Water Resources Research, volume 61, issue 1

Improved Correction of Extreme Precipitation Through Explicit and Continuous Nonstationarity Treatment and the Metastatistical Approach

Cuauhtémoc Tonatiuh Vidrio Sahagún 1
Jianxun He 1
Alain Pietroniro 1
1
 
Civil Engineering Schulich School of Engineering University of Calgary Calgary AB Canada
Publication typeJournal Article
Publication date2025-01-03
scimago Q1
wos Q1
SJR1.574
CiteScore8.8
Impact factor4.6
ISSN00431397, 19447973
Abstract

Climate models simulate extreme precipitation under nonstationarity due to continuous climate change. However, systematic errors in local‐scale climate projections are often corrected using stationary or quasi‐stationary methods without explicit and continuous nonstationarity treatment, like quantile mapping (QM), detrended QM, and quantile delta mapping. To bridge this gap, we introduce nonstationary QM (NS‐QM) and its simplified version for consistent nonstationarity patterns (CNS‐QM). Besides, correction approaches for extremes often rely on limited extreme‐event records. To leverage ordinary‐event information while focusing on extremes, we propose integrating the simplified Metastatistical extreme value (SMEV) distribution into NS‐QM and CNS‐QM (NS‐QM‐SMEV and CNS‐QM‐SMEV). We demonstrate the superiority of NS‐ and CNS‐QM‐SMEV over existing methods through a simulation study and show several real‐world applications using high‐resolution‐regional and coarse‐resolution‐global climate models. NS‐QM and CNS‐QM reflect nonstationarity more realistically but may encounter challenges due to data limitations like estimation errors and uncertainty, particularly for the most extreme events. These issues, shared by existing approaches, are effectively mitigated using the SMEV distribution. NS‐ and CNS‐QM‐SMEV offer lower estimation error, approximate unbiasedness, reduced uncertainty, and improved representation of the entire distribution, especially for samples of ∼70 years, and greater superiority with larger samples. We show existing methods may perform competitively for short samples but exhibit substantial biases in quantile‐quantile matching due to bypassing nonstationarity modeling. NS‐ and CNS‐QM‐SMEV avoid these biases, adhering better to their theoretical functioning. Thus, NS‐ and CNS‐QM‐SMEV enhance the correction of extremes under nonstationarity. Yet, properly identifying nonstationarity patterns is crucial for reliable implementations.

Michalek A.T., Villarini G., Kim T.
Earth's Future scimago Q1 wos Q1 Open Access
2024-03-06 citations by CoLab: 4 Abstract  
AbstractThis study evaluates five bias correction and statistical downscaling (BCSD) techniques for daily precipitation and examines their impacts on the projected changes in flood extremes (i.e., 1%, 0.5%, and 0.2% floods). We use climate model outputs from the Coupled Model Intercomparison Project Phase 6 (CMIP6) to conduct hydrologic simulations across watersheds in Iowa and determine historical and future flood extreme estimates based on generalized extreme value distribution fitting. Projected changes in these extremes are examined with respect to four Shared Socioeconomic Pathways (SSPs) alongside five BCSD techniques. We find the magnitude of the estimates of future annual exceedance probabilities (AEPs) are expected to increase under all SSPs, especially for the emission scenarios with higher greenhouse gases concentrations (i.e., SSP370 and SSP585). Our results also suggest the choice of BCSD impacts the magnitude of the projected changes, with the SSPs that play a more limited role compared to the choice of downscaling method. The variability in projected flood changes across Iowa is similar across the downscaling technique but increases as the AEP increases. Our findings provide insights into the impact of downscaling techniques on flood extremes' projections and useful information for climate planning across the state.
Vidrio-Sahagún C.T., Ruschkowski J., He J., Pietroniro A.
