Statistical Analysis

While the previous layer-wise analysis examined wind structures at each pressure level, this statistical analysis compares how the differences between ERA5 and JRA-55 appear across vertical layers.
Using RMSD, mean bias, correlation, and regression analysis, we evaluate whether the differences are simply local variations or systematic layer-dependent patterns related to different data assimilation systems.

1. RMSD: Magnitude and Timing of Wind-field Differences

The RMSD analysis shows that ERA5–JRA-55 wind-field differences are not uniform across pressure levels. Throughout the analysis period, upper-level winds at 200–300 hPa show larger RMSD than lower-level winds. This indicates that the differences between the two reanalysis datasets are most clearly expressed in the upper troposphere, where AMV-related assimilation effects are expected to be strongest.

The time series also shows that RMSD temporarily decreases near the peak-intensity stage, but increases sharply during the stagnation and recurvature stages. This suggests that the strong typhoon circulation temporarily constrains both reanalysis systems toward a similar structure, while the differences become larger when the storm weakens, stagnates, or interacts with the surrounding synoptic environment.

2. Mean Bias: Direction of Systematic Difference

The mean bias profile shows the systematic direction of the differences between ERA5 and JRA-55.
ERA5 shows a positive wind-speed bias at all pressure levels, meaning that ERA5 generally analyzes stronger wind speeds than JRA-55.
The largest wind-speed bias appears at 200 hPa, where ERA5 exceeds JRA-55 by about 0.70 m/s.

U-wind bias is also positive at all levels, suggesting that ERA5 tends to analyze stronger east–west wind components.
In contrast, V-wind bias is negative across all levels, indicating a systematic difference in the north–south wind component.
Wind-direction bias is largest near 500 hPa, suggesting that the middle troposphere is where the two datasets differ most in wind direction.

3. Correlation and Regression: Pattern Agreement

The correlation profile shows that ERA5 and JRA-55 have high U-wind correlations at all pressure levels, ranging from about 0.917 to 0.985.
The highest correlations appear in the upper troposphere at 200–300 hPa, while the correlation gradually decreases toward the lower troposphere.

This means that ERA5 and JRA-55 capture similar large-scale upper-level wind patterns, even though ERA5 tends to produce stronger wind speeds.
In contrast, lower-level winds show weaker agreement, suggesting that boundary-layer processes, terrain effects, and typhoon inner-core structures are represented differently in the two reanalysis systems.

The scatter plots support this interpretation.
At 200–300 hPa, the regression lines are close to the 1:1 reference line, showing strong linear agreement between ERA5 and JRA-55.
However, at 850–925 hPa, the regression slope becomes smaller than 1 and the scatter becomes wider, indicating weaker consistency in the lower-level wind field.