Fact-checked by Grok 2 weeks ago

Frequency of exceedance

Frequency of exceedance, also known as the rate of exceedance or exceedance probability, is a statistical measure in probability theory and risk analysis that quantifies the expected number of times per unit time (often annually) a random process or variable surpasses a specified threshold value.^[1] It represents the complement of the cumulative distribution function for the variable, expressing the tail probability of exceeding the threshold, and is typically denoted as the annual exceedance probability (AEP) when applied to yearly intervals.^[2] This concept is essential for modeling rare or extreme events, where the frequency is often low (e.g., less than 1 per year), and it underpins decision-making in engineering design by linking probabilistic forecasts to practical risk thresholds.^[3] Closely related to the return period (or recurrence interval), the frequency of exceedance is the reciprocal of the return period; for instance, a 100-year return period corresponds to an annual exceedance probability of 1% or 0.01.^[4] In frequency analysis, it is derived from historical data using distributions such as the log-Pearson Type III, which fits peak values to estimate exceedance curves that plot threshold levels against their probabilities.^[3] These curves enable the prediction of event magnitudes for given frequencies, accounting for uncertainties through confidence intervals, and are adjusted for factors like data length and outliers to improve reliability—longer records (e.g., over 18 years) yield more precise estimates for common events like 10-year floods.^[3] In hydrology and flood risk management, frequency of exceedance is used to determine design flows for infrastructure such as dams, bridges, and stormwater systems, where standards often specify AEPs like 1% for major floods (100-year events) or 10% for intermediate ones (10-year events).^[4] Engineers apply it via flood frequency analysis to forecast peak discharges or volumes, ensuring structures provide adequate protection levels based on regional hydrological data and probabilistic models.^[2] The concept extends to seismic engineering, where it assesses the annual frequency of exceedance for ground motion parameters like peak ground acceleration, informing building codes and hazard maps through probabilistic seismic hazard analysis (PSHA).^[5] In this context, PSHA integrates seismic source models, ground motion prediction equations, and uncertainty quantification to generate hazard curves, targeting low exceedance rates (e.g., 10^{-3} per year for critical facilities) to mitigate earthquake risks.^[6] Beyond natural hazards, frequency of exceedance appears in catastrophe modeling for insurance and financial risk, estimating the probability that losses from events like hurricanes exceed certain amounts over a period, and in environmental assessments to evaluate pollutant exceedances of regulatory limits (e.g., allowing one exceedance per year for air quality standards).^[7]^[1] Its versatility across disciplines highlights its role in fostering resilient systems by translating statistical probabilities into actionable design criteria.

Fundamentals

Definition

The frequency of exceedance (FOE), also known as the upcrossing rate, is a fundamental measure in the analysis of stochastic processes, representing the expected number of times per unit time that a random variable exceeds a specified threshold level. It serves as the reciprocal of the mean inter-exceedance time, providing a rate-based characterization of rare events in continuous or discrete time series. This concept is particularly useful for modeling phenomena where threshold crossings indicate significant occurrences, such as peaks in environmental or engineering systems.^[8] Formally, consider a stationary stochastic process X(t) and a threshold u. The FOE is denoted as \nu(u), defined as \nu(u) = 1 / \mathbb{E}[T], where T denotes the random inter-exceedance time—the duration between consecutive upcrossings of the threshold u (i.e., instances where X(t) transitions from below u to above u). For small exceedance probabilities, the probability of at least one exceedance over a time interval of length t approximates \nu(u) t. This formulation underpins level-crossing analyses in fields like signal processing and reliability engineering.^[8] The term frequency of exceedance originated in the hydrology literature during the 1960s, where it was introduced to quantify the occurrence rates of extreme events like floods in river discharge data. It is distinct from the related concept of return period, which equals $1 / \nu(u) and estimates the average time between exceedances; however, the two are sometimes misused interchangeably, leading to confusion between rates and intervals in risk assessments. In hydrological contexts, FOE often refers specifically to the annual probability of exceedance for peak flows.^[9]^[10]

