Search

The global increase in observed forest dieback, characterised by the death of tree foliage, heralds widespread decline in forest ecosystems. This degradation causes significant changes to ecosystem services and functions, including habitat provision and carbon sequestration, which can be difficult to detect using traditional monitoring techniques, highlighting the need for large-scale and high-frequency monitoring. Contemporary developments in the instruments and methods to gather and process data at large-scales mean this monitoring is now possible. In particular, the advancement of low-cost drone technology and deep learning on consumer-level hardware provide new opportunities. Here, we use an approach based on deep learning and vegetation indices to assess crown dieback from RGB aerial data without the need for expensive instrumentation such as LiDAR. We use an iterative approach to match crown footprints predicted by deep learning with field-based inventory data from a Mediterranean ecosystem exhibiting drought-induced dieback, and compare expert field-based crown dieback estimation with vegetation index-based estimates. We obtain high overall segmentation accuracy (mAP: 0.519) without the need for additional technical development of the underlying Mask R-CNN model, underscoring the potential of these approaches for non-expert use and proving their applicability to real-world conservation. We also find colour-coordinate based estimates of dieback correlate well with expert field-based estimation. Substituting ground truth for Mask R-CNN model predictions showed negligible impact on dieback estimates, indicating robustness. Our findings demonstrate the potential of automated data collection and processing, including the application of deep learning, to improve the coverage, speed and cost of forest dieback monitoring., Comment: 16 pages, 5 figures

Proximally-sensed laser scanning offers significant potential for automated forest data capture, but challenges remain in automatically identifying tree species without additional ground data. Deep learning (DL) shows promise for automation, yet progress is slowed by the lack of large, diverse, openly available labeled datasets of single tree point clouds. This has impacted the robustness of DL models and the ability to establish best practices for species classification. To overcome these challenges, the FOR-species20K benchmark dataset was created, comprising over 20,000 tree point clouds from 33 species, captured using terrestrial (TLS), mobile (MLS), and drone laser scanning (ULS) across various European forests, with some data from other regions. This dataset enables the benchmarking of DL models for tree species classification, including both point cloud-based (PointNet++, MinkNet, MLP-Mixer, DGCNNs) and multi-view image-based methods (SimpleView, DetailView, YOLOv5). 2D image-based models generally performed better (average OA = 0.77) than 3D point cloud-based models (average OA = 0.72), with consistent results across different scanning platforms and sensors. The top model, DetailView, was particularly robust, handling data imbalances well and generalizing effectively across tree sizes. The FOR-species20K dataset, available at https://zenodo.org/records/13255198, is a key resource for developing and benchmarking DL models for tree species classification using laser scanning data, providing a foundation for future advancements in the field.

Population genetics lies at the heart of evolutionary theory. This topic forms part of many biological science curricula but is rarely taught to physics students. Since physicists are becoming increasingly interested in biological evolution, we aim to provide a brief introduction to population genetics, written for physicists. We start with two background chapters: chapter 1 provides a brief historical introduction to the topic, while chapter 2 provides some essential biological background. We begin our main content with chapter 3 which discusses the key concepts behind Darwinian natural selection and Mendelian inheritance. Chapter 4 covers the basics of how variation is maintained in populations, while chapter 5 discusses mutation and selection. In chapter 6 we discuss stochastic effects in population genetics using the Wright-Fisher model as our example, and finally we offer concluding thoughts and references to excellent textbooks in chapter 7.

Quantitative interpretation of transmission electron microscopy (TEM) data of crystalline specimens often requires the accurate knowledge of the local crystal orientation. A method is presented which exploits momentum-resolved scanning TEM (STEM) data to determine the local mistilt from a major zone axis. It is based on a geometric analysis of Kikuchi bands within a single diffraction pattern, yielding the centre of the Laue circle. Whereas the approach is not limited to convergent illumination, it is here developed using unit-cell averaged diffraction patterns corresponding to high-resolution STEM settings. In simulation studies, an accuracy of approximately 0.1mrad is found. The method is implemented in automated software and applied to crystallographic tilt and in-plane rotation mapping in two experimental cases. In particular, orientation maps of high-Mn steel and an epitaxially grown La$_{\text{0.7}}$Sr$_{\text{0.3}}$MnO$_{\text{3}}$-SrTiO$_{\text{3}}$ interface are presented. The results confirm the estimates of the simulation study and indicate that tilt mapping can be performed consistently over a wide field of view with diameters well above 100nm at unit cell real space sampling.

