Article
Environmental Sciences
Laurence Hawker, Peter Uhe, Luntadila Paulo, Jeison Sosa, James Savage, Christopher Sampson, Jeffrey Neal
Summary: This article introduces a method that uses machine learning to remove buildings and forests from the Copernicus Digital Elevation Model, generating a more accurate global map of elevation. By training the algorithm with unique reference elevation data from 12 countries, the method significantly reduces vertical errors in built-up areas and forests. The resulting elevation map is more accurate than existing global elevation maps.
ENVIRONMENTAL RESEARCH LETTERS
(2022)
Article
Forestry
Victor M. Velasco Hererra, Willie Soon, Cesar Perez-Moreno, Graciela Velasco Herrera, Raul Martell-Dubois, Laura Rosique-de la Cruz, Valery M. Fedorov, Sergio Cerdeira-Estrada, Eric Bongelli, Emmanuel Zuniga
Summary: The boreal forests of the Northern Hemisphere, covering the USA, Canada, and Russia, serve as crucial carbon sinks, but an increase in wildfires could disrupt their capacity and have global impacts on wildlife and humans. Forecasting wildfires is essential to mitigate risks, and a novel methodology using Bayesian Machine Learning models can identify climatic variations that influence wildfire activity cycles.
FOREST ECOLOGY AND MANAGEMENT
(2022)
Article
Biodiversity Conservation
Wanjing Li, Qinchuan Xin, Xuewen Zhou, Zhicheng Zhang, Yongjian Ruan
Summary: This study successfully predicted the timing of spring onset for different vegetation types using numerical phenology models and machine learning models with a combination of ground observations and satellite data. Numerical phenology models showed better performance after calibration with ground phenology observations, while machine learning models could also capture spatial variation of satellite data when appropriately trained.
ECOLOGICAL INDICATORS
(2021)
Review
Plant Sciences
Negin Katal, Michael Rzanny, Patrick Maeder, Jana Waeldchen
Summary: Climate change poses one of the most critical threats to biodiversity, affecting species interactions, ecosystem functioning, and biotic community assembly. Plant phenology research has become increasingly important due to the strong impact of seasonal and interannual climate variation on the timing of plant events. The feasibility of phenological monitoring relies on developing tools capable of efficiently analyzing large amounts of data. Deep Neural Networks, known for their impressive accuracy in learning representations from data, have shown significant breakthroughs in fields like image processing. This article presents the first systematic literature review of deep learning approaches in plant phenology research, analyzing 24 peer-reviewed studies published from 2016 to 2021. The methods applied in these studies are categorized based on phenological stages, vegetation types, spatial scales, data acquisition, and deep learning methods. The review identifies research trends and promising future directions, providing a systematic overview for this emerging research field.
FRONTIERS IN PLANT SCIENCE
(2022)
Article
Environmental Sciences
Dario Domingo, Fernando Perez-Rodriguez, Esteban Gomez-Garcia, Francisco Rodriguez-Puerta
Summary: Invasive alien plants are threatening local biodiversity by transforming landscapes. This study used RGB-NIR Sentinel-2 data to discriminate and identify the spatial distribution of Acacia dealbata Link from other species based on phenological spectral peak differences. A random forest machine learning algorithm was trained to discriminate between A. dealbata and native species, achieving an overall accuracy of 94%. The study found that A. dealbata was widely distributed in patches, replacing stands of Pinus pinaster Ait., and also formed continuous monospecific stands. This approach provides valuable information for early detection and mapping of A. dealbata and improving sustainable forest management.
Article
Ergonomics
Yingheng Zhang, Haojie Li, Gang Ren
Summary: This paper introduces generalized random forests (GRF) for the estimation of heterogeneous treatment effects (HTEs) in road safety analysis. The simulation results show that GRF outperforms other causal methods, such as the outcome regression method, propensity score method, and doubly robust estimation method, especially in handling nonlinearity and nonadditivity. The case study on the UK's speed camera program reveals significant reductions in road accidents at speed camera sites, with statistically significant heterogeneity in treatment effects. The study also explores the associations between baseline accident records, traffic volume, local socio-economic characteristics, and the safety effects of speed cameras, providing policy suggestions based on the findings. Overall, the evaluation of HTEs offers comprehensive information to local authorities and policymakers, improving the effectiveness of speed camera programs. GRF shows promise in uncovering treatment effect heterogeneity in road safety analysis.
