Despite the very restrictive laws, Krakow is known as the city with the highest level of air pollution in Europe. It has been proven that, due to its location, air pollutants are transported to this city from neighboring municipalities. In this study, a complex geostatistical approach for spatio-temporal analysis of particulate matter (PM) concentrations was applied. For background noise reduction, data were recorded during the COVID-19 lockdown using 100 low-cost sensors and were validated based on indications from reference stations. Standardized Geographically Weighted Regression, local Moran’s I spatial autocorrelation analysis, and Getis–Ord Gi* statistic for hot-spot detection with Kernel Density Estimation maps were used. The results indicate the relation between the topography, meteorological variables, and PM concentrations. The main factors are wind speed (even if relatively low) and terrain elevation. The study of the PM2.5/PM10 ratio allowed for a detailed analysis of spatial pollution migration, including source differentiation. This research indicates that Krakow’s unfavorable location makes it prone to accumulating pollutants from its neighborhood. The main source of air pollution in the investigated period is solid fuel heating outside the city. The study shows the importance and variability of the analyzed factors’ influence on air pollution inflow and outflow from the city.

Air pollution is an important problem for public health. The spatiotemporal analysis is a crucial step for understanding the complex characteristics of air pollution. Using many sensors and high-resolution time-step observations makes this task a big data challenge. In this study, unsupervised machine learning algorithms were applied to analyze spatiotemporal patterns of air pollution. The analysis was conducted using PM10 big data collected from almost 100 sensors located in Krakow, over a period of one year, with data being recorded at 1-hour intervals. The analysis results using K-means and SKATER clustering revealed distinct differences between average and maximum values of pollutant concentrations. The study found that the K-means algorithm with Dynamic Time Warping (DTW) was more accurate in identifying yearly patterns and clustering in rapidly and spatially varying data, compared to the SKATER algorithm. Moreover, the clustering analysis of data after kriging greatly facilitated the interpretation of the results. These findings highlight the potential of machine learning techniques and big data analysis for identifying hot-spots, cold-spots, and patterns of air pollution and informing policy decisions related to urban planning, traffic management, and public health interventions

Joint inversion is a widely used geophysical method that allows model parameters to be obtained from the observed data. Pareto inversion results are a set of solutions that include the Pareto front, which consists of non-dominated solutions. All solutions from the Pareto front are considered the most feasible models from which a particular one can be chosen as the final solution. In this paper, it is shown that models represented by points on the Pareto front do not reflect the shape of the real model. In this contribution, a collective approach is proposed to interpret the geometry of models retrieved in inversion. Instead of choosing single solutions from the Pareto front, all obtained solutions were combined in one “heat map”, which is a plot representing the frequency of points belonging to all returned objects from the solution set. The conducted experiment showed that this approach limits the problem of equivalence and is a promising way of representing the geometry of the model that was retrieved in the inversion process.

The main goal of this paper was to estimate the heat exchange rock mass volume of a hot dry rock (HDR) geothermal reservoir based on microseismicity location. There are two types of recorded microseismicity: induced by flowing fluid (wet microseismicity) and induced by stress mechanisms (dry microseismicity). In this paper, an attempt was made to extract events associated with the injected fluid flow. The authors rejected dry microseismic events with no hydraulic connection with the stimulated fracture network so as to avoid overestimating the reservoir volume. The proposed algorithm, which includes the collapsing method, automatic cluster detection, and spatiotemporal cluster evolution from the injection well, was applied to the microseismic dataset recorded during stimulation of the Soultz-sous-Forets HDR field in September 1993. The stimulated reservoir volume obtained from wet seismicity using convex hulls is approximately five times smaller than the volume obtained from the primary cloud of located events.

<p>Structures created by hydraulic fracturing can be identified using the location of induced microseismic events. Estimating the effectiveness of stimulation depends on fracture mapping. Event location errors make precise imaging of fractures in a scattered seismic cloud challenging. In order to increase the reliability of the determined structures on the basis of events with location error, we proposed a several-stage procedure. This procedure was demonstrated on microseismic events located during the fracturing of the Wysin-2H/2Hbis horizontal well, an exploration well for shale gas in northern Poland from June 9, 2016 to June 18, 2016. All located events were subjected to a collapsing that allows obtaining new locations of events that are equivalent to original locations in a statistical sense. The creation of such an equivalent point cloud allows us to see certain structures that may reflect, for example, fractures. To validate the results before and after collapsing method, all events were set against the probability of a given brittleness index map.  It is demonstrated that the collapsed events occurred in regions that were more rigid, while the locations of events prior to this procedure showed no relationship with the occurrence of areas with higher susceptibility to fracking. The unsupervised machine learning algorithm HDBSCAN was used on a collapsed cloud to automatically detect clusters of events. The directions of identified clusters agree with the direction of regional maximum horizontal stress.</p>