Quantifying climate change impact on water resources systems at regional or catchment scales is important in water resources planning and management. General circulation models (GCMs) represent our main source of knowledge about future climate change. However, several key limitations restrict the direct use of GCM simulations for water resource assessments. In particular, the presence of systematic bias and the need for its correction is an essential pre-processing step that improves the quality of GCM simulations, making climate change impact assessments more robust and believable. What exactly is systematic bias? Can systematic bias be quantified if the model is asynchronous with observations or other model simulations? Should model bias be sub-categorized to focus on individual attributes of interest or aggregated to focus on lower moments alone? How would one address bias in multiple attributes without making the correction model complex? How could one be confident that corrected simulations for the yet-to-be-seen future bear a closer resemblance to the truth? How can one meaningfully extrapolate correction to multiple dimensions, without being impacted by the ‘Curse of Dimensionality’? These are some of the questions we attempt to address in the paper.

  • Importance of procedures for correcting systematic biases is discussed.

  • Extensive literature is presented on bias correction and its use.

  • The importance of correcting specific attributes for water resources applications is illustrated.

  • Challenges in formulating a bias correction alternative are highlighted.

  • Added information on how correcting these biases is critical before any dynamical or statistical downscaling application.

‘All Models Are Wrong, Some Are Useful’. These famous words from nearly half a century ago (Box 1976) are a reminder of the limits all models are constrained by. The usefulness of models, however, improves, often significantly, if systematic bias can be removed. To do this, however, the bias needs to be quantified, and appropriate corrections needs to be devised. These corrections may be simple re-parameterizations of the model, an update in model state variables, or a more comprehensive re-specification of the model structure. Identifying, characterizing, and removing such bias, however, becomes especially tricky if the model is complex, and its inputs (as well as observed responses used for assessment) uncertain. Such is the scenario that characterizes the biases inherent in climate model simulations of the future, which forms the focus of this article.

Climate change impact studies often require simulated outputs from climate impact models. Coupled atmosphere-ocean general circulation models (GCMs) or regional climate models (RCMs) are our main source of knowledge about likely changes in the climate into the future but are known to contain systematic errors (biases) in simulations when compared with observations (Collins et al. 2013). In particular, regional and local-scale events and extremes are not well represented in GCMs (Mehrotra & Sharma 2015). As a result of this, the majority of the ‘raw’ simulations of climate models do not statistically match with observations (Mehrotra & Sharma 2015). Therefore, it is necessary to remove or minimize the differences between the observed and raw GCM/RCM simulations as these are used to drive downscaling and impact assessment models. Bias correction (BC) has become even more important in climate change impact research due to growing databases of global and regional climate model simulations and relative simplicity and effectiveness of BC approaches. Over recent years, many different methods have been developed to perform BC. These vary from very simplistic methods, such as the delta-change method, which takes the difference between the current and future models and then adds the difference to the current observation to obtain a projected future (Xu 1999; Hay et al. 2000; Lenderink et al. 2007), or correcting only the statistical mean of the simulations (Wilby & Wigley 1997), or the variance correction using simple scaling adjustment (Berg et al. 2012) to more sophisticated methods, for example, based on complex stochastic modelling aimed at fixing bias in various dependence attributes (Eden et al. 2014; Wong et al. 2014). The most common BC approach in hydrological studies uses quantile-mapping (QM) (Panofsky et al. 1958; Haddad & Rosenfeld 1997; Wood et al. 2004; Déqué 2007; Piani et al. 2010; Gudmundsson et al. 2012). QM has been applied to climate model output globally (Haerter et al. 2011; Teutschbein & Seibert 2012; Maurer & Pierce 2014) and its variants have been presented and thoroughly reviewed by researchers (Maraun 2016; Shrestha et al. 2017; Guo et al. 2020).

A number of studies have also highlighted the common issues of the application of QM BC, specifically when assessing projected climate change impacts. It was found that QM can modify long-term changes in simulated series (Hagemann et al. 2011; Themeßl et al. 2012; Maraun 2013; Maurer & Pierce 2014). Maurer & Pierce (2014) evaluated the effect of QM on GCM-simulated precipitation changes between two historical periods and noted that the approach can affect trends in extreme quantiles differently than trends in the mean. They suggested that further work is needed to assess the impact of QM on the tails of the distribution.

Although climate model performance has improved, deficiencies remain in representing regional variability on interannual to decadal time scales (Sheffield et al. 2013; Beadling et al. 2020; Fan et al. 2020; Wang et al. 2021). Biases in the temporal variability of raw climate series can pose problems for impact modelling. For example, a model might have appropriate variability at hourly or daily time scales, however, little or no variability at annual time scales. To deal with this type of bias, a single timescale (e.g., daily or monthly) BC approach such as delta change or QM may not be the best option to be used, since it focuses only on daily or monthly statistics of the time series and therefore cannot necessarily correct for the variability biases at aggregated time scales and the resulting bias-corrected series might show a biased representation of the magnitude and variability of the low-frequency phenomena such as El Niño and La Niña (Ashfaq et al. 2011; Bellenger et al. 2014). Such biases could also modify the simulations and statistics of heat waves, flooding events, agriculture estimates, and reservoir simulations (Pierce et al. 2015).

Considering standard BC approaches ignore the biases in aggregated time series, nested bias correction (NBC) offers a solution as it corrects for the biases in both distribution and persistence at multiple time scales. As the BC works over pre-defined time scales, biases at other time scales not directly corrected may not be eliminated (Johnson & Sharma 2012). To avoid the specification of such pre-defined time scales (e.g., daily, monthly, and annual timescale) researchers have suggested converting the time series into a frequency domain using an appropriate transformation and applying BC in the frequency domain (Pierce et al. 2015; Nguyen et al. 2016) as correction in the frequency space corresponds to the applying NBC at multiple time scales in the time-space. The idea of frequency bias correction (FBC) is based on the logic that when a time series is analysed in the frequency domain, the variance of the time series can be expressed as a function of frequency, making it more suitable for studying variability across different time scales. Similar to FBC, time-frequency-based BC approaches are developed using wavelet transforms. The discrete wavelet-based bias correction (DWBC) (Kusumastuti et al. 2021), employs discrete wavelet transform (DWT) to bias correct climate model simulations in time and frequency domain. The correction is performed across the spectrum recognizing the trend in the time series. Another refined approach, continuous wavelet-based bias correction (CWBC), offers more precise correction than DWBC since it applies a correction to non-dyadic spectrum as well which is missed by DWT (Kusumastuti et al. 2022). The wavelet-based bias correction (WBC) approaches offer the next level of BC with its ability to maintain the climate change signal which QM BC is known to alter (Cannon et al. 2015; Maraun 2016).

Climate change impact assessment studies and both dynamical and statistical downscaling approaches require several climate variables as inputs and these often have strong physical inter-dependencies in observations (Li et al. 2014; Vrac & Friederichs 2015; Mehrotra & Sharma 2015, 2016). The univariate BC, e.g., Delta, Scaling, QM, or time nesting approaches (NBC or recursive nested bias correction, RNBC) which assume independence across variables or GCM grid cells and consider BC independently for each variable or grid cell and do not recognize the spatial and cross-variable dependences of climate variables (White & Toumi 2013; Li et al. 2014; Chen et al. 2015). For example, it was found that univariate QM matched the time series distribution quite well, however, mismatched the observed and bias-corrected spatial and inter-variable dependencies of the time series (White & Toumi 2013). Similarly, Mehrotra & Sharma (2015, 2016) found that univariate BC approaches, namely QM and NBC, were ineffective in correcting space and across-variables dependencies.

A bivariate correction procedure was proposed by Piani & Haerter (2012) to simultaneously correct temperature and precipitation. This required correcting one time series (i.e., precipitation) conditional to the bias-corrected time series for the other variable (i.e., temperature). Alternate bivariate corrections were proposed by Li et al. (2014) by constructing bivariate distributions given univariate marginal based on Copula theory to jointly correct precipitation and temperature biases in climate models. This concept was later extended to multiple variables (Mao et al. 2015; Vrac & Friederichs 2015; Cannon 2016, 2018). The copula theory states that any multivariate distribution can be formulated from the individual marginal cumulative distributions and a copula function describing the dependence between the individual distributions.

A multivariate extension of the linear mean/variance rescaling approach was proposed by Bürger et al. (2011) using a Cholesky decomposition of the covariance matrix, to correct multivariate GCM fields for use in a regression-based downscaling. Bárdossy & Pegram (2012) reproduced the observed precipitation spatial dependence field using the Cholesky decomposition in rank space. They suggested sequential correlation based on either a matrix modification or a sequential regression in order to preserve the spatial correlation. Mehrotra & Sharma (2015) presented a multivariate recursive nesting BC (MRNBC) and a multivariate extension of QM, multivariate recursive quantile nesting BC (MRQNBC) (Mehrotra & Sharma 2016) to correct systematic biases in multiple variables and across multiple time scales. A multivariate correction approach was similarly proposed by Cannon (2018), using the Cholesky decomposition BC rationale (Bürger et al. 2011) and a univariate QM. The approach relies on recursion to achieve convergence (based on the matching of observed and BC time series distributions).

While the multivariate approaches, for example, multivariate recursive nesting bias correction (MRNBC) and QM-based MRNBC (MRQNBC), represent the state of the art to bias correct multiple variables across multiple time scales, they still require specification of time scales of interest by the user. The mathematical complexity of these approaches grows with the increase in the number of variables and grid points/locations. Nguyen et al. (2018) proposed a multivariate extension of the frequency-based BC approach. The approach is aimed at addressing biases in high- and low-frequency variations in the individual time series as well as the biases across multiple variables for each of these variations, all in the frequency space thus avoiding the need for time scale nesting and matrix manipulation involved in MRNBC and MRQNBC procedure. Their results suggest that the approach can reduce both intra- and inter-variable dependence biases in the corrected simulations.

