Includes bibliographical references / Many large, sparsely distributed raptors are threatened by a host of anthropogenic factors, while a minority may actually benefit from some aspects of development and environmental change. Clarity on the size and trajectory of such populations is essential for effective conservation management, but can be difficult to achieve. One solution is to use multivariate habitat association models to derive critical estimates of distribution and abundance. The South African population of Martial Eagle Polemaetus bellicosus is currently estimated at < 800 adult birds , with the bulk of the known population believed to be residing in the larger protected areas. However, Martial Eagles also build nests on pylons that support high voltage transmission lines running through the largely treeless, semiarid landscapes of the Karoo. The main aim of this study was to develop a better understanding of the environmental factors that influence Martial Eagle territory densities and locations along South African transmission lines, and thereby estimate the size of this population and its relative importance to the national conservation status of this globally threatened species. I used habitat association models to d escribe Martial Eagle territory density in relation to eight environmental covariates. Models were first fitted to eagle territory data for the central Karoo regions, collected and pooled over the period 2002 - 2006, and then applied to predict the number of pairs present on each of three adjacent sections of unsurveyed line (northern, southern and eastern lines) . Once these model predictions were verified by a series of aerial and ground surveys, I fitted the models to all the known Martial Eagle territory records for the transmission network and extrapolated from these back to the rest of the network using the fitted relationships. Ultimately, the models predicted 52 additional Martial Eagle territories on the remaining transmission network with a confidence interval ranging from 38 to 67 (based on models that explained up to 39 % of the total variance in terms of only two explanatory terms – rainfall and the proportion of cultivated land). I then examined the role of territoriality and social structure in the eagle population in determining the location and dispersion of pylon nests. To do this I used the location of active nests from the original central Karoo data and a similar number of randomly selected points. I then asked whether I could predict the nest locations from each of the eight environmental covariates and distance to its nearest conspecific active nest or its nearest nest of any other large eagle species. Using a logistic generalised linear model with regression splines for distance to nearest other nest, I found that Martial Eagles strongly avoid proximity to conspecific nests (mean distance to conspecific nest = 28.2 km, range 2.5 - 167.1 km, n = 306). This result shows that minimum spacing should be considered in predicting the distribution of eagles on unsurveyed transmission lines. Lastly, I further investigated the geographical extent of pylon nesting in South African Martial Eagles, with particular focus on variation in the frequency of this behaviour in relation to biome - scale variation in the availability of trees as natural nest sites. To do this, I related Martial Eagle reporting rates generated by citizen - science bird atlas data to the density of transmission lines and biome types across South Africa. While these analyses yielded some suggestive results, such as significant positive and negative relationships between reporting rates and line density in the Desert (P = 0.02) versus the Savanna (P < 0.001) biomes respectively, data sparseness in arid areas and a generally low detection probability limited the conclusiveness of these results. The refined habitat association models developed in this study predict that the South African transmission grid supports 130 - 159 breeding pairs of Martial Eagle. This figure has never been estimated or calculated before, and suggests that 36 % of the national breeding population could reside largely in the commercial ranchland and nest on man - made structures. This result, which is at odds with the generally held belief that the Martial Eagle is increasingly confined to large protected areas, has significant implications for the thinking around the conservation management of this globally threatened species.

