Amazon

Background: The prediction of hollows in standing trees is an expensive operation, but it is essential for decision-making about harvesting in managed forests in the Amazon. The hollow test that is currently used has strong limitations for correct prediction of the presence of hollows in a tree of commercial interest. The objective of this research was to select and validate generalized linear logistic models to estimate the occurrence of hollows in trees of fifteen commercial species and to compare the efficiency of the models to the results from the traditional manual method of hollow testing in the state of Pará, Brazil. A database of 27,380 trees was used to adjust models by species. To validate the equations, 9,915 trees from an independent area were used. Results: Diameter at breast height (DBH), commercial height (hc) and stem quality (SQ) were important predictors of the occurrence of tree hollows, while wood density (WD) did not generate significant gains in the models. Species are determinants of the probability of a tree being hollow. From a DBH of approximately 100 cm, the probability of occurrence of hollows in the trees reaches about 80% for Manilkara bidentata (A. DC.) A. Chev., and for and Mezilaurus itauba (Meisn.) Taub. ex Mez and Astronium lecointei Ducke, for example, hollows occur in diameters of about 120 cm. Logistic equations are more efficient in predicting the presence of a hollow when a tree contains one, compared to the hollow test. Conclusion: It is possible to accurately predict the occurrence of hollows in commercial trees, which may be an alternative to the current hollow test used in managed areas in the Brazilian Amazon. Keywords: hollow trees; commercial trees; generalized linear models; logistic regression; tropical forests

Beneath oceanic spreading centres, the lithosphere–asthenosphere boundary (LAB) acts as a permeability barrier that focuses the delivery of melt from deep within the mantle towards the spreading axis¹. At intermediate-spreading to fast-spreading ridge crests, the multichannel seismic reflection technique has imaged a nearly flat, 1–2-km-wide axial magma lens (AML)² that defines the uppermost section of the LAB³, but the nature of the LAB deeper into the crust has been more elusive, with some clues gained from tomographic images, providing only a diffuse view of a wider halo of lower-velocity material seated just beneath the AML⁴. Here we present 3D seismic reflection images of the LAB extending deep (5–6 km) into the crust beneath Axial volcano, located at the intersection of the Juan de Fuca Ridge and the Cobb–Eickelberg hotspot. The 3D shape of the LAB, which is coincident with a thermally controlled magma assimilation front, focuses hotspot-related and mid-ocean-spreading-centre-related magmatism towards the centre of the volcano, controlling both eruption and hydrothermal processes and the chemical composition of erupted lavas⁵. In this context, the LAB can be viewed as the upper surface of a ‘magma domain’, a volume within which melt bodies reside (replacing the concept of a single ‘magma reservoir’)⁶. Our discovery of a funnel-shaped, crustal LAB suggests that thermally controlled magma assimilation could be occurring along this surface at other volcanic systems, such as Iceland.

Oil palm (Elaeis guineensis Jacq.) is a crop of high relevance in the global economy. In the eastern Amazon, a region with potential for the expansion of this crop, the initiative to plant oil palm together with other regional species of commercial interest, forming agroforestry systems (AFS), is considered ecologically, economically, and socially promising. We evaluated the floristic and structural dynamics of oil palm AFS in the eastern Amazon, Brazil. We analyzed the Shannon–Wiener diversity index (H'), the Importance Value Index (IVI), the Current Annual Increment (CAI), and mortality in six AFS, between 2016 and 2018. We established two types of AFS at each Site, namely, AFS-A—considered less diverse; and AFS-B—considered more diverse. Fabaceae, Arecaceae, Meliaceae, Anacardiaceae, and Malvaceae were the main botanical families occurring in the AFS. The floristic composition showed the greatest differences between AFS A and B from Site 2, whereas the most similar compositions were observed among the AFS from Site 3. The H' index ranged from 0.88 to 2.08. E. guineensis, Theobroma cacao L., and Gliricidia sepium (Jacq.) Kunth ex Walp showed the highest IVIs. In total, 38.46% of the species were wood trees, 38.46% were fruit trees, 7.69% were species planted for green manure purposes, and 15.38% were multiple-use species. Mortality in all AFS was low (< 6%). The largest diameter increments occurred in the 2016–2017 interval. Between 2017 and 2018, E. guineensis, Euterpe oleracea Mart., and T. cacao. showed low diameter increments. Between 2016 and 2018 tree growth slowed down in AFS with lower diversity, but not in AFS with higher diversity, suggesting greater resilience of AFS with higher diversity.

