Understanding External Influences on Target Detection and Classification Using Camera Trap Images and Machine Learning.

Using machine learning (ML) to automate camera trap (CT) image processing is advantageous for time-sensitive applications. However, little is currently known about the factors influencing such processing. Here, we evaluate the influence of occlusion, distance, vegetation type, size class, height, subject orientation towards the CT, species, time-of-day, colour, and analyst performance on wildlife/human detection and classification in CT images from western Tanzania. Additionally, we compared the detection and classification performance of analyst and ML approaches. We obtained wildlife data through pre-existing CT images and human data using voluntary participants for CT experiments. We evaluated the analyst and ML approaches at the detection and classification level. Factors such as distance and occlusion, coupled with increased vegetation density, present the most significant effect on DP and CC. Overall, the results indicate a significantly higher detection probability (DP), 81.1%, and correct classification (CC) of 76.6% for the analyst approach when compared to ML which detected 41.1% and classified 47.5% of wildlife within CT images. However, both methods presented similar probabilities for daylight CT images, 69.4% (ML) and 71.8% (analysts), and dusk CT images, 17.6% (ML) and 16.2% (analysts), when detecting humans. Given that users carefully follow provided recommendations, we expect DP and CC to increase. In turn, the ML approach to CT image processing would be an excellent provision to support time-sensitive threat monitoring for biodiversity conservation.

Stellar clustering shapes the architecture of planetary systems.

Planet formation is generally described in terms of a system containing the host star and a protoplanetary disk1-3, of which the internal properties (for example, mass and metallicity) determine the properties of the resulting planetary system4. However, (proto)planetary systems are predicted5,6 and observed7,8 to be affected by the spatially clustered stellar formation environment, through either dynamical star-star interactions or external photoevaporation by nearby massive stars9. It is challenging to quantify how the architecture of planetary sysems is affected by these environmental processes, because stellar groups spatially disperse within less than a billion years10, well below the ages of most known exoplanets. Here we identify old, co-moving stellar groups around exoplanet host stars in the astrometric data from the Gaia satellite11,12 and demonstrate that the architecture of planetary systems exhibits a strong dependence on local stellar clustering in position-velocity phase space. After controlling for host stellar age, mass, metallicity and distance from the star, we obtain highly significant differences (with p values of 10-5 to 10-2) in planetary system properties between phase space overdensities (composed of a greater number of co-moving stars than unstructured space) and the field. The median semi-major axis and orbital period of planets in phase space overdensities are 0.087 astronomical units and 9.6 days, respectively, compared to 0.81 astronomical units and 154 days, respectively, for planets around field stars. 'Hot Jupiters' (massive, short-period exoplanets) predominantly exist in stellar phase space overdensities, strongly suggesting that their extreme orbits originate from environmental perturbations rather than internal migration13,14 or planet-planet scattering15,16. Our findings reveal that stellar clustering is a key factor setting the architectures of planetary systems.

Fast and inefficient star formation due to short-lived molecular clouds and rapid feedback.

The physics of star formation and the deposition of mass, momentum and energy into the interstellar medium by massive stars ('feedback') are the main uncertainties in modern cosmological simulations of galaxy formation and evolution1,2. These processes determine the properties of galaxies3,4 but are poorly understood on the scale of individual giant molecular clouds (less than 100 parsecs)5,6, which are resolved in modern galaxy formation simulations7,8. The key question is why the timescale for depleting molecular gas through star formation in galaxies (about 2 billion years)9,10 exceeds the cloud dynamical timescale by two orders of magnitude11. Either most of a cloud's mass is converted into stars over many dynamical times12 or only a small fraction turns into stars before the cloud is dispersed on a dynamical timescale13,14. Here we report high-angular-resolution observations of the nearby flocculent spiral galaxy NGC 300. We find that the molecular gas and high-mass star formation on the scale of giant molecular clouds are spatially decorrelated, in contrast to their tight correlation on galactic scales5. We demonstrate that this decorrelation implies rapid evolutionary cycling between clouds, star formation and feedback. We apply a statistical method15,16 to quantify the evolutionary timeline and find that star formation is regulated by efficient stellar feedback, which drives cloud dispersal on short timescales (around 1.5 million years). The rapid feedback arises from radiation and stellar winds, before supernova explosions can occur. This feedback limits cloud lifetimes to about one dynamical timescale (about 10 million years), with integrated star formation efficiencies of only 2 to 3 per cent. Our findings reveal that galaxies consist of building blocks undergoing vigorous, feedback-driven life cycles that vary with the galactic environment and collectively define how galaxies form stars.