Key Concepts

The frequency of exceedance (FOE) relies fundamentally on the assumption of stationarity in the underlying stochastic process, meaning that the statistical properties—such as mean, variance, and higher moments—remain invariant over time. This stationarity ensures that the probability distribution of the process does not change, allowing for consistent estimation of exceedance events across different periods. Without this assumption, non-stationarities like trends or shifts due to climate change or human interventions can invalidate FOE models, leading to unreliable predictions in fields like hydrology and engineering.^[11] A critical aspect of applying FOE is the selection of the threshold level u, which defines the critical value above which exceedances are counted. Thresholds are typically chosen based on established design standards, such as regulatory requirements for infrastructure safety, or targeted risk levels that align with acceptable failure probabilities for a given system. For instance, in flood risk management, u might be set to correspond to a design event with a specified return period, ensuring the analysis addresses practical safety margins without being overly conservative or lax.^[12] FOE is conventionally expressed in units of events per unit time, most commonly as an annual rate (e.g., exceedances per year), reflecting the average number of times the threshold is surpassed within a standard yearly interval. This temporal scaling facilitates comparisons across datasets and supports long-term risk assessments by providing a rate that integrates over time. The probability of exceedance, which measures the likelihood of surpassing u at any single point in time, serves as a related metric but focuses on instantaneous risk rather than event frequency. Distinctions arise in how FOE is conceptualized depending on whether it emphasizes pointwise or upcrossing events. Pointwise FOE tallies instantaneous exceedances, capturing the proportion of time or points where the process value exceeds u at discrete or continuous observations, which is useful for assessing overall exposure duration. In contrast, upcrossing FOE specifically counts the rate of threshold crossings from below to above, treating each crossing as a distinct event and thus better suiting analyses of discrete occurrences like peaks in random processes. This differentiation is essential in extremal theory, where upcrossings align with Rice's formula for level-crossing intensities in stationary Gaussian processes.

Mathematical Foundations

General Formulation

The frequency of exceedance for a stationary stochastic process X(t) quantifies the expected rate at which the process crosses above a specified threshold u from below, known as the upcrossing rate. This rate, denoted \nu(u), represents the mean number of such exceedance events per unit time and forms the foundation for analyzing extremes in fields like engineering reliability and environmental risk assessment. For general stationary processes, the formulation relies on the marginal probability density function f_X(x) of the process values and its derivative, with the upcrossing rate given by Rice's formula:

\nu(u) = \int_{0}^{\infty} \dot{x} \, f_{X, \dot{X}}(u, \dot{x}) \, d\dot{x}

Here, f_{X, \dot{X}}(u, \dot{x}) is the joint probability density function of the process X(t) and its time derivative \dot{X}(t) evaluated at the same instant, and the integration over positive velocities \dot{x} > 0 accounts for upward crossings only. This integral computes the expected intensity of level crossings by weighting the joint density with the crossing speed \dot{x}, ensuring the formula applies to arbitrary distributions without restricting to Gaussian forms.^[13] The derivation of Rice's formula proceeds from first principles by considering the expected number of upcrossings over an interval [0, T] as the time average of an indicator for crossings, leveraging ergodicity to equate it to the ensemble expectation. Specifically, the number of upcrossings N_u(T) can be expressed using a Dirac delta function approximation:

E[N_u(T)] = E\left[ \int_0^T \dot{X}(t) \delta(X(t) - u) \mathbf{1}_{\{\dot{X}(t) > 0\}} \, dt \right]

By stationarity and Fubini's theorem, this simplifies to T \int_0^\infty \dot{x} f_{X, \dot{X}}(u, \dot{x}) \, d\dot{x}, yielding the rate \nu(u) as T \to \infty. For narrow-band processes, where spectral energy is concentrated in a narrow frequency range, the crossings approximate a Poisson point process with independent events, justifying the formula under the assumption that inter-crossing intervals are exponentially distributed; this Poisson approximation holds when dependence decays sufficiently fast. In such cases, the upcrossing rate can be approximated as \nu(u) \approx \nu(0) \times P(\text{peak amplitude} > u), where \nu(0) is the zero-upcrossing rate related to the process's mean frequency, but the general Rice's formula remains the precise expression. Key assumptions underlying these formulations include stationarity (time-invariant statistics), ergodicity (time averages equal ensemble averages), and differentiability of the process (existence of \dot{X}(t)). Additionally, the independence of successive crossings is often invoked for the Poisson interpretation, particularly in narrow-band cases where clusters of crossings are minimal, ensuring the rate \nu(u) reliably predicts long-run exceedance behavior without overcounting dependent events.

