The aim of this research was to improve isotopic models for the geolocation of Cannabis through the integration of machine learning techniques and isotopic analysis. To achieve the intended objectives, isotopic data from isoscapes and real samples of Cannabis from different Brazilian biomes were used. The method involved bootstrap resampling to generate pseudo-samples and the application of classification algorithms, including Naive Bayes, Random Forest, and Neural Networks. The results demonstrated that the algorithms used achieved high accuracy rates in classifying samples by biome, with the combination of methods proving to be an efficient strategy for tracking. It was concluded that among the chosen classifiers, the Support Vector Machine (SVM) showed the best performance both in terms of sample size (from 50 to 500), using the sigmoid kernel function, and in percentage (5% to 95%), in this case using the polynomial kernel. The SVM model with kernel functions achieved the highest average accuracy (0.7970), reinforcing its overall effectiveness even with reduced samples. Meanwhile, the polynomial SVM model reached an average accuracy of 0.8036, the highest among all evaluated models with varying percentages, and stood out for its balance between precision and recall. Thus, the results highlight that the integration of isotopic analysis with machine learning provides significant contributions to forensic science, enabling greater control over drug trafficking and important advances in knowledge about Brazil's biomes.

Atoms are the fundamental building blocks that make up matter and form the basis of everything we observe in the universe. They are composed of protons, neutrons, and electrons. Isotopes are versions of a chemical element (a set of atoms) that share the same number of protons in the atomic nucleus but differ in the number of neutrons. They can be classified into different categories based on their characteristics, including stable, radioactive, natural, artificial, cosmogenic, and industrial. The use of stable isotopes is based on the fact that the isotopic ratio is altered by biogeochemical processes that govern the movement of carbon, nitrogen, and water compounds between soil-plant-atmosphere systems. This alteration in the relative abundance of isotopes is called fractionation or isotopic discrimination, which is usually caused by a kinetic effect due to the small mass difference between the heavy and light isotope. The isotopic methodology consists of creating isoscapes from georeferenced isotopic compositions. Many studies use traditional spatial modeling and geostatistical techniques, such as spatial autoregressive regressions and Kriging methods. However, recent research indicates that machine learning (ML) algorithms, such as Random Forest (RF), provide more accurate results than previous methodologies. The importance of this technique has led to the incorporation of stable isotopes into the routine investigations of various security agencies worldwide and to the creation in 2002 of the Forensic Isotope Ratio Mass Spectrometry Network (FIRMS), which brings together academics, forensic experts, private companies, and government agencies working in this field of application. The objective of this study is to generate isoscapes of δ13C and δ15N from C3 plants for Brazil using geostatistical techniques and, mainly, machine learning to improve the accuracy of geographic isotopic patterns. This work used a dataset with 6,480 leaf samples from trees in 57 georeferenced field plots distributed throughout the Brazilian territory. The δ13C and δ15N isoscapes showed spatial patterns consistent with eco-physiological predictions in ML techniques. There are several potential applications for the isoscapes proposed here. One of them is to track the illegal wildlife trade (IWT), illegal logging, and the certification of natural products related to flora.

Brazil is recognized as one of the world's largest producers and exporters of animal protein. From an environmental perspective, cattle farming faces challenges related to deforestation and greenhouse gas emissions, while poultry and swine farming require greater efficiency in water resource use and waste management. The adoption of sustainable practices has proven essential to reducing environmental impacts and meeting international market demands. This research aimed to investigate animal protein production in Brazil from 1997 to 2023, focusing on beef, chicken, and pork. The data used in the study were obtained from IBGE (SIDRA) and analyzed using the R software, employing statistical models, including the modified Mann-Kendall test and Sen’s Slope test. The results indicated continuous growth in chicken meat production, surpassing beef from 2004 onwards, driven by lower production costs and greater acceptance in the international market. Pork production also expanded, stimulated by increased Chinese demand due to the African Swine Fever outbreak. The analysis revealed fluctuations that significantly impacted meat production and consumption, particularly in 2008, 2014, 2017, and 2020, due to the global financial crisis, adverse climate events, internal economic instabilities, trade wars, and logistical restrictions resulting from the COVID-19 pandemic. Brazil plays a central role in global food security.

Using neural networks to solve differential equations is an advantageous approach for analyzing continuous-time series due to their ability to handle the inherent nonlinearities and complexities of time series data. This approach enables the model to learn and adapt to the underlying properties of the system, enhancing flexibility and efficiency in the analysis and prediction of complex phenomena. In this work, we propose an application of the Euler-Maruyama method using three neural network architectures: Extreme Learning Machines (ELMs), Multi-Layer Perceptrons (MLPs), and Recurrent Neural Networks (RNNs). These networks are employed to solve a set of stochastic differential equations characterized by the Wiener process. The Euler-Maruyama scheme is utilized to discretize the stochastic equations, transforming them into a suitable form for neural network training. Empirical results demonstrate the effectiveness of this approach in accurately and efficiently solving stochastic differential equations.

The analysis of rainfall seasonality is crucial as it enables the detection of trends and climate changes on local or regional scales. This not only facilitates the sustainable management of natural resources but is also essential for mitigating environmental risks and improving the quality of life of affected populations. The main objective of this study is to analyze the rainfall regimes of Paraíba, a state in northeastern Brazil, based on the calculations of the individual seasonality index (SIi) and the general seasonality index (SI), as well as the replicability index (RI). To achieve this goal, daily precipitation data recorded at 130 rain gauge stations over the period from 1994 to 2020 were analyzed. These data were made available by the Executive Agency for Water Management of Paraíba (AESA). The modified Mann-Kendall test and Sen’s slope estimator were applied to investigate trends in the SIi time series. Additionally, the Inverse Distance Weighting (IDW) interpolator was used to perform the spatial analysis of precipitation. The results indicated that the Zona da Mata is the region of the state with the highest rainfall indices, followed by the Agreste, where the rainy season begins in March and extends until June, characterized by strong seasonality and a short dry season. In the Borborema region, the SIi index revealed extreme seasonality, with almost all precipitation concentrated within just one to two months. Meanwhile, the Sertão, in the western portion of the state, exhibited a moderate but markedly seasonal rainfall regime, with a long dry season. Regarding the replicability index, the results suggest that the Sertão has the most consistent rainfall regime in the state. Stations with significant positive trends are predominantly located in the Agreste and Borborema regions. The relationship between the SIi index and longitude was also investigated. The results revealed a significant linear correlation, with a coefficient of r = -0.81. This suggests that the farther a region is from the ocean, the higher its seasonality index.

Forest fires are complex phenomena, influenced by climatic factors, in addition to human interference. In Brazil, fires affect all biomes. Therefore, the study of these fires is necessary due to their negative effects, including not only environmental damage, but also greenhouse gas emissions and economic losses. In this scenario, understanding the temporal distribution of fires is challenging, due to the variation in their natural behavior. This study seeks to address this issue, using fractal techniques to analyze the long-term temporal and spatial correlations between fires and climate variables in Brazilian biomes during the period from 2002 to 2022. The results for the Amazon, Caatinga, Cerrado, and Atlantic Forest biomes showed that in the daily series of fire anomalies and climate variables (relative humidity, maximum temperature, rainfall, and wind speed), there are persistent long-range correlations, in which the persistence of fires was strongest in the Amazon biome and weakest in the Atlantic Forest. Climate variables are more persistent in the Caatinga biome and less persistent in the Atlantic Forest. Furthermore, persistent long-range cross-correlations were observed between the series of climate variables and fires in the four biomes. For the Amazon, Caatinga and Cerrado biomes, the DCCA correlation coefficient values indicated positive correlations between fires and the maximum temperature and wind speed variables, and negative correlations between fires and the relative humidity and rainfall variables. For the Atlantic Forest biome, the correlations between fires and the maximum temperature variable were positive and negative for the other climate variables.

