Differences between Machine Learning and Traditional Downscaling Method in Processing Summer Meteorological Elements in the Yellow River Basin

Expand
  • 1. Key Laboratory for Semi-Arid ClimateChange of the Ministry of EducationSchool of AtmosphericSciencesLanzhou UniversityLanzhou730000GansuChina
    2. Collaborative Innovation Center for Western Ecological SafetyLanzhou UniversityLanzhou730000GansuChina

Online published: 2025-02-24

Supported by


Abstract

Global Climate Models GCMsare the primary tools currently used to predict future climate changehowevertheir coarse spatial resolution limits their ability to assess localized impacts of climate change. To address this issuestatistical downscaling methods based on linear regression equations have been developed to enhance the spatial resolution of GCMs and have continuous improvement and innovation. Meanwhilemachine learning techniques have demonstrated superior performance in various predictive modeling problemsmaking them potential new tools for statistical downscaling. Thereforethis study applied machine learning model-Light Gradient Boosting MachineLightGBMto construct statistical downscaling model for each sitecombined with information flow method to select forecasting factorsand compared with linear regression method stepwise multiple linear regression method based on Empirical Orthogonal Functionto explore the application ability of LightGBM in the field of statistical downscaling. The two methods were applied to downscale the meteorological element of the Yellow River basinan important climate change sensitive area in Chinaestablishing statistical downscaling models for 90 stations within the basin to generate temperature and precipitation data for the summer monthsJuneJulyAugustfrom 1965 to 2014. The performance of both methods is evaluated through an analysis of the correlation coefficientsroot mean square errorsRMSE),and spatial distributions be‐ tween downscaled values and observed values. The results show that both downscaling methods can correct the temperature error of the reanalysis dataERA5in the northern part of the basin. LightGBM shows superior inter-site correlationbut 606452 sites show higher RMSE than regression method in JuneJulyand Augustrespectively. For precipitation downscalingneither of the two downscaling datasets nor ERA5 could accurately represent the spatial distribution of observed valuesbut the downscaling value obtained by LightGBM had a higher inter-site correlation coefficient than the regression methodand only 16714 sites showed higher RMSE than the regression method in JuneJuly and August. Considering the potential of machine learning methods for modeling nonlinear problemsit is still necessary to further improve the algorithm and improve the quality of downscaling data sets in the future. The advantages and disadvantages of machine learning in downscaling work provided a technical reference and support for using statistical downscaling methods to generate high-resolution temperature and precipitation data in the future.

 

Cite this article

CHEN Han, GUAN Xiaodan, MA Tingting . Differences between Machine Learning and Traditional Downscaling Method in Processing Summer Meteorological Elements in the Yellow River Basin[J]. Plateau Meteorology, 0 : 1 . DOI: 10. 7522/j. issn. 1000-0534. 2024. 00118

Outlines

/