Evaluation of some distributional downscaling methods as applied to daily maximum temperature with emphasis on extremes

Abstract

Statistical downscaling methods are extensively used to refine future climate change projections produced by physical models. Distributional methods, which are among the simplest to implement, are also among the most widely used, either by themselves or in conjunction with more complex approaches. Here, building off of earlier work we evaluate the performance of seven methods in this class that range widely in their degree of complexity. We employ daily maximum temperature over the Continental U.S. in a “Perfect Model” approach in which the output from a large‐scale dynamical model is used as a proxy for both observations and model output. Importantly, this experimental design allows one to estimate expected performance under a future high‐emissions climate‐change scenario. We examine skill over the full distribution as well in the tails, seasonal variations in skill, and the ability to reproduce the climate change signal. Viewed broadly, there generally are modest overall differences in performance across the majority of the methods. However, the choice of philosophical paradigms used to define the downscaling algorithms divides the seven methods into two classes, of better versus poorer overall performance. In particular, the bias‐correction plus change‐factor approach performs better overall than the bias‐correction only approach. Finally, we examine the performance of some special tail treatments that we introduced in earlier work which were based on extensions of a widely used existing scheme. We find that our tail treatments provide a further enhancement in downscaling extremes.

The impact of using different probability representations in application of equidistant quantile matching for bias adjustment of daily precipitation over the Daqing River Basin, North China

“…As stated by Wang and Chen (2014), EDCDFm may be meaningless when an unmatched theoretical distribution is used. For the ECDF method, Mcginnis et al (2015) found that the KDE approach performs better than ECDF, whereas Lanzante et al (2019) found no significant difference between their performances. The current study found ECDF method and DKDE method are equally skilful (Sections 3.2 and 3.3), which is also confirmed using different sample sizes (Figure S2).…”

“…In a realistic climate condition, the internal variability, especially the internal low-frequency variability, can bring uncertainty (Maraun and Widmann, 2018;Chen et al, 2020). A better framework would be the pseudo-realistic approach which can provide a more rigorous evaluation of the stationarity hypothesis (Lanzante et al, 2018(Lanzante et al, , 2019. (b) The corrected results largely depend on the output of the raw climate model.…”

“…The extremes were not well corrected and the outliers were shown at the upper tails of the curves for limited data at all stations. This may be due to the climatology and spatial resolution mismatches (Maraun, 2013;Lanzante et al, 2019). The daily performance between the results of all cross-validation at 11 stations and the corresponding observations are displayed in Figure 5.…”

“…Miao et al (2016) used KDE in monthly precipitation QM model; the results showed that it has a slight advantage compared to the parametric approach. Mcginnis et al (2015) and Lanzante et al (2019) applied KDE to QM for temperature and found that it had a performance comparable to other methods at least. However, the KDE method is criticized for its low computational speed and boundary effect if the samples are bounded.…”

Evaluation of some distributional downscaling methods as applied to daily precipitation with an eye towards extremes

“…Recently, this author team, members of the Geophysical Fluid Dynamics Laboratory (GFDL) Empirical Statistical Downscaling (ESD) team (https://www.gfdl.noaa.gov/esd_eval) has focused on evaluating some of these techniques. For an SD overview and our evaluation approach philosophy see our earlier works and cited references (Dixon et al ., 2016; Lanzante et al ., 2018; Lanzante et al ., 2019a, hereafter L19a; Lanzante et al ., 2019b, hereafter L19b). As a caution we note that even the best SD methods will fail to produce credible results if the driving physical climate model is flawed in its representation of circulation features (Hall, 2014; Maraun et al ., 2017).…”