Gaussian Process Case

In the context of the general formulation for the frequency of exceedance (FOE), which quantifies the expected rate of upcrossings of a threshold level by a stochastic process, the case of Gaussian processes admits explicit closed-form expressions due to the joint Gaussianity of the process and its derivative.^[13] These expressions, originally derived by Rice, rely on the properties of stationary Gaussian processes and provide a foundational tool for analyzing exceedance frequencies in fields requiring precise probabilistic modeling.^[13] For a zero-mean, unit-variance stationary Gaussian process X(t), the FOE \nu(u) of a threshold u simplifies to a particularly compact form under the normalization where the standard deviation of the process derivative \sigma_{\dot{X}} = 1:

\nu(u) = \frac{1}{2\pi} \exp\left(-\frac{u^2}{2}\right).

This expression represents the expected number of upcrossings per unit time and follows directly from evaluating the joint density of X(t) and \dot{X}(t) in Rice's integral formula, leveraging the independence of the sign of \dot{X}(t) from X(t) conditional on X(t) = u.^[13] In the more general case of a stationary Gaussian process with mean \mu and variance \sigma_X^2 > 0, the FOE incorporates these parameters along with the standard deviation of the derivative process \sigma_{\dot{X}}:

\nu(u) = \frac{1}{2\pi} \cdot \frac{\sigma_{\dot{X}}}{\sigma_X} \exp\left( -\frac{(u - \mu)^2}{2 \sigma_X^2} \right).

Here, \sigma_{\dot{X}} captures the dynamic behavior of the process and is determined by the second spectral moment of the power spectral density S(\omega), specifically \sigma_{\dot{X}}^2 = \int_0^\infty \omega^2 S(\omega) \, d\omega, assuming a one-sided spectral density normalized such that \sigma_X^2 = \int_0^\infty S(\omega) \, d\omega.^[13] This dependence on spectral moments underscores how the FOE reflects not only the marginal distribution of X(t) but also the correlation structure through the frequency content of the process, with higher-frequency components increasing \sigma_{\dot{X}} and thus the exceedance rate. These derivations assume wide-sense stationarity, ensuring constant mean and covariance, and Gaussian marginals for both X(t) and the mean-square differentiable \dot{X}(t), which guarantees the joint normality required for the closed-form evaluation.^[13] Violations of these conditions, such as non-stationarity or non-Gaussianity, necessitate alternative approaches beyond this Gaussian specialization.

Probability and Time Relations

Probability of Exceedance

The probability of exceedance (POE) quantifies the likelihood that a stochastic process X(t), such as peak discharge or ground motion, surpasses a specified threshold u at least once within a finite time interval [0, T]. Formally, it is defined as P(X(t) > u \text{ for some } t \in [0, T]). Under the assumption of a stationary Poisson process for exceedance occurrences, where events are independent and occur at a constant mean rate \nu(u) (the frequency of exceedance, or FOE), the POE is approximated by $1 - \exp(-\nu(u) T). This formulation arises from the cumulative distribution function of the Poisson process, which models the number of exceedances as a Poisson random variable with parameter \lambda = \nu(u) T, yielding the probability of at least one event as $1 - e^{-\lambda}.^[14] For rare events, where the product \nu(u) T \ll 1, the exact POE simplifies to the linear approximation \nu(u) T. This approximation holds because the higher-order terms in the Taylor expansion of the exponential become negligible, providing a practical estimate when exceedances are infrequent relative to the observation period. In such cases, the POE directly scales with the exposure time T and the underlying FOE \nu(u).^[14] The POE relates closely to the return period R(u) = 1 / \nu(u), which represents the average recurrence interval between exceedances of threshold u. Substituting into the Poisson model gives POE(T) = 1 - \exp(-T / R(u)), linking probabilistic risk directly to the reciprocal of the FOE. This relationship is foundational in hazard analysis, enabling the conversion between annual rates and finite-time probabilities.^[14] In discrete-time settings, such as annual maximum series in flood frequency analysis, the number of exceedances over n years can be modeled using a binomial distribution with parameters n (number of trials) and success probability p = 1 / R(u), yielding POE(n) = 1 - (1 - p)^n. For rare events where p is small and n p is moderate, the binomial converges to the Poisson distribution, justifying the continuous-time approximation. Studies comparing these models in peaks-over-threshold analyses confirm that the Poisson provides a robust fit without significant improvement from the binomial for typical flood data.^[15]