A different approach to the multivariate methods above stemmed from the Schaake Shuffle (SS) (Clark et al. 2004), which proposed a data re-shuffling strategy to alter dependence patterns in multiple ensembles of simulated climate variables. In this approach, the ensemble members for a given time step are ranked and matched with the rank of observation data from other time steps randomly selected from similar dates in the historical record. At the end of the procedure, the ensembles are shuffled following the observed time sequence thereby recovering the observed space–time variability in simulated ensembles. This rank-based shuffling technique with modifications has been applied for the purpose of multivariate BC or downscaling of climate simulations (Cannon 2018; Vrac & Friederichs 2015). A more comprehensive shuffling-based BC method has been introduced by Vrac & Friederichs (2015) and Mehrotra & Sharma (2019) for correcting multivariate structure of climate model outputs. The approach combines univariate QM with SS but attempts to capture the pattern of change suggested for future climates.

The rest of the paper is organized as follows. The next section discusses the biases in a model leading to the development of BC approaches from the simplest to the more sophisticated ones as introduced earlier. Section 3 focuses on the other aspects of BC that need to be considered when applying BC approaches in hydrological studies. In section 4, the limitations and assumptions of BC approaches are discussed. Finally, the discussion is summarized in section 5.

All climate models (both global and regional GCMs, RCMs) have systematic errors (biases) in simulations. For example, climate models often simulate excessive rainy days, creating a drizzle effect. Similarly, systematic errors are present in the timing of the monsoon, the amount of seasonal rainfall, underestimation of rainfall extremes, and consistently too-high or too-low temperatures (Pörtner et al. 2022). The use of uncorrected simulations in impact models or assessments can be unrealistic.

Errors in climate models occur due to a range of factors. Key factors include limited spatial resolution (large grid sizes), uncertain historical forcing, discretization, because of our limited understanding of physical processes, their simplified formulation in models, and imperfect initialization of models (Gohar et al. 2017).

To overcome these biases, a range of BC methods have been developed. For all methods, it is important to realize that the quality of the observational datasets determines the quality of the BC. For an effective BC, it is important to have a good dataset of observations. Long-term datasets are needed if correcting extreme statistics (Mehrotra & Sharma 2015).

The simplest approach used for BC is the so-called delta change of change factor approach (Hay et al. 2000). The delta-change has a long history in climate impact research. It represents not a BC as such, but employs the model's response to climate change to modify observations to represent a plausible future. As it is easy to use, it has formed a useful benchmark for BC in climate impact assessment studies. The approach uses the GCM or RCM response to climate change to modify observations. The delta change assumes that GCMs can accurately simulate relative changes. The observed time series is adjusted by either adding the difference or multiplying the ratio of change parameter between future and present climate as simulated by the GCM, without altering the pattern over time. For rainfall usually, a percentage change is calculated. If the climate model predicts a 20% increase in rainfall, a new time series will be made by multiplying the historic rainfall by 1.2. More complexity in the approach can be introduced by defining different change factors for different months or seasons. For temperature, an additive correction is adopted, for example, if the climate model predicts 3 °C higher temperatures, 3 °C is added to all historic observations to construct a new time series representing the future climate. It may be noted that this method does not take into account change in climate variability such as an increase in extreme rainfall or longer dry or wet spells. The selection of temporal domains, time periods, time scale, and type and number of change factors varies depending on the meteorological parameters and the specific needs of the study (Hansen et al. 2017).

Another popular alternative is quantile mapping (QM) (Panofsky et al. 1958; Haddad & Rosenfeld 1997; Wood et al. 2004; Déqué 2007; Piani et al. 2010; Gudmundsson et al. 2012). The approach modifies a modelled value by mapping quantiles of the model's distribution onto observations. This approach not only adjusts the mean and variance but also the quantiles for the variable of interest. The major limitation of the basic QM method, or for that matter, of almost all BC approaches, is the assumption of stationarity (i.e., the observed cumulative distribution functions (CDFs) represents the baseline period to correct, consequently forcing a model that remains stationary, or creating future distributions that are patterned after the observed). Considering this, many variants of QM have been proposed. For example, the equidistant cumulative function matching (EDCDF) by Li et al. (2010) incorporates information from the model projection CDF rather than the historic model distribution. The method does not assume the observed distribution to apply to the future period. However, it assumes instead that the difference between the observed and model-predicted data remains constant in the future. Similarly, Michelangeli et al. (2009) proposed a cumulative distribution function transformation (CDF-t), which assumes that the historical mapping between the model and observed CDFs applies to the future period as well. EDCDF preserves the GCM-predicted change at each quantile additively (i.e., as future minus observed). However, changes in precipitation are often more usefully evaluated as multiplicative changes, since a fixed amount of precipitation change has different implications in wet and arid regions (Pierce et al. 2015). The quantile delta mapping (QDM) (Cannon et al. 2015) preserves relative changes in simulated precipitation quantiles. This is equivalent to the QM being applied on the detrended climate series and adding back the trends thereafter.

The QM approach is known to alter the magnitude of mean change as projected by the raw GCM simulations (Hagemann et al. 2011; Pierce et al. 2013; Maurer & Pierce 2014). Since BC represents a purely statistical ‘fix’, it does not discriminate between the physical processes associated with anthropogenic forcing and shorter-term fluctuations associated with natural climate variability internal to the model. To reduce the disparity between global modelling studies based on bias-corrected outputs from a GCM/RCM, it is prudent to implement a BC alternative that does not alter the original GCM trend. A simple adjustment to EDCDF can retain the model-predicted future change in mean precipitation that has been evaluated as a ratio (Wang & Chen 2014).

NBC approach

The QM, CDF-t, and EDCDF approaches described in the preceding section focus on either monthly or daily attributes of the climate variables being corrected. However, in some situations and for certain types of variables, longer-term variabilities need better representation to enable aggregated time scale characterization, for example, in the case of modelling drought and the availability of water resources. The missing interannual variability is addressed by using observed time series attributes at multiple time scales and performing time nesting, termed as the NBC (Johnson & Sharma 2012; Mehrotra & Sharma 2012). The NBC approach considers BC at multiple time scales. For example, a three-level NBC starts with BC at daily time scale. The daily observed and bias-corrected series is then aggregated to a monthly time scale and BC is applied. The observed monthly and BC series are aggregated to an annual time scale and BC is applied again. Finally, BC at all time scales is incorporated. As time nesting was found to create artefacts in bias-corrected series attributes at daily time scales, Mehrotra & Sharma (2012) repeats the NBC procedure recursively to minimize the biases at all time scales. This modification was termed as RNBC (Mehrotra & Sharma 2012).

Figure 1 illustrates the applicability of NBC in contrast to the quantile matching approach. As can be noted, there are clear benefits in terms of the representation of leg dependence attributes, of considerable importance from a water management point of view.
Figure 1

Scatter plots of daily, monthly, seasonal and annual means, standard deviations and lag-1 autocorrelations of reanalysis, and raw and bias-corrected GCM data, for the Sydney region, as reported in Mehrotra & Sharma (2012). Points on the plots denote variables. For daily and annual plots 6, for monthly 6*12 and for seasonal 6*4 points are shown. Means and standard deviations of all variables are rescaled to lie between −100 and 100.

Figure 1

Scatter plots of daily, monthly, seasonal and annual means, standard deviations and lag-1 autocorrelations of reanalysis, and raw and bias-corrected GCM data, for the Sydney region, as reported in Mehrotra & Sharma (2012). Points on the plots denote variables. For daily and annual plots 6, for monthly 6*12 and for seasonal 6*4 points are shown. Means and standard deviations of all variables are rescaled to lie between −100 and 100.

Close modal

Multivariate BC approaches

BC approaches are generally applied to a single variable at a location and at a given time scale (Li et al. 2014; Mehrotra & Sharma 2015, 2016; Vrac & Friederichs 2015). As BC is applied to each variable independently, the physical dependencies across variables of interest are ignored (Colette et al. 2012; Maraun 2013). Often, bias-corrected variables are used collectively to estimate quantities such as potential evapotranspiration or their use in impact assessment models, such as rainfall–runoff modelling. Similarly, variable bias-corrected in isolation at different locations might reflect poor observed spatial correlations and lead to improper spatial response (Hnilica et al. 2017; Nahar et al. 2017).

Multivariate BC alternatives have been proposed to address some of these issues. A bivariate correction was proposed by Piani & Haerter (2012) to simultaneously correct temperature and precipitation. This was achieved by correcting one time series (i.e., precipitation) conditionally to the bias-corrected values of the other variable's time series (i.e., temperature). Copula-based methods have also been proposed to consider the joint dependence between variables or the spatial dependence across grids (Mao et al. 2015; Vrac & Friederichs 2015). Mehrotra & Sharma (2015) proposed a parametric multivariate extension of NBC, named as multivariate recursive nested bias correction (MRNBC), while a multivariate and multi-timescale extension of quantile matching-based nonparametric BC, named multivariate recursive quantile nested bias correction (MRQNBC), was suggested by Mehrotra & Sharma (2016).

Two versions of a multivariate BC were proposed by Cannon (2018). These approaches consider quantile mapping to match observed marginal distributions and Pearson correlation and Spearman rank correlations to match dependence structure. François et al. (2020) compared the performances of four parametric and nonparametric multivariate BC approaches and reported that depending on the dimensional configuration, the instability of some methods can possibly affect the bias-corrected simulations. In particular, for multivariate parametric alternatives, increasing the number of variables often led to a deterioration of spatial properties.

With an excessive number of variables and grid points, a parametric option, for example, MRNBC or MRQNBC might struggle with dimensionality issues as it maintains spatiotemporal dependence through statistics of matrices of variables at grid points. This is referred to as the ‘Curse of Dimensionality’ in the literature. With increasing dimensions, stability could be an issue. In such cases, the use of shuffling-based BC provides an effective means of dealing with the dimensionality. In shuffling-based approaches, a univariate BC is used to correct the distribution biases while shuffling (reordering) is used to correct the dependence biases. As both procedures are applied to individual variables, the approach is independent of the number of variables or grid points. An equivalent alternative has been recently proposed by Kusumastuti et al. (2023) using a continuous wavelets transformation, also highlighting the importance of not extending corrections into the far future, as the parameterization of the BC model in the current climate simulation and observed record may not remain valid under extreme change scenarios.

