Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 503 - Climate and Meteorological Statistics
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Statistics and the Environment
Abstract #311139
Title: Constructing daily, high-resolution, bias-corrected climate products: a comparison of methods
Author(s): Maike Holthuijzen* and Brian Beckage and Dave Higdon and Patrick J Clemins and Jonathan Winter
Companies: University of Vermont and University of Vermont and Virginia Tech and University of Vermont and Dartmouth College
Keywords: bias-correction; high-resolution; downscaling; maximum temperature; kriging; inverse distance weighting
Abstract:

High-resolution, bias-corrected climate data is necessary for climate impact studies at local scales. Using gridded historical data for bias correction is convenient but may contain biases resulting from interpolation. Long-term, quality-controlled station data better represent true climatological measurements, but as the spatial distribution of climate stations over the landscape is irregular, station data are challenging to incorporate into downscaling and bias-correction approaches. The use of station data in creating full-coverage, bias-corrected climate products is not well-represented in the literature. In this study, we developed and compared six novel methodologies using station data to produce daily, high-resolution, bias-corrected climate products with maximum temperature simulations from a regional climate model (RCM). The methods differed with respect to interpolation methods and bias-correction techniques. We quantified performance of six methods with the root mean square error (RMSE) and Perkins skill score (PSS) and used two ANOVA models to analyze how performance metrics varied among methods. We temporally validated the six methods using two calibration sets of observed station data (1980-1989 and 1980-2014) and two testing sets of RCM data (1990-2014 and 1980-2014). RMSE for all methods varied considerably throughout the year and was larger in cold seasons, while PSS was more consistent. Quantile-mapping bias-correction techniques performed best in improving PSS, while simple linear transfer functions performed best in improving RMSE. For the 1980-1989 station calibration dataset, simple quantile-mapping techniques outperformed empirical quantile mapping (EQM) in improving PSS; conversely, when the calibration and testing sets represented the same time period, EQM performed best in improving PSS. No one method simultaneously improved RMSE and PSS; however, the simple quantile-mapping based techniques perform as well or better than more sophisticated methods such as empirical quantile mapping.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program