This study aimed to model growth curves of Lionhead rabbits and Zebu x Taurine cattle subjected to diets with different protein levels and feeding regimes, respectively. For the rabbits, 32 animals were distributed into four groups with diets containing 13%, 16%, 19%, and 22% crude protein, evaluated weekly between 73 and 192 days of age. Five nonlinear models (Brody, Gompertz, Logistic, Richards, and von Bertalanffy) were fitted to the data. According to evaluation criteria combined with cluster analysis, the von Bertalanffy model was deemed the most appropriate. Subsequently, a curve identity test and an analysis of critical points (inflection point (IP), maximum acceleration point (MAP), maximum deceleration point (MDP), and asymptotic deceleration point (ADP)) were performed. The curve identity test indicated significant differences between the diets (p < 0.05), except for performance between the 16% and 19% protein diets. The analysis of critical points revealed that rabbits fed the 13% protein diet exhibited slower development, whereas those on the 22% protein diet reached growth stabilization more quickly. The comparison of treatments demonstrated that the 16% protein diet provided the best performance in terms of animal health. In the study with cattle, symmetric nonlinear models were fitted to the accumulated weight gain of 24 animals, which were divided into three feeding regimes: no adaptation protocol (CON), with adaptation protocol (ADP), and extruded diet (EXT). Weight gain was monitored daily over 144 days. The fitted models included Brody, Gompertz, Logistic, Santos, and von Bertalanffy, and the distributions analyzed were Normal, t-Student, Exponential Power, Cauchy, Logistic I, and Logistic II. The von Bertalanffy model fitted with a t-Student distribution with 3 degrees of freedom proved to be the most effective from both statistical and biological perspectives. Regarding feeding regimes, the ADP diet stood out, showing higher absolute growth rates and greater growth acceleration, outperforming the CON and EXT diets. In summary, the analysis of critical points of the growth curves provided valuable insights for more precise nutritional interventions, and the application of symmetric distributions was crucial for handling data with outliers, enhancing the robustness of the analyses. These findings can contribute to more effective nutritional management practices, with implications for both companion animals and meat production.

The combination of different types of vegetation cover significantly changes hydraulic properties, thereby controlling soil erosion. Generalized Linear Models (GLMs) form one of the most popular classes of statistical models and can provide accurate predictions by establishing a link function to linearize the deterministic relationship between the predictor and response variables. Furthermore, Generalized Additive Models for Location, Scale, and Shape (GAMLSS) are more refined distributional regression models that allow flexible regression and smoothing adjustments to data. In the present study, GLM modeling was performed using Gaussian and Gamma distributions, while GAMLSS modeling was conducted with Skew Exponential Power (SEP), Sinh-Arcsinh (SHASH), Johnson's Su (JSU), and Skew t (ST) distributions to determine the relationship between the Reynolds number (Re) and the Froude number (Fr) in the Darcy-Weisbach coefficient (f), which is frequently used to estimate hydraulic resistance to inter-rill erosion under different vegetation cover conditions. For GLM, in the univariate approach, the Froude number demonstrated superiority in explaining hydraulic resistance behavior in shallow surface flows. However, to improve the prediction model's accuracy, other variables were incorporated, such as water flow depth (h), unit discharge (q), and soil loss (SL). Diagnostic analyses were performed to evaluate the goodness-of-fit of the Gamma model with a logarithmic link function to the surface runoff database. In the GAMLSS analysis, the Froude number also demonstrated superiority in predicting hydraulic resistance, along with variables such as sediment concentration (C_s), rainfall intensity (I), infiltration rate (Inf.rate), runoff coefficient (C), mean flow velocity (V_m), h, inter-rill detachment rate (D_i), PS, and vegetation drag coefficient (CD), with an emphasis on the SEP type 4 distribution. The results demonstrated that the predictor variables were significant, and the Gamma modeling with a logarithmic link function (for GLM modeling) and SEP4 (for GAMLSS modeling) were satisfactory in accurately predicting the Darcy-Weisbach resistance coefficient (f) under different types of vegetation cover and bare soil conditions.

University dropout is a complex phenomenon that directly impacts higher education institutions and students, leading to financial resource waste and hindering young people's academic and professional development. At the Federal Rural University of Pernambuco (UFRPE), dropout rates in Agricultural Sciences programs pose a significant challenge, requiring the identification of key factors contributing to early withdrawal. To deepen this analysis, this study adopts a counterfactual inference approach, allowing for the estimation of causal impacts of different variables on dropout rates. By applying counterfactual modeling techniques, alternative scenarios were analyzed to determine which factors increase or decrease students' chances of retention. The results indicate that the number of accumulated course approvals in the first semesters significantly influences academic retention, suggesting that academic support strategies can effectively reduce dropout rates. The proposed approach strengthens university administrators' decision-making processes, providing data-driven insights to enhance educational policies aimed at improving student retention.

Soil has been increasingly studied due to its structure playing a vital role in sustainable food production and the well-being of society, mainly. In this sense, there is a growing search for a more holistic approach to land use and management to deal with the increasing pressure on soil resources for the production of food and, at the same time, reduce the adverse environmental impacts of agricultural practices. Such structure, which can be described as the spatial arrangement or heterogeneity of soil particles, aggregates and voids or pores, determines the functionality and sustainability of this living natural entity essential for the earth to fulfill its function. Recent advances in non-destructive imaging techniques, such as X-ray computed tomography (CT), make it possible to examine pore space characteristics through direct observation of soil structure. A quantitative characterization of the three-dimensional architecture of soil morphology is crucial to understanding soil mechanics as it relates to the control of biological, chemical, and physical processes at all scales. Despite improvements in the resolution of computed tomography equipment and computational power, there is no consensus on data analysis methods that allow revealing the complexity of all elements associated with 3D soil images, especially methods that do not require a threshold to segment images. In this work, we propose an approach to study the morphological properties of soil and analyze changes in soil structure due to the disturbance caused by current sugarcane management techniques, which causes changes in its structure, mainly in the upper layer; for this, we use the Visibility Graph (VG) and Horizontal Visibility Graph (HVG) methods, based on the theory of complex networks, which converts time series into graphs through a geometric visibility criterion that associates each data in the time series to a node in the visibility graph. To date, these two methods have not been used in 3D image analysis. Analysis of the changes occurring in the structure is possible by comparing computed tomography images of the soil with Atlantic Forest and sugar cane coverage. VG and HVG are applied to columns of voxels in the gravity direction (Z) producing the planar (XY) distribution of topological indices. For each column of voxels VG and HVG networks were generated, and the Clustering Coefficient C, the Average Shortest Path Length ⟨d_ij ⟩, and the Average Degree ⟨k⟩ were calculated, which are indices used to describe the network topology. Promising results were obtained with two soil samples, the first of which consists of a computerized image of soil with Atlantic Forest cover, and another of sugar cane, both with a depth of 0-10cm, were used to show the efficiency of the methods applied. In terms of comparing the application of the HVG and VG methods both performed for each vertical column of voxels proved to be efficient and produced very competitive results. The index Average shortest path length of the HVG network and the Average degree of the VG network showed the greatest difference between the samples, showing that these indices are efficient and suitable for quantifying the degradation of soil morphological properties caused by changes in vegetation cover.