Vrac & Friederichs (2015) combined a univariate BC method with the shuffling technique presented by Clark et al. (2004) and referred to the technique as the empirical copula–BC (EC–BC) approach. A further modification to this approach was proposed by Vrac (2018) wherein a multivariate method of rank resampling for distributions and dependences was proposed. The approach was named the R2D2 BC approach. The author claimed that the approach provides some stochasticity since it allows generating as many multivariate corrected outputs as the number of grid cells times the number of climate variables (statistical dimensions) of the simulations to be corrected.

While the above-referenced studies focus on correcting multivariate biases present in GCM simulations mostly, an exciting new domain is the correction of lateral and lower boundary biases in RCM simulations (Rocheta et al. 2017, 2020). Correction of such biases has been shown to improve the consistency with which extreme precipitation is simulated for future climates across model simulations (Kim et al. 2020), with arguments on why multivariate biases are critical to address to improve this further (Kim et al. 2021). This has since been confirmed with RCM lateral and lower boundaries simulated using MRQNBC (Kim et al. 2023) and shown to improve even the simulations of compound extreme events (Kim et al. 2023). This is a promising direction for future applications before any dynamical downscaling is undertaken.

Figure 2 illustrates the benefits any multivariate BC offers over a univariate correction alternative. While the illustration here uses MRQNBC, cross-dependence attributes are represented well in any of the alternatives that have been discussed. Lagged-cross-correlations, however, require the treatment of persistence biases, which the MRQNBC or its variants alone are able to address. The relevance of addressing these biases is felt most in derived streamflow simulations and their use for design or planning activities, such as the dependence between high temperature (evaporation) that can create dryness and later induce an extreme precipitation event. The lack of such consideration may simulate less-dry antecedent conditions, which will distort the derived streamflow hydrograph profile. While this may not matter for extreme flood events (and the rainfall will dominate that simulation), it is of great relevance in frequent flood simulations which drive water supply operations worldwide (Dykman et al. 2023).
Figure 2

Scatter plots of daily, monthly, seasonal, and annual cross and Lag-1 cross-correlations of reanalysis and raw and bias-corrected GCM data for Sydney region as reported in Mehrotra & Sharma (2016).

Figure 2

Scatter plots of daily, monthly, seasonal, and annual cross and Lag-1 cross-correlations of reanalysis and raw and bias-corrected GCM data for Sydney region as reported in Mehrotra & Sharma (2016).

Close modal
Recently, Mehrotra & Sharma (2019) proposed a multivariate three-dimensional bias correction (3DBC) alternative to correct distribution and dependence biases across multiple variables, locations, and time scales. In this, a univariate QM is adopted to remove the distribution biases, and thereafter, resampling of climate data is conducted to follow the observational ranking to match the observed dependence. The changes in the dependence attributes are ascertained by resampling the historical ranks into what these might resemble in the future. The approach is not limited in terms of the number of variables, grid points in space, and the time scale considered. Most importantly, it maintains the shift in dependence and other attributes between the current and the future climate as ascertained by a climate model. Resampling of corrected ranks offers a very simple, flexible, and effective general-purpose multivariate, multi-time, and multi-location BC alternative for current and future climate, highly suited to address the Curse of Dimensionality described earlier. As the approach works in three dimensions, space, time, and variables, it is denoted as 3DBC or three-dimensional BC. More details on the 3DBC approach and results are available in Mehrotra & Sharma (2019). An illustration of the method and the advantages it offers over the other alternatives discussed earlier is available in Figure 3.
Figure 3

Scatter plots of daily and annual LAG0 cross-correlations across grid points of observed, raw, and QM and 3DBC bias-corrected time series for precipitation (top two rows) and daily maximum temperature (bottom two rows).

Figure 3

Scatter plots of daily and annual LAG0 cross-correlations across grid points of observed, raw, and QM and 3DBC bias-corrected time series for precipitation (top two rows) and daily maximum temperature (bottom two rows).

Close modal

Correcting bias in the frequency domain

Daily or monthly QM BC has been used in many climate change impact studies (Wood et al. 2004; Boé et al. 2007; Haerter et al. 2011). However, where low-frequency variability and persistence attributes are important, QM exhibits limited advantage as it corrects for biases at a given time scale and not at higher aggregated time scales (Haerter et al. 2011; Mehrotra & Sharma 2016). One possible solution to this is the nesting of corrections at multiple pre-defined time scales (Haerter et al. 2011; Johnson & Sharma 2012; Mehrotra & Sharma 2016). However, biases at other time scales, not included in the nesting, may not be corrected. Another solution is to look beyond specific time scales. One such possibility is representing the time series in the frequency domain. The use of the frequency space makes the time series more suitable for modulating variability across different time scales (Bloomfield 2004; Maheswaran & Khosa 2012). A Fourier transform or spectral analysis has been widely employed in meteorology and hydroclimatology to examine seasonal, annual, and multi-annual periodicities (Leite & Peixoto 1995; Hegge & Masselink 1996; Fleming et al. 2002). Besides, the Fourier transform is also applied to generate surrogate time series, maintaining the linear properties present in the synthetically generated series. More details on this are found in Prichard & Theiler (1994), Chen et al. (2010) and Keylock (2012). Fast Fourier transform (FFT) has also been used to bias correct the variability for each frequency component (Pierce et al. 2015; Nguyen et al. 2016). Wavelet analysis is a step beyond Fourier analysis, localizing the signal in the frequency domain and providing spectral representation as a function of time. Both discrete and continuous wavelet transform have been used for a range of applications (Farge 1992; Torrence & Compo 1998; Torrence & Webster 1999; Jiang et al. 2020). Discrete wavelet transforms disaggregate the time series (signals) into multiple high- and low-frequency components (Percival & Walden 2000). Therefore, the slow-moving component of a time series which represents the underlying low-frequency behaviour can be extracted and modelled separately (Maheswaran & Khosa 2012; Belayneh et al. 2016; Sang et al. 2016). Additionally, the lowest-frequency component can serve to represent the underlying trend (Adarsh & Janga Reddy 2015). A DWBC alternative was proposed by Kusumastuti et al. (2021) to obtain the underlying trend and to remove bias in variability at varying time scales (frequencies). The principle of discrete analysis provides the information on the disaggregated time series at dyadic frequency components. For some variables, the underlying information of climate variables, such as the low-variability component, may lie in the non-dyadic spectrum. Keeping it in mind, Kusumastuti et al. (2022) proposed a continuous wavelet transform-based bias correction (CWBC). A note of caution for using BC under simulations of extreme change (for a 4xCO2 change scenario) was provided in Kusumastuti et al. (2023), which showed that even while multivariate biases were properly corrected and not affected by the Curse of Dimensionality, the parameterization of the BC model becomes unstable far into the future where the change (and hence bias) is not manifested in the current climate simulations.

Climate models employ fundamental mass and energy balance principles and therefore future changes across climate variables are derived in a physically consistent way. Unfortunately, while reducing in magnitude in each new generation of the climate model, biases continue to exist. For some applications, these biases may present significant obstacles when attempting to project into the future, particularly if the impact or application is sensitive to non-linearities in the system such as univariate or multivariate threshold effects. Statistical methods can correct model bias, though doing so always endangers the physical consistency between the corrected climate variables. This is especially the case with univariate corrections for individual climate variables but continues to impact even with more sophisticated multivariate alternatives. It is recommended to use bias-corrected data if the application only requires precipitation and/or temperature and it is sensitive to non-linearities (such as thresholds) in precipitation and/or temperature. Univariate BC can be used if the application requires variables other than precipitation and temperature but it is not sensitive (or has little sensitivity) to the coupled effect of other variables with precipitation and temperature. Univariate BC should be avoided if the application has strong sensitivities to the combined effect of temperature and/or precipitation with other climate variables such as humidity or wind (in which case physical consistency may be a dominant requirement) (Evans & Argüeso 2014).

Climate models are routinely bias corrected by comparing the modelled simulations for the current climate with the observations. Therefore, the selection of the reference time period warrants greater attention in impact studies. The bias in climate model results cannot be uniquely defined but depends on the reference period used (Maraun 2016). An associated challenge is the presence of a trend within the observed period, which, if significant (as is the case with temperature) needs careful consideration.

Precipitation and temperature (minimum and maximum) are the variables, used commonly in impact assessment applications. These variables are routinely bias corrected as they are the variables with long, reliable observational time series. However, for some studies, other variables are also needed.

Apart from direct statistics of interest, GCMs substantially misrepresent other crucial phenomena as well. For example, key processes governing low-frequency variability such as El Niño/Southern Oscillation (ENSO), Monsoonal systems, or the mid-latitude storm track are biased. Also, for precipitation, GCMs are known to have a ‘drizzling’ bias that is, characterized by high precipitation frequency and low intensity (Dai 2006; DeMott et al. 2007). The ‘drizzling’ bias impedes the realistic representation of precipitation characteristics and hydrologic extremes in the climate models (Trenberth et al. 2003; Trenberth 2011). We discuss here some of these issues and possible solutions suggested.

Choice of reference period in BC

The World Meteorological Organization (WMO) recommends using the 30-year 1961–1990 period as the climate normal when comparing with future periods. It is recommended that this should be maintained as a reference for monitoring long-term climate variability and change and BC (Trewin 2007; WMO 2014). However, a regularly updated 30-year baseline period, for example, 1981–2010, or 1991–2020 can also be employed. The Intergovernmental Panel on Climate Change (IPCC) used the 20-year period 1986–2005 as the baseline in many assessment results in the Fifth Assessment Report (Stocker et al. 2015) and years 1995–2014 in its Sixth Assessment Report (IPCC 2023). Scientists usually measure rising global temperatures against the baseline of the years between about 1850 and 1900 (defined as the ‘pre-industrial period’), ‘when fossil-fuel burning had yet to change the climate’. In principle, ‘pre-industrial levels’ could refer to any period before the industrial revolution. However, the number of direct temperature measurements decreases as one goes back in time. Defining a ‘pre-industrial’ reference is therefore a compromise between the reliability of the temperature information and how well it represents pre-industrial conditions.

