ISSN 1000-0526
CN 11-2282/P
CN 11-2282/P
Study on ECMWF 2 m Temperature Bias Correction Based on Dynamically Approximated Vertical Change Rate
- Article
- Figures
- Metrics
- Preview PDF
- Reference
- Related
- Cited by
- Materials
Abstract:
Taking into account the relationships between model temperature forecasts, model topography, observed temperatures, and actual topography, a dynamically approximated vertical rate correction method for the 2 m temperature that evolves over space and time has been designed. A correction experiment is conducted for Chongqing in 2023. Evaluation of ECMWF model forecasts show that the spatial distribution of forecast skill for maximum and minimum 2 m temperatures is similar, but forecast skill for maximum temperature is significantly worse than for minimum temperature. Additionally, forecast skill declines with increasing forecast lead time. A strong correlation is found between model topographic elevation bias and temperature forecast bias; areas with smaller elevation bias tend to exhibit better forecast skill, and vice versa. Comparative research reveals that the correction scheme based on dynamically approximated vertical rate significantly outperforms the scheme based on constant vertical rate. Both schemes improve the ECMWF model temperature forecasts, with better correction performance for maximum temperatures than for minimum temperatures. Compared to the raw model forecasts, the corrected results based on dynamically approximated vertical rate have increased the 10-day average forecast accuracy of maximum and minimum temperatures by 12.71% and 8.3%, respectively, and reduced the mean absolute error by 0.68°C and 0.3°C, respectively. The monthly average forecast accuracy of the maximum temperature for 24 h forecast lead time has increased by 20.34%. The minimum temperature has increased by 14.44% on average. Overall, the corrected temperature forecasts are more stable. The correction scheme based on dynamically approximated vertical rate can effectively reduce the temperature forecast bias. The greater the model topographic elevation bias, the smaller the amplitude of weather process fluctuations, and the more obvious the correction performance.
Keywords:
Project Supported:
Clc Number:
Mobile website
