Warm-sector torrential rainfall — heavy rain that develops far from synoptic fronts under weak large-scale forcing — is among the most poorly forecast phenomena in the South China monsoon region. The source study (Wu, Zhuang, Min & Meng 2020) used convection-allowing ensemble forecasts of a 19–20 May 2015 coastal warm-sector event to show a large spread in member performance (a practical predictability limit) and an intrinsic predictability limit as short as 6–12 hours, with good members distinguished by stronger nocturnal low-level southerly monsoon flow and stronger near-surface cooling over the northern mountains. The authors close with a coherent set of unresolved issues that remain open as of 2026: what beyond initial-condition errors (horizontal and vertical resolution, boundary-layer and microphysics parameterization, terrain representation, lateral boundary condition errors, model error) limits practical predictability of this event class; why nighttime near-surface temperature over the mountains declines so rapidly (diurnal thermal cycle versus cold pools from daytime precipitation); how predictability varies from case to case across the flow-dependent warm-sector class; how meso- and storm-scale ensemble perturbations should be formulated; and what update frequency and ensemble design a rapidly updating convection-allowing ensemble system tailored to warm-sector events requires. The citing literature shows substantial partial progress on each strand but no study that jointly partitions initial-condition, lateral-boundary, and model-physics error contributions for warm-sector events, no attribution of the nocturnal mountain-cooling mechanism, and no demonstrated warm-sector-tailored rapidly updating ensemble design.

Background

In the pre-summer rainy season (roughly April–June), coastal South China experiences torrential rainfall events that develop within the warm sector of the monsoon circulation, far from fronts and without strong synoptic-scale forcing — so-called warm-sector rainfall. Their prediction is a long-standing failure mode of numerical weather prediction because convection is triggered by subtle mesoscale ingredients: nocturnally strengthening low-level southerly monsoon flow over the South China Sea, its interaction with land and mountain breezes over the coastal orography, marine boundary-layer jets, and thermally driven convergence lines. The Southern China Monsoon Rainfall Experiment (SCMREX), a field campaign with dense rain-gauge, wind-profiler, and radar observations, was designed to study these events. Two concepts frame the source study. Practical predictability is the forecast skill actually achievable given realistic (imperfect) initial conditions (ICs) and models; intrinsic predictability is the hypothetical skill with almost perfect ICs, estimated by shrinking initial differences to far below operational ensemble spread and measuring error growth. Convection-allowing (equivalently convection-permitting) models use ~1–4 km horizontal grid spacing so deep convection is explicitly resolved rather than represented by a cumulus parameterization; an ensemble of such forecasts samples uncertainty from perturbed ICs, lateral boundary conditions (LBCs, the nest-interior forcing from a coarser driving model), and physics. Key parameterizations whose choices matter at these scales are the planetary boundary layer (PBL) scheme, which models turbulent mixing in the lowest ~1 km, and the cloud microphysics scheme, which models condensate and precipitation processes. A cold pool is the cold, dense outflow of air produced by evaporation of precipitation beneath a storm; its interaction with daytime-heated sloping terrain and the nocturnal thermal (diurnal) cooling cycle over mountains controls nighttime near-surface temperature and convergence. Wu, Zhuang, Min & Meng (2020, JGR-Atmospheres 125, e2019JD031313) performed convection-allowing ensemble forecasts of the 19–20 May 2015 SCMREX warm-sector event. Good members had a stronger low-level southerly monsoon flow over the sea and stronger surface cooling over the northern mountains 1–3 hours before convection initiation; ensemble sensitivity analysis tied event occurrence to these two factors. Spatial scaling of initial perturbations had little impact after 3 hours and meso-gamma-scale rainfall decorrelated after 12 hours, indicating an intrinsic predictability limit of only 6–12 hours; with IC differences reduced by nearly an order of magnitude the rainstorm appeared close to a bifurcation point. The study perturbed initial conditions only, taking the model configuration and physics as given. Its closing discussion therefore enumerates what remains unknown: the mechanism of the rapid nighttime temperature decline over the mountains; the contributions of resolution, PBL and microphysics parameterization, terrain representation, LBC errors, and other model errors to practical predictability; generalization from one case to the flow-dependent class of warm-sector events; formulation of proper meso-/storm-scale ensemble perturbation techniques and the associated error-growth dynamics; and the update frequency and ensemble design required by the rapidly updating, probabilistic, convection-allowing ensemble system that both predictability limits motivate. These facets share a common aim: a physically quantified basis for convection-allowing probabilistic forecasting of warm-sector torrential rainfall in the South China monsoon region — knowing which errors limit it, and what system design those errors imply.

Problem Statement

For warm-sector torrential rainfall events in the South China monsoon region, as framed by the closing discussion of Wu et al. (2020) following their 19–20 May 2015 SCMREX case study, determine: (1) the relative contributions to practical forecast error of initial-condition errors versus lateral boundary condition errors, horizontal and vertical resolution, planetary boundary layer and microphysics parameterization, terrain representation, and other model errors; (2) the mechanism of the rapid nighttime decline of near-surface temperature over the northern mountains, adjudicating the two candidate mechanisms identified in the source — nocturnal thermal (diurnal cycle) cooling and the cold-pool effect of daytime precipitation (other factors such as land–air thermal contrast, land/mountain breezes and boundary-layer mixing being optional context rather than required topics) — and its role in conditioning convection initiation; (3) how practical and intrinsic predictability vary flow-dependently from case to case, so that the single-case limits generalize (or fail to generalize) to the wider class of warm-sector heavy rainfall events; and (4) how meso- and storm-scale ensemble perturbations should be formulated, and what update frequency and ensemble design a rapidly updating, probabilistic, convection-allowing ensemble forecast system tailored to warm-sector events in this region requires. Together these questions address a common aim: a physically quantified basis for convection-allowing probabilistic forecasting of warm-sector torrential rainfall in the South China monsoon region. These four questions are jointly posed: the answer is their conjunction.