2024-02-01 citations by CoLab: 8 Abstract  
In flood frequency analysis (FFA), choices of distribution and methods can hinder the reproducibility of results. Besides, changes in climate, land use/cover, and water management can induce nonstationarity. Frameworks to select between stationary FFA (S-FFA) and nonstationary FFA (NS-FFA) are lacking, and NS-FFA tools are limited. Therefore, this paper introduces a systematic and software-supported framework enabling repeatable workflows for both S-FFA and NS-FFA. The framework has three modules to a) process flood series for exploratory data analysis (EDA) and NS-FFA model determination (if needed), b) select the S-FFA or NS-FFA approach underpinned by the EDA, and c) perform FFA including model determination, parameter estimation, uncertainty quantification, and model performance assessment. The framework incorporates various distributions, methods, and metrics, and recent advancements in NS-FFA for model determination and uncertainty quantification and allows for the modeller's intervention while ensuring reproducibility. The software is freely available to the public.
Falkensteiner M., Schellander H., Ehrensperger G., Hell T.
Weather and Climate Extremes scimago Q1 wos Q1 Open Access
2023-12-01 citations by CoLab: 2 Abstract  
The typical approach to account for non-stationarity in the generalized extreme value distribution (GEV) is to model the temporal behavior of the GEV parameters, e. g., with linear relationships between the parameters and e. g., the year. When in addition, seasonality, i. e., sub-yearly patterns are of interest, this relationship needs to be more complex. A natural candidate for nonstationary analysis of precipitation extremes is the metastatistical extreme value distribution (MEVD) assuming the ordinary rainfall events being Weibull distributed. It uses all ordinary events of the underlying rainfall distribution and implicitly considers temporal nonstationary changes in the Weibull parameters by a blockwise estimation. However, the MEVD still does not make use of the sub-yearly evolution of the distribution parameters and thus discards a significant part of information. In this paper a modification of the MEVD is proposed. The temporal MEVD (TMEV) relaxes the assumption of constant coefficients of the underlying distribution. The MEVD in each block estimates the Weibull parameters using the number of wet events in that block, regardless whether an individual wet day contributes much or less to the yearly rainfall sum. The newly proposed TMEV weights the single wet days contributions during parameter estimation and is therefore able to explicitly account for seasonal differences of rainfall events. It is shown that the TMEV provides a very similar error characteristic to the simplified MEVD for the estimation of quantiles with different sample lengths. In addition, we demonstrate the ability of the TMEV to identify longterm trends and seasonal variations of extreme precipitation in Austria and discuss possible implications for general trend estimations. We also present a spatio-temporal TMEV model, which is able to reproduce known patterns of a 50-year return level map of daily rainfall in Austria, thereby lending credence to the TMEV approach.
Lafferty D.C., Sriver R.L.
2023-09-30 citations by CoLab: 29 PDF Abstract  
AbstractEfforts to diagnose the risks of a changing climate often rely on downscaled and bias-corrected climate information, making it important to understand the uncertainties and potential biases of this approach. Here, we perform a variance decomposition to partition uncertainty in global climate projections and quantify the relative importance of downscaling and bias-correction. We analyze simple climate metrics such as annual temperature and precipitation averages, as well as several indices of climate extremes. We find that downscaling and bias-correction often contribute substantial uncertainty to local decision-relevant climate outcomes, though our results are strongly heterogeneous across space, time, and climate metrics. Our results can provide guidance to impact modelers and decision-makers regarding the uncertainties associated with downscaling and bias-correction when performing local-scale analyses, as neglecting to account for these uncertainties may risk overconfidence relative to the full range of possible climate futures.
Vidrio-Sahagún C.T., He J., Pietroniro A.