The objective of this research was to evaluate the height growth of forest species, cultivated in an agroforestry system, including two clones of Eucalyptus (Eucalytpus urophylla × Eucalytpus tereticornis), Aroeira (Myracrodruon urundeuva Allemão) and Angico (Anadenanthera colubrina var. Cebil), in intercrops or monocultures with an agricultural species, Cowpea (Vigna unguiculata (L) Walp.) and a forage species, Tanzânia Grass (Panicum maximum Jacq), and verify the effect of the spacings 3 m × 2 m in monoculture and 4 m × 2 m, both in monocultures and in intercrops with Tanzânia Grass and Cowpea, and also the effectiveness of the application of nutrient sources and a soil conditioner, residue from fish farming tanks, sediments from Itaparica reservoir, biochar and control, each repeated four times. The experiment is located in the municipality of Belém de São Francisco, Pernambuco, and was implemented in March 2014, at the Experimental Station of the Agronomic Institute of Pernambuco. The experimental design is multivariate with repeated measures over time. The plantings were organized in two spacings 4 m × 2 m for tree species, both in monoculture and in consortium with Cowpea and Tanzânia Grass, and with spacing of 3 m × 2 m for tree species in monoculture. The tree species were allocated to 256 plots, each containing 28 plants, 10 of which occupy the useful area. Plots with 3 m × 2 m spacing are 14 m × 12 m in size, with an area of 168 m² per plot, in 4 m × 2 m spacing they have dimensions of 14 m × 16 m, and an area of 224 m² per plot. To carry out the analysis, three statistical methods were used and compared: multivariate with repeated measurements over time, split-plot design and time-series analysis. The 4 m × 2 m spacings combined with Tanzânia Grass and Cowpeas demonstrated promising results, being the most favorable spacing options. The nutrient sources did not show statistical differences between them. The MA 2000 and MA 2001 clones stood out as the species with the greatest growth, however, the MA 2001 clone presented the highest heights in practically all periods analyzed, statistically differentiating itself from all other species. The MA 2001 clone appears to be a highly promising option for cultivation in the region, showing accelerated growth and proving to be suitable to meet local energy demands. The results highlight the relevance of choosing fast-growing species to meet energy needs, especially in semi-arid regions, where illegal logging of native species is a concern. Furthermore, this approach offers opportunities to diversify the sources of income for farmers in the semi-arid region, reducing exclusive dependence on conventional agriculture, since eucalyptus wood has significant value and is used in several sectors.

The present study aimed to evaluate housing systems for breeding rabbits and rabbits during
the fattening period, using 24 F1 crossbreed mother rabbits (1/2 New Zealand White and 1/2
California). The F1 crossbreed breeding rabbits were obtained from crossing New Zealand
White and California rabbits. The experiment was conducted at the Rabbit Farming sector
facilities at the Federal Institute Minas Gerais, Campus Bambuí, during the period from July
2020 to February 2021. The experimental design adopted was completely randomized (CRD),
with three different treatments (housings): T1 - Individual cage, T2 - Cage for two rabbits and
T3 - Collective pen (four rabbits per pen). To apply the analysis of variance (ANOVA), the
normality of the variable of interest (Y) was verified using the Shapiro-Wilk tests, as well as
the homogeneity of variances using the Bartlett test. The productive performance of the
rabbits was evaluated considering the parameters: weight at first insemination, final weight,
weight at 120, 150 and 180 days of age, weight gain in the intervals from 120 to 150 days and
from 150 to 180 days. The completely randomized design (CRD) was used for the initial
parameters, while the other parameters followed a completely randomized split-plot design,
with the types of housing as main treatments and the age of the does as secondary treatments.
Reproductive performance was evaluated according to the following parameters: total number
of births, number of live births, weight of offspring at 35 days of lactation, litter weight of
newborns and litter weight of live births. These parameters followed a completely randomized
split-plot design, with housing types as main treatments and cycles as secondary treatments.
The results revealed that housing did not influence the final weight of the does, but there was
a statistically significant difference in weight at the first insemination between the different
types of housing. In relation to weight gain, the accommodations did not have a significant
influence, except in the period of 120-150 days, where a significant difference in daily weight
gain was observed between the accommodations. With regard to reproductive performance,
housing systems did not demonstrate a significant influence on the analyzed parameters. In
conclusion, housing systems played a significant role in improving the productive
performance of rabbits, highlighting their remarkable influence in this context. However, it is
important to highlight that this influence did not substantially extend to reproductive
performance.

The sustainable use of freshwater resources is a major challenge due to climate change, human practices, the construction of dams and reservoirs. The operation of a dam can be modified to produce a natural-like flow without compromising its main services. The empirical analysis of streamflow data before and after dam construction can therefore provide valuable information to the sector responsible for flow control. The objective of this workis to investigte the nature of the hydrological processes of the São Francisco River basin, and the influence of the Sobradinho dam (BA). In order to detect possible hydrological alterations caused by the construction of this dam, daily streamflow data from the São Francisco River recorded at the fluviometric stations of Juazeiro (BA) and São Francisco (MG) were analyzed for the period from 1929 to 2009. The data are provided by the National Water Agency (ANA). The Recurrence Plot (RP) method, its bivariate extension, the Cross Recurrence Plot (CRP) and its modification, the Ordinal Pattern Recurrence Graph (OPRP), as well as the Recurrence Quantification Analysis (RQA) were used to analyze the non-linear dynamics of streamflow series. The visual analysis of the RP/OPRP patterns and the corresponding set of quantitative measures obtained by the RQA indicate that the operation of the reservoir induced changes in the downstream flow (recorded at the Juazeiro station) towards a less predictable and less stable regime, while the flow dynamics upstream (at the San Francisco station) remained unchanged. The temporal evolution of RQA quatifiers (obtained by applying OPRP in sliding windows) identified the influence of the Três Marias dam on the downstream of São Francisco station. Both dams induced the same downstream flow changes, with a greater degree of change for the Juazeiro station due to the shorter distance from the upstream dam. The OPRP sliding window technique showed sensitivity regarding the dam operation and also the damping process with increasing distance between the dam and the downstream hydrological stations. For the period before construction, the RQA values decrease over time scales, showing that the complexity of the flow dynamics depends on the change in time scale. This suggests that daily flow dynamics exhibit less complexity and therefore less variability than larger-scale flows. From the analysis of the relationship between flow and precipitation, it was observed that during the natural regime, flow dynamics were more predictable than precipitation dynamics, which may be the result of the influence of other factors such as evaporation, soil moisture and vegetation cover. The CRP analysis indicated a decrease in synchronization between hydrological processes in the basin after the construction of Sobradinho. The results are expected to contribute to the literature by providing new aspects of hydrological alterations caused by human activities which can be used to improve meteorological and hydrological modeling and forecasting to help with natural disasters and water resource management.

In this thesis, we present a comparative study of otolith density variations using Topological Data Analysis (TDA). Otoliths are calcium carbonate structures found in the inner ears of fish and are commonly used to study age and growth patterns in fish populations. Traditionally, the analysis of otolith density variations has been a computationally intensive task due to the high-dimensional nature of the data. However, TDA offers a promising approach to reduce the data dimensionality and extract meaningful topological information from otolith images. We applied the Ball Mapper algorithm to a dataset of 3D otolith images from different fish species and ages. The algorithm allowed us to constructo topological graphs representing the density variations in otoliths. We also explored the use of probabilistic sampling techniques to reduce the data and found that a sample size of 5% provided accurate representations of otolith density variations compared to the full dataset, after a Sample Topological Validation procedure developed here to ensure the efficiency and reliability of the sampling process. Topological invariants of the graphs, such as average clustering, node connectivity, assortativity, shortest path length, efficiency, and others, were used to comparizon between graphs. The comparizon of the topological properties of the full dataset with those of the 5% sample found a high degree of similarity, indicating that TDA with a reduced dataset can capture essential density information. Ball Mapper further allowed us to identify and eliminate dirt or anomalies present in otolith images, further enhancing the accuracy of our analysis. Overall, our study demonstrates the efficacy of TDA in studying otolith density variations with significant computational gains over traditional methods. The reduced data size using probabilistic sampling and the robustness of topological invariants provide valuable insights into the density patterns of otoliths. Another TDA technique, Persistent Homology (PH), was applied to the 3D image data with the expectation of unveiling a new classifier for otolith shape. PH demonstrated prominence even in a small sample by effectively separating otolith classes and revealing accurate quantitative separation results, showcasing potential use for otolith classification based on their 3D structure. Finally, a regression analysis demonstrated the possibility of estimating age, length, and radiodensity of otoliths based on the topological features resulting from the classification.

