Hydrological Statistics and Frequency Analysis
Hydrological statistics applies probability and stochastic methods to hydrological data to characterize variability and to estimate the frequency of extremes such as floods and droughts.
Definition
Hydrological statistics and frequency analysis is the application of probability theory and statistics to hydrological data to describe their variability and to estimate the magnitude and probability of events, especially extremes, for design and risk assessment.
Scope
This topic covers probability distributions for hydrological variables, parameter estimation including L-moments, frequency and regional frequency analysis of extremes, and the assumption of stationarity. It provides the statistical backbone for design values used across hydrology, including flood and drought estimation.
Core questions
- How are hydrological variables described by probability distributions?
- How are distribution parameters estimated reliably from short records?
- How is frequency analysis extended regionally and to ungauged sites?
- Is the assumption of stationarity valid under change?
Key concepts
- Probability distributions in hydrology
- Return period and quantiles
- Parameter estimation and L-moments
- Regional frequency analysis
- Stationarity and non-stationarity
- Stochastic hydrology
Key theories
- Frequency analysis of extremes
- Hydrological extremes are modeled with probability distributions whose quantiles give design values; sound practice addresses distribution choice, parameter estimation, and the treatment of outliers and short records.
- Regional frequency analysis with L-moments
- Pooling data from many sites and using L-moments yields more robust estimates of extreme quantiles than at-site analysis, improving estimation at sites with short or no records.
- Non-stationarity
- Climate and land-use change can violate the stationarity assumption underlying traditional frequency analysis, prompting calls to develop methods that account for trends and changing risk.
Clinical relevance
Hydrological statistics supplies the design floods, low flows, and rainfall values used to size and regulate infrastructure, price flood insurance, and plan water resources; the debate over stationarity directly affects how these design values are estimated under a changing climate.
History
Statistical hydrology grew with extreme-value theory and lengthening records through the 20th century; L-moment regional methods improved estimation in the 1990s, and the 2008 argument that 'stationarity is dead' crystallized concern that climate change undermines a core assumption of frequency analysis.
Debates
- Stationarity under climate change
- A central debate is whether the long-standing assumption of stationarity remains tenable for design, and if not, how to incorporate non-stationarity and deep uncertainty into frequency analysis and water management.
Key figures
- Jery R. Stedinger
- Jonathan R. M. Hosking
- P. C. D. Milly
Related topics
Seminal works
- stedinger1993
- hosking1997
- milly2008
Frequently asked questions
- Why use regional frequency analysis?
- Individual sites often have short records, making estimates of rare events unreliable; pooling data from hydrologically similar sites, for example with L-moments, borrows information across the region to produce more stable estimates of extreme quantiles.
- What does 'stationarity is dead' mean for hydrology?
- It expresses the concern that climate and land-use change make the past no longer a reliable guide to the future, so frequency analyses assuming an unchanging probability distribution may misstate risk, motivating non-stationary and scenario-based approaches.