Non-stationary probabilistic downscaling of extreme precipitation

International Journal of Climatology, 2018

Credible information about the properties and changes of extreme events on the regional and local scales is of prime importance in the context of future climate change. Within the EU-COST Action VALUE a comprehensive validation framework for downscaling methods has been developed. Here we present validation results for extremes of temperature and precipitation from the perfect predictor experiment that uses reanalysis-based predictors to isolate downscaling skill. The raw reanalysis output reveals that there is mostly a large bias with respect to the extreme index values at the considered stations across Europe, clearly pointing to the necessity of downscaling. The performance of the downscaling methods is closely linked to their specific structure and setup. All methods using parametric distributions require non-standard distributions to correctly represent marginal aspects of extremes. Also, the performance is much improved by explicitly including a seasonal component, particularly in case of precipitation. With respect to the marginal aspects of extremes the best performance is found for model output statistics (MOS), weather generators (WGs) as well as perfect prognosis (PP) methods using analogues. Spell-length-related extremes of temperature are best assessed by MOS and WGs, spell-length-related extremes of precipitation by MOS and PP methods using analogues. The skill of PP methods with transfer functions varies strongly across the methods and depends on the extreme index, region and season considered.

Journal of Water and Climate Change, 2013

The paper describes downscaling of extreme precipitation in Ireland using a probabilistic method. The method described uses a combined peak-over-threshold (POT) – generalised Pareto distribution (GPD) approach in which the scale parameter of the GPD is allowed to vary with a dominant climate forcing at the location of interest. The dominant climatic forcing is represented by predictors selected from large-scale climatic variables provided by the NCEP/NCAR data. Data from six rainfall stations are used in the study to build the models for each station. The extRemes software is used to build the models as it allows parameters of the fitted distribution to vary as a function of covariate(s). The developed models were tested for goodness-of-fit with the observed data, and model fit was found to be much improved when the scale parameter was assumed to vary with the selected covariates. Return level – return period relations are developed based on the models developed and four future time...

Statistical models based on the scale-invariance (or scaling) concept has increasingly become an essential tool for modeling extreme rainfall processes over a wide range of time scales. In particular, in the context of climate change these scaling models can be used to describe the linkages between the distributions of sub-daily extreme rainfalls (ERs) and the distribution of daily ERs that is commonly provided by global or regional climate simulations. Furthermore, the Generalized Logistic distribution (GLO) has been recommended in UK for modeling of extreme hydrologic variables. Therefore, the main objective of the present study is to propose a scaling GLO model for modeling ER processes over different time scales. The feasibility and accuracy of this model were assessed using ER data from a network of 21 raingages located in Ontario, Canada. Results of this assessment based on different statistical criteria have indicated the comparable performance of the proposed scaling GLO mod...

The Annals of Applied Statistics, 2010

There is substantial empirical and climatological evidence that precipitation extremes have become more extreme during the twentieth century, and that this trend is likely to continue as global warming becomes more intense. However, understanding these issues is limited by a fundamental issue of spatial scaling: that most evidence of past trends comes from rain gauge data, whereas trends into the future are produced by climate models, which rely on gridded aggregates. To study this further, we fit the Generalized Extreme Value (GEV) distribution to the right tail of the distribution of both rain gauge and gridded events. The rain gauge data come from a network of 5873 U.S. stations, and the gridded data from a well-known re-analysis model (NCEP) and on climate model runs from NCAR's Community Climate System Model (CCSM). The results of this modeling exercise confirm, as expected, that return values computed from rain gauge data are typically higher than those computed from gridded data; however, the size of the difference is somewhat surprising with the rain gauge data exhibiting return values sometimes two or three times that of the gridded data. The main contribution of this paper is the development of a family of regression relationships between the two sets of return values that also take spatial variations into account. Based on these results, we now believe it is possible to project future changes in precipitation extremes at the point-location level based on results from climate models.

Extremes, 2010

Present day weather forecast models usually cannot provide realistic descriptions of local and particularly extreme weather conditions. However, for lead times of about 0-5 days, they provide reliable forecasts of the atmospheric circulation that encompasses the sub-scale processes leading to extremes. Hence, forecasts of extreme events can only be achieved through a combination of dynamical and statistical analysis methods, where a stable and significant statistical model based on a-priori physical reasoning establishes a-posterior a statistical-dynamical model between the local extremes and the large scale circulation. Here we present the development and application of such a statistical model calibration on the basis of extreme value theory, in order to derive probabilistic forecasts for (extreme) local precipitation. Besides a semi-parametric approach (censored quantile regression, QR) to derive conditional quantile estimates, we use a Poisson point process representation (PP) with non-stationary parameters but a constant threshold, and a peak-over-threshold representation (POT) using the non-stationary generalized Pareto distribution and a variable threshold. The variable threshold is conditioned on the numerical model output and defined as the 0.95 conditional quantile. The performance of the different approaches is compared using the quantile verification score. The downscaling applies to ERA40 re-analysis, in order to derive estimates of the conditional quantiles of daily precipitation accumulations at German weather stations. In terms of the verification score, the differences between the downscaling approaches are marginal. However, the uncertainty of the quantile estimates is larger for the semiparametric QR approach, particularly for the high quantiles. A constant shape parameter is to be preferred for a stable statistical downscaling model.