Satellite-based remote sensing has revolutionised the way we address global challenges. Huge quantities of Earth Observation (EO) data are generated by satellite sensors daily, but processing these large datasets for use in ML pipelines is technically and computationally challenging. While some preprocessed Earth observation datasets exist, their content is often limited to optical or near-optical wavelength data, which is ineffective at night or in adverse weather conditions. Synthetic Aperture Radar (SAR), an active sensing technique based on microwave length radiation, offers a viable alternative. However, the application of machine learning to SAR has been limited due to a lack of ML-ready data and pipelines, particularly for the full diversity of SAR data, including polarimetry, coherence and interferometry. In this work, we introduce M3LEO, a multi-modal, multi-label Earth observation dataset that includes polarimetric, interferometric, and coherence SAR data derived from Sentinel-1, alongside multispectral Sentinel-2 imagery and auxiliary data describing terrain properties such as land use. M3LEO spans approximately 17M 4x4 km data chips from six diverse geographic regions. The dataset is complemented by a flexible PyTorch Lightning framework configured using Hydra to accommodate its use across diverse ML applications in Earth observation. We provide tools to process any dataset available on popular platforms such as Google Earth Engine for seamless integration with our framework. We show that the distribution shift in self-supervised embeddings is substantial across geographic regions, even when controlling for terrain properties. Data: huggingface.co/M3LEO, Code: github.com/spaceml-org/M3LEO., Comment: 10 pages, 5 figures

To explore the role environment plays in influencing galaxy evolution at high redshifts, we study $2.0\leq z<4.2$ environments using the FourStar Galaxy Evolution (ZFOURGE) survey. Using galaxies from the COSMOS legacy field with ${\rm log(M_{*}/M_{\odot})}\geq9.5$, we use a seventh nearest neighbour density estimator to quantify galaxy environment, dividing this into bins of low, intermediate and high density. We discover new high density environment candidates across $2.0\leq z<2.4$ and $3.1\leq z<4.2$. We analyse the quiescent fraction, stellar mass and specific star formation rate (sSFR) of our galaxies to understand how these vary with redshift and environment. Our results reveal that, across $2.0\leq z<2.4$, the high density environments are the most significant regions, which consist of elevated quiescent fractions, ${\rm log(M_{*}/M_{\odot})}\geq10.2$ massive galaxies and suppressed star formation activity. At $3.1\leq z<4.2$, we find that high density regions consist of elevated stellar masses but require more complete samples of quiescent and sSFR data to study the effects of environment in more detail at these higher redshifts. Overall, our results suggest that well-evolved, passive galaxies are already in place in high density environments at $z\sim2.4$, and that the Butcher-Oemler effect and SFR-density relation may not reverse towards higher redshifts as previously thought., Comment: 11 pages, 5 figures, final version published in PASA

We compare the Frequency-Resolved Frozen Phonon Multislice (FRFPMS) method, introduced in Phys. Rev. Lett. 124, 025501 (2020), with other theoretical approaches used to account for the inelastic scattering of high energy electrons, namely the first-order Born approximation and the quantum excitation of phonons model. We show, that these theories lead to similar expressions for the single inelastically scattered intensity as a function of momentum transfer for an anisotropic quantum harmonic oscillator in a weak phase object approximation of the scattered waves, except for a too small smearing of the scattering potential by the effective Debye-Waller factor (DWF) in the FRFPMS method. We propose that this issue can be fixed by including an explicit DWF smearing into the potential and demonstrate numerically, that in any realistic situation, a FRFPMS approach revised in this way, correctly accounts for the single inelastically scattered intensity and the correct elastic scattering intensity. Furthermore our simulations illustrate that the only requirement for such a revised FRFPMS method is the smallness of mean squared displacements for all atomic species in all frequency bins. The analytical considerations for the FRFPMS method also explain the $1/\omega^2$-scaling of FRFPMS spectra observed in Phys. Rev. B 104, 104301 (2021) by the use of classical statistics in the molecular dynamics simulation. Moreover, we find that the FRFPMS method inherently adds the contributions of phonon loss and gain within each frequency bin. Both of these issues related to the frequency-scaling can be fixed by a system-independent post-processing step.