Exceedance Over Time

In stationary stochastic processes, the frequency of exceedance extends to temporal dynamics by characterizing long-term behaviors such as the intervals between exceedances and the durations of excursions above a threshold u. This contrasts with the short-term probability of exceedance, which focuses on instantaneous likelihoods over finite intervals. For long-term analysis, key metrics include the expected time to the first exceedance and the mean duration of such events, which inform reliability assessments in systems subject to random fluctuations. The expected time to the first exceedance, E[T_{\text{first}}], for a stationary process equals the reciprocal of the mean upcrossing rate \nu(u), yielding E[T_{\text{first}}] = 1 / \nu(u). This follows from renewal theory, where upcrossings form a renewal process in the long run, and the mean inter-arrival time matches the mean recurrence time. For stationary Gaussian processes, \nu(u) is given by Rice's formula, \nu(u) = \frac{1}{2\pi} \sqrt{-\rho''(0)} \exp\left( -\frac{u^2}{2\sigma^2} \right), where \rho(\tau) is the autocorrelation function (normalized so \rho(0) = 1) and the exponential term reflects the Gaussian density tail, but the reciprocal relation holds generally for the first exceedance expectation.^[16] The mean exceedance duration, or average sojourn time above u, is d(u) = P(X > u) / \nu(u) for stationary processes, representing the expected length of time the process remains above the threshold per excursion. In the Gaussian case, this can be expressed as an integral involving the joint exceedance probability: d(u) = \int_0^\infty P(X(t) > u, X(0) > u) / P(X(0) > u) \, dt, conditional on the process starting above u, which leverages the bivariate normal distribution for computation. This formula highlights how correlation structure governs persistence above high thresholds. Over infinite time horizons, the cumulative number of exceedances in stationary processes yields an infinite expected total, as the expected count is \nu(u) \times \infty; however, practical analyses limit observations to finite durations T, resulting in an expected \nu(u) T crossings, which remains finite. For sojourn times above the threshold, a Markov approximation simplifies calculations by modeling the post-upcrossing process as a Markov chain or diffusion, ignoring distant dependencies for high u, where the conditional process behaves like an Ornstein-Uhlenbeck process scaled by the threshold. This approximation proves effective for tail events in Gaussian settings, enabling closed-form estimates of duration distributions.

Applications

Hydrology and Risk Analysis

In hydrology, frequency of exceedance (FOE) plays a central role in water resource management by quantifying the likelihood of extreme hydrologic events, enabling informed decisions on flood control, reservoir operations, and drought mitigation. This approach integrates historical data with probabilistic models to assess risks, where FOE represents the rate at which a specified threshold—such as peak streamflow or low-flow volume—is surpassed, providing a foundation for long-term planning in variable climates.^[17] Flood frequency analysis relies on FOE to estimate annual exceedance probability (AEP), a key metric for designing infrastructure like dams to withstand rare but impactful events. For example, a 1% AEP flood, equivalent to an FOE of ν = 0.01 per year, serves as a benchmark for spillway capacity and embankment stability in many regulatory frameworks.^[18] Empirical methods for this analysis involve fitting FOE curves to records of annual maximum streamflows, commonly using the Gumbel distribution for its simplicity in modeling extreme values or the Log-Pearson Type III distribution for its flexibility with skewed hydrologic data. In the United States, the Log-Pearson Type III is the preferred method for federal flood studies, as it accommodates the log-transformed peaks typical of riverine flows.^[19] FOE extends to drought frequency assessment by evaluating the probability of non-exceedance for minimum flows or volumes, aiding in the sizing of water supply reservoirs and irrigation systems during prolonged dry periods. For instance, thresholds for 95% non-exceedance flows guide allocations to ensure reliability under scarcity.^[20] Climate change complicates these applications by inducing non-stationarity in hydrologic records, prompting adjustments to FOE models—such as incorporating time-varying parameters for precipitation trends—to better project future flood and drought risks.^[21] A prominent application is in U.S. Federal Emergency Management Agency (FEMA) guidelines for the National Flood Insurance Program, established in 1968 and operational with mapping from the early 1970s, which employs 1% AEP derived from FOE to define base flood elevations and delineate special flood hazard areas for insurance rate-setting and community planning.

