Skip to main content
Log in

Latent Process Modelling of Threshold Exceedances in Hourly Rainfall Series

  • Published:
Journal of Agricultural, Biological, and Environmental Statistics Aims and scope Submit manuscript

Abstract

Two features are often observed in analyses of both daily and hourly rainfall series. One is the tendency for the strength of temporal dependence to decrease when looking at the series above increasing thresholds. The other is the empirical evidence for rainfall extremes to approach independence at high enough levels. To account for these features, Bortot and Gaetan (Scand J Stat 41:606–621, 2014) focus on rainfall exceedances above a fixed high threshold and model their dynamics through a hierarchical approach that allows for changes in the temporal dependence properties when moving further into the right tail. It is found that this modelling procedure performs generally well in analyses of daily rainfalls, but has some inherent theoretical limitations that affect its goodness of fit in the context of hourly data. In order to overcome this drawback, we develop here a modification of the Bortot and Gaetan model derived from a copula-type technique. Application of both model versions to rainfall series recorded in Camborne, England, shows that they provide similar results when studying daily data, but in the analysis of hourly data the modified version is superior.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6

Similar content being viewed by others

References

  • Bortot, P., and Gaetan, C. (2014), “A latent process model for temporal extremes,” Scandinavian Journal of Statistics, 41, 606–621.

  • Bortot, P., and Tawn, J. (1998), “Models for the extremes of Markov chains,” Biometrika, 85, 851–867.

  • Casciani, M. (2015), Analisi dei valori estremi di serie storiche: un approccio bayesiano,, Master’s thesis, Facoltà di Economia, Università degli Studi di Roma “La Sapienza”, Rome, Italy.

  • Chavez-Demoulin, V., and Davison, A. C. ( 2012), “Modelling time series extremes,” REVSTAT - Statistical Journal, 10, 109–133.

  • Coles, S. G. (2001), An Introduction to Statistical Modeling of Extreme Values, New York: Springer.

  • Coles, S. G., Tawn, J. A., and Smith, R. L. ( 1994), “A seasonal Markov model for extremely low temperatures,” Environmetrics, 5, 221–239.

  • Davis, R., and Yau, C.-Y. (2011), “Comments on pairwise likelihood in time series models,” Statistica Sinica, 21, 255–277.

  • Davison, A. C., and Smith, R. L. ( 1990), “Models for exceedances over high thresholds,” Journal of Royal Statistical Society: Series B, 3, 393–442.

  • de Haan, L. (1984), “A spectral representation for max-stable processes,” The Annals of Probability, 12, 1194–1204.

  • Eastoe, E. F., and Tawn, J. A. (2009) , “Modelling non-stationary extremes with application to surface level ozone,” Journal of the Royal Statistical Society: Series C (Applied Statistics), 58, 25–45.

  • Ferro, C. A. T., and Segers, J. ( 2003), “Inference for clusters of extreme values,” Journal of Royal Statistical Society: Series B, 65, 545–556.

  • Gaver, D., and Lewis, P. (1980), “First-order autoregressive Gamma sequences and point processes,” Advances in Applied Probability, 12, 727–745.

  • Huser, R., and Davison, A. C. (2014) , “Space-time modelling of extreme events,” Journal of the Royal Statistical Society: Series B, 76, 439–461.

  • Jonathan, P., and Ewans, K. (2013), “Statistical modelling of extreme ocean environments for marine design: a review,” Ocean Engineering, 62, 91–109.

  • Katz, R. W., Parlange, M. B., and Naveau, P. ( 2002), “Statistics of extremes in hydrology,” Advances in Water Resources, 25, 1287–1304.

  • Koutsoyiannis, D. (2004), “Statistics of extremes and estimation of extreme rainfall: II. Empirical investigation of long rainfall records / Statistiques de valeurs extrêmes et estimation de précipitations extrêmes: II. Recherche empirique sur de longues séries de précipitations,” Hydrological Sciences Journal, 49, 591–610.

  • Leadbetter, K. R., Lindgren, G., and Rootzén, H. ( 1983), Extremes and Related Properties of Random Sequences and Processes, Berlin: Springer.

  • Ledford, A. W., and Tawn, J.A. ( 1997), “Modelling dependence within joint tail regions,” Journal of the Royal Statistical Society: Series B, 59, 475–499.

  • —— 2003. “Diagnostics for dependence within time series extremes,” Journal of Royal Statistical Society: Series B, 65, 521–543.

  • Lindsay, B. (1988), “Composite likelihood methods,” Contemporary Mathematics, 80, 221–239.

  • Marin, J.-M., Pudlo, P., Robert, C. P., and Ryder, R. J. (2011), “Approximate Bayesian computational methods,” Statistics and Computing, 22, 1167–1180.

  • Pickands, J. (1975), “Statistical inference using extreme order statistics,” The Annals of Statistics, 3, 119–131.

