Evaluation of the UMASEP-10 Version 2 Tool for Predicting All >10 MeV SEP Events of Solar Cycles 22, 23 and 24

The prediction of solar energetic particle (SEP) events may help to improve the mitigation of adverse effects on humans and technology in space. UMASEP (University of Málaga Solar particle Event Predictor) is an empirical model scheme that predicts SEP events. This scheme is based on a dual-model approach. The first model predicts well-connected events by using an improved lag-correlation algorithm for analyzing soft X-ray (SXR) and differential proton fluxes to estimate empirically the Sun–Earth magnetic connectivity. The second model predicts poorly connected events by analyzing the evolution of differential proton fluxes. This study presents the evaluation of UMASEP-10 version 2, a tool based on the aforementioned scheme for predicting all >10 MeV SEP events, including those without associated flare. The evaluation of this tool is presented in terms of the probability of detection (POD), false alarm ratio (FAR) and average warning time (AWT). The best performance was achieved for the solar cycle 24 (i.e., 2008–2019), obtaining a POD of 91.1% (41/45), a FAR of 12.8% (6/47) and an AWT of 2 h 46 min. These results show that UMASEP-10 version 2 obtains a high POD and low FAR mainly because it is able to detect true Sun–Earth magnetic connections.

Using LSTM Neural Networks for Onsite Earthquake Early Warning

Onsite earthquake early warning (EEW) systems determine possible destructive S waves solely from initial P waves and issue alarms before heavy shaking begins. Onsite EEW plays a crucial role in filling in the blank of the blind zone near the epicenter, which often suffers the most from disastrous ground shaking. Previous studies suggest that the peak P-wave displacement amplitude (Pd) may serve as a possible indicator of destructive earthquakes. However, the attempt to use a single indicator with fixed thresholds suffers from inevitable errors because the diversity in travel paths and site effects for different stations introduces complex nonlinearities. In addition, the short warning time poses a threat to the validity of EEW. To conquer the aforementioned problems, this study presents a deep learning approach employing long short-term memory (LSTM) neural networks, which can produce a highly nonlinear neural network and derive an alert probability at every time step. The proposed LSTM neural network is then tested with two major earthquake events and one moderate earthquake event that occurred recently in Taiwan, yielding the results of a missed alarm rate of 0% and a false alarm rate of 2.01%. This study demonstrates promising outcomes in both missed alarms and false alarms reduction. Moreover, the proposed model provides an adequate warning time for emergency response.

Flash Flood Disaster Reconstruction for Estimating the Available Warning Time, the Case in Sempor River on 21st of February 2020, Mt. Merapi Slope, Yogyakarta Special Region

Flash floods are hazardous events characterized by short response times. The occurrences of flash flood disasters have increased significantly in the last few years, producing a remarkable casualty number globally. On February 21st, 2020, a flash flood occurred in the Sempor River of the Mount Merapi slope, Special Region of Yogyakarta, Indonesia, causing a significant death of high school students. This study aims to reconstruct the river’s hydrologic and hydraulic conditions and identify the available warning time based on the flash flood event. This study extracted the catchment configuration from a Digital Elevation Model (DEM) using the GIS technique. A simulation of flood hydrograph at a control point used the HEC-HMS model, the SCS Curve Number Loss method, and the SCS Unit Hydrograph. The simulated flood hydrograph was inputted into the one-dimensional unsteady flow model of HEC-RAS to simulate water depth and flow velocity. The calibration process adjusts both models’ parameters by comparing the simulated peak discharge with the surveyed data. The modelling results provide warning time components. The results of this study can support the decision-making in flash flood risk mitigation for the local communities.

Retrospective Study of FinDer Algorithm During the 2019, Ridgecrest Earthquake Sequence

Real-time characterization of evolving rupture is crucial for mitigating against seismic hazards exposed to potentially devastating earthquake events in EEWs (Earthquake Early Warning system). Currently, FinDer (Finite Fault Rupture Detector) algorithm explicitly utilizes observed ground motion pattern to solve for the evolving rupture to generate alerts for early warning purpose, which is currently contributing to ShakeAlert EEW system in West Coast of United States, within the area covered by the Advanced National Seismic System (ANSS) network. Here we implement FinDer offline to explore its feasibility assuming ideal field telemetry on a database of real earthquakes with magnitude M ≥5.0 occurring in Ridgecrest, Southern California in 2019. We specially focus on evaluating the performance of FinDer through end-user-orientated analysis in terms of warning time and accuracy of ground shaking prediction. Overall, FinDer classifies alerts with a rate of success over 74% across a broad range of alert criteria, substantial fraction of sites can be successfully alerted including the most difficult cases with high ground motion intensities regardless of invariable few seconds of warning time. FinDer can be configured to generate more useful alerts with higher cost savings by applying lower alert threshold during the Ridgecrest earthquake sequence. Furthermore, although large fractions of sites would have been timely alerted, it is significantly challenging for predicting accurately the moderate or worse intensities (Modified Mercalli Intensity > 5.5) in advance even if applying lower alert threshold and higher damage threshold. Nonetheless, FinDer performs well in an evolutionary manner to guarantee reliable alerts by resorting to a consistent description of point source or occurring rupture.