Engineering and Reliability

In seismic engineering, the frequency of exceedance (FOE) is integral to probabilistic seismic hazard analysis (PSHA), where it quantifies the annual likelihood of peak ground acceleration (PGA) surpassing specified thresholds, informing building code requirements for structural resilience. For instance, the ASCE 7 standard employs hazard curves derived from PSHA to establish design ground motions, typically targeting a 2% probability of exceedance in 50 years (equivalent to an annual FOE of approximately 4 × 10^{-4}) for PGA and spectral accelerations, ensuring buildings can withstand rare but intense events without collapse.^[22] This approach allows engineers to map site-specific risks, balancing safety against overdesign by integrating fault data, attenuation models, and uncertainty quantification.^[23] In wind and fatigue analysis for structures like bridges, FOE evaluates the recurrence of stress levels exceeding material limits, often combined with rainflow counting to tally cyclic loads from gusts and turbulence, predicting cumulative damage over the structure's lifespan. Rainflow methods extract equivalent stress cycles from time histories of wind-induced responses, enabling fatigue life estimates where high-cycle, low-amplitude events contribute to long-term degradation; for example, AASHTO guidelines use such counts to verify that the FOE of critical stress ranges aligns with allowable limits for infinite fatigue life, typically below 10^{-6} annually.^[24] In performance-based wind engineering, FOE curves describe the mean annual rate of exceeding wind speed or pressure thresholds, facilitating risk-informed designs for slender elements prone to vortex shedding or buffeting.^[25] Gaussian processes may model spatial wind load variability briefly in these assessments, capturing correlation in turbulent fields across bridge spans. The reliability function in engineering systems frames failure rates directly as the FOE of applied loads exceeding structural capacity, providing a probabilistic measure of dependability under variable demands. This formulation integrates load and resistance factor design (LRFD) principles, where the annual FOE of the limit state function g = capacity - load < 0 determines the system's safety index, often calibrated to achieve target reliabilities like β = 3.0 for a 10^{-3} annual failure probability.^[26] Such metrics guide redundancy and ductility provisions, ensuring that the convolution of load FOE distributions with capacity variabilities yields acceptable downtime risks for critical infrastructure.^[27] The 1994 Northridge earthquake prompted significant reassessments of seismic risk models, incorporating observed ground motions to refine FOE estimates for blind thrust faults in urban areas. Post-event analyses revealed that actual PGA levels exceeded pre-1994 PSHA predictions by factors of 1.5–2.0 in the epicentral region, leading to updated USGS hazard maps that increased annual FOE values for 0.2g accelerations from ~10^{-3} to higher rates in southern California, influencing subsequent code revisions for enhanced collapse prevention.^[28] This reassessment underscored the need for empirical validation of FOE curves, improving long-term reliability projections for seismically active regions.^[29]