Probability distribution models have been widely studied by researchers and scientists for describing and explaining natural phenomena. In recent years, several more flexible probability distribution models have been proposed to describe data, for example in Survival Analysis. Among the models proposed in the literature, Generalized Transmuted Families stand out as they are efficient in describing and explaining data of this nature. Despite many published works that deal with new models or classes of distributions, few are aimed at developing efficient goodness-of-fit measures. The objectives of this work are: (i) to propose a new estimation criterion and goodness-of-fit measures - considering qualitative and quantitative aspects - based on the Mellin Transform. Such goodness-of-fit measures were developed using Hotelling statistics and applied, initially, to four submodels of the Transmuted-G class: Transmuted Weibull, Transmuted Fréchet, Transmuted log-logistics and Transmuted Pareto; and, later, in two submodels of the New Transmutada-G class: New Transmutada Weibull and New Transmutada Fréchet. To verify the performance of the proposed methodology, a survival analysis database and SAR image data were used; (ii) Explore properties relating to information measurements for the New Transmuted-G class: Shannon Entropy; Fisher information and mean Gini divergence. Furthermore, divergence measures such as Kullback-Leibler divergence and Chi-Square divergence were obtained to measure the closeness between two probability distributions. The density of Nova Transmutada-G is written as a mixture between the base distribution and the exponentiated base distribution, which are called its components. Furthermore, divergence measures, such as Kullback-Leibler and Chi-Square, were obtained to measure the proximity between two probability distributions within such a class, that is, between the generator's density function and its components. It was found that the Kullback-Leibler divergence between a New Transmuted-G model and its components is free from the base distribution. Furthermore, the mean Gini difference is expressed as the sum of the mean Gini difference of its components.

The present work deals with a new methodology for pricing European options in the Brazilian market: the resolution of the Black-Scholes Equation through a Neural Network Multi Layer Perceptron. It is a supervised learning problem in which the response of the Neural Network is given by the resolution of the Differential Equation and the circumference conditions are the prices practiced in the Brazilian derivative market. It was verified that the Neural Network can learn from the data and presented better results than the analytical solution of Black-Scholes. The learning capacity of the Neural Network was also tested in comparison with ARIMA modeling and this was more accurate for short prediction horizons, but for larger horizons the neural network showed more satisfactory results.

This study investigated the potential and complexity of time series for wind speed in the period from 19 to 2097 in the Northeast region of Brazil. To understand the temporal behavior of the variable in studies, hierarchical clustering and multiscale entropy methods were used. It was observed that there were no occurrences of any of the longest months or throughout the months that were observed in the winter seasons and in the spring territory. And, in the series of better use of time or energy, for a better use, for a better time scale, that is, for a greater intensity of resources, for decision-making over time, for decision-making better quality, for better use, for decision making and direction. pursuit of sustainable development. The results also show the importance of exploratory techniques and information theory in studies of meteorological/climatological events.

Many studies shown that the increase in the planet’s average temperature causes the hydrological cycle to intensify. This could cause changes in rainfall patterns, such as na increase in the frequency and intensity of extreme events resulting in severe and prolonged droughts in some locations and excessive rainfall in others, which would significantly impact the hydrological availability of a region and the quality of life of its inhabitants. Thus, it is necessary to study the variability and impacts of climate change, enabling a better understanding of the area’s climate in order to adapt and mitigate these climatic conditions. In this context, the present work aimed to analyze the trends and magnitudes of 11 extreme weather indices recommended by the Expert Team on Climate Change Detection and Indices for the state of Pernambuco. For this, 809 grid points were used, which contain information regarding daily rainfall from 1961 to 2020 and three non-parametric methods were employed: the Mann-Kendall test for trend detection, sen’s slope for estimating the magnitude of the trend, and the Mann-Whitney-Wilcoxon test to assess whether there are significant differences in the index values for each region of Pernambuco: Zona da Mata, Agreste, and Sertão. In addition, the Inverse Distance Weighting interpolator was used to perform the spatial analysis of precipitation and extreme weather indices. The results indicated na intensification of drought over much of the state, with significant reductions in total annual precipitation, consecutive wet days, and an increase in consecutive dry days. The trends show an acceleration in the desertification process in the Sertão region, which is part of the semi-arid Northeast and already suffers from scarce and poorly distributed rainfall. In relation to the Zona da Mata, the extreme rainfall indices showed significant increases, alerting us to the natural disasters that affect this region. The Agreste region showed similar results to the Zona da Mata, but with less intensity. Based on the results obtained it is possible to infer that the study area tends to become drier, with rainfall increasingly concentrated in shorter periods of time, and the dry periods interspersed between these rainfall events are becoming longer.

This work is dedicated to investigating the characteristics of drought in the state of Pernambuco based on precipitation data from the period 1962 to 2012. Since the phenomenon of drought is recognized as offering the most dangerous natural risks, which can affect the most varied regions of the world and generate economic and social damage, the increase in the search for understanding the dynamics of droughts and the number of methods of analysis created with the objective of identifying, monitoring and quantifying them. The drought indicator recommended by the World Meteorological Organization (WMO), however, is the Standardized Precipitation Index (SPI), the method adopted here. Since the SPI is an index capable not only of identifying whether there has been a drought, but also of classifying its category (mild, moderate, severe and extreme), it has been widely used in the study of this very important phenomenon. Based on data from 133 rainfall stations spread across the entire territorial extension of the state, the SPI was calculated here on an annual (SPI-12) and seasonal (SPI-3) scale. From the series obtained, the drought characteristics were investigated: frequency, affected area and intensity. Together with the results obtained for drought characteristics, the spatial interpolation method Inverse Distance Weighting (IDW) was used. In addition, based on the results of the spatial distributions obtained, the Wilcoxon-Mann-Whitney test was applied in order to identify statistical differences between the Sertão, Agreste and Zona da Mata regions of Pernambuco regarding drought characteristics (frequency, intensity and affected area). For the temporal analysis, the non-parametric Mann-Kendall tests were used to identify the trend, and the Sen's slope test for the magnitude of the trend. Results obtained for the spatial distribution of the frequency of drought on an annual scale indicated that the drought and the mild type was the one that occurred with greater frequency, mainly in the Sertão region, where it was also verified that there was a greater proportion of stations with a significant trend. On a seasonal scale, the winter and spring seasons showed more pronounced concentrations between types of drought and greater statistical differences between regions. With regard to the results of the area affected by drought, it was observed that, both in the annual scale and for the seasonal scale, the Mann-Kendall test points to the presence of a significant positive trend. As for the intensity of annual drought, the spatial distributions showed higher concentrations in the south of the Zona da Mata for general (total) and moderate droughts, and in the east of the Sertão and southwest of the Agreste for severe and extreme droughts; on the seasonal scale, there was a lower intensity of severe and extreme drought in the Sertão during winter and spring. The average intensity of drought in both scales showed a significant positive trend. Finally, the relationship between the average intensity and the area affected by the drought was obtained, which was positive for all types of droughts and in the two scales considered.

Given the uncertainty in which real systems operate, especially when it involves, by its nature, unpredictable human actions or machine malfunctions. It becomes necessary to search for determinı́stic models, which contribute to the understanding of the dynamic behavior of a system, at the basic level. Such systems can be described by probabilı́stic models, taking advantage of certain features of regularity that they exhibit. Thus, one can resort to Stochastic Processes as a way to treat these phenomena quantitatively, depending on certain characteristics one can resort to Markov Processes. Given a dynamical system, whose dynamics can be given in continuous or discrete time, a study of its evolution of states over time is necessary. In some applications, the distinction between continuous and discrete systems is not critical, and the choice is made for convenience. This work seeks the description of an algorithm that is able to discretize the system studied, in a natural way, thus forming a connected graph, in order to study the dynamics of states of this same graph over time. Thus, given the input, it can generate the appropriate output, thus answering the questions pertinent to the system.