Advances in Water Resources scimago Q1 wos Q1
2023-06-01 citations by CoLab: 5 Abstract  
The nonstationary hydrological frequency analysis (NS-HFA) aids in assessing the recurrence of hydrological extremes under nonstationarity, but its reliability is often questioned due to relatively limited record lengths. The Metastatistical extreme value (MEV) distribution, which harnesses the ordinary event records along with the extremes, has proven advantageous under stationarity and in instances with limited records. Yet, nonstationary applications of the MEV distribution are lacking, and only modified versions, such as the simplified MEV (SMEV) and block-based MEV (MEVBB) distributions, have been proposed. This paper develops the nonstationary version of the MEV distribution for NS-HFA (called the MEV-based model), which incorporates an explicit nonstationary structure and preserves the stochastic interannual variability of ordinary events. The developed model is assessed using several benchmark models in both simulation studies and real applications for both in-sample fitting and out-of-sample prediction from the perspectives of uncertainty, accuracy, and fitting efficiency. The benchmark models for comparison included the MEVBB as well as the SMEV- and Generalized Extreme Value distribution (GEV)-based models. The results demonstrated that the proposed MEV-based model outperformed the MEVBB regarding overfitting and captured the underlying process more efficiently, accurately, and with less uncertainty. In addition, the MEV-based model was superior to the GEV-based model due to its higher accuracy, equivalent or better fitting efficiency, and lower uncertainty. Furthermore, although the MEV-based model performed overall equivalently to the SMEV-based, the MEV-based model was shown advantageous for adopting nonstationary stochastic physical covariates and facilitating out-of-sample predictions. Overall, these results demonstrated that the proposed MEV-based model has distinct advantages for the NS-HFA, and consequently promotes its implementation.
Gründemann G.J., Zorzetto E., Beck H.E., Schleiss M., van de Giesen N., Marani M., van der Ent R.J.
Journal of Hydrology scimago Q1 wos Q1
2023-06-01 citations by CoLab: 15 Abstract  
Quantifying the magnitude and frequency of extreme precipitation events is key in translating climate observations to planning and engineering design. Past efforts have mostly focused on the estimation of daily extremes using gauge observations. Recent development of high-resolution global precipitation products, now allow estimation of global extremes. This research aims to quantitatively characterize the spatiotemporal behavior of precipitation extremes, by calculating extreme precipitation return levels for multiple durations on the global domain using the Multi-Source Weighted-Ensemble Precipitation (MSWEP) dataset. Both classical and novel extreme value distributions are used to provide insight into the spatial patterns of precipitation extremes. Our results show that the traditional Generalized Extreme Value (GEV) distribution and Peak-Over-Threshold (POT) methods, which only use the largest events to estimate precipitation extremes, are not spatially coherent. The recently developed Metastatistical Extreme Value (MEV) distribution, that includes all precipitation events, leads to smoother spatial patterns of local extremes. For durations of 5 and 10 days, however, there are less events per year to fit the distribution (37 and 22 on average, respectively), leading to larger inter-annual variability and possible overestimation of the extremes. While the GEV and POT methods predict a consistent shift from heavy to thin tails with increasing duration, the MEV method predicts a relatively constant heaviness of the tail for any precipitation duration, opening up an important research question on what is the ‘correct’ tail behavior of extreme precipitation for different durations. The generated extreme precipitation return levels and corresponding parameters are provided as the Global Precipitation EXtremes (GPEX) dataset. These data can be useful for studying the underlying physical processes causing the spatiotemporal variations of the heaviness of extreme precipitation distributions.
Marra F., Amponsah W., Papalexiou S.M.
Advances in Water Resources scimago Q1 wos Q1
2023-03-01 citations by CoLab: 24 Abstract  
The exceedance probability of extreme daily precipitation is usually quantified assuming asymptotic behaviours. Non-asymptotic statistics, however, would allow us to describe extremes with reduced uncertainty and to establish relations between physical processes and emerging extremes. These approaches are still mistrusted by part of the community as they rely on assumptions on the tail behaviour of the daily precipitation distribution. This paper addresses this gap. We use global quality-controlled long rain gauge records to show that daily precipitation annual maxima are samples likely emerging from Weibull tails in most of the stations worldwide. These non-asymptotic tails can explain the statistics of observed extremes better than asymptotic approximations from extreme value theory. We call for a renewed consideration of non-asymptotic statistics for the description of extremes.
