Explore the vast range of titles available.

Birds are important indicators for monitoring both biodiversity and habitat health; they also play a crucial role in ecosystem management. Declines in bird populations can result in reduced ecosystem services, including seed dispersal, pollination and pest control. Accurate and long-term monitoring of birds to identify species of concern while measuring the success of conservation interventions is essential for ecologists. However, monitoring is time-consuming, costly and often difficult to manage over long durations and at meaningfully large spatial scales. Technology such as camera traps, acoustic monitors and drones provide methods for non-invasive monitoring. There are two main problems with using camera traps for monitoring: (a) cameras generate many images, making it difficult to process and analyse the data in a timely manner; and (b) the high proportion of false positives hinders the processing and analysis for reporting. In this paper, we outline an approach for overcoming these issues by utilising deep learning for real-time classification of bird species and automated removal of false positives in camera trap data. Images are classified in real-time using a Faster-RCNN architecture. Images are transmitted over 3/4G cameras and processed using Graphical Processing Units (GPUs) to provide conservationists with key detection metrics, thereby removing the requirement for manual observations. Our models achieved an average sensitivity of 88.79%, a specificity of 98.16% and accuracy of 96.71%. This demonstrates the effectiveness of using deep learning for automatic bird monitoring.

The Endangered giant pangolin Smutsia gigantea is rare and elusive across its Central African range. Because of its solitary and nocturnal nature, the species is difficult to study and so its ecology is little known. Pangolins are considered the most trafficked mammals in the world. Therefore, confirming presence accurately and monitoring trends in distribution and abundance are essential to inform and prioritize conservation efforts. Camera traps are popular tools for surveying rare and cryptic species. However, non-targeted camera-trap surveys yield low camera-trapping rates for pangolins. Here we use camera-trap data from surveys conducted within three protected areas in Uganda to test whether targeted placement of cameras improves giant pangolin detection probability in occupancy models. The results indicate that giant pangolin detection probability is highest when camera traps are targeted on burrows. The median number of days from camera deployment to first giant pangolin detection event was 12, with the majority of events captured within 32 days from deployment. The median interval between giant pangolin events at a camera-trap site was 33 days. We demonstrate that camera-trap surveys can be designed to improve the detection of giant pangolins and we outline a set of recommendations to maximize the effectiveness of efforts to survey and monitor the species.

Camera traps offer enormous new opportunities in ecological studies, but current automated image analysis methods often lack the contextual richness needed to support impactful conservation outcomes. Integrating vision–language models into these workflows could address this gap by providing enhanced contextual understanding and enabling advanced queries across temporal and spatial dimensions. Here, we present an integrated approach that combines deep learning-based vision and language models to improve ecological reporting using data from camera traps. We introduce a two-stage system: YOLOv10-X to localise and classify species (mammals and birds) within images and a Phi-3.5-vision-instruct model to read YOLOv10-X bounding box labels to identify species, overcoming its limitation with hard-to-classify objects in images. Additionally, Phi-3.5 detects broader variables, such as vegetation type and time of day, providing rich ecological and environmental context to YOLO’s species detection output. When combined, this output is processed by the model’s natural language system to answer complex queries, and retrieval-augmented generation (RAG) is employed to enrich responses with external information, like species weight and IUCN status (information that cannot be obtained through direct visual analysis). Combined, this information is used to automatically generate structured reports, providing biodiversity stakeholders with deeper insights into, for example, species abundance, distribution, animal behaviour, and habitat selection. Our approach delivers contextually rich narratives that aid in wildlife management decisions. By providing contextually rich insights, our approach not only reduces manual effort but also supports timely decision making in conservation, potentially shifting efforts from reactive to proactive.

The major explosive eruption of Chaitén volcano, Chile, in May 2008 provided a rare opportunity to track the long‐range dispersal and deposition of fine volcanic ash. The eruption followed ∼10,000 years of quiescence, was the largest explosive eruption globally since Hudson, Chile, in 1991, and was the first explosive rhyolitic eruption since Novarupta, Alaska, in 1912. Field examination of distal ashfall indicates that ∼1.6 × 1011 kg of ash (dense rock equivalent volume of ∼0.07 km3) was deposited over ∼2 × 105 km2 of Argentina during the first week of eruption. The minimum eruption magnitude, estimated from the mass of the tephra deposit, is 4.2. Several discrete ashfall units are identifiable from their distribution and grain size characteristics, with more energetic phases showing a bimodal size distribution and evidence of cloud aggregation processes. Ash chemistry was uniform throughout the early stages of eruption and is consistent with magma storage prior to eruption at depths of 3–6 km. Deposition of ash over a continental region allowed the tracking of eruption development and demonstrates the potential complexity of tephra dispersal from a single eruption, which in this case comprised several phases over a week‐long period of intense activity.

Pangolins are some of the most overexploited but least studied mammals. The giant pangolin Smutsia gigantea is the largest of the eight pangolin species, measuring up to 180 cm in length and weighing up to 40 kg. It is a nocturnal, solitary species that occurs at low densities and little is known regarding its biology and ecology. It is distributed widely across the rainforests and forest savannah mosaics of equatorial Africa but its exact range extent is unknown. Apart from a single record in Kenya predating 1971, the eastern limit of its range was thought only to extend to central Uganda and western Tanzania. Here we present three spatially separate records confirming the presence of this species in Kenyan Afromontane forests. The three records are c. 120 km apart and c. 500 km east of the nearest confirmed giant pangolin population in Uganda. These records represent a significant range extension for the species and highlight the biodiversity and conservation importance of the Afromontane forests of western Kenya.