Nous présentons une nouvelle méthode pour étudier les relations entre la structure génétique des populations et l'environnement. Cette méthode repose sur des modèles hiérarchiques bayésiens qui utilisent conjointement des données génétiques multi-locus et des données spatiales, environnementales et/ou culturelles. Elle permet d'estimer la structure génétique des populations, d'évaluer ses liens avec des covariables non génétiques, et de projeter la structure génétique des populations en fonction de ces covariables. Dans un premier temps, nous avons appliqué notre approche à des données de génétique humaine pour évaluer le rôle de la géographie et des langages dans la structure génétique des populations amérindiennes. Dans un deuxième temps, nous avons étudié la structure génétique des populations pour 20 espèces de plantes alpines et nous avons projeté les modifications intra spécifiques qui pourront être causées par le réchauffement climatique. / We introduce a new method to study the relationships between population genetic structure and environment. This method is based on Bayesian hierarchical models which use both multi-loci genetic data, and spatial, environmental, and/or cultural data. Our method provides the inference of population genetic structure, the evaluation of the relationships between the structure and non-genetic covariates, and the prediction of population genetic structure based on these covariates. We present two applications of our Bayesian method. First, we used human genetic data to evaluate the role of geography and languages in shaping Native American population structure. Second, we studied the population genetic structure of 20 Alpine plant species and we forecasted intra-specific changes in response to global warming. STAR

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2017-04-18T14:36:39Z No. of bitstreams: 2 Tese - Odilon Peixoto de Morais Júnior - 2016.pdf: 4169553 bytes, checksum: 1841a99cece6656c30b62fbd8fda9da5 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2017-04-18T14:36:54Z (GMT) No. of bitstreams: 2 Tese - Odilon Peixoto de Morais Júnior - 2016.pdf: 4169553 bytes, checksum: 1841a99cece6656c30b62fbd8fda9da5 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2017-04-18T14:36:54Z (GMT). No. of bitstreams: 2 Tese - Odilon Peixoto de Morais Júnior - 2016.pdf: 4169553 bytes, checksum: 1841a99cece6656c30b62fbd8fda9da5 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2016-12-15 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Genetic gains for quantitative traits associated with the maintenance of genetic variability are important factors in recurrent selection programs. With advances in the area of statistical genomics, selection strategies potentially faster to achieve genetic gains are being developed, such as genomic selection. Using a subtropical population of irrigated rice (CNA12S), conducted during three cycles of recurrent selection, this study had as general objective to evaluate the potential of use of genomic recurrent selection (GRS) in a rice breeding program. Three specific studies were developed. In the first chapter, the efficiency of the genotypic recurrent selection (RS) used in the Embrapa’s rice breeding program was evaluated, in order to obtain genetic gains and maintain the population genetic variability. Ten yield trials of S1:3 progenies were used in the analyses. The evaluated traits were grain yield, plant height and days-to-flowering. Variance and covariance components were obtained using Bayesian approach. Using single nucleotide polymorphisms (SNP) markers, the population diversity and genetic structure also were estimated. Adjusted means of progenies in each cycle were computed and, genetic progress was estimated by generalized linear regression using frequentist approach. The magnitudes of effective population size and genetic variance indicated maintenance of genetic variability over selection cycles. The genetic progress achieved for grain yield was 760 kg ha-1 per cycle (1.95% per year), and for days-to-flowering, it was -6.3 days per cycle (-1.28% per year). It was concluded that the genetic progress already achieved and the genetic variability available in the population demonstrate the efficiency of RS in the improvement of rice populations. In the second chapter, in the context of genomic selection, the relative efficiency of GRS on RS was assessed, as well as the accuracy of different models of genomic prediction, in order to propose a GRS scheme for population breeding of self-pollinating species such as rice. In this study, the genetic material was the S1:3 progenies yield trial of the third selection cycle. From a group of 196 progenies that were phenotyped for eight traits with different heritabilities and genetic architectures, a group of 174 progenies was genotyped for SNP markers. Ten predictive models were fitted to the data set. The proposed GRS scheme, when compared to the RS method, showed higher efficiency, especially in genetic gain per unit of time. From the predictive models assessed, HBLUP (hybrid best linear unbiased prediction, using hybrid relationship matrix based in pedigree and SNP markers) and RForest (random forest) have greater potential for genomic prediction in irrigated rice, given the high accuracy of their predictions for a number of traits. The HBLUP model was notoriously superior for more complex traits, such as grain yield, while RForest stood out for less complex traits. The high extent of linkage disequilibrium in the population suggests that the marker density employed (approximately one SNP per 60 kb) is enough for the practice of genomic selection in populations with similar genetic structure. In the third chapter, the objective was to extend a class of HBLUP models based on reaction norm, in context of multi-environmental trials with genotype x environment interaction, for accommodation of hybrid genetic relationship and information of the assessed environments. The accuracy of alternative models for multi-environmental predictions was evaluated, as well as the relative importance of structures of additive and multiplicative components, using genetic relationship information and environmental covariates. This strategy allowed to evaluate the influence of different approaches to group the genetic-environmental information on the accuracy of models for prediction of breeding value of progenies for agronomic