Vidrio-Sahagún C.T., He J.
Journal of Hydrology scimago Q1 wos Q1
2022-09-01 citations by CoLab: 14 Abstract  
• The performance-based NS-FFA is challenged theoretically and practically. • A new decomposition procedure for the decomposition-based NS-FFA is proposed. • The proposed decomposition procedure preserves the stationary component. • The proposed decomposition-based NS-FFA captures the nonstationary process better. • This approach overcomes the previous issues of the NS-FFA and performs competitively. The nonstationary flood frequency analysis (NS-FFA) is conducted when the assumption of stationarity in hydrologic extremes is violated. The commonly used approach is the performance-based NS-FFA, which jointly determines the distribution and the nonstationary structure based upon the model performance. However, this approach is challenged from both theoretical and practical perspectives. An alternative is the herein named decomposition-based NS-FFA, which determines the two NS-FFA model components separately and explicitly uses the available knowledge of the nonstationarity. However, it has been barely implemented in practice. This paper proposed a novel decomposition procedure that strictly follows the theoretical decomposition of nonstationary stochastic processes to advance the decomposition-based NS-FFA. The proposed decomposition procedure was compared with a previously reported method in both an analytical deduction and a simulation study. The proposed decomposition-based NS-FFA was further compared to the performance-based NS-FFA using both synthetic and real datasets from North America, which exhibit different patterns of nonstationarity. The Particle Filter was adopted for uncertainty quantification and parameter estimation. The results revealed that the proposed decomposition-based approach was advantageous in preserving the moments of underlying stochastic component, particularly the higher-order moment (i.e., skewness). In addition, the comparison of the two NS-FFA approaches demonstrated the superiority of the proposed decomposition-based approach in capturing the underlying (known) nonstationary stochastic process and being competitive with the performance-based approach from the performance perspective in real applications. The results from the simulation study and the real application also revealed several caveats of the performance-based approach, including the potential overfitting and equifinality problems as well as the selection of distinct models when adopting different performance metrics. In addition, differing from the performance-based approach, the decomposition-based NS-FFA avoided/alleviated the ergodicity violation. All these results demonstrated the advancements of the proposed decomposition-based NS-FFA and advocated its application in the NS-FFA.
Vidrio-Sahagún C.T., He J.
Advances in Water Resources scimago Q1 wos Q1
2022-08-01 citations by CoLab: 14 Abstract  
• The RF-PL method is extended for the MEV and SMEV models. • The MEV distribution is found prone to overfit the samples under nonstationarity. • The SMEV outperforms MEV and GEV for both fitting and prediction purposes. • The use of SMEV can advance the NS-HFA, especially by reducing the uncertainty. The implementation of nonstationary hydrological frequency analysis (NS-HFA) has often been hampered by the relatively short datasets and the resulting high uncertainty. Most recently, the non-asymptotic Metastatistical extreme value (MEV) and simplified MEV (SMEV) distributions, which rely on the ordinary events rather than the extremes only, have attracted attention in the HFA of both rainfall and streamflow. Despite their use for trend detection/attribution and producing future projections, their practical implementation for the NS-HFA is absent in the literature. This paper therefore implemented these models (called MEV and SMEV-based models) in the NS-HFA and comprehensively assessed their performance from the perspectives of fitting efficiency, accuracy, and uncertainty for both in-sample fitting and out-of-sample prediction purposes. The asymptotic models based on the generalized extreme value (GEV) distribution were used as the benchmark. The assessment employed synthetic and real rainfall datasets that exhibit stationarity in the number of events per year. All the nonstationary ordinary-event datasets followed the Weibull distribution with linearly changing parameters, while their standardized annual maximum series aligned with the GEV distribution. Thus, the MEV, SMEV- and GEV-based models could be fairly assessed and compared. The regula-falsi profile likelihood method was extended to quantify the uncertainty of the MEV and SMEV-based models. The results demonstrated that the MEV model was not advantageous over other models in terms of all three evaluation perspectives. Whereas the SMEV-based models demonstrated superiority due to their higher accuracy, equivalent or better fitting efficiency, as well as lower uncertainty compared to all other models. Therefore, this paper advocates the use of the SMEV distribution to advance the NS-HFA by harnessing the information from the ordinary events.