This thesis employs crystal-specific techniques, combined with field observations, petrology, geochemistry and numerical modelling to reconstruct the magmatic system associated with the ~ 340 ka Whakamaru supereruption, New Zealand. Comparisons are drawn with the ~ 74 ka Youngest Toba Tuff (YTT) supereruption. Whakamaru Group Ignimbrites contain five pumice types, characterised by different mineralogies and crystal contents. Pumice petrography and geochemistry indicate that basaltic magma mixing occurred, possibly triggering eruption. Geothermobarometers suggest an eruption temperature of ~ 770°C and magma storage at ~ 5 km depth. High-resolution thermal records from Ti-in-quartz analysis indicate a thermal pulse of ~ 100°C prior to eruption. Diffusion timescales show multiple recharge events with the most significant event occurring ~ 35 y prior to eruption. Zircon U-Pb data show that most crystallisation occurred at ~ 400 ka, with antecrysts and xenocrysts incorporated. Zircon trace-element data suggest multiple recharge events and complex mixing over ~ 100 ky, consistent with an incrementally growing reservoir. Oxygen-isotope data illustrate that zircon, quartz and feldspar crystallised together in equilibrium, with isotopically homogenous magma sources feeding the reservoir over time. Whakamaru and YTT tephra thickness and grain-size data were used in ash dispersal modelling. Results indicate the YTT eruption had a ~ 35 km column height and erupted volumes of 1500 – 1900 km³, with deposition from a co-ignimbrite phase; whereas Whakamaru had a Plinian column ~ 45 km high with SE dispersal and a minimum volume of ~ 400 km³. The widespread dispersal of large volumes of fine ash from both eruptions would have had global environmental consequences. The data are integrated to reconstruct a new Whakamaru magma reservoir model. The complex crystal records indicate the system was characterised by long periods of incremental assembly, mixing, recycling of material, and reactivation during multiple recharge episodes which perturbed the system and primed the magma for eruption.

Camera traps offer enormous new opportunities in ecological studies, but current automated image analysis methods often lack the contextual richness needed to support impactful conservation outcomes. Here we present an integrated approach that combines deep learning-based vision and language models to improve ecological reporting using data from camera traps. We introduce a two-stage system: YOLOv10-X to localise and classify species (mammals and birds) within images, and a Phi-3.5-vision-instruct model to read YOLOv10-X binding box labels to identify species, overcoming its limitation with hard to classify objects in images. Additionally, Phi-3.5 detects broader variables, such as vegetation type, and time of day, providing rich ecological and environmental context to YOLO's species detection output. When combined, this output is processed by the model's natural language system to answer complex queries, and retrieval-augmented generation (RAG) is employed to enrich responses with external information, like species weight and IUCN status (information that cannot be obtained through direct visual analysis). This information is used to automatically generate structured reports, providing biodiversity stakeholders with deeper insights into, for example, species abundance, distribution, animal behaviour, and habitat selection. Our approach delivers contextually rich narratives that aid in wildlife management decisions. By providing contextually rich insights, our approach not only reduces manual effort but also supports timely decision-making in conservation, potentially shifting efforts from reactive to proactive management.

Birds are important indicators for monitoring both biodiversity and habitat health; they also play a crucial role in ecosystem management. Decline in bird populations can result in reduced eco-system services, including seed dispersal, pollination and pest control. Accurate and long-term monitoring of birds to identify species of concern while measuring the success of conservation interventions is essential for ecologists. However, monitoring is time consuming, costly and often difficult to manage over long durations and at meaningfully large spatial scales. Technology such as camera traps, acoustic monitors and drones provide methods for non-invasive monitoring. There are two main problems with using camera traps for monitoring: a) cameras generate many images, making it difficult to process and analyse the data in a timely manner; and b) the high proportion of false positives hinders the processing and analysis for reporting. In this paper, we outline an approach for overcoming these issues by utilising deep learning for real-time classi-fication of bird species and automated removal of false positives in camera trap data. Images are classified in real-time using a Faster-RCNN architecture. Images are transmitted over 3/4G cam-eras and processed using Graphical Processing Units (GPUs) to provide conservationists with key detection metrics therefore removing the requirement for manual observations. Our models achieved an average sensitivity of 88.79%, a specificity of 98.16% and accuracy of 96.71%. This demonstrates the effectiveness of using deep learning for automatic bird monitoring.

[Donald McCain] should also hit the target of a is in with Alegralil, who defends an unbeaten record in the Durham City Of Culture 2013 Mares' Novices' Hurdle (2.20). The four-year-old hosed up in bumpers at Market Rasen and Bangor in the spring and jumped with aplomb on her hurdling introduction at Towcester to beat a decent sort in Miss Overdrive by 13 lengths. She is clearly an above average mare and she could well run up a sequence in ordinary northern events such as this. Tomorrow's tips LINGFIELD: GAZETTE BET: 3.10 Spear Thistle. NAOMI MATTHEW: 12.40 Sebastiano, 1.10 Call At Midnight, 1.40 Kanad, 2.10 Hellfire Club, 2.40 Sound Stage, 3.10 Rileyev, 3.40 Mr Tallyman. SOUTHWELL: GAZETTE BET: 1.30 Budva. NAOMI MATTHEW: 12.00 Lord Of The Dance, 12.30 Castle Myth, 1.00 Taper Jean Girl, 1.30 Rock Of Eire, 2.00 Kinigi, 2.30 Inflammable (nap), 3.00 Shadows Lengthen (nb), 3.30 Great Charm.