ACCIDENT ANALYSIS AND PREVENTION
(2022)
Review
Computer Science, Information Systems
Guangsheng Chen, Hailiang Lu, Weitao Zou, Linhui Li, Mahmoud Emam, Xuebin Chen, Weipeng Jing, Jian Wang, Chao Li
Summary: Remote sensing images have been widely used in Earth observation tasks, but a single sensor cannot provide observational images with both high spatial and temporal resolution. The spatiotemporal fusion (STF) method has been proposed to overcome this constraint. Many STF methods have been proposed based on different principles and strategies. A new review is needed to reflect the current research status. This review provides a comprehensive overview of current advances, discusses the basic principles and limitations, and collects recent applications. It also introduces publicly available resources and quantitative metrics, and discusses open problems and challenges for future attention.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2023)
Article
Biodiversity Conservation
Lydia K. D. Katsis, Andrew P. Hill, Evelyn Pina-Covarrubias, Peter Prince, Alex Rogers, C. Patrick Doncaster, Jake L. Snaddon
Summary: Unsustainable hunting is a major driver of global biodiversity loss, but monitoring this activity is challenging. Researchers developed a two-stage detection pipeline using acoustic detection and convolutional neural networks to detect gunshots in tropical forests. The method showed high accuracy and recall, providing a more manageable dataset for human verification.
ECOLOGICAL INDICATORS
(2022)
Article
Ecology
Sarab S. Sethi, Robert M. Ewers, Nick S. Jones, Jani Sleutel, Adi Shabrani, Nursyamin Zulkifli, Lorenzo Picinali
Summary: Accurate occurrence data is crucial for conservation, and automated acoustic monitoring can provide an efficient alternative to manual surveys. By using local soundscapes and a convolutional neural network, it is possible to predict the occurrence of various species with high accuracies, especially for those with strong temporal patterns. Soundscapes were shown to be a better predictor of species occurrence than above-ground carbon density, opening up new possibilities for large-scale habitat suitability assessments.
Review
Environmental Sciences
Richard Dein D. Altarez, Armando Apan, Tek Maraseni
Summary: This study provides a systematic review of the applications of spaceborne remote sensing in tropical montane forests (TMFs). The review reveals an increasing number of published papers in this field between 1997 and 2021. Most studies were conducted in the Americas, while Asia, Africa, and Oceania had fewer studies. Optical sensors with low to medium spatial resolution were commonly used, while synthetic aperture radar received less attention. The main research themes included forestry, climate science, and disaster management.
GEOCARTO INTERNATIONAL
(2022)
Article
Biodiversity Conservation
Nestor Rendon, Susana Rodriguez-Buritica, Camilo Sanchez-Giraldo, Juan M. Daza, Claudia Isaza
Summary: This article introduces a new Acoustic Heterogeneity Index (AHI) that quantifies the acoustic heterogeneity related to landscape transformation. By analyzing sound recordings of different habitats, the AHI can estimate the acoustic dissimilarity between sites. The effectiveness of the method was validated using sample data, and AHI was used to analyze soundscape similarities on geographic maps.
ECOLOGICAL INDICATORS
(2022)
Article
Environmental Sciences
Ali R. Al-Aizari, Yousef A. Al-Masnay, Ali Aydda, Jiquan Zhang, Kashif Ullah, Abu Reza Md Towfiqul Islam, Tayyiba Habib, Dawuda Usman Kaku, Jean Claude Nizeyimana, Bazel Al-Shaibah, Yasser M. Khalil, Wafaa M. M. AL-Hameedi, Xingpeng Liu
Summary: This study assesses flood susceptibility in the desert environment of Yemen using remote sensing devices and machine learning algorithms. The results show that all models have a high capacity to predict floods, with the tree-based ensemble algorithms performing the best. This research is important for assessing disaster susceptibility and reducing the risk of natural disasters.
