Journal Article

The effects of atypical diurnal temperature cycles on regression-based downscaling of daily temperature extrema in the Central United States

Pallardy, Quinn

2023

Overview

The effects of variations in time of day of daily temperature extrema on regression-based statistical downscaling of daily temperature extrema were examined. These effects were analyzed by evaluating the performance of a regression-based downscaling model with multiple approaches to the incorporation of the relevant temperature data. The differing approaches included which predictor variables were selected for inclusion in the model, as well as variations in model methodology. Three different versions of the downscaling model were evaluated: (i) standard multiple linear regression, (ii) a weather classification scheme combined with multiple linear regression, and (iii) a weather classification scheme combined with multiple linear regression using dynamic time-step predictors. Bias and accuracy were measured on days with atypical and typical times of temperature extrema. The performance of regression models had the potential to be greatly degraded by days with atypical times of temperature extrema. The degree to which these atypical days were affected was dependent on which predictors were included in the regression models, with the temperature extrema derived from reanalysis data playing the most important role. Implementation of the weather classification scheme also improved downscaling performances for atypical days in a number of situations. For typical days, the improvements to RMSE values were smaller and were only present under certain predictor combinations.

• Publisher: Springer Nature B.V
• Subject: Classification / Classification schemes / Climate science / Daily / Daily temperatures / Modelling / Performance evaluation / Regression analysis / Regression models / Statistical analysis / Temperature cycles / Temperature data / Time of use / Weather