Van de Velde J., Demuzere M., De Baets B., Verhoest N.E.
2022-05-03 citations by CoLab: 15 Abstract  
Abstract. Climate change is one of the biggest challenges currently faced by society, with an impact on many systems, such as the hydrological cycle. To assess this impact in a local context, regional climate model (RCM) simulations are often used as input for rainfall-runoff models. However, RCM results are still biased with respect to the observations. Many methods have been developed to adjust these biases, but only during the last few years, methods to adjust biases that account for the correlation between the variables have been proposed. This correlation adjustment is especially important for compound event impact analysis. As an illustration, a hydrological impact assessment exercise is used here, as hydrological models often need multiple locally unbiased input variables to ensure an unbiased output. However, it has been suggested that multivariate bias-adjusting methods may perform poorly under climate change conditions because of bias nonstationarity. In this study, two univariate and four multivariate bias-adjusting methods are compared with respect to their performance under climate change conditions. To this end, a case study is performed using data from the Royal Meteorological Institute of Belgium, located in Uccle. The methods are calibrated in the late 20th century (1970–1989) and validated in the early 21st century (1998–2017), in which the effect of climate change is already visible. The variables adjusted are precipitation, evaporation and temperature, of which the former two are used as input for a rainfall-runoff model, to allow for the validation of the methods on discharge. Although not used for discharge modeling, temperature is a commonly adjusted variable in both uni- and multivariate settings and we therefore also included this variable. The methods are evaluated using indices based on the adjusted variables, the temporal structure, and the multivariate correlation. The Perkins skill score is used to evaluate the full probability density function (PDF). The results show a clear impact of nonstationarity on the bias adjustment. However, the impact varies depending on season and variable: the impact is most visible for precipitation in winter and summer. All methods respond similarly to the bias nonstationarity, with increased biases after adjustment in the validation period in comparison with the calibration period. This should be accounted for in impact models: incorrectly adjusted inputs or forcings will lead to predicted discharges that are biased as well.
Miniussi A., Marra F.
Journal of Hydrology scimago Q1 wos Q1
2021-12-01 citations by CoLab: 19 Abstract  
• At-site and spatial Simplified Metastatistical Extreme Value in Germany. • Light-tails for precipitation in Germany may lead to dangerous underestimation. • We identify a correlation between orography and parameters of daily rainfall. • Inverse distance weighting of parameters minimizes error in ungauged locations. Estimating extreme precipitation return levels at ungauged locations is key for hydrological applications and risk management, and demands improved techniques to decrease the large uncertainty of traditional methods. Here, we leverage the perks of the simplified metastatistical extreme value (SMEV) approach with a twofold aim: we show how it can be effectively used in situations in which the ordinary daily precipitation events cannot be fully described using a two-parameter distribution, and we examine the performance of different interpolation techniques for the estimation of return levels in ungauged locations. SMEV proved adequate at representing at-site extremes for a set of 4000+ stations in Germany, with a general tendency to underestimate the probability of the largest annual maxima. At the same time SMEV tends to overestimate with respect to the design return levels currently adopted in the country, suggesting that these might actually underestimate the distribution tail. Among the investigated methods, the inverse distance weighted interpolation of SMEV parameters provides the most accurate estimates of extreme return levels for ungauged locations, with typical standard errors of 0.79 (0.83) for rain gauge densities of 1/500 km −2 (1/1000 km −2 ). Albeit only less than 10% of the variance in estimation errors is explained by elevation, the correlation between SMEV parameters and orography (up to 43% explained variance) suggests that future applications should test the inclusion of such information in spatial estimates.
