No abstract

No abstract

Soil nitrogen isotope composition (δ15N) is an essential tool for investigating ecosystem nitrogen balances, plant–microbe interactions, ecological niches, animal migration, food origins, and forensics. The advancement of these applications is limited by a lack of robust geospatial models that are capable of capturing variation in soil δ15N (i.e., isotopic landscapes or isoscapes). Due to the complexity of the nitrogen cycle and general scarcity of isotopic information, previous approaches have reconstructed regional to global soil δ15N patterns via highly uncertain linear regression models. Here, we develop a new machine learning approach to ascertain a finer‐scale understanding of geographic differences in soil δ15N, using the South American continent as a test case. We use a robust training set spanning 278 geographic locations across the continent, spanning all major biomes. We tested three different machine learning methods: cubist, random forest (RF), and stochastic gradient boosting (GBM). 10‐fold cross‐validation revealed that the RF method outperformed both the cubist and GBM approaches. Variable importance analysis of the RF framework pointed to biome type as the most crucial auxiliary variable, followed by soil organic carbon content, in determining the model performance. We thereby created a biogeographic boundary map, which predicted an expected multiscale spatial pattern of soil δ15N with a high degree of confidence, performing substantially better than all previous approaches for the continent of South America. Therefore, the RF machine learning framework showed to be a great opportunity to explore a broad array of ecological, biogeochemical, and forensic issues through the lens of soil δ15N.

Spatial patterns of stable isotope ratios can be represented in spatial models called isoscapes, and have been widely used to track biogeochemical processes in natural and anthropic systems. Isoscapes have the potential to improve isotope dissemination and interpretation of spatial patterns, increase scientific results appropriation by non-specialists and improve natural resource management. However, the isoscape approach has not commonly been used in studies performed in the Brazilian context. Isoscapes with oxygen, hydrogen, nitrogen and carbon stable isotopes contribute in areas such as animal migration, forensics, hydrology, and studies on population, community and ecosystem level, among others. Here, we show the well-known global use and applications of isoscapes in different studies worldwide as a background to point out the potential for more Brazilian researchers to employ this approach in their studies, taking advantage of existing methods and filling spatial and methodological gaps. The incorporation of isoscapes may broaden the understanding of mechanisms and processes of major biogeochemical cycles in Brazil, assist in solving crimes, track illicit drug origins, help to detect wild animal trafficking, and increase Brazilian knowledge about the hydrological cycle and animal migration patterns in the Neotropics.

High δ 13 C in human tissues in Brazil indicate high consumption of C 4 -based sources due to the consumption of highly processed food and animal protein. The significant positive correlation between the human developed index (HDI) developed by the United Nations Development Program, and fingernail δ 13 C at the county level proved to be useful as a new proxy in tracking human nutrition. Regions with higher HDI are those with higher consumption of highly processed food.

No abstract

A classificação supervisionada por assinaturas geomorfométricas é um procedimento que pode auxiliar no mapeamento de formas de terreno a partir da utilização de medidas de similaridade ou distância. Este trabalho tem como objetivo comparar os métodos supervisionados de classificação a partir de medidas de similaridade e distância para o mapeamento do relevo. A comparação foi realizada no Campo de Instrução Militar de Formosa (GO), seguindo os seguintes passos: aquisição de dados HydroSHEDS, geração de imagens de curvatura, seleção de assinaturas geomorfométricas; classificação formas de terreno usando o método de classificação por ângulo espectral (SAM) e a distância euclidiana (DE); comparação da classificação através da matriz de tabulação cruzada, análise de modelo de elevação em 3D, avaliação da média e do desvio padrão das curvaturas para cada classe mapeada e observação em campo. A seleção de assinaturas geomorfométrica considerou as seguintes etapas: (a) redução dos atributos geomorfométricos pela transformação de Fração Mínima de Ruído (MNF); (b) redução espacial pelo Índice de Pureza de Pixel (PPI); e (c) a seleção manual pelo n-Dimensional Visualizador. O processo de classificação adotou 14 AGs que descrevem dois comportamentos: Tipo 1 quando a curvatura longitudinal tem um valor maior que a curvatura transversal; e Tipo 2 quando ocorre o inverso. O processo foi simplificado para seis classes de formas de terreno (FT): Convexo/Convexo (Cx/Cx); Côncavo/Convexo (Cc/Cx); Côncavo/Côncavo (Cc/Cc); Côncavo/Retilíneo (Cc/Rt); Convexo/Retilíneo (Cx/Rt); Retilíneo/Retilíneo (Rt/Rt). No mapeamento SAM, as formas de relevo predominantes são Cc/Rt, Cx/Rt e Cc/Cx, indicando uma heterogeneidade com muitas áreas de transição e côncavas. A classificação a partir da DE mostrou prevalência de feições retilíneas (Rt/Rt). Apesar disso, essa FT apresentou os menores desvios e valores médios próximos de zero para todas as curvaturas, indicando que a DE foi o método mais eficiente para o mapeamento das áreas retilíneas.

No abstract