Research on Risk Measurement and Early Warning of Electronic Banking Business Based on GMDH Algorithm

In order to deal with the problems of traditional e-banking risk measurement and early warning methods, such as low accuracy of e-banking risk measurement and longer early warning time, an e-banking risk measurement and early warning method based on the GMDH algorithm is proposed. This scheme mines the e-banking risk measurement and early warning indicators by the GMDH algorithm, and it will input the influencing factors and risk factors as independent variables into the GMDH modeling network and then input the e-banking business growth rate as the dependent variable into the GMDH modeling network which is standardized by the normative method of processing the e-banking business risk measurement and early warning index data. According to the processing results, it calculates the weight of the measurement and early warning index by the entropy method, and it constructs the e-banking risk measurement model with the genetic algorithm which can help to calculate the optimal solution of the parameters, formulate the risk measurement interval, and determine the risk in order to realize the risk warning of electronic banking business. The simulation results show that the proposed method has a higher accuracy of e-banking risk measurement and a shorter warning time.

Risk-Based Warning Decision Making of Cascade Breaching of the Tangjiashan Landslide Dam and Two Smaller Downstream Landslide Dams

Mega earthquakes or serious rainfall storms often cause crowded landslides in mountainous areas. A large part of these landslides are very likely blocking rivers and forming landslide dams in series along rivers. The risks of cascading failure of landslide dams are significantly different from that of a single dam. This paper presented the work on risk-based warning decision making on cascading breaching of the 2008 Tangjiashan landslide dam and two small downstream landslide dams in a series along Tongkou River. The optimal decision was made by achieving minimal expected total loss. Cascade breaching of a series of landslide dams is more likely to produce a multi-peak flood. When the coming of the breaching flood from the upstream dam perfectly overlaps with the dam breaching flood of the downstream dam, a higher overlapped peak flood would occur. When overlapped peak flood occurs, the flood risk would be larger and evacuation warning needs to be issued earlier to avoid serious life loss and flood damages. When multi-peak flood occurs, people may be misled by the warning of the previous peak flood and suddenly attacked by the peak flood thereafter, incurring catastrophic loss. Systematical decision making needs to be conducted to sufficiently concern the risk caused by each peak of the breaching flood. The dam failure probability Pf linearly influences the expected life loss and flood damage but does not influence the evacuation cost. The expected total loss significantly decreases with Pf when the warning time was insufficient. However, it would not change much with Pf when warning time is sufficient.

Vigilance in reindeer (Rangifer tarandus); evolutionary history, predation and human interference

To elucidate genetic variability in vigilance behaviour for reindeer with historical differences in their interactions with predators and humans, we measured vigilance frequency and duration for grazing reindeer in Southern Norway (Rondane and Norefjell-Reinsjøfjell), Svalbard (Edgeøya and Nordenskiöld Land) and Barf/Royal Bay and Busen in the southern Hemisphere (South Georgia). Averaged for all areas, frequency and duration of vigilance bouts were less than 0.5 and 2.5 s, respectively. Frequency was insignificantly 1.3 times higher in Rondane than Edgeøya, and significantly 2.0, 3.5, 5.2 and 12.4 times higher than Norefjell, Nordenskiöld Land, Barf/Royal Bay and Busen, respectively. Duration per vigilance bout was not different amongst the areas. Thus, while frequency varied considerably, duration remained constant, supporting a hard-wired adaptation to, among other suggestions, an open landscape. Plasticity in frequency allows for flexible behavioral responses to environmental factors with predation, domestication and hunting key drivers for reindeer. Other factors include (1) the open, treeless alpine/Arctic environment inhabited by Rangifer subspecies allowing warning time, (2) grouping behaviour, (3) relative low density of predators and (4) the anatomy and physiology of ungulate vision.

Early warning of hand, foot, and mouth disease transmission: A modeling study in mainland, China

Background Hand, foot, and mouth disease (HFMD) is a global infectious disease; particularly, it has a high disease burden in China. This study was aimed to explore the temporal and spatial distribution of the disease by analyzing its epidemiological characteristics, and to calculate the early warning signals of HFMD by using a logistic differential equation (LDE) model. Methods This study included datasets of HFMD cases reported in seven regions in Mainland China. The early warning time (week) was calculated using the LDE model with the key parameters estimated by fitting with the data. Two key time points, “epidemic acceleration week (EAW)” and “recommended warning week (RWW)”, were calculated to show the early warning time. Results The mean annual incidence of HFMD cases per 100,000 per year was 218, 360, 223, 124, and 359 in Hunan Province, Shenzhen City, Xiamen City, Chuxiong Prefecture, Yunxiao County across the southern regions, respectively and 60 and 34 in Jilin Province and Longde County across the northern regions, respectively. The LDE model fitted well with the reported data (R2 > 0.65, P < 0.001). Distinct temporal patterns were found across geographical regions: two early warning signals emerged in spring and autumn every year across southern regions while one early warning signals in summer every year across northern regions. Conclusions The disease burden of HFMD in China is still high, with more cases occurring in the southern regions. The early warning of HFMD across the seven regions is heterogeneous. In the northern regions, it has a high incidence during summer and peaks in June every year; in the southern regions, it has two waves every year with the first wave during spring spreading faster than the second wave during autumn. Our findings can help predict and prepare for active periods of HFMD.