The choice of the baseline period (as a reference for assessing future changes for any projected variable) can play an important role in the ensuing assessment. In regional climate impact studies, well-established or arbitrarily chosen baselines are often used without questioning. Also, BC methods require sufficient observational data to characterize the reference climatology. The use of 30 years of data to include some variations at the decadal timescale (Li et al. 2010; Mehrotra & Sharma 2016) is common.

It has been suggested (Maraun 2016) that the bias in climate model results cannot be uniquely defined but depends on the reference period adopted. While the 1961–1990 period is frequently used as a reference, CORDEX climate simulations were bias corrected using 1989–2010 as a reference time window. Hence, the climate simulated during 1961–1990 is likely to be different from observations, although the magnitude of the differences is unknown. Only a few studies have investigated this issue in detail. For example, Zhang et al. (2018) in a study found that baseline period choice is an important source of uncertainty when studying the impacts of land use and climate change on hydrology. Similarly, Kotlarski et al. (2019) found that the observational uncertainty (i.e., uncertainty related to the differences between reference observation-based datasets) is smaller than RCM uncertainty at a European scale, although the observational uncertainty can be larger in some regions. A possibly large effect of observational uncertainty was corroborated also in the study by Sun et al. (2018) who reported a variation of annual precipitation estimates of as much as 300 mm/year among 30 climate datasets.

BC approaches with trend correction

Some variables, for example, temperature, show clear increasing trends in the future and the BC method applied should preserve the trend of the raw data.

It is argued that trends of raw climate series should be preserved so that the BC does not impact climate sensitivity (Hempel et al. 2013). For atmospheric moisture and associated variables (precipitation), preserving the relative change signal is also important for maintaining physical scaling relationships with model-projected temperature change (e.g., Clausius–Clapeyron scaling of 7% suggests an equivalent increase in atmospheric moisture or column of water vapour for each degree Centigrade rise in temperature (Wasko et al. 2016). A trend-preserving BC (Hempel et al. 2013) was applied to CMIP5 representative concentration pathway (RCP) GCM experiments (Taylor et al. 2012). The approach was found to preserve monthly trends, but trends in the daily extremes were modified by the QM. Similarly, a detrended quantile mapping (DQM) algorithm was applied by Bürger et al. (2013), and results indicated that while the approach tends to maintain the modelled long-term trend in the mean, it does not guarantee that trends in precipitation extremes, as governed by the tails of the distribution, are also preserved.

Another variant of QM, the quantile delta change or quantile perturbation method, was used for precipitation projections (Olsson et al. 2009; Willems & Vrac 2011; Sunyer et al. 2015). The approach modifies historical observations by superimposing relative trends in quantiles from the climate. In this case, modelled trends in all quantiles, including in the tails, are preserved. It may be noted that as the approach adjust observed series, it does not explicitly bias correct the raw time series of the climate model.

A detrended QM was used by Hempel et al. (2013) and Bürger et al. (2013) to preserve monthly trends. A modification to the basic QM was presented by Cannon et al. (2015), termed DQM. Relative to QM, DQM incorporates additional but limited information about the climate model simulation corresponding to the projected period, in this case in the form of the projected mean. Depending on the degree of extrapolation still required after detrending (and the extrapolation algorithm used), the climate change signal from DQM will tend to match that of the underlying climate model. This applies to the mean but not necessarily to all quantiles, such as those representing the tails of the distribution that define climate extremes.

Similarly, Cannon et al. (2015) modified the QM method and outlined an approach called QDM and claimed that it is not constrained by the stationarity assumption. They also showed that QDM is not very different from equidistant-quantile matching (EQM) if BC is applied additively. Switanek et al. (2017) proposed a nonstationary QM correction and named it ‘Scaled Distribution Mapping’, which is similar to QDM but explicitly accounts for the number of rainy days and wet spells.

Recent applications of time-frequency-based BC approaches, DWBC and CWBC, have shown that these approaches allow to maintain the underlying trend of the time series while correcting for the biases in mean and standard deviation across the spectrum. The delta function introduced in the approach maintains the continuity of the trend from current to future climate (Kusumastuti et al. 2021, 2022).

Retaining model-predicted mean changes in the bias-corrected time series

Previous studies have noted that traditional QM modifies the mean of climate change signals (Hagemann et al. 2011; Pierce et al. 2013; Maurer & Pierce 2014). This can introduce inconsistencies in the BC results. QM is also known to alter the variability and trend at all time scales (Pierce et al. 2013; Maurer & Pierce 2014). If a climate variable shows too much variability, QM reduces it and if the variable has too little variability, QM increases it.

The model-predicted mean changes in the bias-corrected series can be retained by further adjusting the QM bias-corrected results. The QM approach preserves the magnitude of error in the current climate model value for the same future climate value. EQM assumes the magnitude of error in a current climate model value at a given quantile is preserved in the future. CDF-t assumes the magnitude of error in the current climate model quantile at a given quantile is preserved in the future. Thus, one needs to consider the procedure which would preserve the magnitude of error in the current climate model quantile to be preserved in the future. EQM preserves model-predicted mean changes for variables where the correction is applied additively. However, the approach requires modification for variables where the correction is applied multiplicatively. Pierce et al. (2015) proposed an extension to the EQM approach to retain model-predicted mean changes. They suggested multiplying the EQM corrected values by a correction factor, , where y is the change (calculated as a ratio) in the value of the variable and is the change in the mean value following BC. The angle brackets indicate that the mean is taken over a time period, for example, a month.

One way in which extrapolation can be avoided to some extent and the trend in the raw GCM data can also be maintained is to first detrend the series, apply BC, and reimpose the trend afterward. The trend removal is likely to shift the future climate distribution closer to that of the current climate. Adding the trend afterward will retain the trend in the time series.

Adjustment of the number of wet days in the precipitation time series

If the number of wet days of a raw climate model precipitation simulation does not match the observed ones, an adjustment of the values is needed as the BC method might not be able to correct precipitation intensity appropriately. There could be two cases. The first one relates to the situation where the raw data have too many wet days compared to the observations, a common case with the majority of GCMs. In the second case, raw data have fewer wet days in comparison to the observations.

When the raw model simulation has too many wet days, a simple approach is to define a threshold (TH), such that the frequency of days with model precipitation > TH is the same as the frequency of rainy days in the observed dataset (Ines & Hansen 2006; Schmidli et al. 2006; Lavaysse et al. 2012; Nguyen et al. 2018; Mehrotra & Sharma 2019). Before the application of BC, model precipitation values below TH are set to zero. As an alternative, it is also possible to apply a BC model directly on the complete time series including irrespective of dry or wet days (Piani et al. 2010; Vrac et al. 2012; Mehrotra & Sharma 2016). When QM is applied on the full-time series, in order to match the number of zeros, in case of an excessive number of wet days, some days will be assigned negative values and setting them to zero will result in an appropriate number of wet days.

When the raw model simulation has fewer numbers of wet days than observations, the threshold approach cannot be applied. Mehrotra & Sharma (2019) suggested adding more rainy days by picking the dry days at the start or end of dry spells and adding a minimum rainfall threshold. Adding the threshold modifies the selected dry days into wet days and achieves the correct number of wet days. As with the case of an excessive number of wet days, QM can adjust precipitation values and match the number of wet days. This adjustment procedure is applied by comparing the current climate data with the observations. Future climate data are adjusted in the same proportion.

Climate variables used in BC

Precipitation and temperature are the two primary climate variables frequently considered in the BC approaches. The majority of studies have considered monthly or daily precipitation (Haddad & Rosenfeld 1997; Wood et al. 2004; Boé et al. 2007; Déqué 2007; Graham et al. 2007; Moore et al. 2008; Olsson et al. 2009; Piani et al. 2010; Willems & Vrac 2011; Gudmundsson et al. 2012; Lafon et al. 2013; Cannon et al. 2015). Precipitation is more difficult to bias correct than temperature because it is intermittent in nature (rainy days followed by days with no rainfall), has a skewed distribution and zero lower bounds, and often behaves differently at aggregated spatial and temporal scales.

In addition to precipitation and temperature, other climate variables and phenomena including wind, humidity, streamflow, evaporation, bushfire, surface pressure, geopotential height, mean sea level pressure, and air temperature have also been bias-corrected and used in climate change impact studies (Bürger 1996; Matyasovszky & Bogardi 1996; Enke & Spekat 1997; Mehrotra et al. 2004; Mehrotra & Sharma 2010; Iizumi et al. 2017; Dowdy et al. 2019).

Please note that statistical and dynamical downscaling techniques are routinely used to enhance the regional information provided by GCMs by combining large-scale climatic information with small-scale behaviour and dynamics. These large-scale variables often require some kind of BC before their use in the downscaling models.

There is no perfect BC approach. All existing BC approaches offer some limitations in terms of their application coupled with their methodological limitations. These limitations and assumptions are a source of uncertainty in bias-corrected RCM rainfall and the resulting hydrological impact assessments and evaluations estimated using them (Teutschbein & Seibert 2012; Maraun & Widmann 2018; Potter et al. 2018). It is important to note that no BC approach can correct for errors inherited from the RCMs or GCMs in temporal sequencing, seasonality, or biases in large-scale circulation patterns that could lead to unrealistic and non-physical climate projections (Stocker et al. 2015).

In general, a BC is subject to the following limitations:

  • BC methods assume that the magnitude and nature of BC are stationary and will not change into the future. This assumption, however, has been questioned by researchers from time to time (Ehret et al. 2012; Maraun 2016). Recent studies (e.g., Nahar et al. 2017; Hui et al. 2019) have reported nonstationary biases in the climate model precipitation. However, realizing that using raw model simulations will perhaps be even more questionable, many studies choose to bias correct the raw climate data before use. Some modifications in the BC approaches trying to account for the non-stationarity in the biases have been proposed (Cannon et al. 2015; Switanek et al. 2017):

  • BC is a purely statistical alteration to the raw model simulation. It cannot discriminate between the physical processes determining trends associated with climate change, or with the natural climatic variability internal to the model.

