Tunnel failures and severe underground losses are documented in many official investigations and case histories, but cross-case learning is limited by inconsistent terminology, uneven source quality, and incomplete reporting of mechanisms and consequences. This Data Descriptor presents an evidence-graded database of reported tunnel failures and severe-loss events. Corpus Release 3.1 contains 78 incidents from 33 countries or cross-border settings: 19 Grade A records supported by official investigations, 23 Grade B records supported by peer-reviewed or technical sources, and 36 Grade C contextual records; the primary A/B subset contains 42 incidents, each linked to the source on which its grade rests. Every incident is linked to a source register, field-level uncertainty table, six-step causal chain, chronology, and audit log; 35 records of the previous release were removed after source verification, and every coded consequence value was checked against a source. The schema separates immediate mechanisms from geotechnical triggers, structural or system failures, operational and management contributors, warning status, and consequences; these explanatory fields are coded for the Grade A records. Results are reported by evidence stratum rather than as a global census. The Mendeley Data package supports reuse, audit, correction, and versioned expansion.
Urban grasslands are widespread elements of cities, yet plot-level records combining taxon-specific cover with explicit geographic coordinates remain unevenly available across the Baltic region. This Data Descriptor presents vegetation data collected across Jelgava, Latvia, on 12 survey dates from 4 June to 25 August 2025. A two-resolution spatial grid, comprising 500 × 500 m cells in central Jelgava and 1 × 1 km cells in peripheral areas, defined 104 candidate sampling locations. One 1 × 1 m quadrat was planned per location; vegetation surveys were completed at 102 plots, while two locations (Plot IDs 57 and 66) remained unsurveyed. For each completed plot, vascular plant taxa were recorded, and their projective cover was visually estimated as a percentage. The dataset integrates plot identifiers, location descriptions, geographic coordinates, survey dates, taxonomic information, and taxon-level cover values. It contains 131 taxon rows, of which 129 have at least one recorded occurrence, and 1042 positive taxon-by-plot observations. Taxon richness ranges from 5 to 19 taxa per completed plot, with a median of 10. Data validation included checks of survey status, plot-identifier uniqueness, geographic coordinates, and taxon-cover values. The dataset provides a location-referenced baseline for urban vegetation inventories, descriptive spatial and comparative analyses, teaching, and future location-based comparisons. For reuse, Plot IDs 57 and 66 should be treated as unsurveyed observations rather than as zero-cover vegetation records, and the spatial, seasonal, and taxonomic limitations of the dataset should be considered.
This dataset supports the development of non-invasive methods for assessing the freshness of rainbow trout (Oncorhynchus mykiss) fillets by combining digital images, gas sensor responses, environmental measurements, and microbiological reference data. Data was collected from 80 fillet portions during seven days of refrigerated storage at 4 °C. Of these, 59 fillets were longitudinally monitored, while the remaining 21 fillets were used for destructive Total Viable Count (TVC) analysis. The dataset contains 820 RAW images and their corresponding JPEG derivatives, together with 210 sensor acquisition sessions comprising 120 time points from five MQ-series gas sensors and two environmental sensors. Based on the daily mean TVC and a threshold of 7 log10 CFU/g, days 1–5 were classified as Fresh and days 6–7 as Spoiled. These classifications represent storage-day microbiological status. The dataset can support computer vision, electronic-nose analysis, and multimodal learning.
Remote-access Virtual Private Networks (VPNs) are a fundamental component of modern enterprise infrastructure, providing secure connectivity for distributed users and devices. Despite the widespread deployment of WireGuard, OpenVPN, and IPsec, publicly available datasets supporting reproducible evaluation of their behavior under realistic operating conditions remain limited. This paper introduces VPNBench, an open dataset designed to support the study of VPN performance, resilience, and operational behavior in controlled enterprise remote-access environments. The dataset was collected using an automated experimentation framework comprising a cloud-hosted VPN gateway and three heterogeneous client platforms, namely, a Raspberry Pi 4, a VirtualBox-based Ubuntu virtual machine, and an Ubuntu environment running under Windows Subsystem for Linux. Measurements were collected across baseline, CPU-stress, latency, packet-loss, MTU-variation, wireless-band, and mobility scenarios, resulting in 189 run-level observations spanning 63 experimental configurations. Each observation includes structured metadata, operating-system and network-state snapshots, TCP and UDP performance measurements, latency traces, resource-utilization telemetry, routing information, interface configurations, and VPN runtime-state records. The dataset is accompanied by deployment scripts, experimental automation tools, and analysis utilities, enabling reproducible experimentation and extension of the measurement campaign. The dataset provides a reproducible benchmark resource for studying VPN behavior under the specific operating conditions represented in the testbed and may support comparative analyses of performance, resilience, and mobility-related characteristics. By combining network-performance metrics with environmental metadata and protocol-state information, the dataset enables investigations that extend beyond conventional throughput-oriented VPN evaluations.
Seafloor mapping data derived from acoustic measurements remain scarce for many coastal systems, including locations of active ecological research such as Bahía Almirante, Bocas del Toro, Panama. We present the first openly accessible sonar and ground-truth seafloor dataset for this area. Sonar data were collected from June 2019 to April 2022 using a commercially available Humminbird HELIX 10 CHIRP single-beam echosounder, providing water depth and two relative acoustic bottom indices (E1 and E2) associated with seafloor roughness and hardness, respectively. Survey lines covered the bay at a spacing of approximately 800 m, densified to ~400 m in areas of higher bathymetric relief and ~200 m within the channels connecting the bay to the open ocean. Ground-truth data were collected at 90 sites from October 2021 to July 2023 following a standardized, decision-based field protocol, and include site-level sonar data and qualitative substrate assessment (90 sites), photoquadrat imagery (58 sites), and sediment grain-size analysis where substrate and field conditions permitted (19 sites). All data are provided in standard, analysis-ready formats with full documentation to facilitate reuse and to support ongoing oceanographic, ecological, geological, and coastal management research in the Bocas del Toro Archipelago.