This study aims to evaluate the use of weighing tape in pregnant mares as well as to determine the most appropriate equation for estimating the body weight of pregnant mares. Since equine scales are not readily available due to their high cost, measuring tapes are the most common method of measuring body weight in horses. However, these measurements are not always representative of the animal's actual value, with different factors potentially affecting the tape readings. The experiment was carried out in the Equideoculture Productive Didactic Module, Campus II, Center for Agricultural Sciences of the UFPB, located in the municipality of Areia. Four pregnant mares were used, evaluated weekly during the end of pregnancy. Weighing was carried out weekly on a commercial scale, measurement with a3 weight tape for horses; in addition, the collection of biometric measurements was carried out: G1—thoracic circumference; L1—body length from shoulder to ischium; L2—length from elbow to ischium; H—height at withers and N—neck circumference. The use of a weighing tape and the calculation of metabolic weight are effective methodologies in horses. However, for pregnant mares in the final third of gestation, the results demonstrate that the alternative weighing methods differ statistically from the actual weight. It can be concluded that none of the alternative methods proved to be as efficient as the scale for measuring the weight of pregnant mares in the final third. Since Martinson et al. (2014)c the most appropriate method, compared to the other methods studied in this experiment, for pregnant mares in the final third.

The need for access to computational resources grows as the increasing complexity in the development of computational algorithms becomes frequent in different sectors of the scientific community. The search for these resources has stimulated the development of several cloud platforms, which abstract the complexity of a computational infrastructure while offering its users access to the resources needed for their simulations. However, the cost of accessing these resources can limit the profile of users who can access them, setting aside various studies that could be carried out in a simpler infrastructure. Furthermore, with the complexity increase of the problems to be solved in research activities through the development of computer simulations, advanced features of parallel and distributed programming have become a requirement for these simulations to be executed promptly. The existing gap in the statistics undergraduate courses syllabus opens space for the development of a solution that allows an abstraction of the complexity of this type of programming, allowing written codes to be executed sequentially. Run in parallel with minimal tweaks to the original code. In this thesis proposal, we present the Parallel Experiment for Sequential Code (PESC), a platform for distributing simulations on computers available in a network, packaging the user code in containers that abstract all complexity required to configure an execution environment and allow any user to benefit from this existing infrastructure. With an easy-to-use web module and a client module installed on the computers that will run the simulations, it is possible to run simulations in several programming languages, scripts, and frameworks (Python, Java, C, R, PyTorch, Tensorflow, among others). The results are consolidated through the user’s page with runtime statistics and a distribution map of the simulations by the computers. We will present results obtained in simulations that required more than 1000 runs with different initial parameters and various other studies that benefited from using PESC.

Sporotrichosis, a fungal disease of the genus Sporothix, present in social life, affecting human and non-human animals, characterized as an endemic disease in Latin America, eventually causing regional outbreaks in the country. Responsible for causing various injuries to tissues, being a disease with infectious potential and transmissibility. Incisive presence in felines, whose contamination occurs by traumatic inoculation via scratches, plant thorns and fights between animals, contact with contaminated soil and animal bites, thus considered an occupational disease. In the process of treatment and care, the veterinarian assumes a relevant role, from the accurate diagnosis to the appropriate care and treatment, aiming to avoid abandonment and thus contributing to minimize transmission to humans. The treatment of sporotrichosis has been carried out with the use of the drug itraconazole, in some cases associated with other drugs such as iodine, with an average duration of six months of treatment until the disease remission, the difficulty of which lies not only in the oral administration of the drug, need to isolate the animal and lack of knowledge about the disease and care on the part of guardians. The research was submitted to the Research Ethics Committee - CEP, approved by the opinion consubstantiated under the registration CAAE: 44330921.2.0000.9547 and carried out in Greater Recife, in the municipalities of Recife, Cabo de Santo Agostinho and Camaragibe, in addition to the municipality of Timbaúba, State from Pernambuco. In methodological principle, it was decided to divide the study into three chapters: I - Survey of Geospatialization and Statistics of Sporothrix spp in Academic Digital Platforms in Brazil. Aiming to reflect on postgraduate research in public and private Brazilian universities. Using the State of the Art and bibliographic methodologies, considering the time frame 2011 to 2020, in the databases, CAPES and BDBTD platforms. Whose data analysis made use of statistical analysis through ANOVA and geospatialization with the QGIS software, version QGIS-OSGeo4W-3.24.0-2 (18.02.2022). Resulting in the recovery of 49 scientific productions, 18 from the CAPES platform and 31 from the BDBTD platform. Concluding that there are records that the disease has been present in the country since 1980, as well as the treatment of the disease caused by the pathogenic agent is possible and depends on the therapy. II - Geospatialization of Feline Sporotrichosis, from the Evidence-Based Health Perspective, in the Metropolitan Mesoregion of Recife in Pernambuco/Brazil. It aimed to verify the behavior of cases of feline sporotrichosis, registered at the Veterinary Clinic in the municipality of Abreu e Lima, Pernambuco/Brazil, striving to find clinical evidence in the use of combined drugs or substitutes for itraconazole. The method used was the PICO strategy from the perspective of evidence-based health. As a result, feline sporotrichosis tends to increase in the number of confirmed cases, coexisting with various forms of treatment. Concluding that the diagnosis prevails in male felines, whose treatment and therapeutic care for healing depends on the animal's organism to the therapeutic response. III - Geospatialization and Statistical Analysis of the Evolution of Sporotrichosis Cases in Timbaúba and Greater Recife/PE - Brazil: The Relation of the Disease with Domestic Animals. The objective was to propose a management model based on geospatialization, with statistical methods applied to preventive and corrective care in order to control the evolution of cases of sporotrichosis in care at animal health establishments in Greater Recife-PE. QGIS software was used as a methodology for special analysis, specifically MMQGIS for geocoding. The addresses being converted into geographic coordinates using the Google Maps® algorithm and support of the SIRGAS 2000 UTM coordinate systems. As well as the GeoDa software (version 1.2) was used. The statistical analysis was descriptive, given the proposed variables in absolute and relative (categorical) frequencies, seeking to present the data of summary measures in: mean, median, minimum, maximum and standard deviation (numerical). The respective analyzes were performed after tabulating the data, in spreadsheets (Excel), in Jamovi Software (version 2.3), with R Core Team language (version 4.1), in R Studio Software (4.2.2) and GeoDa Software (version 1.2) , with a time frame from 2011 to 2020. Resulting in the lack of records at three collection points, which may characterize the underreporting of cases of sporotrichosis. Concluding that the lack of adequate records, registration information, animal care, at the collection points, combined with the conclusive completion of the medical record, define the evidence of underreporting of zoonosis. The diagnosis of sporotrichosis prevails in felines, males and adults, whose form of treatment follows the standard other (itraconazole). With the implementation of the Veterinary Registration System, it is possible to manage information consistent with the identification of focus areas, possible outbreaks, socio-environmental conditions, as well as making it possible to build a database with mapping of areas that aim at efficient public policies on the part of of human and non-human health services.

The financial market is a highly dynamic environment, characterized by significant volatility in the data associated with its processes. Consequently, modeling and predicting time series derived from these markets pose substantial challenges. However, comprehending the behavior of financial time series plays a pivotal role in making more informed decisions in the business domain. Consequently, numerous studies aim to develop sophisticated methodologies for forecasting series, with a particular emphasis on financial time series prediction. Notably, studies that employ multiple models to perform forecasts have garnered attention. The combination of time series forecasting models has consistently yielded more accurate results than individual models, as demonstrated by several works in the literature. As a result, numerous techniques promoting the combination of forecasting models have been introduced since the previous century. Research efforts have focused on devising accurate combination models that effectively weight all the involved models. In this study, our objective is to showcase the potential of utilizing copula functions to combine deep learning techniques for predicting financial time series. Specifically, we employ established individual forecasting models, such as ARIMA, Artificial Neural Networks (ANN), and recurrent deep learning networks (Long Short Term Memory - LSTM), to predict five financial time series. Performance metrics, including Rooted Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Prediction of Change in Direction (POCID), are used to compare the results. The ensemble techniques employed in this article include simple mean, simple median, copula functions, and MLPs. The findings of this study demonstrate that combining copula functions with deep learning approaches yields superior results compared to other approaches documented in the literature. Overall, we conclude that, in the context of financial time series, combining deep learning techniques using copula functions generally leads to more accurate predictions in terms of accuracy.