  • Raillard, N., Ailliot, P., and Yao, J. ( 2014), “Modeling extreme values of processes observed at irregular time steps: Application to significant wave height,” The Annals of Applied Statistics, 8, 622–647.

  • Reich, B. J., Shaby, B. A., and Cooley, D. ( 2014), “A hierarchical model for serially-dependent extremes: a study of heat waves in the Western US,” Journal of Agricultural, Biological, and Environmental Statistics, 19, 119–135.

  • Reiss, R., and Thomas, M. (2007), Statistical Analysis of Extreme Values, third edn, Basel: Birkhäuser.

  • Robinson, M. E., and Tawn, J. A. ( 2000), “Extremal analysis of processes sampled at different frequencies,” Journal of the Royal Statistical Society. Series B (Statistical Methodology), 62, 117–135.

  • Smith, R. L. (1990), “Regional estimation from spatially dependent data,” Preprint, University of North Carolina.

  • Smith, R., Tawn, J. A., and Coles, S. ( 1997), “Markov chain models for threshold exceedances,” Biometrika, 84, 249–268.

  • Varin, C., and Vidoni, P. (2005), “A note on composite likelihood inference and model selection,” Biometrika, 52, 519–528.

  • Walker, S. (2000), “A note on the innovation distribution of a Gamma distributed autoregressive process,” Scandinavian Journal of Statistics, 27, 575–576.

  • Warren, D. (1992), “A multivariate Gamma distribution arising from a Markov model,” Stochastic Hydrology and Hydraulics, 6, 183–190.

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Carlo Gaetan.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary material 1 (pdf 27 KB)

Appendix

Appendix

Let \(LP_{a}^{(1)}(v)=\text {E}\left( e^{-v\Lambda _t}\right) \) and \( LP_{a;t'-t}^{(2)}(v_1,v_2)= \text {E}\left( e^{-v_1 \Lambda _{t}-v_2\Lambda _{t'}}\right) \), \(t' > t\) be the univariate and bivariate Laplace transform, respectively, of \(\{\Lambda _t\}\) under specification a, with a = GL or W. Bortot and Gaetan (2014) show that, for \(\alpha =1/\xi \) and \(\beta =\sigma /\xi \),

$$\begin{aligned} LP_{a}^{(1)}(v)=\left( \frac{\beta }{\beta +v}\right) ^{\alpha }, \end{aligned}$$

regardless of a, while

$$\begin{aligned} LP_{\tiny {\text{ GL }}; t'-t}^{(2)}(v_1,v_2)=\left( \frac{(\beta +\rho ^{t'-t}v_{2})\beta }{(\beta +v_{2})(\beta +v_1+\rho ^{t'-t}v_{2})}\right) ^\alpha , \end{aligned}$$

and

$$\begin{aligned} LP_{\tiny {\text{ W }};t'-t}^{(2)}(v_1,v_2)=\left[ 1+(v_1+v_2)/\beta +(1-\rho ^{t'-t})v_1v_2/\beta ^2\right] ^{-\alpha }. \end{aligned}$$

Let \(f_a(y_t,y_{t'};\psi )\), with \(\psi =(\xi ,\sigma ,\rho ,\kappa )\), be the contribution of \((y_t,y_{t'})\) to the censored pairwise likelihood under model \(M_a\). We have

Let \(f^*_a(y_t,y_{t'};\psi ^*)\), with \(\psi ^*=(\xi ^*,\sigma ^*,\rho ,\kappa )\), be the censored pairwise likelihood contribution of \((y_t,y_{t'})\) under model \(M^*_a\), and let \(h_a(y,y')=f_a(y,y';\psi _0)\), with \(\psi _0=(1,1,\rho ,\kappa )\). Then,

$$\begin{aligned} f^*_a(y_t,y_{t'};\psi ^*)=\left\{ \begin{array}{lc} h_a(s(y_t),s(y_{t'}))s'(y_t)s'(y_{t'})&{} y_t> 0, y_{t'}>0\\ h_a(s(y_t),0)s'(y_t)&{} y_t> 0, y_{t'} = 0\\ h_a(0,s(y_{t'}))s'(y_{t'})&{} y_t = 0, y_{t'} >0\\ h_a(0,0)&{} y_t = 0, y_{t'} = 0\\ \end{array} \right. , \end{aligned}$$

where

$$\begin{aligned} s(y)= (\kappa +1)\left\{ \left( 1+\frac{\xi ^* y}{\sigma ^*}\right) ^{1/\xi ^*}-1\right\} , \end{aligned}$$

and

$$\begin{aligned} s'(y)=\frac{\kappa +1}{\sigma ^*}\left( 1+\frac{\xi ^* y}{\sigma ^*}\right) ^{1/\xi ^*-1}. \end{aligned}$$

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Bortot, P., Gaetan, C. Latent Process Modelling of Threshold Exceedances in Hourly Rainfall Series. JABES 21, 531–547 (2016). https://doi.org/10.1007/s13253-016-0254-5

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s13253-016-0254-5

Keywords

Navigation