  • BC methods require sufficient observational data to characterize the reference climatology. Usually, 30 years of data is considered as a minimum requirement, mostly to include low-frequency variability or variations across years at the decadal time scales (Mehrotra & Sharma 2016, 2019).

  • The quality of the observed data affects the quality of bias-corrected data and how well the climate model is able to represent the relevant physical processes that govern the variable of interest.

  • The physical consistency of the different climate variables may not be maintained if they are bias-corrected independently. For example, bias-correcting temperature may result in sub-zero values, whereas rainfall does not convert to snowfall in simulations. Negative values for diurnal temperature range may be generated from daily minimal and maximal temperatures. This will be important in some applications but not others. Similarly, it may rain when the temperature is above 35°, the occurrence of which has been documented as rare (Roderick et al. 2019, 2020).

  • BC is not recommended if the biases are large (Murphy et al. 2018). Also, the GCMs or ensembles that show large biases in the analysis should be dropped (Fung 2018). Interpretation of any bias-corrected data needs to take into account the assumptions stated above.

  • Simple daily rainfall BC approaches offer limited ability to correct for biases in the wet and dry spells (continuous sequence of wet or dry days) and wet spell totals (Chen et al. 2013; Addor & Seibert 2014; Evans et al. 2017).

  • When developing a BC method to adjust the simulations, the different sources of uncertainties contributing to the biases are not differentiated.

BC is a reasonable choice that allows using the output of existing climate models and conducting analyses to understand the climate change impacts. However, the assumption of stationarity requires careful consideration as there is evidence that a warming world can change the climate response in the future, for example, extreme precipitation, temperature, and evapotranspiration (Maraun 2016; Nahar et al. 2017; Hui et al. 2019). Given that the climate continues to change, traditional BC methods may become less effective, and new approaches will need to be developed to account for these changes. The use of innovative techniques, such as deep learning, may also play an increasingly important role in the future of BC. These methods have shown promise in other fields and may prove useful in the BC field as well. Overall, the future of BC approaches for climate variables is likely to involve the development of more advanced statistical and machine learning techniques that can address the many challenges associated with working with climate data.

Outputs of GCMs and RCMs are routinely used as inputs to impact assessment models. Despite the modelling and scientific developments, climate simulations are often biased when compared to observations, leaving modellers with questions as to how they can rely on projections issued into the future. Using raw outputs can significantly affect the results of impact studies, necessitating the use and development of a range of alternatives that correct systematic climate model biases.

The usefulness of climate models improves, often significantly, if systematic bias can be removed. To do this, however, the bias needs to be quantified, and appropriate corrections need to be devised. These corrections may be simple re-parameterizations of the model or a re-specification of the model structure. Identifying, characterizing, and removing such bias, however, become especially tricky if the model is complex, and its inputs (as well as observed responses used for assessment) uncertain. Such is the scenario that characterizes the biases inherent in climate model simulations of the future.

BC is routinely used to correct the statistical biases of the raw simulations of climate models before using them as input in climate change assessment studies. It aims to adjust selected statistics of a climate model simulation to better match observed statistics over the current climate reference period. Depending upon the complexity, it may adjust marginal statistics only, or also multi-time scale, multi-variable, and multi-locational aspects. A fundamental assumption of BC is that the chosen climate model produces meaningful output, including a credible representation of climate change with consistent biases. BC cannot fix the fundamental problems of a climate model.

If the main interest of a study is to assess the relative change in a future climate, bias is not a big problem, and BC can be ignored. BC is important for studies involving a threshold analysis (crop behaviour at certain temperatures, rainfall, and temperature extremes, heat waves, etc.), and water availability analysis as any bias in rainfall can have a large impact on the outcome and in the estimation of extremes as future changes in flood risk can go wrong.

There are trade-offs between bias-corrected climate data versus directly modelled climate data used in climate change impact assessment studies. These trade-offs should be looked into when modelling impact assessment frameworks for specific applications; for example, it makes sense to use bias-corrected data to assess the climate change impacts on reservoir storage and operation than using it to investigate the impacts on emissions and air quality (Liu et al. 2014). Do not bias-correct if the biases are large, e.g., the Met Office Hadley Centre Model (HadGEM3-GC3.05) subset of the global projections shows significant bias in winter surface air temperature in large parts of the Northern Hemisphere, although not over the UK (see Section 3.4 of Land Science Report (Murphy et al. 2018). You should disregard the members that show these large biases in your analysis.

The results discussed in the paper are based on the output of CMIP5 models, although the BC approach is independent of model type, group, or phase. The CMIP6 modelling project is based on an improved understanding of the Earth's climate system. The improvements in representations of processes such as clouds, aerosols, and biogeochemistry, spatial resolution, and data sharing in CMIP6 are expected to lead to more accurate and reliable projections of future climate change, as well as a better understanding of the underlying processes driving these changes, leading to reduced biases.

Currently used BC methods largely neglect feedback mechanisms, and it is unclear whether they are time-invariant under climate change conditions. Applying BC increases agreement of climate model output with observations in the current climate and hence narrows the uncertainty range of simulations and predictions without, however, providing a satisfactory physical justification. In any case, it should be acknowledged that a successful BC relies on a sound understanding not only of the statistical model but also the relevant climatic processes and their representation of the considered climate model.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.