Article
Engineering, Multidisciplinary
Gabriela Baban, Mihai Daniel Nita
Summary: The study aims to test the integration of forest canopy height satellite measurements with a Random Forest algorithm to obtain continuous sets of data. The performance of two LiDAR missions was compared to field measurements taken with a mobile LiDAR scanner. The compatibility of the sensors as an input for an RF model was also tested. The results showed significant correlation between MLS height and GEDI height, but not for ICEsat-2 height measurements.
Article
Biodiversity Conservation
Yongshuo H. Fu, Xinxi Li, Shouzhi Chen, Zhaofei Wu, Jianrong Su, Xing Li, Shuaifeng Li, Jing Zhang, Jing Tang, Jingfeng Xiao
Summary: Autumn phenology in subtropical forests is mainly influenced by soil moisture, with solar radiation playing a larger role in northern forests. Under future climate warming conditions, the autumn photosynthetic transition dates are predicted to be delayed, but the delay is smaller compared to the current trend. Machine learning methods outperform process-based models in predicting these dates.
GLOBAL CHANGE BIOLOGY
(2022)
Article
Forestry
Saeideh Karimi, Mehdi Heydari, Javad Mirzaei, Omid Karami, Brandon Heung, Amir Mosavi
Summary: Wildfire has a significant impact on plant phenology and can be monitored using time series satellite data to identify the growing season. This study investigated the use of remote sensing data and land surface phenology parameters to evaluate the impacts of fire in semi-arid oak forests of Iran. The results showed that the fire had a negative effect on land surface phenology, but there were signs of forest restoration after two years.
Article
Computer Science, Artificial Intelligence
Hugo Oliveira, Caio Silva, Gabriel L. S. Machado, Keiller Nogueira, Jefersson A. dos Santos
Summary: In traditional semantic segmentation, the lack of ability to recognize unknown classes in an open set scenario poses a challenge. This paper introduces two fully convolutional approaches, OpenFCN and OpenPCS, to effectively address open set semantic segmentation. OpenPCS shows promising results by outperforming OpenFCN and SoftMax thresholding methods. The experiments also demonstrate the effectiveness and robustness of OpenPCS in improving the recognition of unknown class pixels without reducing the accuracy on known class pixels.
Article
Ecology
Carlos A. Ordonez-Parra, Roberta L. C. Dayrell, Daniel Negreiros, Antonio C. S. Andrade, Leticia G. Andrade, Yasmine Antonini, Leilane C. Barreto, Fernanda de V. Barros, Vanessa da Cruz Carvalho, Blanca Auxiliadora Dugarte Corredor, Antonio Claudio Davide, Alexandre A. Duarte, Selma Dos Santos Feitosa, Alessandra F. Fernandes, G. Wilson Fernandes, Maurilio Assis Figueiredo, Alessandra Fidelis, Leticia Couto Garcia, Queila Souza Garcia, Victor T. Giorni, Vanessa G. N. Gomes, Carollayne Goncalves-Magalhaes, Alessandra R. Kozovits, Jose P. Lemos-Filho, Soizig Le Stradic, Isabel Cristina Machado, Fabiano Rodrigo Maia, Andrea R. Marques, Clesnan Mendes-Rodrigues, Maria Cristina T. B. Messias, Leonor Patricia Cerdeira Morellato, Moemy Gomes de Moraes, Bruno Moreira, Flavia Peres Nunes, Ademir K. M. Oliveira, Yumi Oki, Alba R. P. Rodrigues, Carolina Pietczak, Jose Carlos Pina, Silvio Junio Ramos, Marli A. Ranal, Joao Paulo Ribeiro-Oliveira, Flavio H. Rodrigues, Denise G. Santana, Fernando M. G. Santos, Ana Paula M. S. Senhuk, Rodrigo A. Silveira, Natalia Costa Soares, Olivia Alvina Oliveira Tonetti, Vinicius Augusto da Silveira Vieira, Leticia Cristiane de Sena Viana, Marcilio Zanetti, Heloiza L. Zirondi, Fernando A. O. Silveira
Summary: The Rock n' Seeds database provides functional trait data and germination experiments from Brazilian rock outcrop vegetation, including 16 functional traits for 383 taxa and 48 germination experiments for 281 taxa. This database will be valuable for synthesizing germination data, advancing comparative functional ecology, and guiding seed-based restoration and biodiversity conservation in tropical megadiverse ecosystems.
