Key measures of socioeconomic indicators are essential for making informed policy decisions, but due to the high costs and operational difficulties of traditional data collection efforts, obtaining reliable socioeconomic data remains a challenge, particularly in developing countries. This work presents a deep learning methodology to estimate socioeconomic indicators using satellite imagery. The neural network model developed was trained at the Brazilian region of Vale do Ribeira with the goal of analyzing the socioeconomic indicator of income. The preliminary results showed that models using nightlight (NL) or multispectral daytime (MS) imagery performed better than models trained only on RGB bands and that models trained exclusively on NL or MS imagery performed similar to one another and nearly as well as the combined model MS+NL. Finally, the model yielded a low performance (R2 = 0.289), but it is still promising once the dataset employed was considerably smaller than the one used in the original study that attempted to replicate.

Large urban centers like the Metropolitan Region of São Paulo (MASP) are impacted by air pollution, especially by Inhalable particle matter (PM10). Persistent exceedance events (PEE) are defined as exceedance events that last for many consecutive days and occur simultaneously at many air quality monitoring stations across the MASP. This study aims to develop a predictive model for the occurrence of PEE in the MASP based on surface meteorological variables. Hourly PM10 concentrations from 12 air quality monitoring stations in the MASP between 2005 and 2021 were provided by the São Paulo State Environmental Agency (CETESB). Daily data on surface meteorological variables were provided by the IAG/USP meteorological station. Persistent exceedance events (PEE) were identified using the criteria: exceedance events that occurred simultaneously in at least 50% monitoring stations, persisting for at least 5 consecutive days. PEE occurrence was represented as a timeseries of a binary variable. The resulting daily dataset had 6204 lines and 13 attributes, without missing values. The dataset was divided into a training set (80%) and a test set (20%). A logistic regression model was applied, having the PEE occurrence (positive = 1) as the target value. The Variance Inflation Factor and the Stepwise Feature Selection method was applied to obtain an optimized subset of predictors. Model accuracy was accessed by the ROC curve and by a confusion matrix. Results indicate that PEE can be satisfactorily predicted by surface meteorological variables using a logistic regression. As for the next steps, we intend to extract easy-tocommunicate classification rules, aiming to support the development of warnings systems for air quality poor conditions in the MASP.

The renewable energies have confirmed their potential to reshape the energy matrix of several countries, in the last two decades, reinforcing the set of actions to better protect the environment. For instance, the solar and wind sources of energy production can dim the production and release of pollutants associated to electricity production. Unfortunately, the atmospheric temperature rise has been linked to the increase of the gas emissions, such as Carbon Dioxide (CO2) and Methane (CH4), from manifold sources, such as the thermal power stations for electricity generation. The Amazon basin has caught special importance for Brazil and abroad, and more recently, the use of satellites, fixed instrumented stations and airborne surveys has provided data focused on studies of the environment impact. This work deals with the assessment of the solar irradiation on the Manaus city, the largest city in the Western Amazon region, using Data Science tools, in a way to help the evaluation of the renewable energy in that region.

The Amazon rainforest has a great influence on the global energy balance and carbon fluxes, responsible for the net removal of approximately 4 million tons of carbon per year, via photosynthetic activity. Climate change and deforestation have impacts on the carbon budget in Amazonia, transforming CO2 sink areas into sources. Given the complexity of the factors that govern the carbon exchange in the Amazon and its influence on biological processes, the use of Data science strategies can promote a better understanding about the main environmental factors for different scenarios, and also, assist in public policies to mitigate the global warming effects. This study aims to identify the environmental factors that determine the temporal variability of carbon exchanges between the biosphere and the atmosphere in the Tapajós National Forest, in the Amazon, applying Data Science strategies in an integrated set of environmental data from energy and carbon fluxes and remote sensing data. The specific objective is to assess the influence of a selected set of environmental variables on the variability of carbon exchanges, with the use of an artificial neural networks classification model to identify the variables with great impact on source, sink and neutrality scenarios in Tapajós National Forest. Data Science strategies were applied to an integrated dataset of ground-based carbon flux measurements and remote sensing data, considering the period between 2002 and 2006. An artificial neural network (ANN) classification model was developed to identify the environmental variables with great impact on carbon source, sink and neutrality conditions. The average global score of ANN model was 65%. It was possible to identify the predictor variables with greatest impact to the carbon sink condition: radiation at the top of the atmosphere, sensible and latent energy fluxes and leaf area index. Thus, the ANN model with an ensemble of Data Science strategies can improve a better understanding of variability CO2 fluxes and be a powerful tool to promote new knowledge.

The complex network is a method with a high flexibility and easy application. Complex Network allows extracting relevant information from the system, like its organization and dynamics, as well as different indices that allow obtaining particular characteristics. This work studies the communities present on the rain network in the Amazon basin for the austral summer. Summer was used due to the presence of the South American monsoon system (SAMS), since this is the greatest mechanism for modulating precipitation over South America. Once the communities were obtained, the minimum correlation value (MCV) was varied in order to verify the spatial variations of the communities. Where it was verified how certain communities are composed of subcommunities while others simply disappear. Finally, it is shown how the spatial distribution of the subcommunities shows a relationship with the presence of SAMS. However, more detailed analyzes are needed for each of these communities.

In this work, machine learning techniques were applied to detect clusters present in satellite and weather radar images. The technique used was the unsupervised clustering algorithm DBSCAN. This algorithm was used to extract the morphological characteristics of atmospheric systems that occurred between February 1 and March 30, 2014 (rainy season) and September 15 to October 15, 2014 (dry season). The morphological characteristics are extracted from different thresholds (235K, 220K and 210K) of cloud top brightness temperature observed in the infrared channel of GOES-13 satellite, and also the precipitation estimated at the reflectivity thresholds (20dBZ, 30dBZ and 40dBZ) of the SIPAM meteorological radar in the city of Manaus. The results present the number of clusters identified by the algorithm and described the characteristics of the clusters during the diurnal cycle and in both seasons.

Brazil is one of countries highest incidence of lightning in the world and the characterization of thus event can help in the development of public polices and decision-making by authorities to mitigate the socio-economic damage that may be caused. This work presents some analysis of spatio-temporal patterns of lightnings in Brazil in 2020, generated from Self-Organizing Map (SOM) technique. This analysis considers the activity of the lightning in the hourly, daily and monthly periods accumulated in the different Brazilian states. The seasonal variation of lightning was also evaluated, considering the four seasons of 2020. The results showed that the self-organizing maps were efficient in identifying spatio-temporal patterns of lightning, which are highly variability events. Thus, theses results can support the development of new tools or analysis in which the spatio-temporal information lightning is important, for example, in warning and forecasting systems.

AbstractFeature interaction is a crucial concept when it comes to machine learning interpretability. These interactions make it easier for the person using a machine learning tool to know how the model made its prediction. Intrinsic models such as the Generalized linear models pull their interpretability aspects from detecting feature interactions on the data. However, these models do not search the whole sample space for interactions and assume all interactions to the target feature are identical. A hybrid model combining Rough Set and Generalized linear models is proposed. Rough Set uses the concept of information granulation and approximations of regions with meaningful information to find feature interactions in the whole sample space. The detected features will then be modeled onto a Generalized linear model for prediction purposes. The dataset used in the experiment was drawn from wunderground.com online weather site, specifically the Kariki Farm online weather station. The proposed methodology for the research is the CRISP-DM method.Keywords: Machine interpretability, Rough Set, Generalized linear model, weather prediction.