Atmósfera, 2015

El presente estudio analiza los posibles impactos futuros del cambio climático sobre los eventos meteorológicos de sequía en Turquía, utilizando para ello una nueva técnica estadística de reducción de escala (downscaling) basada en regresión logística politómica. Esta técnica, conocida como "estadísticas de salida del modelo" (model output statistics, MOS), está diseñada para la reducción de escala de las categorías de sequía del índice estandarizado de precipitación (SPI, por sus siglas en inglés). El principal objetivo de un procedimiento de reducción de escala es determinar la influencia de la variabilidad climática de gran escala y los cambios proyectados en las variables a nivel regional y local. Los predictores de gran escala utilizados en este estudio se obtuvieron a partir de simulaciones del modelo canadiense de circulación general acoplado de segunda generación (CGCM2, por sus siglas en inglés), las cuales abarcan de 1940 a 2100 e incluyen tres escenarios socioeconómicos: control, con los límites de la concentración atmosférica de gases de efecto invernadero en el siglo XX, y los escenarios A2 y B2 del Informe especial sobre escenarios de emisiones del Panel Intergubernamental de Cambio Climático. Se utilizaron observaciones de 96 estaciones meteorológicas para calcular los valores anuales del SPI para el periodo 1940-2010, dejando los últimos 10 años para validación contra los resultados simulados por el CGCM2. Los resultados del MOS derivados de la simulación del clima denominada control coincidieron con los patrones observados en el clima actual. Los resultados del MOS derivados de escenarios climáticos futuros llevan a concluir que hay una probabilidad disminuida de que se presenten condiciones muy húmedas o extremadamente húmedas. Adicionalmente, las probabilidades de que las condiciones sean cercanas a lo normal disminuirán en la costa del Mar Negro, y aumentarán en la transición del Mármara y en Anatolia oriental.

Advances in Geosciences, 2005

Daily precipitation is recorded as the total amount of water collected by a rain-gauge in 24 h. Events are modelled as a Poisson process and the 24 h precipitation by a Generalized Pareto Distribution (GPD) of excesses. Hazard assessment is complete when estimates of the Poisson rate and the distribution parameters, together with a measure of their uncertainty, are obtained. The shape parameter of the GPD determines the support of the variable: Weibull domain of attraction (DA) corresponds to finite support variables, as should be for natural phenomena. However, Fréchet DA has been reported for daily precipitation, which implies an infinite support and a heavy-tailed distribution. We use the fact that a log-scale is better suited to the type of variable analyzed to overcome this inconsistency, thus showing that using the appropriate natural scale can be extremely important for proper hazard assessment. The approach is illustrated with precipitation data from the Eastern coast of the Iberian Peninsula affected by severe convective precipitation. The estimation is carried out by using Bayesian techniques.

Natural Hazards and Earth System Science, 2006

Daily precipitation is recorded as the total amount of water collected by a rain-gauge in 24 h. Events are modelled as a Poisson process and the 24 h precipitation by a Generalised Pareto Distribution (GPD) of excesses. Hazard assessment is complete when estimates of the Poisson rate and the distribution parameters, together with a measure of their uncertainty, are obtained. The shape parameter of the GPD determines the support of the variable: Weibull domain of attraction (DA) corresponds to finite support variables as should be for natural phenomena. However, Fréchet DA has been reported for daily precipitation, which implies an infinite support and a heavy-tailed distribution. Bayesian techniques are used to estimate the parameters. The approach is illustrated with precipitation data from the Eastern coast of the Iberian Peninsula affected by severe convective precipitation. The estimated GPD is mainly in the Fréchet DA, something incompatible with the common sense assumption of that precipitation is a bounded phenomenon. The bounded character of precipitation is then taken as a priori hypothesis. Consistency of this hypothesis with the data is checked in two cases: using the raw-data (in mm) and using logtransformed data. As expected, a Bayesian model checking clearly rejects the model in the raw-data case. However, logtransformed data seem to be consistent with the model. This fact may be due to the adequacy of the log-scale to represent positive measurements for which differences are better relative than absolute.

EPiC series in engineering, 2018

This paper proposes an efficient spatio-temporal statistical downscaling approach for estimating IDF relations at an ungauged site using daily rainfalls downscaled from global climate model (GCM) outputs. More specifically, the proposed approach involves two steps: (1) a spatial downscaling using scaling factors to transfer the daily downscaled GCM extreme rainfall projections at a regional scale to a given ungauged site and (2) a temporal downscaling using the scale-invariance GEV model to derive the distribution of sub-daily extreme rainfalls from downscaled daily rainfalls at the same location. The feasibility and accuracy of the proposed approach were evaluated based on the climate simulation outputs from 21 GCMs that have been downscaled by NASA to a regional 25-km scale for two different RCP 4.5 and 8.5 scenarios and the observed extreme rainfall data available from a network of 15 raingauges located in Ontario, Canada. The jackknife technique was used to represent the ungauged site conditions. Results based on different statistical criteria have indicated the feasibility and accuracy of the proposed approach.

Natural Hazards and Earth System Science, 2011

In Mediterranean regions, climate studies indicate for the future a possible increase in the extreme rainfall events occurrence and intensity. To evaluate the future changes in the extreme event distribution, there is a need to provide non-stationary models taking into account the nonstationarity of climate. In this study, several climatic covariates are tested in a non-stationary peaks-over-threshold modeling approach for heavy rainfall events in Southern France. Results indicate that the introduction of climatic covariates could improve the statistical modeling of extreme events. In the case study, the frequency of southern synoptic circulation patterns is found to improve the occurrence process of extreme events modeled via a Poisson distribution, whereas for the magnitude of the events, the air temperature and sea level pressure appear as valid covariates for the Generalized Pareto distribution scale parameter. Covariates describing the humidity fluxes at monthly and seasonal time scales also provide significant model improvements for the occurrence and the magnitude of heavy rainfall events. With such models including climatic covariates, it becomes possible to asses the risk of extreme events given certain climatic conditions at monthly or seasonal timescales. The future changes in the heavy rainfall distribution can also be evaluated using covariates computed by climate models.