Extensions and Limitations

Non-Stationary Processes

In non-stationary processes, the frequency of exceedance (FOE) evolves over time due to trends in the underlying distribution, such as shifting means or variances driven by climate change or other covariates, extending the stationary baseline where exceedances occur at a constant rate. In non-stationary processes, the frequency of exceedance varies over time due to changes in the underlying distribution, such as time-varying parameters in extreme value models or point process approaches in peaks-over-threshold frameworks.^[30] This adjustment ensures that the rate of threshold exceedances reflects non-homogeneous conditions, as opposed to the constant \nu(u) in stationary cases.^[31] To handle autocorrelation in non-stationary series, declustering methods identify independent extreme events by applying time-based rules, such as runs declustering with a fixed interval (e.g., 7 days for hydrological extremes) to separate clusters, thereby enabling valid extreme value analysis despite temporal dependence.^[30] Bayesian updating further addresses non-stationarity by incorporating covariates into regression models for distribution parameters (e.g., location and scale in generalized extreme value distributions), allowing posterior distributions to evolve with new observations and quantify predictive uncertainties in hydrological applications like flood risk assessment.^[32] An illustrative example is the increasing FOE of high river flows attributed to anthropogenic climate change, where observed trends show a high-confidence rise in flood magnitude and frequency in regions like northern mid-latitudes and the tropics, with global flood exposure increasing by 20–24% from 2000–2018.^[33] These shifts, projected to intensify under warming scenarios, highlight the need for non-stationary models in water resource management. Recent studies as of 2025 continue to emphasize the role of non-stationarity in extreme event modeling, with ongoing refinements in IPCC assessments.^[34] Key challenges in extending FOE to non-stationary processes include the breakdown of the Poisson assumption for independent exceedances, as trends introduce clustering and dependence that stationary models overlook, potentially underestimating risks.^[30] Additionally, incorporating covariates like the El Niño-Southern Oscillation (ENSO) is essential to capture oscillatory influences on extremes, yet selecting appropriate ones remains complex due to interactions with long-term trends and limited record lengths that hinder robust attribution.^[31]

Computational Methods

Monte Carlo simulation provides an empirical approach to estimate the frequency of exceedance, denoted as \nu(u), by generating multiple realizations of the underlying stochastic process and counting the number of upcrossings of the level u across each realization. This method is particularly useful for complex, non-linear, or non-stationary processes where analytical solutions are intractable, allowing practitioners to approximate \nu(u) as the average number of upcrossings per unit time over a large number of simulations. For instance, in estimating extreme response statistics of dynamical systems, Monte Carlo techniques exploit the mean level upcrossing rate to compute tail distributions efficiently.^[35]^[36] Integration of extreme value theory enhances tail estimation of \nu(u) by applying the block maxima method, where data are divided into non-overlapping blocks, the maximum value in each block is extracted, and these maxima are fitted to a generalized extreme value (GEV) distribution to model the tail behavior. This approach indirectly informs the frequency of exceedance by relating the GEV parameters to the rate of extreme events, especially for long-term predictions in environmental data where block sizes are chosen based on physical timescales, such as annual maxima in hydrology. The resulting GEV fit allows extrapolation of \nu(u) for rare high thresholds beyond observed data.^[37] Software tools facilitate practical implementation of these methods. The R package extRemes (version 2.2-1 as of 2024), which was redesigned in version 2.0 in 2016, supports block maxima fitting via the GEV distribution and peaks-over-threshold modeling, enabling estimation of exceedance frequencies through functions like fevd for parameter inference and return.level for related metrics. In Python, while SciPy does not provide a built-in function for the Rice formula, users can implement it numerically using scipy.stats.norm for Gaussian joint densities to compute upcrossing rates for stationary Gaussian processes. These tools streamline simulations and fittings, with extRemes particularly suited for univariate extreme value analysis in climate and weather applications. Validation of computational methods often involves comparing Monte Carlo estimates of \nu(u) to analytical results derived from the Rice formula for benchmark Gaussian processes, confirming accuracy in controlled settings. For example, simulations of stationary Gaussian fields show that Monte Carlo upcrossing counts converge to Rice formula predictions as the number of realizations increases, with relative errors typically below 5% for moderate sample sizes. Such comparisons underscore the reliability of empirical methods for more general cases.^[38]^[39]