traits. The data consisted of the same ten trial of S1:3 progenies, carried out during three recurrent selection cycles. Six predictive HBLUP models of reaction norm were considered, using genetic and environmental covariates, as well as interactions between these effects. Genomic information was derived from SNP markers obtained for the 174 progenies of the third selection cycle. The 401 environmental covariates, the genetic information (hybrid genetic relationship) and the interactions among these effects explained an important portion of the phenotypic variance, allowing an increase in the predictive accuracy of models. The use of genetic information and environmental covariates only from the respective selection cycle is enough for accurate predictions of unphenotyped progenies, even in non-sampled environments. This is the first study to take into account simultaneously hybrid genetic relationship, stemming from pedigree information plus SNP markers, and environmental covariates in multi-environmental models based on reaction norm for breeding value prediction in target environments of a recurrent selection program. / A obtenção de ganhos genéticos para caracteres quantitativos associada à manutenção da variabilidade genética são fatores importantes em programas de seleção recorrente. Com os avanços no campo da estatística genômica, estratégias de seleção potencialmente mais rápidas para alcance de ganhos genéticos estão sendo desenvolvidas, como a seleção genômica. Partindo-se de uma população subtropical de arroz irrigado (CNA12S), conduzida durante três ciclos de seleção recorrente, este estudo teve como objetivo geral avaliar o potencial de emprego do esquema de seleção recorrente genômica (GRS) em programas de melhoramento genético de arroz. Três estudos específicos foram desenvolvidos. No primeiro deles, avaliou-se a eficiência do esquema de seleção recorrente genotípica (RS) utilizado no programa de melhoramento de arroz da Embrapa, na obtenção de ganhos genéticos e manutenção da variabilidade genética populacional. O material experimental utilizado constituiu-se de dez ensaios de rendimento de progênies S1:3 associadas a cada ciclo de seleção. Os caracteres avaliados foram produtividade de grãos, altura de planta e número de dias até o florescimento. Componentes de variância e covariância foram obtidos via abordagem Bayesiana e, com uso de marcadores SNP (single nucleotide polymorphisms) associados às progênies, também a diversidade e a estrutura genética populacional. Médias ajustadas de progênies em cada ciclo foram computadas e, por regressão linear generalizada, estimou-se o progresso genético, via abordagem frequentista. As magnitudes do tamanho efetivo populacional e da variância genética indicaram manutenção da variabilidade genética ao longo dos ciclos de seleção. O progresso genético alcançado para produtividade de grãos foi de 760 kg ha-1 por ciclo (1,95 % ao ano) e para dias para florescimento, -6,3 dias por ciclo (-1,28 % ao ano). Concluiu-se que, o progresso genético já alcançado e a variabilidade genética disponível na população demonstram a eficiência de RS no melhoramento de populações de arroz. Num segundo estudo, no contexto de seleção genômica, avaliou-se a eficiência relativa de GRS sobre o esquema de RS; além da acurácia de diferentes modelos de predição genômica, buscando-se propor um esquema de GRS para melhoramento populacional de espécies autógamas como o arroz. Nesse estudo, o material genético foi composto por um ensaio de rendimento de progênies S1:3 do terceiro ciclo de seleção. Do grupo de 196 progênies fenotipadas para oito caracteres, com herdabilidades e arquiteturas genéticas diferentes, um grupo de 174 progênies foi genotipado para marcadores SNP. Dez modelos preditivos foram ajustados ao conjunto de dados. O esquema de GRS, quando comparado ao de RS, apresentou maior eficiência, sobretudo em ganho genético por unidade de tempo. Dos modelos preditivos avaliados, HBLUP (hybrid best linear unbiased prediction, com uso de matriz híbrida de parentesco baseada em pedigree e marcadores SNP) e RForest (random forest) apresentaram maior potencial para predição genômica, haja vista a elevada acurácia de suas predições para maior número de caracteres. O modelo HBLUP foi notoriamente superior para caracteres mais complexos, como produtividade de grãos, enquanto RForest destacou-se para caracteres menos complexos. A alta extensão do desequilíbrio de ligação na população sugere que a densidade de marcadores empregada (aproximadamente um SNP por 60 kb) é suficiente para a prática de predição genômica em populações com estrutura genética similar. No terceiro estudo buscou-se estender uma classe de modelos preditivos HBLUP baseados em norma de reação (contexto de ensaios multiambientais com interação genótipos × ambientes), para acomodar informações de parentesco e de covariáveis associadas aos ambientes de avaliação. Assim, avaliouse a acurácia preditiva de modelos alternativos para predições multiambientais, bem como a importância relativa de estruturas de componentes aditivos e multiplicativos; além da influência de diferentes abordagens de agrupamento de informações genético-ambientais sobre a acurácia dos modelos. O material genético constituiu-se nos mesmos dez ensaios de rendimento de progênies S1:3, conduzidos durante três ciclos de seleção recorrente. Foi considerada uma sequência de seis modelos preditivos de norma de reação, do tipo HBLUP, com uso de covariáveis genéticas e ambientais, além de interações entre esses efeitos. A informação genômica foi proveniente de marcadores SNP obtidos por genotipagem de 173 progênies do terceiro ciclo de seleção. As covariáveis ambientais (num total de 401), informações genéticas (parentesco híbrido) e as interações entre esses efeitos explicaram importante porção da variância fenotípica, o que possibilitou aumento da acurácia preditiva dos modelos. O emprego de informações genéticas e de covariáveis ambientais apenas do respectivo ciclo de seleção mostrou-se suficiente para predições acuradas do desempenho de progênies não fenotipadas, mesmo em ambientes não amostrados. Este estudo é pioneiro em considerar conjuntamente parentesco híbrido, oriundo de informações de pedigree mais marcadores SNP, e covariáveis ambientais em modelos multiambientais baseados em norma de reação, para predição de valor genético em ambientes-alvo de programas de seleção recorrente.