Dozens of definitions of soil can be found in literature, ranging from the most straightforward concepts, where it is asserted, for example, that soil is a heterogeneous mixture of air, water, inorganic and organic solids, and microorganisms, to more complex concepts, where the soil is considered a living, four-dimensional natural Entity. However, regardless of the adopted definition, the importance of soil is unquestionable, as it provides nutrients for plant growth essential for human and animal nutrition. Moreover, history has frequently shown that its misuse can lead to poverty, hunger, drought, and ecological and economic disasters. This great importance given to soil generates a need for ongoing studies searching for methods and tools that contribute to new knowledge. A powerful tool that can observe the elements of soil in a non-destructive way is computerized tomography (CT). Despite advances in the resolution of CT equipment and computer power, there is no consensus on data analysis methods that can reveal the complexity of all elements associated with 3D soil images, especially methods that do not require a threshold to segment images. In this context, this work employs two methods originally developed for the analysis of complex signals, Detrended Fluctuation Analysis (DFA) and Fisher-Shannon (FS), to bring a new understanding of the complexity of morphological properties of soil based on the analysis of 3D CT images. Up to date, these two methods have not been used in 3D image analysis. In this work, 3D soil tomographic images were analyzed using DFA in its original form and its generalization for 2D and 3D data. The results of DFA exponents were found to be smaller than 0.5 indicating antipersistence of local density fluctuations, which are consistently stronger (lower exponent value) for the sugar cane plantation sample, than for the Atlantic Forest. Furthermore, a new complexity measure is defined as the distance from the isocomplexity line in the normalized FS plane, which may be seen as a quantifier of soil degradation level. This novel approach resulted in a high grouping success rate (91.7%) between soil covered by native vegetation (Atlantic Forest) and soil that was the subject of the degradation process as the consequence of land use change (from native Atlantic Forest to sugarcane cultivation).

The evolution of methods for data collection is undeniable, leading to a growing and accelerated availability of these in the most varied areas of knowledge. This context, the need for a proportional evolution of the tools associated with data governance is evident, since the prospecting and analysis process depends entirely on it. Therefore, considering the issue of data governance associated with communication/visualization, it is necessary to basically follow the steps of import, organization, transformation, filtering, modeling and finally presentation/communication. Evidently, in the execution of the mentioned steps, there is the application of statistical tools aiming at multiple interests and, therefore, demanding the use of computational power, both in terms of hardware and software, and in general, software for data analysis has a significant cost. In this sense, this work will demonstrate that interactive-responsive systems, focused on the data governance stages of import, transformation, filtering, modeling, and communication/visualization, when implementing relevant analysis tools and outputs, constitute efficient, easily accessible, and low cost tools for analyzing the significant data demand present in the present time. The fields of education and healthcare are among the priorities for strengthening sustainable development in Brazil and many other countries, and for this reason, they were the focus of the present thesis for the application of the developed systems. Thus, in this work, systems of diverse interests were developed to elucidate the implementation methods, each considering its own particularities. Firstly, the source code was implemented to access, process, and model the registration data for the selection process at Colégio Dom Agostinho Ikas of the Universidade Federal Rural de Pernambuco. This includes generating the ranking list in a .pdf document, as well as presenting the registration data and classification results through an interactive-responsive via the web. Additionally, the source code was developed for the analysis of Covid-19 pandemic data provided by the Ministry of Health of Brazil for the state of Pernambuco. This was motivated by the significant relevance of the topic associated with the data, as well as the opportunity to analyze the official data provided by the Brazilian Ministry of Health, regardless of their quality. In summary, the research involving the aforementioned systems are complementary in what the work proposes to demonstrate, since the first has a relative degree of sophistication in relation to data processing and treatment, while the second deals with a robust database.

The study of the constructors of probabilistic distribution classes by composition of functions aims to present six methods that generalize and generate probability distributions from the composition of functions. The methods proposed here allow the construction of classes of distributions, using monotonic functions of predefined distributions. With these methods it is possible to find new probabilistic distributions, classes of probabilistic distributions and also families of probability distributions, including some already present in the literature. Propositions are presented for the construction of the methods of this work, after that, tables are displayed with some values and pairs of increasing and decreasing continuous functions, together with their derivatives to feed the expressions of the constructor methods, generating some subcases of these methods. Initially, it is presented in this study, the methods that generate classes of probabilistic distributions, via compositions of cumulative distribution functions – cdf, in which baseline mixtures will be used from the construction of connection functions, together with the distribution construction methods. Next, methods for generating classes of probabilistic distributions by probability density function composition – pdf, based on the construction of linkage functions with baseline mixtures are presented. Studies of the supports of the functional generators of probabilistic distributions and the identifiability of the classes generated by the proposed methods are also developed. Some applications of these methods on simulated or real data are explained. In addition, for the classes and distributions generated by the functional models of the cdf and the pdf, there is the development of the expansions of the cdf and the pdf, the risk function, expansions for moments of order m, and for the moment generating function, as well as for the characteristic function, the expansions of central moments m and for the general coefficient, for the mean deviation and quantile deviation, the derivatives of the log-likelihood function and, the entropies of Shannon and Renyi. As a result, applications of a case of the methods are developed in which they are used in simulated data and/or in real data, thus pointing out the viability and generalization that serve as an anchorage for other researches in the different branches of the agricultural sciences.

Changes in the behavior of rainfall evolution affected the hydrological cycle and consequently water resources. In cases of extreme climate change, the impact is related to changes in water resources, the occurrence of more severe and frequent floods and droughts. In this sense, studying the impacts of climate change on water resources is strategic for the elaboration, implementation and strengthening of public policies involved in the management of water resources. The present research aimed to characterize the rainfall economy in Sertão do Pajeú - PE, as well as provide subsidies for public policies aimed at water scarcity through the study and analysis of space/time. To this end, information was collected from rainfall levels in the region and surrounding areas, corresponding to the period 1993 and 2022, and statistical and geostatistical methodologies were used. data were initially processed using the regression method to complement missing data. Then, a series was checked for trend, consistency and disruption of information. In all cases, the results of the tests showed that there was no statistically significant trend, the consistency of the data was validated and the ruptures identified were not significant at the 5% level. Therefore, we approached the statistical characterization of the Pajeú variation where the climatological normal was 80.62mm. Furthermore, the classification of years was carried out using the quantile technique. The results were overwhelming, showing that some years classified as Very Dry or Dry (1993, 1998, 2015 and 2016) coincided with the occurrence of ENSO events. In relation to the geostatistical analysis, adjustments were made to the spherical, exponential and gaussian models. The models were selected using the cross-validation method. All models were well adjusted, however the gaussian model showed greater goodness of fit R2 = 90.5%. Therefore, the gaussian model proved to be more suitable for adjusting the precipitation in Sertão do Pajeú. Finally, simple spatial interpolation techniques (Thiessen, Spline and IDW), the Spline method, showed better fit and accuracy in interpolation. In relation to the other techniques, kriging considering the
parameters of the gaussian model presented better results in terms of both adjustment (R2 = 95.1%) and accuracy (lower error measurements). Thus, it is possible to conclude that the methodology used to study the average annual rainfall in the Pajeú microregion allowed satisfactory results to be obtained in the assessment of its spatial variability, being able to determine and express the spatial continuity of rainfall and support policies public issues related to water issues in the region.