References

[1]
Exceedance - an overview | ScienceDirect Topics
Exceedance refers to the probability that a natural extreme event will exceed a specified intensity over a given period, highlighting the relationship ...
[2]
Probability Density Function - Hydrologic Engineering Center
Exceedance means the probability that flow is greater than a value. The complement of the flow frequency curve has notation F(x) and is called a non-exceedance ...
[3]
[PDF] Peak-Flow Frequency Estimates for U.S. Geological Survey ...
Exceedance probabilities define the average length of time that separates peak flows of a given magnitude, but it is possible that such flows could occur during ...
[4]
Section 2: Probability of Exceedance
The probability of exceedance describes the likelihood of a specified flow rate (or volume of water with specified duration) being exceeded in a given year.
[5]
[PDF] OVERVIEW OF THE PROBABILISTIC SEISMIC HAZARD ANALYSES
Nov 30, 2006 · For ground motion, the target hazard for. Category-1 events has been established at 10-3 annual frequency of exceedance. For Category-2.
[6]
[PDF] Chapter R20 Probabilistic Seismic Hazard Analysis
uncertainties is a quantification of the uncertainty in the frequency of exceedance of earthquake ground motions (McCann, JBA). In the same manner there is ...
[7]
[PDF] Exceedance Probability in Catastrophe Modeling
Jul 14, 2020 · Exceedance Probability (EP) is the probability that a certain loss value will be exceeded in a predefined future time period.
[8]
[PDF] Safety Margins for Flight Through Stochastic Gusts
May 9, 2014 · interarrival time is the reciprocal of the frequency of exceedance, so. Table 1. Estimated and simulated values of various airspeed statistics ...
[9]
Recurrence intervals between exceedances of selected river levels ...
Analysis of the frequency of exceedance of low river levels requires the examination of conditional probabilities. A first-order Markov model is selected as ...
[10]
[PDF] U.S. Geological Survey Scientific Investigations Report 2008-5119
tics, to their expected frequency of exceedance, is termed a flood-frequency analysis. The frequency of peak discharges generally is expressed in terms of ...Missing: origin | Show results with:origin
[11]
Flood-frequency estimation for very low annual exceedance ...
Aug 25, 2020 · A particularly difficult assumption to evaluate for flood-frequency analysis is the underlying assumption that the flood series is stationary— ...
[12]
[PDF] CHAPTER 810 – HYDROLOGY - Topic 811 – General - Caltrans
Jul 1, 2020 · Application of traditional predetermined design flood frequencies implies that an acceptable level of risk was considered in establishing the ...
[13]
Why do seismic hazard models worldwide appear to overpredict ...
May 1, 2024 · ... per year—a dimensional quantity with the unit of 1/yr—and RP has ... frequency of exceedance) is equal to the reciprocal of RP with the ...
[14]
Mathematical analysis of random noise - Nokia Bell Labs
THIS paper deals with the mathematical analysis of noise obtained by passing random noise through physical devices. The random noise considered is that ...
[15]
[PDF] Open-File Report 2008–1128 - USGS Publications Warehouse
To obtain a probability from an annual frequency of exceedance we apply the Poisson equation (time independent). ... probability of exceedance in 50 years for the ...
[16]
Full article: Comparison between the peaks-over-threshold method ...
The POT method gave better results than the AM method. The binomial distribution did not offer any noticeable improvement over the Poisson distribution for ...
[17]
[PDF] NO, NsG-466 - NASA Technical Reports Server
For stationary processes the first-occurrence time density is closely ... time density po(a, t) are given below for some stationary Gaussian processes ...
[18]
[PDF] ESTIMATING MAGNITUDE AND FREQUENCY OF FLOODS USING ...
FLOOD-FREQUENCY ANALYSIS. Flood-frequency analysis provides information about the magnitude and frequency of selected flood discharges.
[19]
[PDF] Probabilistic Extreme Flood Hydrographs That Use PaleoFlood Data ...
An extreme flood is considered to have an Annual Exceedance Probability (AEP) of 0.005 or less. There are many methods of estimating extreme flood runoff ...
[20]
[PDF] Guidelines for Determining Flood Flow Frequency Bulletin 17C