Article
Computer Science, Artificial Intelligence
Pedro Barros, Fabiane Queiroz, Flavio Figueiredo, Jefersson A. Dos Santos, Heitor Ramos
Summary: We propose a novel deep metric learning method using an autoencoder to define a new latent space called S-space. We locate markers in S-space to identify similar and dissimilar objects and estimate their similarities using a kernel-based Cauchy distribution. Our approach simultaneously estimates the markers' positions and represents the objects in the same space while preventing collapsing of similar markers. We demonstrate the effectiveness of our method on various datasets.
ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY
(2023)
Article
Plant Sciences
Francisco Javier Jimenez-Lopez, Montserrat Arista, Maria Talavera, Leonor Patricia Cerdeira Morellato, John R. Pannell, Juan Viruel, Pedro L. Ortiz Ballesteros
Summary: The genetic divergence between species depends on reproductive isolation (RI) caused by traits reducing interspecific mating (prezygotic isolation) or reduced hybrid fitness (postzygotic isolation). Prezygotic barriers were found to be generally stronger than postzygotic barriers, but previous studies mostly examined F-1 hybrid fitness in early life cycle stages. This study combined field and experimental data to assess the strength of 17 prezygotic and postzygotic reproductive barriers between co-occurring Lysimachia species. The results showed near complete RI between the two species, with prezygotic barriers contributing more in reducing gene flow in allopatry, while their contributions were more similar in sympatry. The strength of postzygotic RI was underestimated when effects on late stages of the life cycle were disregarded.
Article
Environmental Sciences
Jing Wang, Guangqin Song, Michael Liddell, Patricia Morellato, Calvin K. F. Lee, Dedi Yang, Bruna Alberton, Matteo Detto, Xuanlong Ma, Yingyi Zhao, Henry C. H. Yeung, Hongsheng Zhang, Michael Ng, Bruce W. Nelson, Alfredo Huete, Jin Wu
Summary: In tropical forests, leaf phenology varies greatly and plays a crucial role in regulating biogeochemical cycles and climate. We developed an index-guided, ecologically constrained autoencoder (IG-ECAE) method to accurately derive a deciduousness metric from PlanetScope data, allowing for monitoring phenology variability at both fine and ecosystem scales.
REMOTE SENSING OF ENVIRONMENT
(2023)
Article
Engineering, Biomedical
Fabio Giuliano Caetano, Paulo Roberto Pereira Santiago, Ricardo da Silva Torres, Sergio Augusto Cunha, Felipe Arruda Moura
Summary: The purpose of this study was to analyse the interplayer coordination between opponents during offensive sequences and to determine if different coordination patterns exist for offensive sequences ending in shots to goal compared to those ending in defensive tackles. A total of 580 offensive sequences were analysed, with 172 ending in shots to goal and 408 ending in defensive tackles. The results showed that in-phase coordination was the most common pattern for all displacement directions and offensive sequence outcomes, while antiphase coordination was the least frequent. For lateral displacements, offensive sequences ending in shots to goal had a lower frequency of in-phase coordination and a higher frequency of offensive player phase compared to those ending in defensive tackles. This study provides valuable insights for future research and helps coaches understand the behaviors of successful and unsuccessful attacks.
SPORTS BIOMECHANICS
(2023)
Review
Computer Science, Information Systems
Wajeeha Nasar, Ricardo Da Silva Torres, Odd Erik Gundersen, Anniken T. Karlsen
Summary: Whenever disasters occur, search and rescue services are crucial for proper response. Decision support systems using data management solutions and artificial intelligence technologies have improved the efficiency and effectiveness of search and rescue operations. This paper presents findings from a study that identified existing search and rescue processes, analyzed their research contributions, and explored knowledge transfer potential. The review highlights the use of unconventional data management solutions and integration of geographical information systems with machine learning in land rescue operations, but suggests a research gap in search and rescue decision support at sea.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2023)
Article
Environmental Sciences
Thais Pereira de Medeiros, Leonor Patricia Cerdeira Morellato, Thiago Sanna Freire Silva
Summary: Modern UAS or drones with high-resolution imagery have proven effective in classifying heterogeneous vegetation using computer vision and machine learning techniques, providing valuable data for the study of complex vegetation systems.
