1,721,017 research outputs found
How severe wildfires and climate change could drive post-fire recovery of low-elevation vegetation: data from the first field campaign of a monitoring survey in the Karts (North-East Iatly)
No full textWildfires are a major ecological factor shaping vegetation and landscape, and their impacts are projected to
escalate due to global warming: the intensity, frequency and extent of fires are increasing all over the world,
with dramatic consequences on habitats. In summer 2022, severe fires burnt over 4000 hectares of the
western Karst between Italy and Slovenia, a submediterranean low-hilly area near the coast. In spring 2023, we
started a survey to analyze the consequences of the 2022 fires on plant communities and to monitor post-fire
vegetation dynamics. This study aims to investigate the possible effects of interactions of severe fires, climate
change and alien species on the floristic composition of habitats and on the typical processes of post-fire
vegetation recovery in a low-elevated area.
The study was focused on 4 major habitats of the western Karst, 3 of which dynamically related: the
thermophilous karst grassland Centaureo cristatae-Chrysopogonetum grylli, the thermophilous shrubland
Pruno mahaleb-Paliuretum spina-christi, and the karst downy oak wood Aristolochio luteae-Quercetum
pubescentis. Black pine plantations were also included due to their large extent. Permanent plots were
installed in the most intensively burnt areas mapped by satellite remote sensing data using a stratified random
sampling, by placing 7 x 7 m2 squared-plots in the four major habitat types identified on the basis of available
habitat maps and photo-interpretation. In each plot the percent cover of total vegetation, bare soil and of all
species was recorded.
At the habitat level, the highest total species richness and the lowest one for alien species were both found
in the dry karst grassland, which also exhibited excellent quantitative and qualitative recovery, confirming itself
as a highly resilient habitat. Shrubland showed a strong recovery of native shrub species, a rather high number
of total species and alien species compared to the investigated habitats, however with alien species occurring
with low cover values. The downy oak woodland had similar species richness values to shrubland, but higher
abundance of alien species, esp. Robinia pseudoacacia and Ailanthus altissima, and of native ruderal species:
therefore strong modifications of the floristic structure with deviations from the typical secondary succession
are possible. Black pine plantations were found to be characterized by the lowest total species richness, the
highest number of native ruderal and alien species, poor recovery of native species and unclear dynamic
trajectories.
The study is meant to provide information i) to identify interventions to support and eventually correct the
post-fire recovery of habitats, ii) to support land management policies to enhance the resilience and resistance
of the Karst landscape to wildfires
A Hybrid System for Systematic Generalization in Simple Arithmetic Problems
Full text linkSolving symbolic reasoning problems that require compositionality and systematicity is considered one of the key ingredients of human intelligence. However, symbolic reasoning is still a great challenge for deep learning models, which often cannot generalize the reasoning pattern to out-of-distribution test cases. In this work, we propose a hybrid system capable of solving arithmetic problems that require compositional and systematic reasoning over sequences of symbols. The model acquires such a skill by learning appropriate substitution rules, which are applied iteratively to the input string until the expression is completely resolved. We show that the proposed system can accurately solve nested arithmetical expressions even when trained only on a subset including the simplest cases, significantly outperforming both a sequence-to-sequence model trained end-to-end and a state-of-the-art large language model
Assessing the Emergent Symbolic Reasoning Abilities of Llama Large Language Models
No full textLarge Language Models (LLMs) achieve impressive performance in a wide range of tasks, even if they are often trained with the only objective of chatting fluently with users. Among other skills, LLMs show emergent abilities in mathematical reasoning benchmarks, which can be elicited with appropriate prompting methods. In this work, we systematically investigate the capabilities and limitations of popular open-source LLMs on different symbolic reasoning tasks. We evaluate three models of the Llama 2 family on two datasets that require solving mathematical formulas of varying degrees of difficulty. We test a generalist LLM (Llama 2 Chat) as well as two fine-tuned versions of Llama 2 (MAmmoTH and MetaMath) specifically designed to tackle mathematical problems. We observe that both increasing the scale of the model and fine-tuning it on relevant tasks lead to significant performance gains. Furthermore, using fine-grained evaluation measures, we find that such performance gains are mostly observed with mathematical formulas of low complexity, which nevertheless often remain challenging even for the largest fine-tuned models
Benchmarking GPT-4 on Algorithmic Problems: A Systematic Evaluation of Prompting Strategies
Full text linkLarge Language Models (LLMs) have revolutionized the field of Natural Language Processing thanks to their ability to reuse knowledge acquired on massive text corpora on a wide variety of downstream tasks, with minimal (if any) tuning steps. At the same time, it has been repeatedly shown that LLMs lack systematic generalization, which allows to extrapolate the learned statistical regularities outside the training distribution. In this work, we offer a systematic benchmarking of GPT-4, one of the most advanced LLMs available, on three algorithmic tasks characterized by the possibility to control the problem difficulty with two parameters. We compare the performance of GPT-4 with that of its predecessor (GPT-3.5) and with a variant of the Transformer-Encoder architecture recently introduced to solve similar tasks, the Neural Data Router. We find that the deployment of advanced prompting techniques allows GPT-4 to reach superior accuracy on all tasks, demonstrating that state-of-the-art LLMs constitute a very strong baseline also in challenging tasks that require systematic generalization
