Treffer: Diagnosing Errors in Climate Forecast Models Using Forced Autoregressive Models.

Climate models initialized near the observed state typically drift toward their own climatology as the forecast evolves. This drift is commonly corrected through a lead‐time and start‐month dependent bias adjustment, derived from a hindcast data set. While widely used, this traditional correction has well‐known limitations: it is statistically inefficient, prone to introducing artificial discontinuities, and offers little insight into the underlying causes of forecast error. This paper presents an alternative framework that addresses these limitations and provides a more process‐oriented diagnostic. The proposed method fits separate autoregressive models with exogenous input (ARX models) to both forecasts and observations. Forecast errors are then predicted and removed using the difference between the two ARX models. The method is demonstrated and compared to traditional methods using seasonal forecasts of global mean temperature from the SPEAR model, a contributor to the North American Multi‐Model Ensemble (NMME). The ARX approach outperforms traditional methods in independent data even when traditional approaches include a linear trend correction. The analysis reveals that SPEAR exhibits an exaggerated response to radiative forcing, leading to significant trend errors. Notably, these errors are already present in the first month. These initial trend errors can be reproduced by a one‐dimensional data assimilation system, indicating that they originate from SPEAR's exaggerated response to radiative forcing, which is carried forward into the first‐guess fields used in the data assimilation system. Plain Language Summary: Weather and climate centers around the world routinely use physics‐based dynamical models to predict monthly mean variables such as temperature and precipitation. While dynamical models provide valuable forecasts, they often contain systematic errors, or biases. The traditional way to address these biases is to calculate the average forecast error over a recent period for each lead time and start month, and then subtract this average error from the forecast. However, this method suffers from statistical inefficiencies and offers little insight into the source of the errors. This paper introduces a new method for reducing forecast biases that addresses these issues. The approach uses statistical models, called ARX models. The ARX models are fitted to observations and forecasts separately, and then used to predict and remove forecast errors. In addition to improving forecasts, the ARX models can be examined to better understand the sources of biases in the original model. This new method offers a more comprehensive way to reduce errors in seasonal or climate predictions and provides insights into the underlying causes of forecasting errors. Key Points: Forced autoregressive models are fitted independently to forecasts and observations to predict and remove forecast errorsThis method outperforms traditional methods, including those with trend correction, while using fewer parametersComparing the forecast‐ and observation‐based models provides a novel, process‐oriented diagnostic for identifying sources of model error [ABSTRACT FROM AUTHOR]

Copyright of Journal of Advances in Modeling Earth Systems is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)