L’étude des rendements en blé a mis en évidence une stagnation apparue dans les années 1990, notamment en France, et principalement lié aux stress hydrique et thermique. Dans ce contexte, améliorer la tolérance du blé européen à ces stress est de première importance. Cette étude avait pour but d’étudier le déterminisme génétique de la tolérance à ces stress chez le blé. Pour ce faire, trois populations de blé tendre du CIMMYT combinant des caractères d’adaptation à ces stress ont été cultivées en conditions irriguée, sèche et stress thermique irriguée plusieurs années. Des caractères physiologiques et agronomiques ont été mesurés sur un réseau de 15 essais. Une méthodologie de caractérisation environnementale a été développée et a permis l’identification de six scenarii de stress au sein du réseau. Une covariable environnementale représentative de chacun a été extraite. L’utilisation des modèles de régression factorielles a permis la décomposition de l’interaction génotype x environnement ainsi que la mise en évidence d’une sensibilité différentielle au stress dans le germplasm. Une recherche de QTL multi-environnementale a conduit à la détection de régions génomiques contrôlant les caractères physiologiques et agronomiques ainsi que leurs interactions avec l’environnement. De la caractérisation environnementale à la détection de QTL, cette étude a abouti au développement d’un outil pour les sélectionneurs permettant l’évaluation du potentiel des génotypes face à une gamme d’environnement, mais aussi à l’identification de régions génomiques impliquées dans le contrôle de la tolérance aux stress hydrique et thermique chez le blé tendre. Ceci pourrait améliorer la tolérance à ces stress au sein du germplasm européen. / A stagnation of wheat yield was reported in France and other countries worldwide since the 1990’s, which incriminated mainly drought and heat stress. Improving the European wheat tolerance to them is of first importance. This study aimed to investigate the genetic determinism of the tolerance to such stresses. Three CIMMYT bread wheat populations combining complementary heat and drought adaptive habits were grown in Northern Mexico under irrigated, drought and heat-irrigated treatments from 2011 to 2013. The trial network comprised 15 trials and both physiological and agronomic traits were scored. First, an environmental characterization methodology was developed and resulted in the identification of six main environmental scenarios in the network. A representative environmental covariate was extracted from each of them. Then, a factorial regression model leaded to the dissection of the genotype-by-environment interaction and highlighted differential stress sensitivity of the germplasm. Finally, a multi-environmental QTL detection resulted in the discovery of genomic regions involved in the control of both physiological and agronomic traits and the study of their sensitivity to the environment. From the environmental characterization to the QTL detection, this study resulted in the development of a tool for breeders which may enable the evaluation of the potential of any genotypes in front of a range of environment, but also the identification of genomic regions involved in the control of the tolerance to drought and heat stress in bread wheat. This may help in improving the tolerance of the European bread wheat germplasm to drought and heat stress.