Review of Invasive Plant Functional Traits and Management Using Remote Sensing in Sub-Saharan Africa
Full text linkBiodiversity and sustainable development in Sub-Saharan Africa (SSA) are considerably impacted by invasive alien plants (IAPs). Increasing plant invasions in SSA threaten agricultural productivity, biodiversity conservation, and other socioeconomic activities, which in turn put the United Nations Sustainable Development Goals (SDGs) in peril. In order to effectively combat IAPs, understanding their functional traits (morphological, physiological, and phenological traits) and integrating them into remote sensing (RS) is vital. While functional traits influence IAPs’ fitness to invade and establish in a new geographical range, RS aids in studying them remotely, delineating and mapping them, and predicting their potential invasions. The information on this study topic was gathered by reviewing various existing studies published between 2000 and 2024. Based on this review, it was deduced that the majority of IAPs are fast-growing (or acquisitive), with a shorter leaf lifespan, bigger leaves, and higher plant height, ultimately resulting in a higher resource acquisition ability. We established further that in SSA, there are limited studies on IAP functional traits and their integration in RS. Many studies conducted in the region focus mostly on IAP distribution. Evidence from prior studies revealed that functional trait remote sensing (FTRS)-based research not only improves detection and mapping but also predicts whether a certain alien plant can become invasive or expand its distribution range. Thus, using the FTRS approach could help IAP management in SSA, ultimately achieving the SDGs. Our review discusses IAP implications in SSA (e.g., Angola, Tanzania, Benin, Kenya, Uganda, Rwanda, Zambia, Burundi, Zimbabwe, Botswana, Malawi, etc.) and for the achievement of SDGs; functional traits and their impact on alien invasions; and the importance of incorporating functional traits into RS
First record of naturalization of Erechtites hieraciifolius (L.) Raf. ex DC. (Asteraceae) in Italy
Full text linkThe plant species Erechtites hieraciifolius (Asteraceae) is here reported for the first time in Italy as a naturalized neophyte in the Classical Karst. The species was observed in 2023 in post-fire forest areas burnt by wildfires in the summer 2022. The features of findings suggest for a naturalization of the species with putative invasive character. This novel occurrence highlights the need for additional research to better understand its colonization and expansion, suggesting the need of early eradication actions
A simplified framework for fast and reliable measurement of leaf turgor loss point
No full textDrought tolerance shapes the distribution of plant species, and it is mainly determined by the osmotic potential at full turgor (π 0 ) and the water potential at turgor loss point (Ψ tlp ). We provide a simplified framework for π 0 and Ψ tlp measurements based on osmometer determination of π 0 (π 0_osm ). Specifically, we ran regression models to i) improve the predictive power of the estimation of π 0 from π 0_osm and morpho-anatomical traits; ii) obtain the most accurate model to predict Ψ tlp on the basis of the global relationship between π 0 and Ψ tlp. The inclusion of the leaf dry matter content (LDMC), an easy-to-measure trait, in the regression model improved the predictive power of the estimation of π 0 from π 0_osm . When π 0_osm was used as a simple predictor of Ψ tlp , discrepancies arose in comparison with global relationship between π 0 and Ψ tlp. Ψ tlp values calculated as a function of the π 0 derived from π 0_osm and LDMC (π 0_fit ) were consistent with the global relationship between π 0 and Ψ tlp . The simplified framework provided here could encourage the inclusion of mechanistically sound drought tolerance traits in ecological studies
ITV-net: leveraging intraspecific trait variability to bridge vegetation science and trait-based research in Italy
No full textVegetation science is a branch of community ecology that relies on species identities and abundance to classify vegetation in coherent units and to explore species coexistence and turnover dynamics. The advent of trait-based ecology has expanded vegetation science, providing a framework that allows for a better understanding of plant strategies and the functional structure of communities. These complementary disciplines have remained largely independent among Italian plant ecologists. Therefore, in 2021, we launched the ITV-net initiative, a national collaborative effort for bringing together vegetation plots and field-measured plant trait data to develop a national platform that can serve both vegetation and trait-based ecologists. In the first data call, we were able to gather trait data on two key leaf traits (i.e., Leaf Area and Specific Leaf Area) for >700 species across 1,043 georeferenced vegetation plots, complemented with species relative abundances, across eight different EUNIS habitat types. Despite this remarkable first milestone, we aim to enlarge the scope of this initiative to include more vegetation plots and functional traits across more habitat types in Italy. Here, we provide an overview of the ITV-net initiative and its underlying methodological details as a ‘manifesto’ to spread the data call to other potential contributors in the Italian community of plant ecologists. Our ultimate objective is to bridge the vegetation science and trait-based ecological research in Italy towards developing a national database of vegetation plots and plant functional traits. We believe this effort will contribute to building a solid network among Italian plant ecologists to cross the artificial boundaries of different, yet complementary, disciplines
- …