Clients rely on database systems to be correct, which requires the system not only to implement transactions’ semantics correctly but also to provide isolation guarantees for the transactions. This paper presents a client-centric technique for checking both semantic correctness and isolation-level guarantees for black-box database systems based on observations collected from running transactions on these systems. Our technique verifies observational correctness with respect to a given set of transactions and observations for them, which holds iff there exists a possible correct execution of the transactions under a given isolation level that could result in these observations. Our technique relies on novel symbolic encodings of (1) the semantic correctness of database transactions in the presence of weak isolation and (2) isolation-level guarantees. These are used by the checker to query a Satisfiability Modulo Theories solver. We applied our tool Troubadour to verify observational correctness of several database systems, including PostgreSQL and an industrial system under development, in which the tool helped detect two new bugs. We also demonstrate that Troubadour is able to find known semantic correctness bugs and detect isolation-related anomalies.

Strongly-consistent replicated data stores are a popular foundation for many kinds of online services, but their implementations are very complex. Strong replication is not available under network partitions, and so achieving a functional degree of fault-tolerance requires correctly implementing consensus algorithms like Raft and Paxos. These algorithms are notoriously difficult to reason about, and many data stores implement custom variations to support unique performance tradeoffs, presenting an opportunity for automated verification tools. Unfortunately, existing tools that have been applied to distributed consensus demand too much developer effort, a problem stemming from the low-level programming model in which consensus and strong replication are implemented—asynchronous message passing—which thwarts decidable automation by exposing the details of asynchronous communication. In this paper, we consider the implementation and automated verification of strong replication systems as applications of weak replicated data stores. Weak stores, being available under partition, are a suitable foundation for performant distributed applications. Crucially, they abstract asynchronous communication and allow us to derive local-scope conditions for the verification of consensus safety. To evaluate this approach, we have developed a verified-programming framework for the weak replicated state model, called Super-V. This framework enables SMT-based verification based on local-scope artifacts called stable update preconditions , replacing standard-practice global inductive invariants. We have used our approach to implement and verify a strong replication system based on an adaptation of the Raft consensus algorithm.

Model interpretability has emerged as a critical factor in the successful implementation of machine learning and time series forecasting systems within demand planning. As organizations increasingly adopt sophisticated forecasting models, the need to balance prediction accuracy with explainability becomes paramount. The tension between model complexity and transparency presents significant challenges for stakeholders who must understand and trust these systems. While advanced neural networks and ensemble methods offer improved forecasting capabilities, their black-box nature often hinders effective decision-making. This document explores the multifaceted aspects of model interpretability, from fundamental challenges to strategic advantages, and presents a comprehensive framework for building and implementing interpretable forecasting systems. By focusing on stakeholder communication, continuous improvement mechanisms, and practical implementation strategies, organizations can develop forecasting solutions that combine technical excellence with business utility

This paper studies the problem of action noise in model-based offline reinforcement learning, i.e., the actions recorded in the transition trajectories are polluted with noise. Though this is a relatively new problem in the literature, it has become a relevant issue since offline reinforcement learning has become more and more widely used. This is particularly important in applications where an offline dataset collected without the intention to improve the policy is repurposed for reinforcement learning. This paper presents an error analysis for the value function due to the action noise and provides numerical studies. This work also prepares for further developments of novel algorithms for addressing action noise.

Illumination estimation from a single indoor image is a promising yet challenging task. Existing indoor illumination estimation methods mainly regress lighting parameters or infer a panorama from a limited field-of-view image. Nevertheless, these methods fail to recover a panorama with both well-distributed illumination and detailed environment textures, leading to a lack of realism in rendering the embedded 3D objects with complex materials. This paper presents a novel multi-stage illumination estimation framework named IllumiDiff. Specifically, in Stage I, we first estimate illumination conditions from the input image, including the illumination distribution as well as the environmental texture of the scene. In Stage II, guided by the estimated illumination conditions, we design a conditional panoramic texture diffusion model to generate a high-quality LDR panorama. In Stage III, we leverage the illumination conditions to further reconstruct the LDR panorama to an HDR panorama. Extensive experiments demonstrate that our IllumiDiff can generate an HDR panorama with realistic illumination distribution and rich texture details from a single limited field-of-view indoor image. The generated panorama can produce impressive rendering results for the embedded 3D objects with various materials.