The isotopic mass of constituent elements of materials has a well-known effect on the energy of vibrational modes. By means of monochromated scanning transmission electron microscopy we have experimentally studied the phonon bandstructure of hexagonal BN, where a phonon band gap appears between in-plane optical phonon modes and the lower energy part of the phonon spectrum. The size of the phonon band gap can be manipulated by the isotopic mass of the boron. While in $^{11}$BN the phonon band gap is about 7 meV wide, in $^{10}$BN the gap nearly closes, being an order of magnitude smaller (below 0.5 meV). This opens exciting options for manipulating terahertz wave propagation through isotopically structured devices having otherwise no interfaces between chemically distinct components., Comment: 5 pages, 3 figures

Forecasting disease spread is a critical tool to help public health officials design and plan public health interventions.However, the expected future state of an epidemic is not necessarily well defined as disease spread is inherently stochastic, contact patterns within a population are heterogeneous, and behaviors change. In this work, we use time-dependent probability generating functions (PGFs) to capture these characteristics by modeling a stochastic branching process of the spread of a disease over a network of contacts in which public health interventions are introduced over time. To achieve this, we define a general transmissibility equation to account for varying transmission rates (e.g. masking), recovery rates (e.g. treatment), contact patterns (e.g. social distancing) and percentage of the population immunized (e.g. vaccination). The resulting framework allows for a temporal and probabilistic analysis of an intervention's impact on disease spread, which match continuous-time stochastic simulations that are much more computationally expensive.To aid policy making, we then define several metrics over which temporal and probabilistic intervention forecasts can be compared: Looking at the expected number of cases and the worst-case scenario over time, as well as the probability of reaching a critical level of cases and of not seeing any improvement following an intervention.Given that epidemics do not always follow their average expected trajectories and that the underlying dynamics can change over time, our work paves the way for more detailed short-term forecasts of disease spread and more informed comparison of intervention strategies., Comment: 20 pages, 5 figures

We study a one-dimensional sluggish random walk with space-dependent transition probabilities between nearest-neighbour lattice sites. Motivated by trap models of slow dynamics, we consider a model in which the trap depth increases logarithmically with distance from the origin. This leads to a random walk which has symmetric transition probabilities that decrease with distance $|k|$ from the origin as $1/|k|$ for large $|k|$. We show that the typical position after time $t$ scales as $t^{1/3}$ with a nontrivial scaling function for the position distribution which has a trough (a cusp singularity) at the origin. Therefore an effective central bias away from the origin emerges even though the transition probabilities are symmetric. We also compute the survival probability of the walker in the presence of a sink at the origin and show that it decays as $t^{-1/3}$ at late times. Furthermore we compute the distribution of the maximum position, $M(t)$, to the right of the origin up to time $t$, and show that it has a nontrivial scaling function. Finally we provide a generalisation of this model where the transition probabilities decay as $1/|k|^\alpha$ with $\alpha >0$., Comment: 17 pages, revised version accepted for J. Stat. Mech

Gravitational lenses can magnify distant galaxies, allowing us to discover and characterize the stellar populations of intrinsically faint, quiescent galaxies that are otherwise extremely difficult to directly observe at high redshift from ground-based telescopes. Here, we present the spectral analysis of two lensed, quiescent galaxies at $z\gtrsim 1$ discovered by the ASTRO 3D Galaxy Evolution with Lenses survey: AGEL1323 ($M_*\sim 10^{11.1}M_{\odot}$, $z=1.016$, $\mu \sim 14.6$) and AGEL0014 ($M_*\sim 10^{11.5}M_{\odot}$, $z=1.374$, $\mu \sim 4.3$). We measured the age, [Fe/H], and [Mg/Fe] of the two lensed galaxies using deep, rest-frame-optical spectra (S/N $\gtrsim 40$~$\mathring {\mathrm A}$$^{-1}$) obtained on the Keck~I telescope. The ages of AGEL1323 and AGEL0014 are $5.6^{+0.8}_{-0.8}$~Gyr and $3.1^{+0.8}_{-0.3}$~Gyr, respectively, indicating that most of the stars in the galaxies were formed less than 2~Gyr after the Big Bang. Compared to nearby quiescent galaxies of similar masses, the lensed galaxies have lower [Fe/H] and [Mg/H]. Surprisingly, the two galaxies have comparable [Mg/Fe] to similar-mass galaxies at lower redshifts, despite their old ages. Using a simple analytic chemical evolution model connecting the instantaneously recycled element Mg with the mass-loading factors of outflows averaged over the entire star formation history, we found that the lensed galaxies may have experienced enhanced outflows during their star formation compared to lower-redshift galaxies, which may explain why they quenched early., Comment: 18 pages, 11 figures. Accepted for publication in ApJ. The broadband SED fitting and stellar mass measurements updated. A new Section 5.4 is added to discuss the differences between the measured parameters from the SED and full-spectrum fitting