data with known acquisition parameters. The method is then applied to experimental XPS data and a strong correlation between the estimated alignment parameter and the typically used alignment proxy is shown.