This series of manuals on Techniques and Methods (TM) describes approved scientific and data-collection procedures and standard methods for planning and ...Missing: Poisson | Show results with:Poisson
[21]
On the need for streamflow drought frequency guidelines in the U.S.
Nov 12, 2021 · We provide a justification for the need for developing national guidelines for streamflow drought frequency analysis as an analog to the ...
[22]
Chapter: 4 Climate and Floods: Role of Non-Stationarity
Flood frequency analysis, as traditionally practiced, is marked by an assumption that annual maximum floods conform to a stationary, independent, ...
[23]
[PDF] Guidelines for Performance- Based Seismic Design of Tall Buildings
May 3, 2017 · Hazard Curve – A plot of the mean annual frequency of exceedance of a ground motion intensity parameter as a function of the ground motion ...
[24]
[PDF] Earthquakes and Seismic Design
The vertical axis of this plot presents the annual frequency of exceedance of peak ground accelerations of different amounts, shown along the horizontal axis.
[25]
[PDF] Manual for Repair and Retrofit of Fatigue Cracks in Steel Bridges
Table 4 Stress Ranges Collected Through Rainflow Counting for ... AASHTO LRFD provisions clearly has a much higher frequency of exceedance than 0.01%.
[26]
[PDF] A probabilistic framework for Performance-Based Wind Engineering
A modern approach to wind engineering should consider performances as key objectives of structural ... of the mean annual frequency of exceedance: γ. IP. IM γ. IM.
[27]
[PDF] Earthquake Reliability of Structures
Building design codes achieve a preselected reliability by the use of partial load and capacity reduction ... Defining HD as the annual frequency of exceedance of ...
[28]
[PDF] A Technical Framework for Probability-Based Demand and Capacity ...
Nov 8, 2003 · frequency of exceedance (MAF) for the displacement-based demand or the drift hazard and (2) by deriving the conditional probability that ...
[29]
Earthquake Hazards Assessment - USGS.gov
The Northridge earthquake had two primary effects on seismic-hazards maps for California. First, it emphasized the importance of blind thrust faults to seismic ...
[30]
Preliminary seismic hazard assessment for Los Angeles, Ventura ...
Jul 14, 2017 · The seismic ground motion hazard is assessed for a 10% probability of exceedance in 50 years for the three counties (Los Angeles, Ventura, ...
[31]
Nonstationary weather and water extremes: a review of methods for ...
Jul 7, 2021 · Here we provide a review of the drivers, metrics, and methods for the detection, attribution, management, and projection of nonstationary hydroclimatic ...
[32]
https://link.springer.com/chapter/10.1007/978-94-007-4479-0_3
[33]
Bayesian Methods for Non-stationary Extreme Value Analysis
Bayesian inference offers an attractive framework to estimate non-stationary models and, importantly, to quantify estimation and predictive uncertainties.Missing: updating hydrology
[34]
Chapter 4: Water | Climate Change 2022: Impacts, Adaptation and ...
Climate change or variability is the main contributor to changes in basin-scale trends for 75% of rivers, while direct human effects on streamflow dominate for ...
[35]
Monte Carlo Methods for Estimating the Extreme Response of ...
Apr 22, 2025 · The key quantity for calculating the statistical distribution of extreme response is the mean level upcrossing rate function. By exploiting the ...
[36]
Prediction of Extreme Response Statistics of Narrow-Band Random ...
One of the widely used approaches for extreme response estimation is the Poisson assumption, that is, independence of upcrossing events above high thresholds.
[37]
Extreme-value analysis for the characterization of extremes in water ...
For the stationary fit with threshold = 6 mm, the frequency of exceedance is about 4 days per season. ... extRemes 2.0: an extreme value analysis package in R. J.
[38]
[PDF] Gaussian integrals and Rice series in crossing distributions to ...
Jul 11, 2018 · Rice's formula for the expected number, and higher moments, of level crossings by a Gaussian process stand up to contemporary numer- ical ...
[39]
Fig. 4. Comparison between analytical estimates and Monte Carlo...
The accuracy of the extended VanMarcke formula in estimating this probability is similar to that of VanMarcke formula for the marginal probability estimates.