Data Spotlight: Downscaled Climate Projections to Inform Climate Research in the South-Central U.S. Region

Sources/Usage: Public Domain. View Media Details

What’s Up with Downscaled Data? 

Characterizing how the earth’s climate may change is challenging for decision makers looking to proactively address the impacts of climate change on the resources they manage. Global Climate Models (GCM) are an important tool for researching climate and developing future climate projections. However, because these models are global in scale, they’re not always adequate for understanding changes in regional and local-scale climates.  Statistical downscaling (SD) is a common technique used to translate the large-scale GCM data into smaller spatial scales (such as a single watershed) which can be better utilized by regional and local stakeholders to address their specific needs.  

Along with different statistical downscaling methods, the projections produced can be affected by the differences between multiple components used to create them, including the GCMs and emissions scenarios. It is important for scientists and managers to fully understand the nuances of different downscaling methods, when using downscaled climate data to make decisions. To address the variability among downscaled projections, a pair of studies funded by the South Central CASC (Study 1 & Study 2) aimed to provide information that can assist managers in navigating the use of the appropriate downscaled projections for their specific management questions. Led by South Central CASC Climate Scientist Adrienne Wootten at the University of Oklahoma and co-investigators from University of Oklahoma and NOAA - Geophysical Fluid Dynamics Laboratory, this project produced 81 sets of SD climate projections of daily high temperature, daily low temperature, and daily total precipitation for the South Central US region to guide stakeholders in addressing potential climate change impacts on a range of systems, from watersheds to agriculture.  

Details of the Dataset 

The dataset is called South Central Climate Projections Evaluation Project (C-PrEP) and presents future projections of temperature and precipitation produced from a combination of GCMs, emissions scenarios, downscaling techniques, and training data. These data are provided in the NetCDF format (i.e. a file format for displaying multi-dimensional data such as temperature, wind speed and direction). 

Scenarios Used 

Three representative concentration pathways (RCPs), representing a low to high emission scenario in the 21st century, were used in the creation of 81 sets of SD climate projections for the South-Central U.S. region. These scenarios were: 

1. RCP 2.6 Scenario – A scenario in which temperatures would peak by mid-century and then decline to the end of the 21st century. 

1. RCP 4.5 Scenario - This can be called the “stabilization” scenario, meaning, by the year 2100, radiative forcing levels will remain steady at 4.5 W/m2, resulting in a moderate overall increase in temperature. 

1. RCP 8.5 Scenario – This is the highest emissions scenario and would result in the largest increases in temperature. 

Global Climate Models Used 

Three GCMs were used representing a range of future changes among the global models. These three GCMs were: 

1. Community Climate System Model version 4 (CCSM4) 

1. Model for Interdisciplinary Research on Climate version 5 (MIROC5) 

1. Max-Planck-Institute Earth System Model running on a low-resolution grid (MPI-ESM-LR) 

Statistical Downscaling (SD) Techniques 

To provide a robust dataset and highlight the sensitivity of statistical downscaling, three different SD methods were used (outlined below) in conjunction with the GCMs and RCPs in the analysis. Each of the techniques used three gridded observational data products (Daymet Version 2.1, Livneh Version 1.2, and PRISM Version AN81d) to calibrate the methods for accuracy. 

1. Delta Method (DeltaSD) – A simple method that is widely used and assumes that changes occur over larger, regional scales and that relationships between climate variables will remain the same in future scenarios. (Pourmokhtarian et al., 2016) 

1. Equi-distant Quantile Mapping Method (EDQM) – This is a modified version of the quantile mapping technique to account for shifts in time with bias-correction specifically for monthly GCM outputs. (Li et al., 2010) 

1. Piecewise Asynchronous Regression Method (PARM) – Similar to the EDQM, this method utilizes bias-correction techniques but also addresses the random irregularities from sampling variables by fitting a series of lines to the point data on order to smooth out the small-scale noise. (Lanzante et al., 2019) 

Due to the variability in the SD methods’ performance characteristics, using these three SD techniques and the various observational data products demonstrates that the choice in SD method can influence the variability of the climate projections produced.  

Where and When 

Each of the 81 sets of projections are produced on a 10 km by 10 km grid covering the South-Central U.S. region, including AR, KS, LA, NM, OK, TX, and portions of CO and MO. The dataset includes both historical data representing conditions from 1981-2005 and future projections representing conditions from 2006 -2099. 

How to Access the Data 

The data can be accessed through the USGS GeoData Portal (GDP). The GDP provides easy access to large-scale datasets, such as gridded climate and land use data in the NetCDF or geotiff format, for managers and the public to utilize. For more information on navigating the USGS GDP, click here. 

Results and Applications  

The various climate trajectories demonstrated by the different RCP scenarios in these downscaled climate data all project that temperatures will rise in the future. Although not comprehensive, these crucial data can serve as an important tool in the future of climate impacts research in the South-Central U.S. region. In the resulting report, researchers provided some important insights into the variability among SD methods that should be considered by managers and applied researchers when using SD data. For example, the methods by which the observational training data are translated to a gridded format can result in differences in the future projections for wet‐day frequency, intensity of precipitation extremes, and the length of multi‐day wet and dry periods. So, it is important for stakeholders to choose the downscaled projections that are best suited for their climate variable of interest. An example provided by the research results for “stakeholders in the southern Rockies with concerns over changes to temperature extremes should carefully consider the training data and downscaling technique as these two components reflect 30% or more of the variability for projections of temperature extremes”.  For more details on published results related to this study please see here. 

This work was part of the projects “Characterizing Uncertainties in Climate Projections to Support Regional Decision-Making,” supported by the South Central CASC.  

<< Back to the Climate Adaptation Science Centers' website