FRONTIERS IN ENVIRONMENTAL SCIENCE
(2023)
Article
Plant Sciences
Maria Gabriela Gutierrez Camargo, Montserrat Arista, Klaus Lunau, Pedro Luis Ortiz, Soizig Le Stradic, Nathalia Miranda Walter Bretas Rocha, Leonor Patricia Cerdeira Morellato
Summary: Within a community, co-occurring plant species can both diverge and benefit from floral signal standardisation, depending on the flower colour display and flowering phenology. In highly diverse tropical and temperate vegetation types, the visual similarity of rewarding flowers among co-occurring species was investigated. Flower colour was generally not distinguishable within groups by bees, and the flowering periods overlapped in Mediterranean species but tended to be segregated in Brazilian campo rupestre species. The standardisation of floral colour signal within these two species-rich plant communities is advantageous for most of the species studied, despite different flowering phenologies.
Article
Environmental Sciences
Priscilla P. P. Loiola, Leonor Patricia Cerdeira Morellato, Maria Gabriela Gutierrez Camargo, Vitor A. A. Kamimura, Jacqueline S. S. Mattos, Annia Susin Streher, Soizig Le Stradic
Summary: This study investigated the effects of environmental variables on plant diversity along an old tropical mountain in southeastern Brazil. The results showed that the richness of graminoids and herbaceous species increased with elevation and nutrient-impoverished soils, while woody richness showed the opposite pattern. The study highlighted the importance of elevation, soil, and vegetation types in driving plant diversity.
JOURNAL OF MOUNTAIN SCIENCE
(2023)
Article
Computer Science, Information Systems
Pedro H. T. Gama, Hugo Oliveira, Jose Marcato Jr, Jefersson A. dos Santos
Summary: This paper presents two novel meta-learning methods, WeaSeL and ProtoSeg, for few-shot semantic segmentation with sparse annotations. Extensive evaluation of the proposed methods in different fields (12 datasets) including medical imaging and agricultural remote sensing demonstrates their potential in segmenting coffee/orange crops and anatomical parts of the human body compared to full dense annotation.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Geochemistry & Geophysics
Thiago Carvalho, Jorge A. Chamorro Martinez, Hugo Oliveira, Jefersson A. dos Santos, Raul Queiroz Feitosa
Summary: When it comes to technology in agriculture, crop monitoring is a crucial aspect. A semantic segmentation model is proposed to identify plantations and classify main crops, while also automatically recognizing unknown crops. The framework of outlier exposure is adapted for open set image segmentation, resulting in significant improvement in semantic segmentation of crop imagery.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Computer Science, Information Systems
Gabriel Machado, Matheus B. B. Pereira, Keiller Nogueira, Jefersson A. Dos Santos
Summary: In some cases, a single input image may not be sufficient for object classification. In such scenarios, exploring complementary information from multiple perspectives of the same object becomes crucial to enhance scene understanding and improve performance. However, the task of multi-view image classification faces a major challenge of missing data. This paper proposes a novel technique that addresses this problem using deep learning-based approaches and metric learning, which can be easily applied to other domains and applications. Experimental results on two distinct multi-view aerial-ground datasets demonstrate that the proposed algorithm outperforms state-of-the-art methods in terms of accuracy.
Article
Computer Science, Information Systems
Zhicheng Hu, Amirabbas Hojjati, Amirashkan Haghshenas, Agus Hasan, Ricardo Da Silva Torres
Summary: This paper introduces a novel method called Temporal Topology Density Map (TTDM) to represent 2D spatial data with temporal variations in a 2D continuous spatial space constrained by a topology. The method combines topological density maps with Change Frequency Heatmap (CFH) to provide an intuitive visualization for analyzing spatiotemporal data changes over time. The effectiveness of the proposed method is demonstrated through two case studies.
Article
Biodiversity Conservation
Renan Borgiani, Maria Tereza Grombone-Guaratini, Betania da Cunha Vargas, Amanda Eburneo Martins, Maria Gabriela Gutierrez Camargo, Leonor Patricia Cerdeira Morellato
Summary: The remnant area of cerrado in Itirapina municipality, Sao Paulo, Brazil, holds significant plant species diversity of environmental and ecological importance. After 12 years of inventory, a checklist of 195 plant species from 54 families and 131 genera was compiled.