2011年-世界发展银行全球_Uncertainty_and_Climate_Variability_in_the_Design_and_Operation_of_Water_Resources_Projects___Examples_and_Case_Studies_Volume_2_Annexes_36页_2mb

Summary of UNCERTAINTY AND CLIMATE VARIABILITY IN THE DESIGN AND OPERATION OF WATER RESOURCES PROJECTS

Core Content

This document explores the challenges of uncertainty and climate variability in the design and operation of water resources projects, focusing on statistical methods and tools used for flood frequency analysis and climate reconstruction.

Main Topics and Key Points

1. Statistical Methods for Extreme Values

• Common Problems in Flood Hydrology:

  • Estimate the return period for a given flood.
  • Estimate the flood for a given return period.

• Statistical Distributions for Maxima:

  • Gumbel PDF: Used for modeling extreme events. Its form is:
    $$
    F(x_T) = \exp \left{- \exp \left{- \alpha (x_T - u) \right} \right}
    $$
    with parameters $ \alpha $ (shape) and $ u $ (location).
  • Pearson Type III (P3) PDF: Also known as the log-Pearson Type III (LP3) distribution. It is a three-parameter distribution with parameters $ \xi $ (location), $ \alpha $ (scale), and $ k $ (shape). It is widely used in the US for flood frequency analysis.
  • Generalized Extreme Value (GEV) PDF: Combines the three extreme value distributions (Type I, II, and III). It has three parameters: $ \xi $ (location), $ \alpha $ (scale), and $ k $ (shape). The GEV PDF is:
    $$
    F(x) = \exp \left[ -1 \cdot \left( 1 - \frac{k(x - \xi)}{\alpha} \right)^{1/k} \right], \quad \text{if } k \neq 0
    $$
    • If $ k = 0 $, it reduces to the Gumbel PDF.
    • If $ k < 0 $, it corresponds to the Weibull distribution.
    • If $ k > 0 $, it corresponds to the Fréchet distribution.
  • Other Distributions: Includes Generalized Logistic (GLO), Generalized Pareto (GPO), lognormal, and gamma distributions, though they are less commonly used compared to the above.

2. Parameter Estimation Methods

• Method of Moments (MoM):

  • Used for estimating parameters of distributions.
  • For Gumbel PDF:
    $$
    \alpha = \mu - 0.5772 \cdot \beta, \quad \beta = \frac{\sqrt{6} \cdot \sigma}{\pi}
    $$
  • For LP3:
    • If location parameter $ \xi $ is known, parameters $ \alpha $ and $ k $ can be estimated from the mean and standard deviation of $ \ln(x - \xi) $.
    • If $ \xi $ is unknown, the following equations are used:
      $$
      \widehat{k} = 4 G_s^{-2}, \quad \hat{a} = s G_s / 2, \quad \hat{\xi} = m - \hat{k} \cdot \hat{a}
      $$
      where $ m $ is the sample mean, $ s $ is the standard deviation, and $ G_s $ is the sample skewness.

• L-Moments Method (LMoM):

  • A robust method for estimating distribution parameters using linear combinations of order statistics.
  • L-moments are defined as:
    $$
    \lambda_r = \frac{1}{r} \sum_{k=0}^{r-1} (-1)^k \binom{r-1}{k} \cdot E{X_{r-k:r}}
    $$
    for $ r = 1, 2, 3, \dots $
  • For Gumbel distribution:
    $$
    \alpha = \lambda_2 / \log 2, \quad \xi = \lambda_1 - \gamma \cdot \alpha
    $$
    where $ \gamma $ is Euler’s constant (0.5772).

3. Mann-Kendall Non-Parametric Test

• Purpose: Detects trends in time series data without assuming a particular distribution.
• Procedure:
  • The test statistic $ S $ is calculated as:
    $$
    S = \sum_{k=1}^{n-1} \sum_{j=k+1}^{n} \operatorname{sign}(x_j - x_k)
    $$
  • If $ S > 0 $, there is an increasing trend; if $ S < 0 $, a decreasing trend.
  • If $ S = 0 $, no trend is detected.
  • Variance Calculation:
    $$
    \operatorname{VAR}(S) = \frac{n(n - 1)(2n + 5) - \sum_{t} t(t - 1)(2t + 5)}{18}
    $$
  • Z-Score Calculation:
    • If $ S > 0 $: $ Z = \frac{S - 1}{\sqrt{\operatorname{VAR}(S)}} $
    • If $ S = 0 $: $ Z = 0 $
    • If $ S < 0 $: $ Z = \frac{S + 1}{\sqrt{\operatorname{VAR}(S)}} $
  • Significance Testing: The Z-score is compared to a significance level (e.g., 95%), with negative Z indicating a decreasing trend and positive Z indicating an increasing trend.

4. Dendrochronology Reconstruction of Hydrologic Records

• Purpose: Extends hydrologic records using tree ring data (dendroclimatology), which provides high-resolution, annually resolved climate data.
• Methodology:
  • Principal Component Analysis (PCA): Used to reduce dimensionality and identify dominant climate patterns.
  • Orthogonal Spatial Regression (OSR): A method used to relate tree ring data to hydrologic variables like floods, droughts, and streamflow.
  • Reconstruction Process:
    • A vector $ \overline{Y} $ represents the standardized predictand.
    • A matrix $ \overline{TR} $ contains tree ring chronologies.
    • The correlation matrix $ \overline{\overline{C}} $ is analyzed to derive principal component vectors $ \overline{\overline{E}} $.
    • The OSR model is expressed as:
      $$
      \bar{Y} = \overline{\bar{U}} \cdot \bar{\beta} + \bar{\varepsilon}
      $$
      where $ \overline{\bar{U}} $ are the principal components, $ \bar{\beta} $ are the regression coefficients, and $ \bar{\varepsilon} $ is the error vector.
• Limitations:
  • Reconstructed indices have less variability than original indices (e.g., PDSI).
  • Only 47% of dry years are captured simultaneously by original and reconstructed PDSI data.
  • Regional skewness data is outdated (last updated in 1982), limiting its applicability in other regions.

Key Information

• The Gumbel PDF is widely used for modeling both maxima and minima.
• The GEV PDF is a more flexible distribution that includes the Gumbel, Weibull, and Fréchet distributions.
• The L-Moments method is preferred over traditional moments due to its resistance to outliers and unbiasedness.
• The Mann-Kendall test is a non-parametric tool for trend detection in hydrologic data.
• Dendrochronology provides a high-resolution, annual reconstruction of climate patterns, which is useful for hydrologic analysis but has limitations in capturing full variability.

References

• Gilbert, R. O. (1987). Statistical Methods for Environmental Pollution Monitoring.
• Greenwood, J. A., et al. (1979). Probability Weighted Moments.
• Hosking, J. R. M. (1990). L-moments Analysis and Estimation.
• Kendall, M. G. (1975). Rank Correlation Methods.
• Water Resources Council Hydrology Sub-Committee (1982). Bulletin 17B: Guidelines for Determining Flood Flow Frequency.
• Cook, E. R., et al. (1994). Dendrochronology.
• González, A., and Valdés, J. B. (2003). Methodology for converting reconstructed data into a form comparable with instrumental data.

Conclusion

The document emphasizes the importance of statistical methods in analyzing extreme hydrologic events and climate variability. It outlines the use of Gumbel, Pearson Type III, and Generalized Extreme Value distributions, along with parameter estimation techniques like MoM and L-Moments. Additionally, it introduces the Mann-Kendall test for trend detection and dendrochronology for climate reconstruction, highlighting the benefits and limitations of each approach in the context of water resources planning.