Chen D., Dai A., Hall A.
2021-08-04 citations by CoLab: 54 Abstract  
Overestimation of precipitation frequency and duration while underestimating intensity, that is, the “drizzling” bias, has been a long-standing problem of global climate models. Here we explore this issue from the perspective of precipitation partitioning. We found that most models in the Climate Model Intercomparison Project Phase 5 (CMIP5) have high convective-to-total precipitation (PC/PR) ratios in low latitudes. Convective precipitation has higher frequency and longer duration but lower intensity than non-convective precipitation in many models. As a result, the high PC/PR ratio contributes to the “drizzling” bias over low latitudes. The PC/PR ratio and associated “drizzling” bias increase as model resolution coarsens from 0.5° to 2.0°, but the resolution's effect weakens as the grid spacing increases from 2.0° to 3.0°. Some of the CMIP6 models show reduced “drizzling” bias associated with decreased PC/PR ratio. Thus, more reasonable precipitation partitioning, along with finer model resolution should alleviate the “drizzling” bias within current climate models.
Slater L.J., Anderson B., Buechel M., Dadson S., Han S., Harrigan S., Kelder T., Kowal K., Lees T., Matthews T., Murphy C., Wilby R.L.
2021-07-07 citations by CoLab: 159 Abstract  
Abstract. Hydroclimatic extremes such as intense rainfall, floods, droughts, heatwaves, and wind or storms have devastating effects each year. One of the key challenges for society is understanding how these extremes are evolving and likely to unfold beyond their historical distributions under the influence of multiple drivers such as changes in climate, land cover, and other human factors. Methods for analysing hydroclimatic extremes have advanced considerably in recent decades. Here we provide a review of the drivers, metrics, and methods for the detection, attribution, management, and projection of nonstationary hydroclimatic extremes. We discuss issues and uncertainty associated with these approaches (e.g. arising from insufficient record length, spurious nonstationarities, or incomplete representation of nonstationary sources in modelling frameworks), examine empirical and simulation-based frameworks for analysis of nonstationary extremes, and identify gaps for future research.
Hernanz A., García‐Valero J.A., Domínguez M., Ramos‐Calzado P., Pastor‐Saavedra M.A., Rodríguez‐Camino E.
2021-07-05 citations by CoLab: 35 Abstract  
The Spanish Meteorological Agency (AEMET) is responsible for the elaboration of downscaled climate projections over Spain to feed the Second National Plan of Adaptation to Climate Change (PNACC-2). The main objective of this article is to establish a comparison among five statistical downscaling methods developed at AEMET: (1) Analog, (2) Regression, (3) Artificial Neural Networks, (4) Support Vector Machines and (5) Kernel Ridge Regression. This comparison has been carried out under present conditions and with perfect predictors, based on the framework established by the VALUE network, in particular, on its perfect predictor experiment. In this experiment, we evaluate the marginal aspects of the distributions of daily maximum/minimum temperatures and daily accumulated precipitation analysed by seasons, on a high resolution observational grid (0.05°) over mainland Spain and the Balearic Islands. This is the first of a set of three experiments aimed to allow us to decide which methods, and under what configuration, is more appropriate for the generation of downscaled climate projections over our region. For maximum/minimum temperatures, all methods display a similar behaviour. They capture very satisfactorily the mean values although slight biases are detected on the extremes. In general, results for maximum temperature appear to be more accurate than for minimum temperature, and the nonlinear methods display certain added value. For precipitation, remarkable differences are found among all methods. Most of the methods are capable of reproducing the total precipitation amount quite satisfactorily, whereas other aspects such as intense precipitations and the precipitation occurrence are captured with more accuracy by the Analog method.

Are you a researcher?

Create a profile to get free access to personal recommendations for colleagues and new articles.
Share
Cite this
GOST | RIS | BibTex
Found error?