Adarsh
S.
&
Janga Reddy
M.
2015
Trend analysis of rainfall in four meteorological subdivisions of southern India using nonparametric methods and discrete wavelet transforms
.
International Journal of Climatology
35
(
6
),
1107
1124
.
https://doi.org/10.1002/joc.4042.
Addor
N.
&
Seibert
J.
2014
Bias correction for hydrological impact studies – beyond the daily perspective
.
Hydrological Processes
28
(
17
),
4823
4828
.
https://doi.org/10.1002/hyp.10238.
Ashfaq
M.
,
Skinner
C. B.
&
Diffenbaugh
N. S.
2011
Influence of SST biases on future climate change projections
.
Climate Dynamics
36
(
7
),
1303
1319
.
doi:10.1007/s00382-010-0875-2
.
Bárdossy
A.
&
Pegram
G.
2012
Multiscale spatial recorrelation of RCM precipitation to produce unbiased climate change scenarios over large areas and small
.
Water Resources Research
48
(
9
).
https://doi.org/10.1029/2011WR011524.
Beadling
R. L.
,
Russell
J. L.
,
Stouffer
R. J.
,
Mazloff
M.
,
Talley
L. D.
,
Goodman
P. J.
,
Sallée
J. B.
,
Hewitt
H. T.
,
Hyder
P.
&
Pandde
A.
2020
Representation of Southern Ocean properties across coupled model intercomparison project generations: CMIP3 to CMIP6
.
Journal of Climate
33
(
15
),
6555
6581
.
doi:10.1175/jcli-d-19-0970.1
.
Belayneh
A.
,
Adamowski
J.
,
Khalil
B.
&
Quilty
J.
2016
Coupling machine learning methods with wavelet transforms and the bootstrap and boosting ensemble approaches for drought prediction
.
Atmospheric Research
172–173
,
37
47
.
doi:10.1016/j.atmosres.2015.12.017
.
Bellenger
H.
,
Guilyardi
E.
,
Leloup
J.
,
Lengaigne
M.
&
Vialard
J.
2014
ENSO representation in climate models: From CMIP3 to CMIP5
.
Climate Dynamics
42
(
7
),
1999
2018
.
doi:10.1007/s00382-013-1783-z
.
Berg
P.
,
Feldmann
H.
&
Panitz
H. J.
2012
Bias correction of high resolution regional climate model data
.
Journal of Hydrology
448–449
,
80
92
.
https://doi.org/10.1016/j.jhydrol.2012.04.026
Bloomfield
P.
2004
Fourier Analysis of Time Series: An Introduction
.
John Wiley & Sons
,
New York, USA
.
Boé
J.
,
Terray
L.
,
Habets
F.
&
Martin
E.
2007
Statistical and dynamical downscaling of the Seine basin climate for hydro-meteorological studies
.
International Journal of Climatology
27
(
12
),
1643
1655
.
https://doi.org/10.1002/joc.1602.
Box
G. E. P.
1976
Science and statistics
.
Journal of the American Statistical Association
71
(
356
),
791
799
.
doi:10.1080/01621459.1976.10480949.
Bürger
G.
,
Schulla
J.
&
Werner
A. T.
2011
Estimates of future flow, including extremes, of the Columbia River headwaters
.
Water Resources Research
47
(
10
).
https://doi.org/10.1029/2010WR009716.
Bürger
G.
,
Sobie
S. R.
,
Cannon
A. J.
,
Werner
A. T.
&
Murdock
T. Q.
2013
Downscaling extremes: An intercomparison of multiple methods for future climate
.
Journal of Climate
26
(
10
),
3429
3449
.
doi:10.1175/jcli-d-12-00249.1
.
Cannon
A. J.
,
Sobie
S. R.
&
Murdock
T. Q.
2015
Bias correction of GCM precipitation by quantile mapping: How well do methods preserve changes in quantiles and extremes?
Journal of Climate
28
(
17
),
6938
6959
.
doi:10.1175/jcli-d-14-00754.1
.
Chen
J.
,
Brissette
F. P.
&
Leconte
R.
2010
A daily stochastic weather generator for preserving low-frequency of climate variability
.
Journal of Hydrology
388
(
3
),
480
490
.
https://doi.org/10.1016/j.jhydrol.2010.05.032.
Chen
J.
,
Brissette
F. P.
,
Chaumont
D.
&
Braun
M.
2013
Finding appropriate bias correction methods in downscaling precipitation for hydrologic impact studies over North America
.
Water Resources Research
49
(
7
),
4187
4205
.
doi:10.1002/wrcr.20331
.
Chen
J.
,
Brissette
F. P.
&
Lucas-Picher
P.
2015
Assessing the limits of bias-correcting climate model outputs for climate change impact studies
.
Journal of Geophysical Research: Atmospheres
120
(
3
),
1123
1136
.
https://doi.org/10.1002/2014JD022635.
Clark
M.
,
Gangopadhyay
S.
,
Hay
L.
,
Rajagopalan
B.
&
Wilby
R.
2004
The Schaake shuffle: A method for reconstructing space–time variability in forecasted precipitation and temperature fields
.
Journal of Hydrometeorology
5
(
1
),
243
262
.
doi:10.1175/1525-7541(2004)005 < 0243:Tssamf > 2.0.Co;2
.
Colette
A.
,
Vautard
R.
&
Vrac
M.
2012
Regional climate downscaling with prior statistical correction of the global climate forcing
.
Geophysical Research Letters
39
(
13
).
https://doi.org/10.1029/2012GL052258.
Collins
M.
,
Knutti
R.
,
Arblaster
J.
,
Dufresne
J.-L.
,
Fichefet
T.
,
Friedlingstein
P.
,
Gao
X.
,
Gutowski
W. J.
,
Johns
T.
&
Krinner
G.
2013
Climate Change 2013 - The Physical Science Basis: Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
.
Cambridge University Press
,
Cambridge
, pp.
1029
1136
.
Dai
A.
2006
Precipitation characteristics in eighteen coupled climate models
.
Journal of Climate
19
(
18
),
4605
4630
.
doi:10.1175/jcli3884.1
.
DeMott
C. A.
,
Randall
D. A.
&
Khairoutdinov
M.
2007
Convective precipitation variability as a tool for general circulation model analysis
.
Journal of Climate
20
(
1
),
91
112
.
doi:10.1175/jcli3991.1
.
Dowdy
A. J.
,
Ye
H.
,
Pepler
A.
,
Thatcher
M.
,
Osbrough
S. L.
,
Evans
J. P.
,
Di Virgilio
G.
&
McCarthy
N.
2019
Future changes in extreme weather and pyroconvection risk factors for Australian wildfires
.
Scientific Reports
9
(
1
),
10073
.
doi:10.1038/s41598-019-46362-x
.
Dykman
C.
,
Sharma
A.
,
Wasko
C.
&
Nathan
R.
2023
Can annual streamflow volumes be characterised by flood events alone?
.
Journal of Hydrology
617
,
128884
.
doi:https://doi.org/10.1016/j.jhydrol.2022.128884.
Eden
J. M.
,
Widmann
M.
,
Maraun
D.
&
Vrac
M.
2014
Comparison of GCM- and RCM-simulated precipitation following stochastic postprocessing
.
Journal of Geophysical Research: Atmospheres
119
(
19
),
11040
11053
.
https://doi.org/10.1002/2014JD021732.
Ehret
U.
,
Zehe
E.
,
Wulfmeyer
V.
,
Warrach-Sagi
K.
&
Liebert
J.
2012
HESS opinions ‘Should we apply bias correction to global and regional climate model data?’
.
Hydrology and Earth System Sciences
16
(
9
),
3391
3404
.
doi:10.5194/hess-16-3391-2012
.
Evans
J. P.
&
Argüeso
D.
2014
Guidance on the Use of Bias Corrected Data
.
NARCliM Technical Note 3, 7pp. NARCliM Consortium
,
Sydney, Australia
.
Evans
J. P.
,
Argueso
D.
,
Olson
R.
&
Di Luca
A.
2017
Bias-corrected regional climate projections of extreme rainfall in south-east Australia
.
Theoretical and Applied Climatology
130
(
3
),
1085
1098
.
Fan
X.
,
Miao
C.
,
Duan
Q.
,
Shen
C.
&
Wu
Y.
2020
The performance of CMIP6 versus CMIP5 in simulating temperature extremes over the global land surface
.
Journal of Geophysical Research: Atmospheres
125
(
18
),
e2020JD033031
.
https://doi.org/10.1029/2020JD033031.
Farge
M.
1992
Wavelet transforms and their applications to turbulence
.
Annual Review of Fluid Mechanics
24
(
1
),
395
458
.
doi:10.1146/annurev.fl.24.010192.002143
.
Fleming
S. W.
,
Marsh Lavenue
A.
,
Aly
A. H.
&
Adams
A.
2002
Practical applications of spectral analysis to hydrologic time series
.
Hydrological Processes
16
(
2
),
565
574
.
doi:10.1002/hyp.523
.
François
B.
,
Vrac
M.
,
Cannon
A. J.
,
Robin
Y.
&
Allard
D.
2020
Multivariate bias corrections of climate simulations: Which benefits for which losses?
Earth System Dynamics
11
(
2
),
537
562
.
doi:10.5194/esd-11-537-2020
.
Fung
F.
2018
How to Bias Correct. UKCP18 Guidance
.
Met Office
,
Exeter
,
UK
.
Gohar
L. K.
,
Lowe
J. A.
&
Bernie
D.
2017
The impact of bias correction and model selection on passing temperature thresholds
.
Journal of Geophysical Research: Atmospheres
122
(
22
),
12,045
012,061
.
https://doi.org/10.1002/2017JD026797
.
Gudmundsson
L.
,
Bremnes
J. B.
,
Haugen
J. E.
&
Engen-Skaugen
T.
2012
Technical note: Downscaling RCM precipitation to the station scale using statistical transformations – a comparison of methods
.
Hydrology and Earth System Sciences
16
(
9
),
3383
3390
.
doi:10.5194/hess-16-3383-2012
.
Guo
Q.
,
Chen
J.
,
Zhang
X. J.
,
Xu
C.-Y.
&
Chen
H.
2020
Impacts of using state-of-the-art multivariate bias correction methods on hydrological modeling over North America
.
Water Resources Research
56
(
5
),
e2019WR026659
.
https://doi.org/10.1029/2019WR026659.
Haddad
Z. S.
&
Rosenfeld
D.
1997
Optimality of empirical Z-R relations
.
Quarterly Journal of the Royal Meteorological Society
123
(
541
),
1283
1293
.
https://doi.org/10.1002/qj.49712354107.
Haerter
J. O.
,
Hagemann
S.
,
Moseley
C.
&
Piani
C.
2011
Climate model bias correction and the role of timescales
.
Hydrology and Earth System Sciences
15
(
3
),
1065
1079
.
doi:10.5194/hess-15-1065-2011
.
Hagemann
S.
,
Chen
C.
,
Haerter
J. O.
,
Heinke
J.
,
Gerten
D.
&
Piani
C.
2011
Impact of a statistical bias correction on the projected hydrological changes obtained from three GCMs and two hydrology models
.
Journal of Hydrometeorology
12
(
4
),
556
578
.
doi:10.1175/2011jhm1336.1
.
Hansen
C. H.
,
Goharian
E.
&
Burian
S.
2017
Downscaling precipitation for local-scale hydrologic modeling applications: Comparison of traditional and combined change factor methodologies
.
Journal of Hydrologic Engineering
22
(
9
),
04017030
.
doi:10.1061/(ASCE)HE.1943-5584.0001555
.
Hay
L. E.
,
Wilby
R. L.
&
Leavesley
G. H.
2000
