The classification of animal vocalizations through bioacoustic analysis has become a crucial tool in wildlife monitoring and conservation. This study presents a Systematic Literature Review (SLR) focused on the use of Convolutional Neural Networks (CNNs) for animal sound classification, synthesizing insights from 21 peer-reviewed journal articles published between 2016 and 2024. The review investigates the effectiveness of CNN models, the impact of different architectures (e.g., ResNet, VGG, MobileNet, Xception), commonly used evaluation metrics, and the advantages and limitations of CNN-based approaches. Results show that CNNs consistently outperform traditional classifiers by leveraging spectrogram-based feature extraction techniques and deep feature learning, achieving high classification accuracy across birds, mammals, amphibians, and marine species. Lightweight CNNs provide viable alternatives for real-time and resource-constrained applications, while ensemble learning and transfer learning further enhance model performance. Nonetheless, challenges persist, including computational cost, interpretability, and domain-specific generalization. This review underscores CNNs' growing role in ecological research and highlights areas for future improvement, particularly in data-scarce environments. The findings contribute to guiding future implementations and innovations in automated animal sound classification systems.
29 December 2025 in ZERO: Jurnal Sains, Matematika dan Terapan
Modeling rainfall is crucial for hydrological studies and climate adaptation, especially in regions with complex topography such as the Toba Lake area, North Sumatra. Classical probability distributions often struggle to represent skewness, heavy tails, and variability observed in tropical rainfall. This study explores APTXL distribution as a flexible two-parameter model. Through the alpha power transformation, APTXL extends the X-Lindley distribution by introducing an additional shape parameter, allowing better accommodation of asymmetrical and extreme values while maintaining analytical tractability. Statistical properties are derived, and parameters are estimated using maximum likelihood. The model is applied to a long-term dataset from 13 meteorological stations, covering 408 monthly observations per station. Comparative analysis against Gamma, Lognormal, and Generalized Extreme Value distributions using multiple goodness-of-fit criteria indicates that APTXL provides consistently improved performance. These results suggest APTXL as a practical tool for rainfall modeling and water-resource applications in climate-sensitive regions.
Wildfires pose increasing environmental and socio-economic risks, particularly in climate-sensitive and tropical regions, necessitating reliable satellite-based monitoring and predictive frameworks. This study presents a comprehensive survey of satellite-derived wildfire indicators and spatiotemporal modeling approaches, covering their historical development, current methodologies, and emerging research directions. We review major active fire and hotspot datasets derived from MODIS, VIIRS, and related platforms, along with key environmental drivers such as vegetation indices, meteorological variables, and land-surface with a specific case study for the Indonesian region. Modeling approaches are synthesized from classical statistical regression and time-series analysis to contemporary machine learning and deep learning architectures, including convolutional neural networks, recurrent neural networks, and transformer-based models. The analysis highlights the transition toward multi-source data integration and spatiotemporal deep learning frameworks capable of capturing complex wildfire dynamics. Finally, we identify future research challenges, including hybrid physical–AI modeling, uncertainty quantification, and scalable real-time wildfire intelligence systems. This survey provides a structured reference for researchers and practitioners seeking to advance satellite-based wildfire monitoring and prediction.
29 July 2026 in ZERO: Jurnal Sains, Matematika dan Terapan
Shamir's (k, n)-threshold secret sharing scheme provides information-theoretic secrecy against unauthorized subsets. However, it does not verify whether submitted shares are genuine during reconstruction. This study proposes a parity-based modification of Shamir's scheme. In the proposed construction, each share is extended into a triplet over Zp (the integers modulo p). The additional component ci serves as an authenticated verification parameter bound to the participant identity and share value. Under the authenticated parity model, the proposed scheme detects and identifies forged shares before reconstruction. Once a forged share is identified, its original value can be restored locally from the authenticated parity value. The cheating success probability is bounded by 1/p. The construction also attains the Ogata-Kurosawa-Stinson lower bound on share size and preserves the threshold reconstruction property. The secret is embedded as the leading coefficient ak-1 and recovered using a recursive divided-difference formulation, requiring O(k2) field operations and O(k) memory after verification. Runtime evaluation over a 256-bit prime field shows that reconstruction remains practical for large reconstruction sets. The additional parity component yields information rate rho = 1/2. It also requires only one extra field element per participant. A blockchain wallet key-distribution case study is included to illustrate how authenticated parity values can support share verification and correction in institutional key recovery. Compared with prior cheating-detection schemes, the proposed construction achieves an OKS-bound cheating probability within a Shamir-based framework. It also supports recursive coefficient recovery and post-identification share correction.