Empirical Statistical Downscaling Evaluations

---

Our Statistical Downscaling Research Team’s efforts

Overview:

Here at NOAA GFDL, we view our Empirical Statistical Downscaling (ESD) team’s research efforts as occupying a niche akin to that of a Consumer Reports^® for statistical downscaling and bias correction techniques that are used to refine climate forecasts and projections. Our main goal is to perform evaluations and share the results so that others can make better informed decisions about the suitability of different data products created using bias correction and statistical downscaling methods for various applications relevant to decision-making and resiliency planning. Our team’s efforts to systematically evaluate the performance characteristics of commonly used statistical refinement methods are potentially valuable to three different communities interested in climate variability and change. They are…

Applied Science Researchers & Decision-Makers

For many climate impacts studies, large-scale dynamical climate model output is viewed as having inadequate spatial resolution and/or unacceptably large biases that make it unsuitable for direct use in applications of interest. (For an example , see our Downscaled Climate Projections for Heat & Health Applications page.) When used in climate impacts analyses that support planning and decision-making, statistically downscaled and bias corrected climate projections (hereafter referred to collectively as ‘ESD data products’) typically are viewed as being value-added products that address climate model biases and provide finer spatial scale detail. A common question asked by users of ESD data products is “which of the publicly available downscaled climate projection data products is the best one for me to use?“‘  Unfortunately, there is no one-size-fits-all answer to that question, because different climate impacts studies have different data requirements and sensitivities -and- different ESD data products exhibit characteristics linked to the choice of statistical refinement methodology and the choice of observation-based data used to ‘train’ the ESD method.

Our research can benefit users of ESD products in two main ways. First, information gleaned from our various sensitivity analyses, inter-comparisons, and “perfect model” experimental design research efforts increase our knowledge about the performance of different bias correction and statistical downscaling methods when applied to changing climatic conditions. Our analyses of ESD data products distributed by other research groups and ones we generate in-house uncover variations that exist across geographical regions, by time of year, variable of interest, and amount of climate change. Knowledge gained from our studies can be used to help applied researchers who employ ESD data products as input to their studies better assess to what extent different ESD data products match their applications’ needs and sensitivities (i.e., supporting “fit for purpose” considerations.)  Secondly, as outlined below, our analysis efforts can improve the understanding of elements of the multi-step process that generates ESD products, thereby helping to inform the development of the next generation of statistically refined climate projection data sets.

Empirical-Statistical Downscaling Community

Lacking observations of the future, it is difficult to assess whether and to what extent the skill of ESD techniques may degrade when applied to future climate projections.  One analysis approach we have used – a “perfect model experimental design” – has allowed us to conduct rigorous quantitative assessments of ESD skill both for the contemporary climate and for future projections. The results of our analyses have proven useful in isolating weakness in some ESD methods, and have led to ESD method improvements.

We also examine how different ESD products can vary in factors such as their representation of a variable’s projected central tendencies, the tails of the distribution (e.g., extreme events), the temporal sequence of events, and the extent to which a dynamical climate model’s climate change signal is altered during the statistical refinement process.  Some publicly available ESD products were developed with particular end uses in mind, and thus may have prioritized some of these factors more than others. However, once downloaded, the publicly available ESD products are used in a much wider range of applications. We investigate the underlying reasons for some of the method-to-method variations. Communicating results of our analyses that reveal performance characteristics and trade-offs of present in different ESD products can assist developers of ESD methods, as well as users of ESD data products, in considerations of an ESD data product’s suitability for different end use applications.

Global Climate Modeling Community

Empirical Statistical Downscaling can be thought of as an additional post-processing step applied to Global Climate Model (GCM) output. In effect, an ESD method challenges GCM results with observations as it seeks to address shortcomings in the GCM-simulated climate based on information gleaned from the observational record. Our team’s efforts can promote GCM development by helping to identify GCM weaknesses and strengths, with an emphasis on surface climate variables of particular value to adaptation planners, resource managers, and other decision-makers.

SELECT JOURNAL PUBLICATIONS :

Le Roy B, KW Dixon, and D Adams-Smith (2024) High-resolution urban climate simulations for heat and health applications in Philadelphia. Urban Climate, doi: 10.1016/j.uclim.2024.102114.

Wootten AM, KW Dixon, DJ Adams‐Smith, RA McPherson, (2021) Statistically Downscaled Precipitation Sensitivity to Gridded Observation Data and Downscaling Technique, International Journal of Climatology, doi:10.1002/joc.6716

Lanzante JR, KW Dixon, MJ Nath, CE Whitlock, D Adams-Smith, (2018) Some Pitfalls in Statistical Downscaling of Future Climate. Bulletin of the American Meteorological Society, doi:10.1175/BAMS-D-17-0046.1

Dixon KW, JR Lanzante, MJ Nath, K Hayhoe, A Stoner, A Radhakrishnan, V Balaji, CF Gaitán (2016) Evaluating the stationarity assumption in statistically downscaled climate projections: is past performance an indicator of future results? Climatic Change. doi:10.1007/s10584-016-1598-0

---

PEOPLE
Team Members & Collaborators

---

SELECT JOURNAL PUBLICATIONS
3 Papers of Interest

---

PROJECTS & LINKAGES
Past & Present

---

NEWS & PRESENTATIONS

SELECT CONFERENCE PRESENTATIONS: