Evaluation of Real-Time Forecasts during DECODE 2022 in China

Convection initiation (CI) is a major challenge in weather forecasting worldwide. As an important lifting mechanism of CI, boundary layer convergence lines (boundaries here after) may occur due to differential of heating from heterogeneous underlying surface. The Hetao area in North China contains highly heterogeneous vegetation where near-surface boundaries parallel to the oasis–desert border often appear and trigger convection. Aiming at examine the CI mechanisms due to vegetation variability across oasis-desert transition zone in Hetao area by high-resolution observations and numerical simulations, the DEsert-oasis COnvergence line and Deep convection Experiment (DECODE) is performed from 5 July to 9 August in 2022. This current study will introduce the real-time forecasts in support of DECODE operations, and provide the evaluation of forecast skills on boundaries, CI and related convective systems, as well as detailed investigation on forecasts of typical cases during the field campaign.

During the DECODE period, the forecast team provides deterministic forecasts for the occurrences of boundary and CI and qualitatively predicted the location, timing and strength of the boundary based on convection-permitting WRF forecasts. WRF-ARW model is configured with three two-way nested domains and the innermost 1-km domain fully covered the IOP region. GFS forecasts with 0.25°×0.25° horizontal grid spacing and 3-h time interval are used to provide the initial and boundary conditions. For DECODE operations, the forecast is issued daily at 1500 LST Day 1, providing the forecast with a valid time window of 0800–2000 LST Day 2 (integration of 6–18 h). The primary model outputs include composite radar reflectivity, horizontal convergence on the bottom model level, surface wind and temperature fields available every 10 min, which effectively guide the deployment of field experiment.

One of the key challenges the forecast team faces during the campaign is how to identify the thermally-induced boundary out of the multiple convergence lines overpredicted by the model, which is likely due to other forcing such as the terrain effects and previous convection. Aided by our previous knowledge on the unique features and synoptic environment for the boundary in this area, the boundary occurrence is more confidently predicted if the convergence line locates in a weakly forced synoptic environment with southerly near surface wind serving as background. The near-surface temperature gradient over vegetation-contrast regime during daytime, and the enhancement of the local thermal circulation wind from oasis are also prominent harbingers in the boundary forecasts. During the 36-day field experiment, there are 23 days with the occurrence of boundary, and 11 days wherein are observed with CI associated with boundary. Though with perceptible and varied forecast errors in timing and location, our real-time forecasts decently predict the occurrences of boundary within IOP region in 21 out of 23 observed days and an impressive equitable threaten score of 0.4559. By contrast, the hit rate of CI associated with boundary is relatively lower and 6 out of 11 days are predicted, suggesting higher predictability of boundary over this area than that of CI associated with boundary. Typical hit, miss, and false alarm examples of boundary formation and CI will also be further evaluated and investigated in details.