Forecast Uncertainty

PV Perspective

Although the skill of numerical weather models has increased significantly over the last decades (e.g. Bauer et al., 2015), very poor weather forecasts do still occur (e.g. Rodwell et al., 2013). Therefore, it is important to gain a better understanding of how forecast errors and uncertainties are growing with increasing forecast time.

Our main focus lies on the physical processes governing the synoptic-scale amplification of forecast errors and uncertainties near the tropopause. We use a potential vorticity (PV)-perspective as PV provides a useful variable to analyze error growth from a dynamical perspective (e.g., Davies and Didone, 2013). The evolution of the PV error (difference between the PV in the forecast and analysis) is clearly related to a misrepresentation of the Rossby wave packets near the tropopause.

PV error on 320 K for a) forecast day 2, b) forecast day 4, and c) forecast day 6. The black solid (dashed) contour denotes the dynamical tropopause of the analysis (forecast) on 320 K.
We derived a tendency equation to quantify PV error growth (cf., Davies and Didone, 2013) and partitioned it into individual contributions using a Helmholtz partitioning and piecewise PV inversion (cf., Teubler and Riemer, 2016). This partitioning allows us to quantify the relative importance of four processes to error growth near the tropopause: tropopause-near dynamics, tropospheric-deep interaction, upper-tropospheric divergence, and direct diabatic PV modification. The clear separation between upper- and lower errors motivates the partitioning into tropopause-near dynamics and tropospheric-deep interaction.

3D picture of the PV error between 850 and 250 hPa and the Θ error on 875 hPa for forecast day 4. The black solid (dashed) contour denotes the dynamical tropopause of the analysis (forecast) on 250 hPa.