Agribusiness is one of the most important economic activities developed in Brazil, its share in the national GDP is 26.6%. Fruit farming also stands out in this context, in recent decades the fruit trade in Brazil has grown significantly, serving the domestic and foreign markets. Among the main fruits produced and marketed, mangoes and grapes stand out, occupying the first and third positions of the most exported fruits by Brazil, respectively. In this work, the time series of returns and weekly price volatility of two mango varieties, “Palmer” and “Tommy Atkins”, and two types of grape varieties, “Itália” and “Benitaka”, produced in the São Francisco Valley, were analyzed. , a region with relevant participation in the production and export of these fruits. The Detrended Fluctuation Analysis (DFA) and Detrended Cross Correlation Analysis (DCCA) methods were used to calculate scale exponents of long-range correlations and cross-correlations between the analyzed series. The results showed that the volatility series have stronger persistence than the return series that presented two regimes of scale invariance with anti-persistent correlations on the larger temporal scales. The cross-correlations between the series of returns also presented two scaling regimes, obtaining, for mango, exponents similar to the series of returns of the “Tommy Atkins” variety. The values of the correlation coefficient obtained by the Detrended Cross Correlation Coefficient ρ_DCCA the method showed that for the returns and volatility of the two fruits the correlations between the series are positive, increasing with time scale and are stronger for the return series.

In this work we present three classes of probability distributions build upon an innovative method that considers the Lebesgue integration of a function denoted by H(t). For the three classes we set H(t) to be the standard normal cumulative distribution function, differing only in the limits of integration. The upper limit of integration of the second class is unprecedented in the statistical literature and the third class admits two different baselines, being a competing alternative to twocomponent mixture models. The classes have the advantage of demanding no extra parameters, besides those of the baseline(s), providing thereby parsimonious models. Under certain conditions, the distributions generated by the new classes are identifiable. We present some mathematical properties of the classes, like the linear representation of the probability density function, the moments and the moment generating function. Monte Carlo simulation studies are performed to investigate the behavior of the maximum likelihood estimates of the parameters for some submodels of the classes. The same submodels examined in the simulations are also used in applications to real datasets. Information criteria and formal goodness-of-fit statistics (Cramér-von Mises and Anderson-Darling) are used as criteria for model selection in comparisons considering the studied distributions emerged from the classes and some well-known distributions commonly employed in modelling similar data. The results suggest that the distributions from the proposed classes outperform the competing models

The dengue is compulsorily notifiable disease and are present on National Compulsory Notification List of Diseases, Aggravation and Public Health Events. In Brazilian Northeast, Pernambuco is the fourth state with the highest detection rate (per 100,000 inhab.) and lethality. With that, the present research aimed to analyze the number of weekly cases of dengue in the state of Pernambuco through georeferencing analysis. The research it's about epidemiological study of the ecological type, carried out with data from the Citizen Information Service and the Pernambuco Agency for Water and Climate, in the period 2000 to 2018. Used the TerraView 4.2.2 and Qgis 3.4.11 software for spatial analysis and the free software R version 3.6.1 and IBM SPSS Statistics 20, in which statistical tests were applied. To assess the correlation between the variables, the Pearson correlation coefficient was calculated and the generalized linear models were proceeded. Have been confirmed 508,948 cases of dengue in the state of Pernambuco, in the period 2000 and 2018. The spatial distribution of cases showed a high concentration in the Metropolitan Mesoregion of Recife, with greater frequency from Recife municipality. The results indicate the formation of clusters with positive spatial autocorrelation between municipalities in the Metropolitan Mesoregion of Recife. Regarding the Pearson correlation coefficient, positive correlations were observed in 75.54% of the municipalities. From the estimates of the models, it was verified that the Negative Binomial model proved to be more satisfactory to analyze the distribution of dengue cases. Therefore, the computerized processing of georeferenced data allowed the identification of municipalities with greater vulnerabilities, providing subsidies for the regulatory agencies of epidemics and endemics in the state of Pernambuco to plan their actions.

Semi-arid climate variability combined with ongoing climate changes affects agricultural systems in the world with impacts on crops, and crop yields and threatens global food security. The region of the valley of the sub middle of the São Francisco River has been suffering from these changes that have brought significant losses arising from the fall in the productivity of the orchards and the disqualification of fruits for commercialization, either by the increase in rainfall during the harvest of the table grape or by the harmful effect that high temperatures and low relative humidity have caused during flowering/beginning of fruiting in mango trees cv. Palmer. Therefore, as important as quantifying is predicting extreme events related to high temperature and low relative humidity in order to understand the response mechanism of natural vegetation and agricultural crops to climate change. In this context, inference and statistical analysis are common tools to estimate the risk of events of interest, and the use of copula has been a natural choice as it is flexible to build multivariate distribution and capture the structure of interdependence between the multiple variables involved. Thus, the objective of this work was to select copulas for joint analysis of temperature and relative humidity and spatialization of probabilities in the region of the sub middle valley of the São Francisco River. The study was carried out with daily data on temperature and relative air humidity obtained between 2003 and 2018 at 17 meteorological stations in the region. For each month of data, 12 probability models and 10 copula models were fitted, which presented the best performances for univariate and bivariate estimates. For most months, the Log-Logistics-Exponenced, Generalized Extreme Values distributions and the Plackett and Frank copulas were the ones that were best suited to model the temperature and relative humidity of the air in the region of the sub middle of the São Paulo River. Francis. October and November are the months in that most of the São Francisco River sub-medium valley region is subject to maximum and average temperatures, respectively, above the thresholds of 31.6°C and 26.1°C with 90% probability. October and November are also the months with the highest risk of maximum temperatures above 33°C associated with the simultaneous occurrence of minimum relative humidity below 30%. Therefore, November is the month with the highest probability of occurrence of temperature and relative humidity unsuitable for the flowering of 'Palmer' mango trees and, therefore, more favorable to the occurrence of stenospermocarpic fruits at harvest in April or May, considering the cycle of the variety.

In this thesis, we provide results in two fields. In the first one, the main idea is to make stochastic comparisons using the majorization theory. We investigate conditions on the parameters of the Exponentiated Generalized class with equal and different baselines. In the second one, to describe the fits of models to the data by using Goodness-of-Fit (GoF) measures, we use the Transmuted Inverse Weibull distribution and derive the Mellin Transform from that model. We estimate the parameters by Moments, Maximum Likelihood, and Log-Cumulants (LC) methods. An interesting analysis is carried out with Hotelling's T2 Statistic. In this step, we construct an LC diagram and furnish ellipses of confidence for the LCs. The applications in this part were based on lifetime datasets such as mechanical components' survival time, electrical insulator films' failure times, and censored data from times of bladder cancer patients. We present some simulation results to verify the performance of the Moments, Maximum Likelihood, and LC methods. Kolmogorov-Smirnov and Cramér-von Mises as GoF criterias are used. Monte Carlo methods such as bootstrap and Jackknife are used to estimate population characteristics. The Mellin-based GoF test's performance is compared with Anderson-Darling and Kolmogorov-Smirnov tests.

The aim of this work was to evaluate the fit of nonlinear models in describing the growth of New Zealand rabbits fed different diets. The models Santos et al. (2018), Richards, Gompertz, Brody, Von Bertallanfy and Logistic, with the objective of identifying the best fit model and, later, verifying the performance of the treatments. For the experiment, rabbits weaned at 35 days of age were used, receiving different types of feed, and their weights were measured every 5 days, up to 75 days of age. The 88 animals were divided into eleven groups: Reference Diet (REF), consisting of conventional ingredients, semi-simplified diet based on Hay from the Upper Third of Cassava Rama (FTSRM) - (SSM), semi-simplified diet based on Hay Alfalfa (FAL) - (SSA), semisimplified diet based on Cassava Leaf Flour (FFM) - (SSF), simplified diet based on mixtures of FFM and FAL - (SFA), semi-simplified diet based mixture of FFM and FAL - (SSFA), semi-simplified diet based on the mixture of FTSRM and FFM - (SSMA), SSFA diet with carbohydrates and phytase enzymes - (SSFAE), SFA diet with carbohydrates and phytase enzymes - (SFAE), SSM diet with added carbohydrates and phytase enzymes - (SSME) and SSF with added carbohydrates and phytase enzymes - (SSFE). The fit quality evaluators used were: Adjusted determination coefficient (R2aj.), Mean residual square (QMR), Mean absolute deviation (DMA), Akaike information criterion (AIC) and Bayesian information criterion (BIC), being the cluster analysis performed with the aid of the Ratkowsky index to determine clusters of the models according to the average values of the evaluation criteria. After selecting the most appropriate model, the curves identity test was performed to check for a possible divergence between the diets, in order to assess their performance in the growth of rabbits. The model that showed the best fit was Santos et al. (2018), who from the cluster analysis, joined the group that presented, on average, lower values for QMR, DMA, AIC and BIC and higher values for R2aj. When comparing the growth curves at the 5% significance level, it was observed that the reference diet (REF) performed better than the others. Diets based on Cassava Leaf Flour (SSFE, SSF, SFAE and SFE) provided inferior performance mainly due to the presence of a high level of antinutritional factors that impair the digestive process. The hay-based diet of the upper third of the cassava branch (SSM) provided satisfactory performance, with relatively low cost, and could be an economically viable alternative for the breeding ofx rabbits. The use of the Santos et al. (2018), may contribute to future studies in the description of animal growth.