Soil carbon is the largest sink and source of the global carbon cycle and is disturbed by several natural, anthropogenic, and environmental factors. The global increase of atmospheric CO2 affects soil carbon cycling through varied biogeochemical processes. The first chapter is a compilation of current information on potential factors triggering soil acidification and weathering mechanisms under elevated CO2 and their consequences on soil inorganic carbon (SIC) pool and quality. Soil water content and precipitation were critical factors influencing elevated CO2 effects on the SIC pool. The second chapter examines a detailed column experiment in which six soils from the state of Mississippi, USA, representing acidic, neutral, and alkaline pH, were exposed to different CO2 enrichments (100%, 10%, and 1%) for 30 days. The leachates’ pH tended to attain an equilibrium state (neutral) with time under CO2 saturation. SIC increased under CO2 saturation, whereas cation exchange capacity (CEC) showed a decreasing pattern in all soils. In the third chapter, an eXplainable artificial intelligence (XAI) was performed to visualize the different forms of soil carbon variability across the Mississippi River Basin area. This model explains key insights and local discrepancies, suggesting a solution to the “Black-Box” issue. The best performing model, stack ensemble, showed improved RMSE (3 to 8%) and spatial variability for soil carbons than other ML models, especially after adding the residuals from regression analyses. Land cover type > soil pH > total nitrogen, > NDVI were identified as the top four crucial factors for predicting SOC when bulk density > precipitation, soil pH > mean annual temperature described SIC. The proposed automatic machine learning (AutoML) model with model agnostic interpolations might be a hallmark to mitigate the C loss under adverse climate change conditions and allow diverse knowledge groups to adopt a new interpretable ML algorithm more confidently. Findings from this study help predict the impact of elevated atmospheric CO2 on soil pH, acidification, and nutrient availability and develop strategies for sustainable land management practices under a changing climate.

<p dir="ltr">Digital soil mapping (DSM) combines field and laboratory data with environmental factors to predict soil properties. The accuracy of these predictions depends on factors such as model selection, data quality and quantity, and landscape characteristics. In our study, we investigated the impact of sample density and the use of various environmental covariates (ECs) including slope, topographic position index, topographic wetness index, multiresolution valley bottom flatness, and multiresolution ridge top flatness, as well as the spatial resolution of these ECs on the predictive accuracy of four predictive models; Cubist (CB), Random Forest (RF), Regression Kriging (RK), and Ordinary Kriging (OK). Our analysis was conducted at three sites in Indiana: the Purdue Agronomy Center for Research and Education (ACRE), Davis Purdue Agriculture Center (DPAC), and Southeast Purdue Agricultural Center (SEPAC). Each site had its unique soil data sampling designs, management practices, and topographic conditions. The primary focus of this study was to predict the spatial distribution of soil properties, including soil organic matter (SOM), cation exchange capacity (CEC), and clay content, at different depths (0-10cm, 0-15cm, and 10-30cm) by utilizing five environmental covariates and four spatial resolutions for the ECs (1-1.5 m, 5 m, 10 m, and 30 m).</p><p dir="ltr">Various evaluation metrics, including R², root mean square error (RMSE), mean square error (MSE), concordance coefficient (pc), and bias, were used to assess prediction accuracy. Notably, the accuracy of predictions was found to be significantly influenced by the site, sample density, model type, soil property, and their interactions. Sites exhibited the largest source of variation, followed by sampling density and model type for predicted SOM, CEC, and clay spatial distribution across the landscape.</p><p dir="ltr">The study revealed that the RF model consistently outperformed other models, while OK performed poorly across all sites and properties as it only relies on interpolating between the points without incorporating the landscape characteristics (ECs) in the algorithm. Increasing sample density improved predictions up to a certain threshold (e.g., 66 samples at ACRE for both SOM and CEC; 58 samples for SOM and 68 samples for CEC at SEPAC), beyond which the improvements were marginal. Additionally, the study highlighted the importance of spatial resolution, with finer resolutions resulting in better prediction accuracy, especially for SOM and clay content. Overall, comparing data from the two depths (0-10cm vs 10-30cm) for soil properties predications, deeper soil layer data (10-30cm) provided more accurate predictions for SOM and clay while shallower depth data (0-10cm) provided more accurate predictions for CEC. Finally, higher spatial resolution of ECs such as 1-1.5 m and 5 m contributed to more accurate soil properties predictions compared to the coarser data of 10 m and 30 m resolutions.</p><p dir="ltr">In summary, this research underscores the significance of informed decisions regarding sample density, model selection, and spatial resolution in digital soil mapping. It emphasizes that the choice of predictive model is critical, with RF consistently delivering superior performance. These findings have important implications for land management and sustainable land use practices, particularly in heterogeneous landscapes and areas with varying management intensities.</p>