Model interpretability has emerged as a critical factor in the successful implementation of machine learning and time series forecasting systems within demand planning. As organizations increasingly adopt sophisticated forecasting models, the need to balance prediction accuracy with explainability becomes paramount. The tension between model complexity and transparency presents significant challenges for stakeholders who must understand and trust these systems. While advanced neural networks and ensemble methods offer improved forecasting capabilities, their black-box nature often hinders effective decision-making. This document explores the multifaceted aspects of model interpretability, from fundamental challenges to strategic advantages, and presents a comprehensive framework for building and implementing interpretable forecasting systems. By focusing on stakeholder communication, continuous improvement mechanisms, and practical implementation strategies, organizations can develop forecasting solutions that combine technical excellence with business utility

AI-augmented database migration represents a paradigm shift in how organizations approach the challenging process of moving data from legacy systems to modern platforms. This article explores how artificial intelligence technologies are transforming traditional migration workflows through automated schema analysis, intelligent data mapping, predictive testing, and autonomous execution. By examining case studies from financial services, healthcare, and retail sectors, the article demonstrates how AI tools significantly reduce conversion time, improve mapping accuracy, decrease testing cycles, and shorten overall migration timelines. The technical implementation considerations highlight the importance of appropriate model selection, seamless integration with existing workflows, and platform-specific optimizations. While challenges remain in areas such as training data requirements, explainability, edge case handling, and security, the future of AI-augmented database migrations promises fully autonomous operations, cross-platform optimizations, continuous synchronization models, and self-learning systems that could transform what has historically been viewed as a necessary technical burden into a strategic advantage for digital transformation.

Despite the impressive performance obtained by recent single-image hand modeling techniques, they lack the capability to capture sufficient details of the 3D hand mesh. This deficiency greatly limits their applications when high-fidelity hand modeling is required, e.g. , personalized hand modeling. To address this problem, we design a frequency split network to generate 3D hand meshes using different frequency bands in a coarse-to-fine manner. To capture high-frequency personalized details, we transform the 3D mesh into the frequency domain, and proposed a novel frequency decomposition loss to supervise each frequency component. By leveraging such a coarse-to-fine scheme, hand details that correspond to the higher frequency domain can be preserved. In addition, the proposed network is scalable, and can stop the inference at any resolution level to accommodate different hardware with varying computational powers. To feed the scalable frequency network with frequency split image features, we proposed an image-graph ring feature mapping strategy. To train our network with per-vertex supervision, we use a bidirectional registration strategy to generate a topology-fixed ground-truth. To quantitatively evaluate the performance of our method in terms of recovering personalized shape details, we introduce a new evaluation metric named Mean-frequency Signal-to-Noise Ratio (MSNR) to measure the mean signal-to-noise ratio of mesh signal on each frequency component. Extensive experiments demonstrate that our approach generates fine-grained details for high-fidelity 3D hand reconstruction, and our evaluation metric is more effective than traditional metrics for measuring mesh details.

Sugarcane is a vital agricultural crop in Brazil, playing a crucial role in both the national economy and bioenergy production. To enhance productivity and operational efficiency across the sugarcane production chain, the adoption of advanced technologies, such as machine learning (ML) algorithms, has become increasingly essential. This study aims to assess the current state of scientific research on the application of ML algorithms in Brazilian sugarcane cultivation through a comprehensive review of national and international publications. The findings reveal a predominance of supervised learning algorithms, particularly Random Forest, Support Vector Machine, Decision Tree, and Artificial Neural Networks. In contrast, unsupervised learning methods have been applied less frequently, while reinforcement learning techniques are notably absent from the reviewed studies. The analysis also highlights significant regional disparities, with a strong concentration of research in the Southeast region, Brazil’s primary sugarcane-producing area. Meanwhile, the North region—despite its considerable potential for expansion—exhibits limited research activity. This review underscores the need to explore underutilized ML approaches, such as semi-supervised and reinforcement learning, to better address data scarcity and the complexities of dynamic agricultural management. Additionally, the incorporation of Explainable Artificial Intelligence techniques, such as SHAP and LIME, is recommended to enhance model transparency and interpretability for decision-makers. These findings highlight the importance of expanding both methodological diversity and geographical coverage in Brazilian sugarcane research while encouraging the integration of emerging ML techniques to foster sustainable agricultural practices and improve crop productivity.