Manjuba (Opisthonema oglinum) is the most common sardine in Santa Cruz Channel’s
(CSC) landings, located on the northern coast of Pernambuco state. The species is classified
as visiting marine origin, as it spends part of its life cycle at the sea and enters estuaries
at strategic times. The initial research’s issue focuses on speculating some hypotheses
related to the juveniles’ over-catching from O. oglinum species, that are relevant to fisheries
professionals on decision-making. The high exploitation of the species and its importance
to the economy in the municipalities surrounding the CSC led us to reflect on the need
for investigations to promote the sustainable fishing of O. oglinum and, consequently,
expand the understanding of the species in the CSC, once the comprehension of the length
structure in catches is valid not only to expand knowledge but also to raise hypotheses about
the fishing pressure on the resource. The investigation was performed by estimating the
standard length of the specimens, using generalized linear models (GLM) and generalized
additive models for location, scale, and shape (GAMLSS). After performing the modeling,
we observed that the models estimated via GAMLSS enabled a better fit to describe the
standard length of O. oglinum compared to those obtained via GLM, according to the
GAIC criterion and residue analysis. The Box Cox t Original distribution was chosen to
adjust the response variable with the modeling of position and scale parameters. Therefore,
we realized the existence of seasonality in the standard length of O. oglinum captured
in the Santa Cruz Channel, with the presence of larger individuals in the rainy season,
and greater concentration in the center-south region of the channel. On the other hand,
the smaller sardines were concentrated near the outflow of the Botafogo River, an area
of secondary channels with greater mangrove coverage. We have found evidence that few
adults enter the channel and the vast majority of specimens captured in the CSC were
young. Thus, it would be pertinent to explore specimens that have completed at least the
sexual maturity, since the capture of young individuals can cause an imbalance in the
sustainability of the species.

Rain is an element that causes many concerns, because its intensity and frequency of occurrence, can cause harmful effects to society, the rain of a given location can be estimated, in probabilistic terms, by theoretical distribution models adjusted to a historical series. Precipitation data from the state of Pernambuco were used, formed by 133 monthly time series distributed over the entire state during the period from 1950 to 2012. Adherences of Probability distributions were analyzed: exponential, gamma,beta, lognormal, Weibull, normal, Gumbel Marshall-Olkin, log-logistics, exponentiated log-logistics, based on the Kolmogorov-Smirnov test at the 5% significance level . The distributions with less rejection during the adherence test were Weibull, gamma and beta, october had the lowest number of distributions considered adequate to model monthly precipitation and in march more than 99% of the pluviometric stations had some adequate probabilistic distribution to model precipitation monthly. The Weibull distribution was the best suited to model monthly precipitation.

Solar and wind energy play a strategic role in Brazil’s efforts for sustainable development. The implementation of new, more efficient technologies for solar and wind energy conversion will facilitate energy supply in remote areas, such as the Amazon region and islands, and help reduce greenhouse gas emissions to the atmosphere, by reducing the fossil fuel consumption. The high solar irradiation levels with small seasonal variation and wind regime make the coastal areas of northeastern Brazil highly valuable for alternative energy developing programs. Fernando de Noronha archipelago, located close to the Brazilian coast in the Atlantic Ocean, belongs to the state of Pernambuco, and there are preservation rules established by federal and state government with a purpose to preserve natural resources that can be achieved through sustainable development. However, the island energy supply currently comes ma inly from diesel generators, and there is a continuous effort in developing efficient technological solutions for energy supply based on wind and solar resources. The evaluation of wind power potential requires careful statistical analysis of mean wind speed and its frequency distribution. However, the biggest challenge in integrating wind power into the electric grid is its intermittency due to temporal and spatial variability of wind in large range of scales. As a natural process of turbulence wind is the most complex weather variable with specific properties as long-range spatial and temporal correlations and fractal and multifractal dynamics. In order to contribute to better understanding of wind speed and solar radiation dynamics at the location of Fernando de Noronha island, in this work we investigate long-term correlations in temporal series of these variables recorded during the period 2003-2011. We use Detrended Fluctuation Analysis (DFA) method which was designed to detect and quantify long-term correlations in non stationary temporal series. Our results show that both wind speed and solar radiation belong to the class of persistence stochastic processes characterized by long-term correlations. The value of scaling exponent is higher for wind sp eed than for solar radiation, indicating stronger persistence. We also compare temporal variation of scaling exponents for wind and solar radiation by applying DFA on three year intervals, and find that during the studied period the scaling exponent is mor e stable for solar radiation than for wind speed. The results of this work provide new information about wind speed and solar radiation dynamics at the location of Fernando de Noronha island, and should be considered in evaluation of renewable energy potential.

It was objectified to correlate of the variable-response (production, productivity, final weight and tax of survival), with the water physic (temperature in the deep one, temperature in the surface and water level) and the chemical (salinity, oxygen in the deep one, oxygen in the surface, pH and transparency of the water) and the involved variable and the adopted handling (amount of ration, period of winter, period of summer, days of culture, initial density, supplying laboratory of after-larva, area of the fishery, number of cycles, initial population, aeration and days of preparation) of the culture of the marine shrimp Litopenaeus vannamei. The data base of the shrimp farm was composed for 68 comments, correspondents to the period of 2003 the 2005 Esteem the parameters of the models it was used technique of the square minimums. The election of variable was carried through using the process of Stepwise Backward, associated the Box and Cox, to minimize the sum of square errors. The adequacy of the esteem equations and the hypothesis of linearity, normality and homogeneous variance for the errors had been analyzed on the basis of the analysis of variance and in the analysis of residue. According to the results, it was possible to conclude that the water physic and chemical influenced (p<0.05) the vanableness of the variable- response

O estudo sobre curvas de lactação pode contribuir para o melhor entendimento do sistema de produção, pois o conhecimento da forma da curva e suas implicações sobre a produção de leite pode auxiliar o produtor na previsão da produção de leite de suas vacas em determinado estádio de lactação e, também, na tomada de decisões quanto ao descarte ou manejo apropriado. Entretanto, estudos desta natureza com vacas da raça Sindi praticamente não existem no Brasil. Em função disto, o objetivo do presente trabalho foi estudar as curvas de lactação e de gordura de um rebanho da raça Sindi no Estado da Paraíba, avaliando a influência da estação do parto, ano e da ordem de parto sobre a forma das curvas de lactação e gordura médias de vacas da raça Sindi, usando a função Quadrática Logarítmica. Dados referentes a 1165 lactações de produção de leite e gordura de 87 vacas da raça Sindi, controladas no período de 1987 a 1997, de propriedade da fazenda Carnaúba, pertencente à AMDA (Agropecuária Manoel Dantas Ltda), situada no município de Taperoá, microrregião do Cariri Ocidental do Estado da Paraíba, foram utilizados. Os coeficientes de determinação das curvas médias de lactação e gordura foram de 35,1% e 21%, respectivamente. Observou-se efeito significativo dos fatores ambientais (ano, estação e ordem de parto) sobre todos os parâmetros da função, mostrando que existe diferença na forma das curvas e que estas devem ser apresentadas para cada efeito, separadamente.