A comparison of delta change and downscaled GCM scenarios for three mountainous basins in the United States
.
JAWRA Journal of the American Water Resources Association
36
(
2
),
387
397
.
doi:10.1111/j.1752-1688.2000.tb04276.x
.
Hegge
B. J.
&
Masselink
G.
1996
Spectral analysis of geomorphic time series: Auto-spectrum
.
Earth Surface Processes and Landforms
21
(
11
),
1021
1040
.
doi:10.1002/(SICI)1096-9837(199611)21:11 < 1021::AID-ESP703 > 3.0.CO;2-D
.
Hempel
S.
,
Frieler
K.
,
Warszawski
L.
,
Schewe
J.
&
Piontek
F.
2013
A trend-preserving bias correction – the ISI-MIP approach
.
Earth System Dynamics
4
(
2
),
219
236
.
doi:10.5194/esd-4-219-2013
.
Hnilica
J.
,
Hanel
M.
&
Puš
V.
2017
Multisite bias correction of precipitation data from regional climate models
.
International Journal of Climatology
37
(
6
),
2934
2946
.
https://doi.org/10.1002/joc.4890.
Hui
Y.
,
Chen
J.
,
Xu
C.-Y.
,
Xiong
L.
&
Chen
H.
2019
Bias nonstationarity of global climate model outputs: The role of internal climate variability and climate model sensitivity
.
International Journal of Climatology
39
(
4
),
2278
2294
.
https://doi.org/10.1002/joc.5950.
Iizumi
T.
,
Furuya
J.
,
Shen
Z.
,
Kim
W.
,
Okada
M.
,
Fujimori
S.
,
Hasegawa
T.
&
Nishimori
M.
2017
Responses of crop yield growth to global temperature and socioeconomic changes
.
Scientific Reports
7
(
1
),
7800
.
doi:10.1038/s41598-017-08214-4
.
Ines
A. V. M.
&
Hansen
J. W.
2006
Bias correction of daily GCM rainfall for crop simulation studies
.
Agricultural and Forest Meteorology
138
(
1
),
44
53
.
https://doi.org/10.1016/j.agrformet.2006.03.009.
IPCC 2023 Climate Change 2021 – The Physical Science Basis: Working Group I Contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change
.
Cambridge University Press
,
Cambridge
.
Jiang
Z.
,
Sharma
A.
&
Johnson
F.
2020
Refining predictor spectral representation using wavelet theory for improved natural system modeling
.
Water Resources Research
56
(
3
),
e2019WR026962
.
doi:10.1029/2019wr026962
.
Kim
Y.
,
Rocheta
E.
,
Evans
J. P.
&
Sharma
A.
2020
Impact of bias correction of regional climate model boundary conditions on the simulation of precipitation extremes
.
Climate Dynamics
55
(
11
),
3507
3526
.
doi:10.1007/s00382-020-05462-5.
Kim
Y.
,
Evans
J. P.
,
Sharma
A.
&
Rocheta
E.
2021
Spatial, temporal, and multivariate bias in regional climate model simulations
.
Geophysical Research Letters
48
(
11
),
e2020GL092058. doi:https://doi.org/10.1029/2020GL092058.
Kim
Y.
,
Evans
J. P.
&
Sharma
A.
2023
Multivariate bias correction of regional climate model boundary conditions
.
Climate Dynamics
doi:10.1007/s00382-023-06718-6.
Kotlarski
S.
,
Szabó
P.
,
Herrera
S.
,
Räty
O.
,
Keuler
K.
,
Soares
P. M.
&
Pianko-Kluczyńska
K.
2019
Observational uncertainty and regional climate model evaluation: A pan-European perspective
.
International Journal of Climatology
39
(
9
),
3730
3749
.
https://doi.org/10.1002/joc.5249.
Kusumastuti
C.
,
Jiang
Z.
,
Mehrotra
R.
&
Sharma
A.
2021
A signal processing approach to correct systematic bias in trend and variability in climate model simulations
.
Geophysical Research Letters
48
(
13
),
e2021GL092953
.
https://doi.org/10.1029/2021GL092953.
Kusumastuti
C.
,
Jiang
Z.
,
Mehrotra
R.
&
Sharma
A.
2022
Correcting systematic bias in climate model simulations in the time-frequency domain
.
Geophysical Research Letters
49
(
19
),
e2022GL100550
.
https://doi.org/10.1029/2022GL100550.
Kusumastuti
C.
,
Mehrotra
R.
&
Sharma
A.
2023
Rectifying low-frequency variability in future climate sea surface temperature simulations: Are corrections for extreme change scenarios realistic? Environmental Research Letters 18 (5), 1–7
.
doi:10.1088/1748-9326/accdf1.
Lafon
T.
,
Dadson
S.
,
Buys
G.
&
Prudhomme
C.
2013
Bias correction of daily precipitation simulated by a regional climate model: A comparison of methods
.
International Journal of Climatology
33
(
6
),
1367
1381
.
doi:10.1002/joc.3518
.
Lavaysse
C.
,
Vrac
M.
,
Drobinski
P.
,
Lengaigne
M.
&
Vischel
T.
2012
Statistical downscaling of the French Mediterranean climate: Assessment for present and projection in an anthropogenic scenario
.
Natural Hazards and Earth System Sciences
12
(
3
),
651
670
.
doi:10.5194/nhess-12-651-2012
.
Leite
S. M.
&
Peixoto
J. P.
1995
Spectral analysis of climatological series in Duero Basin
.
Theoretical and Applied Climatology
50
(
3
),
157
167
.
doi:10.1007/BF00866114
.
Li
C.
,
Sinha
E.
,
Horton
D. E.
,
Diffenbaugh
N. S.
&
Michalak
A. M.
2014
Joint bias correction of temperature and precipitation in climate model simulations
.
Journal of Geophysical Research: Atmospheres
119
(
23
),
13,153
113,162
.
https://doi.org/10.1002/2014JD022514.
Liu
M.
,
Rajagopalan
K.
,
Chung
S.
,
Jiang
X.
,
Harrison
J.
,
Nergui
T.
,
Guenther
A.
,
Miller
C.
,
Reyes
J.
&
Tague
C.
2014
What is the importance of climate model bias when projecting the impacts of climate change on land surface processes?
Biogeosciences
11
(
10
),
2601
2622
.
Maheswaran
R.
&
Khosa
R.
2012
Comparative study of different wavelets for hydrologic forecasting
.
Computers & Geosciences
46
,
284
295
.
https://doi.org/10.1016/j.cageo.2011.12.015.
Mao
G.
,
Vogl
S.
,
Laux
P.
,
Wagner
S.
&
Kunstmann
H.
2015
Stochastic bias correction of dynamically downscaled precipitation fields for Germany through copula-based integration of gridded observation data
.
Hydrology and Earth System Sciences
19
(
4
),
1787
1806
.
doi:10.5194/hess-19-1787-2015
.
Maraun
D.
2013
Bias correction, quantile mapping, and downscaling: Revisiting the inflation issue
.
Journal of Climate
26
(
6
),
2137
2143
.
doi:10.1175/jcli-d-12-00821.1
.
Maraun
D.
2016
Bias correcting climate change simulations – a critical review
.
Current Climate Change Reports
2
(
4
),
211
220
.
doi:10.1007/s40641-016-0050-x
.
Maraun
D.
&
Widmann
M.
2018
Cross-validation of bias-corrected climate simulations is misleading
.
Hydrology and Earth System Sciences
22
(
9
),
4867
4873
.
doi:10.5194/hess-22-4867-2018
.
Matyasovszky
I.
&
Bogardi
I.
1996
A space-time stochastic climatological approach to daily global solar radiation
.
Climate Research
7
(
1
),
11
20
.
Maurer
E. P.
&
Pierce
D. W.
2014
Bias Correction can Modify Climate Model Simulated Precipitation Changes Without Adverse Effect on the Ensemble Mean
.
Mehrotra
R.
&
Sharma
A.
2010
Development and application of a multisite rainfall stochastic downscaling framework for climate change impact assessment
.
Water Resources Research
46
(
7
).
https://doi.org/10.1029/2009WR008423.
Mehrotra
R.
&
Sharma
A.
2012
An improved standardization procedure to remove systematic low frequency variability biases in GCM simulations
.
Water Resources Research
48
(
12
).
doi:10.1029/2012wr012446
.
Mehrotra
R.
&
Sharma
A.
2015
Correcting for systematic biases in multiple raw GCM variables across a range of timescales
.
Journal of Hydrology
520
,
214
223
.
doi:10.1016/j.jhydrol.2014.11.037
.
Mehrotra
R.
&
Sharma
A.
2019
A resampling approach for correcting systematic spatiotemporal biases for multiple variables in a changing climate
.
Water Resources Research
55
(
1
),
754
770
.
https://doi.org/10.1029/2018WR023270.
Mehrotra
R.
,
Sharma
A.
&
Cordery
I.
2004
Comparison of two approaches for downscaling synoptic atmospheric patterns to multisite precipitation occurrence
.
Journal of Geophysical Research: Atmospheres
109
(
D14
).
https://doi.org/10.1029/2004JD004823.
Michelangeli
P.-A.
,
Vrac
M.
&
Loukos
H.
2009
Probabilistic downscaling approaches: Application to wind cumulative distribution functions
.
Geophysical Research Letters
36
(
11
).
https://doi.org/10.1029/2009GL038401.
Moore
K.
,
Pierson
D.
,
Pettersson
K.
,
Schneiderman
E.
,
Samuelsson
P.
,
2008
Effects of warmer world scenarios on hydrologic inputs to Lake Mälaren, Sweden and implications for nutrient loads
. In:
European Large Lakes Ecosystem Changes and Their Ecological and Socioeconomic Impacts
(
Nõges
T.
,
Eckmann
R.
,
Kangur
K.
,
Nõges
P.
,
Reinart
A.
,
Roll
G.
,
Simola
H.
&
Viljanen
M.
, eds).
Springer Netherlands
,
Dordrecht
, pp.
191
199
.
Murphy
J.
,
Harris
G.
,
Sexton
D.
,
Kendon
E.
,
Bett
P.
,
Clark
R.
,
Eagle
K.
,
Fosser
G.
,
Fung
F.
&
Lowe
J.
2018
UKCP18 Land Projections: Science Report
.
Nahar
J.
,
Johnson
F.
&
Sharma
A.
2017
Assessing the extent of non-stationary biases in GCMs
.
Journal of Hydrology
549
,
148
162
.
doi:10.1016/j.jhydrol.2017.03.045
.
Nguyen
H.
,
Mehrotra
R.
&
Sharma
A.
2016
Correcting for systematic biases in GCM simulations in the frequency domain
.
Journal of Hydrology
538
,
117
126
.
doi:10.1016/j.jhydrol.2016.04.018
.
Nguyen
H.
,
Mehrotra
R.
&
Sharma
A.
2018
Correcting systematic biases across multiple atmospheric variables in the frequency domain
.
Climate Dynamics
52
(
1–2
),
1283
1298
.
doi:10.1007/s00382-018-4191-6
.
Olsson
J.
,
Berggren
K.
,
Olofsson
M.
&
Viklander
M.
2009
Applying climate model precipitation scenarios for urban hydrological assessment: A case study in Kalmar City, Sweden
.
Atmospheric Research
92
(
3
),
364
375
.
https://doi.org/10.1016/j.atmosres.2009.01.015.
Panofsky
H. A.
,