References

  1. Wu, N., Zhuang, X., Min, J., and Meng, Z. (2020). Practical and Intrinsic Predictability of a Warm‐Sector Torrential Rainfall Event in the South China Monsoon Region. Journal of Geophysical Research: Atmospheres 125, e2019JD031313. doi:10.1029/2019jd031313, https://doi.org/10.1029/2019jd031313
  2. Xu, D. S., Chen, H. W., Leung, J. C., Huang, H., and Zhang, B. L. (2023). Sensitivity of Nocturnal Warm Sector Rainfall Simulation to the Configuration of Initial and Lateral Boundary Conditions: A Case Study in Southern China Based on the Operational TRAMS Model. Journal of Geophysical Research: Atmospheres 128, e2022JD038452. doi:10.1029/2022JD038452, https://doi.org/10.1029/2022JD038452
  3. Shen, Y., and Du, Y. (2023). Sensitivity of boundary layer parameterization schemes in a marine boundary layer jet and associated precipitation during a coastal warm-sector heavy rainfall event. Frontiers in Earth Science 10, 1085136. doi:10.3389/feart.2022.1085136, https://doi.org/10.3389/feart.2022.1085136
  4. Huang, L., Luo, Y., and Bai, L. (2022). An Evaluation of Convection‐Permitting Ensemble Simulations of Coastal Nocturnal Rainfall Over South China During the Early‐Summer Rainy Season. Journal of Geophysical Research: Atmospheres 127, e2021JD035656. doi:10.1029/2021JD035656, https://doi.org/10.1029/2021JD035656
  5. Huang, L., Bai, L., and Zhang, Y. (2024). Analysis of Uncertainties in Convection-Permitting Ensemble Simulations of Land Breeze and Nocturnal Coastal Rainfall in South China. Journal of Meteorological Research 38, 1047–1063. doi:10.1007/s13351-024-4075-0, https://doi.org/10.1007/s13351-024-4075-0
  6. Rao, X., Zhu, K., Zhao, K., Chen, X., Hu, S., Liu, X., and Zhou, A. (2023). Evaluation and Error Source Analysis of Convection‐Permitting Forecasts for Localized Nocturnal Rainfall Over a Complex Mountainous Region in Pearl River Delta, South China. Journal of Geophysical Research: Atmospheres 128, e2023JD039065. doi:10.1029/2023JD039065, https://doi.org/10.1029/2023JD039065
  7. Zhou, L., Lin, X., Gao, C., Liu, Z., Xu, D., Jian, Y., Liang, J., Feng, Y., Li, Y., and Zhang, B. (2025). Increased Vertical Resolution of Initial Field in TRAMS Model Leads to Spurious Convection Over Sea Surface in Simulating a Typical Warm Sector Rainfall Event in the Southern China. Meteorological Applications 32, e70098. doi:10.1002/met.70098, https://doi.org/10.1002/met.70098
  8. Zhang, X., Li, J., and Zhang, G. (2025). Multiscale Interactions of Initial Condition Perturbations for Convection-Permitting Ensemble Forecasting over South China during the Rainy Season. Monthly Weather Review 153, 543–563. doi:10.1175/mwr-d-24-0181.1, https://doi.org/10.1175/mwr-d-24-0181.1
  9. Zhang, L., Min, J., Zhuang, X., Wang, S., and Qiao, X. (2023). The Lateral Boundary Perturbations Growth and Their Dependence on the Forcing Types of Severe Convection in Convection-Allowing Ensemble Forecasts. Atmosphere 14, 176. doi:10.3390/atmos14010176, https://doi.org/10.3390/atmos14010176
  10. Bao, X., Xia, R., Luo, Y., and Yue, J. (2023). Efficiently Improving Ensemble Forecasts of Warm-Sector Heavy Rainfall over Coastal Southern China: Targeted Assimilation to Reduce the Critical Initial Field Errors. Journal of Meteorological Research 37, 486–507. doi:10.1007/s13351-023-2140-8, https://doi.org/10.1007/s13351-023-2140-8
  11. Wang, J., Chen, J., Ma, Y., Li, X., and Ma, R. (2026). Sensitivity analysis of singular vectors (SVs) with diverse configurations and application of multiscale SV initial perturbations in convection-permitting ensemble prediction. Atmospheric Research 331, 108683. doi:10.1016/j.atmosres.2025.108683, https://doi.org/10.1016/j.atmosres.2025.108683
  12. Xia, Y., Zhang, H., Liu, X., Zhang, S., and Huang, X.-Y. (2026). Study on the reconstruction of initial and lateral boundary conditions for a convective-allowing ensemble forecast system. Journal of Atmospheric and Solar-Terrestrial Physics 278, 106697. doi:10.1016/j.jastp.2025.106697, https://doi.org/10.1016/j.jastp.2025.106697
  13. Chen, S., Hu, S., Li, H., Xiao, H., He, J., Zhang, Y., and Weng, F. (2026). Targeted All-Sky FY-3 microwave assimilation using localized ensemble transform sensitivity for warm-sector heavy rainfall prediction. Atmospheric Research 109291. doi:10.1016/j.atmosres.2026.109291, https://doi.org/10.1016/j.atmosres.2026.109291
  14. Zhang, H., Xia, Y., Zhang, S., and Ji, Y. (2026). Brief Introduction of Operational Convection‐Allowing Ensemble Prediction System of North China. Meteorological Applications 33, e70210. doi:10.1002/met.70210, https://doi.org/10.1002/met.70210