A team of global researchers is set to transform weather forecasting with their latest advancement in machine learning. Despite the maturity of ensemble numerical weather prediction (NWP), forecasts often remain under-dispersed, prompting the need for improved calibration tools.
Enter the non-crossing quantile regression neural network (NCQRNN), a groundbreaking model designed to enhance the accuracy and reliability of weather predictions. Unlike traditional quantile regression (QR) methods, which face challenges like quantile crossing, the NCQRNN ensures that lower quantiles remain consistently below higher ones, enhancing interpretability without sacrificing precision.
\"Our NCQRNN model maintains the natural order of forecast values, ensuring lower quantiles stay smaller than higher ones. This boosts accuracy and significantly improves forecast interpretability,\" explains Yang Dazhi, a professor at Harbin Institute of Technology.
Published in Advances in Atmospheric Sciences, the study is a collaborative effort involving institutions from around the world, including the Karlsruhe Institute of Technology (KIT) in Germany, National University of Singapore, UK Power Networks, and Budapest University of Technology and Economics (BME).
Dr. Martin J. Mayer from BME highlights the model's versatility, stating, \"This non-crossing layer can be added to a wide range of different neural network structures, ensuring the wide applicability of the proposed technique.\"
Dr. Sebastian Lerch of KIT adds that the NCQRNN's general framework allows it to be adapted for various weather and climate applications beyond solar irradiance forecasting. This adaptability promises clearer and more reliable predictions across multiple weather variables.
The researchers emphasize the significant role of machine learning in advancing weather and climate science. Xia Xiang'ao from the Institute of Atmospheric Physics of the Chinese Academy of Sciences notes, \"This study provides an instructive case study on how to apply advanced machine learning methods to numerical weather prediction models to improve the accuracy of weather forecasts and climate predictions.\"
With its innovative approach, the NCQRNN model is poised to become a vital tool in the quest for more precise and dependable weather forecasting, benefiting industries and communities worldwide.
Reference(s):
Scientists enhance weather forecasts reliability with machine learning
cgtn.com