Brier
G. W.
&
Best
W. H.
1958
Some Application of Statistics to Meteorology
.
Percival
D. B.
&
Walden
A. T.
2000
Wavelet Methods for Time Series Analysis
.
Cambridge University Press
,
Cambridge
.
Piani
C.
&
Haerter
J. O.
2012
Two dimensional bias correction of temperature and precipitation copulas in climate models
.
Geophysical Research Letters
39
(
20
).
https://doi.org/10.1029/2012GL053839.
Piani
C.
,
Haerter
J. O.
&
Coppola
E.
2010
Statistical bias correction for daily precipitation in regional climate models over Europe
.
Theoretical and Applied Climatology
99
(
1
),
187
192
.
doi:10.1007/s00704-009-0134-9
.
Pierce
D. W.
,
Das
T.
,
Cayan
D. R.
,
Maurer
E. P.
,
Miller
N. L.
,
Bao
Y.
,
Kanamitsu
M.
,
Yoshimura
K.
,
Snyder
M. A.
,
Sloan
L. C.
,
Franco
G.
&
Tyree
M.
2013
Probabilistic estimates of future changes in California temperature and precipitation using statistical and dynamical downscaling
.
Climate Dynamics
40
(
3
),
839
856
.
doi:10.1007/s00382-012-1337-9
.
Pierce
D. W.
,
Cayan
D. R.
,
Maurer
E. P.
,
Abatzoglou
J. T.
&
Hegewisch
K. C.
2015
Improved bias correction techniques for hydrological simulations of climate change
.
Journal of Hydrometeorology
16
(
6
),
2421
2442
.
doi:10.1175/jhm-d-14-0236.1
.
Pörtner
H.-O.
,
Roberts
D.C.
,
Poloczanska
E.
,
Mintenbeck
K.
,
Tignor
M.
,
Alegría
A.
,
Craig
M.
,
Langsdorf
S.
,
Löschke
S.
&
Möller
V.
2022
Potter
N. J.
,
Ekström
M.
,
Chiew
F. H. S.
,
Zhang
L.
&
Fu
G.
2018
Change-signal impacts in downscaled data and its influence on hydroclimate projections
.
Journal of Hydrology
564
,
12
25
.
https://doi.org/10.1016/j.jhydrol.2018.06.018
.
Prichard
D.
&
Theiler
J.
1994
Generating surrogate data for time series with several simultaneously measured variables
.
Phys Rev Lett
73
(
7
),
951
954
.
doi:10.1103/PhysRevLett.73.951
.
Rocheta
E.
,
Evans
J. P.
&
Sharma
A.
2017
Can bias correction of regional climate model lateral boundary conditions improve low-frequency rainfall variability?
Journal of Climate
30
(
24
),
9785
9806
.
doi:https://doi.org/10.1175/JCLI-D-16-0654.1.
Rocheta
E.
,
Evans
J. P.
&
Sharma
A.
2020
Correcting lateral boundary biases in regional climate modelling: the effect of the relaxation zone
.
Climate Dynamics
55
(
9
),
2511
2521
.
doi:10.1007/s00382-020-05393-1.
Roderick
T. P.
,
Wasko
C.
&
Sharma
A.
2019
Atmospheric moisture measurements explain increases in tropical rainfall extremes
.
Geophysical Research Letters
46
(
3
),
1375
1382
.
Roderick
T. P.
,
Wasko
C.
&
Sharma
A.
2020
An improved covariate for projecting future rainfall extremes?
.
Water Resources Research
56
(
8
),
e2019WR026924
.
Sang
Y.-F.
,
Singh
V. P.
,
Sun
F.
,
Chen
Y.
,
Liu
Y.
&
Yang
M.
2016
Wavelet-based hydrological time series forecasting
.
Journal of Hydrologic Engineering
21
(
5
),
06016001
.
doi:10.1061/(ASCE)HE.1943-5584.0001347
.
Schmidli
J.
,
Frei
C.
&
Vidale
P. L.
2006
Downscaling from GCM precipitation: A benchmark for dynamical and statistical downscaling methods
.
International Journal of Climatology
26
(
5
),
679
689
.
https://doi.org/10.1002/joc.1287
.
Sheffield
J.
,
Camargo
S. J.
,
Fu
R.
,
Hu
Q.
,
Jiang
X.
,
Johnson
N.
&
Zhao
M.
2013
North American climate in CMIP5 experiments. part II: Evaluation of historical simulations of intraseasonal to decadal variability
.
Journal of Climate
26
(
23
),
9247
9290
.
doi:10.1175/JCLI-D-12-00593.1
.
Shrestha
M.
,
Acharya
S. C.
&
Shrestha
P. K.
2017
Bias correction of climate models for hydrological modelling – are simple methods still useful?
Meteorological Applications
24
(
3
),
531
539
.
doi:10.1002/met.1655
.
Stocker
T. F.
,
Dahe
Q.
,
Plattner
G.-K.
&
Tignor
M.
2015
IPCC workshop on regional climate projections and their use in impacts and risk analysis studies. Paper presented at the workshop report
.
Sun
Q.
,
Miao
C.
,
Duan
Q.
,
Ashouri
H.
,
Sorooshian
S.
&
Hsu
K.-L.
2018
A review of global precipitation data sets: Data sources, estimation, and intercomparisons
.
Reviews of Geophysics
56
(
1
),
79
107
.
https://doi.org/10.1002/2017RG000574
.
Sunyer
M. A.
,
Hundecha
Y.
,
Lawrence
D.
,
Madsen
H.
,
Willems
P.
,
Martinkova
M.
,
Vormoor
K.
,
Bürger
G.
,
Hanel
M.
,
Kriaučiūnienė
J.
,
Loukas
A.
,
Osuch
M.
&
Yücel
I.
2015
Inter-comparison of statistical downscaling methods for projection of extreme precipitation in Europe
.
Hydrology and Earth System Sciences
19
(
4
),
1827
1847
.
doi:10.5194/hess-19-1827-2015
.
Switanek
M. B.
,
Troch
P. A.
,
Castro
C. L.
,
Leuprecht
A.
,
Chang
H. I.
,
Mukherjee
R.
&
Demaria
E. M. C.
2017
Scaled distribution mapping: A bias correction method that preserves raw climate model projected changes
.
Hydrology and Earth System Sciences
21
(
6
),
2649
2666
.
doi:10.5194/hess-21-2649-2017
.
Taylor
K. E.
,
Stouffer
R. J.
&
Meehl
G. A.
2012
An overview of CMIP5 and the experiment design
.
Bulletin of the American Meteorological Society
93
(
4
),
485
498
.
doi:10.1175/bams-d-11-00094.1
.
Teutschbein
C.
&
Seibert
J.
2012
Bias correction of regional climate model simulations for hydrological climate-change impact studies: Review and evaluation of different methods
.
Journal of Hydrology
456–457
,
12
29
.
https://doi.org/10.1016/j.jhydrol.2012.05.052
Themeßl
M. J.
,
Gobiet
A.
&
Heinrich
G.
2012
Empirical-statistical downscaling and error correction of regional climate models and its impact on the climate change signal
.
Climatic Change
112
(
2
),
449
468
.
doi:10.1007/s10584-011-0224-4
.
Torrence
C.
&
Compo
G. P.
1998
A practical guide to wavelet analysis
.
Bulletin of the American Meteorological Society
79
(
1
),
61
78
.
doi:10.1175/1520-0477(1998)079 < 0061:APGTWA > 2.0.CO;2
.
Torrence
C.
&
Webster
P. J.
1999
Interdecadal changes in the ENSO–monsoon system
.
Journal of Climate
12
(
8
),
2679
2690
.
Trenberth
K. E.
2011
Changes in precipitation with climate change
.
Climate Research
47
(
1–2
),
123
138
.
Trenberth
K. E.
,
Dai
A.
,
Rasmussen
R. M.
&
Parsons
D. B.
2003
The changing character of precipitation
.
Bulletin of the American Meteorological Society
84
(
9
),
1205
1218
.
doi:10.1175/bams-84-9-1205
.
Trewin
B. C.
2007
The Role of Climatological Normals in a Changing Climate
.
World Meteorological Organization
,
Geneva, Switzerland
.
Vrac
M.
&
Friederichs
P.
2015
Multivariate – intervariable, spatial, and temporal – bias correction
.
Journal of Climate
28
(
1
),
218
237
.
doi:10.1175/jcli-d-14-00059.1
.
Vrac
M.
,
Drobinski
P.
,
Merlo
A.
,
Herrmann
M.
,
Lavaysse
C.
,
Li
L.
&
Somot
S.
2012
Dynamical and statistical downscaling of the French Mediterranean climate: Uncertainty assessment
.
Natural Hazards and Earth System Sciences
12
(
9
),
2769
2784
.
doi:10.5194/nhess-12-2769-2012
.
Wang
Z.
,
Han
L.
,
Zheng
J.
,
Ding
R.
,
Li
J.
,
Hou
Z.
&
Chao
J.
2021
Evaluation of the performance of CMIP5 and CMIP6 models in simulating the Victoria Mode–El Niño relationship
.
Journal of Climate
34
(
18
),
7625
7644
.
doi:10.1175/jcli-d-20-0927.1
.
Wasko
C.
,
Sharma
A.
&
Westra
S.
2016
Reduced spatial extent of extreme storms at higher temperatures
.
Geophysical Research Letters
43
(
8
),
4026
4032
.
https://doi.org/10.1002/2016GL068509.
White
R. H.
&
Toumi
R.
2013
The limitations of bias correcting regional climate model inputs
.
Geophysical Research Letters
40
(
12
),
2907
2912
.
https://doi.org/10.1002/grl.50612.
Wilby
R. L.
&
Wigley
T. M. L.
1997
Downscaling general circulation model output: A review of methods and limitations
.
Progress in Physical Geography: Earth and Environment
21
(
4
),
530
548
.
doi:10.1177/030913339702100403
.
Willems
P.
&
Vrac
M.
2011
Statistical precipitation downscaling for small-scale hydrological impact investigations of climate change
.
Journal of Hydrology
402
(
3
),
193
205
.
https://doi.org/10.1016/j.jhydrol.2011.02.030.
WMO
2014
Scientists urge more frequent updates of 30-year climate baselines to keep pace with rapid climate change (Press Release Number: 997, 9 July 2014). Available from: https://public.wmo.int/en/media/press-release/no-997-scientists-urge-more-frequent-updates-of-30-year-climate-baselines-keep.
Wong
G.
,
Maraun
D.
,
Vrac
M.
,
Widmann
M.
,
Eden
J. M.
&
Kent
T.
2014
Stochastic model output statistics for bias correcting and downscaling precipitation including extremes
.
Journal of Climate
27
(
18
),
6940
6959
.
doi:10.1175/jcli-d-13-00604.1
.
Wood
A. W.
,
Leung
L. R.
,
Sridhar
V.
&
Lettenmaier
D. P.
2004
Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs
.
Climatic Change
62
(
1
),
189
216
.
doi:10.1023/B:CLIM.0000013685.99609.9e
.
Xu
C.-y.
1999
From GCMs to river flow: A review of downscaling methods and hydrologic modelling approaches
.
Progress in Physical Geography: Earth and Environment
23
(
2
),
229
249
.
doi:10.1177/030913339902300204
.
Zhang
L.
,
Nan
Z.
,
Yu
W.
,
Zhao
Y.
&
Xu
Y.
2018
Comparison of baseline period choices for separating climate and land use/land cover change impacts on watershed hydrology using distributed hydrological models
.
Science of The Total Environment
622–623
,
1016
1028
.
https://doi.org/10.1016/j.scitotenv.2017.12.055
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).