Hidden Warning Signs Detected Before China’s Sudden Front-Free Downpours

Some of the most dangerous rainstorms in eastern China never announce themselves with an approaching weather front. They erupt deep inside the warm, humid air mass hundreds of kilometers ahead of any frontal boundary, under weak large-scale forcing, and they are notoriously hard to forecast. A new five-year study published in Atmospheric Chemistry and Physics has now quantified just how much these elusive storms matter for regional rainfall, and has captured, minute by minute, the dynamic changes in the lowest kilometer or two of the atmosphere that unfold in the hour before the first raindrops hit the ground.

The research team, led by Weilong Deng and Jianping Guo of the Chinese Academy of Meteorological Sciences and Lanzhou University, focused on the eastern China plains, a densely populated belt spanning the North China Plain and the middle-and-lower Yangtze Plain. Warm-sector heavy rainfall, as this phenomenon is known, was first documented in South China during the pre-summer rainy season, but it has since been implicated in disasters across the country, including the localized Shijiazhuang rainstorm of July 2017 and the extreme Beijing rainstorm of late July 2025. Because organized frontal lifting is absent, these storms tend to be highly localized, develop abruptly, move slowly, and dump disproportionate amounts of water on small areas, producing urban waterlogging, flash floods, and river flooding that operational forecast models consistently struggle to anticipate.

To study the phenomenon rigorously, the team assembled an unusually rich observational arsenal covering the warm seasons of 2021 through 2025. Minute-resolution records from 493 surface rain gauges were combined with ERA5 reanalysis fields and a mesonet of radar wind profilers, instruments that measure wind speed and direction continuously at vertical resolutions of 120 meters and temporal resolutions of six minutes. The researchers built an objective identification framework that integrates automated front detection based on the thermal front parameter applied to pseudo-equivalent potential temperature at 850 hectopascals, a warm-sector mask requiring warm advection at both 925 and 1000 hectopascals with southerly flow, and a station-specific threshold set at the 99th percentile of hourly rainfall. Events had to occur between 200 and 400 kilometers from the detected front, and any rainfall influenced by a tropical cyclone was excluded to keep the sample clean.

The scheme identified 477 warm-sector heavy rainfall events alongside 6002 other heavy rainfall events across the study region. The contrast between the two populations was striking. Warm-sector events were spatially scattered and weakly organized: 196 of the 493 stations recorded none at all, while 297 recorded at least one, with no clear relationship to terrain elevation within the deliberately low-relief domain. Other heavy rainfall, by contrast, was ubiquitous and well organized, increasing from north to south, with every station recording at least five events. Temporally, warm-sector rainfall concentrated in midsummer, peaking in July and August, and showed a strong nocturnal preference, with a dominant onset mode between 20:00 and 03:00 Beijing Time and only a weaker afternoon mode around 13:00 to 16:00. Other heavy rainfall peaked sharply in the late afternoon, a signature of thermally driven convection.

The hazard posed by these storms becomes clear in the event statistics. Warm-sector heavy rainfall accounted for only about 5 to 10 percent of annual heavy rainfall occurrences, yet its events were shorter, with a median duration of 5.7 hours compared with 6.8 hours for other heavy rainfall, while delivering comparable or greater water. Median accumulations reached 70.3 millimeters for warm-sector events against 56.2 millimeters for the others, and the upper tail was markedly heavier, with a 90th percentile of 156.1 millimeters versus 120.2 millimeters. Among the 297 stations that recorded at least one such event, the mean contribution of warm-sector rainfall to total heavy-rainfall accumulation was 13.8 percent, with a maximum of 48.1 percent at a single station. In other words, at some sites, nearly half of all the heavy rain that fell over five warm seasons arrived in brief, front-free bursts.

Composite reanalysis fields revealed the thermodynamic stage on which these storms play out. At the hour of onset, event stations sat within a moist tongue at 850 hectopascals, with specific humidity reaching 14.8 grams per kilogram, near the northern leading edge of warm, humid air characterized by pseudo-equivalent potential temperatures of roughly 350 to 352 kelvin, immediately south of the tightly packed gradient marking the frontal zone. Critically, low-level moisture flux convergence between 1000 and 700 hectopascals intensified progressively in the hours before onset, deepening from a modest signal six hours out to a pronounced convergence core of less than minus five times ten to the minus four kilograms per square meter per second sitting essentially on the station at onset. This progressive moistening and convergence preconditions the column for deep convection, but under weak synoptic forcing it is not by itself sufficient to trigger the storms.

That is where the radar wind profilers came in. Aligning profiler observations with the precisely timed onset of rainfall at 37 co-located stations, the team composited 62 warm-sector heavy rainfall events over the 180 minutes preceding onset, and compared them with 271 warm-sector events that never reached heavy-rainfall intensity. The composites showed a marked kinematic evolution in the lowest levels of the atmosphere. Wind kinetic energy below three kilometers intensified during the final hour before onset. Low-level vertical wind shear, concentrated in shallow layers below about 1.5 kilometers and reaching values up to about fourteen times ten to the minus three per second, strengthened as onset approached. The winds veered clockwise with height, a signature of warm-air advection consistent with the moist southerly flow feeding the storms.

The most distinctive precursor involved the low-level jet, a band of fast wind in the lowest few kilometers of the atmosphere. A low-level jet appeared at least once during the analysis window in 46.8 percent of the heavy-rainfall events, compared with 39.5 percent of the non-heavy events. More telling was the timing: in the heavy-rainfall composites, jet occurrence frequency rose from roughly 5 to 11 percent at earlier lead times to a peak of 21.0 percent at 36 minutes before onset, whereas the non-heavy composites remained comparatively steady at 9 to 13 percent. The event-mean increase in jet occurrence between the early and late windows was 7.0 percentage points, though the authors caution that the bootstrap confidence interval of minus 0.25 to 14.18 percentage points includes zero, meaning the tendency carries appreciable uncertainty. Enhanced ascent also emerged, with the strongest upward air movements near 1 to 2 kilometers intensifying during roughly the final 30 minutes before onset, and the upper tail of ascent speeds growing more pronounced than in the non-heavy comparison sample.

The authors are careful about what these composites can and cannot establish. The distributions of jet properties and ascent speeds overlap substantially between the heavy and non-heavy samples, so the presence of a low-level jet or strong low-level shear alone does not guarantee a downpour. The analysis does not locate jet termini, resolve the horizontal convergence they might generate, or test the nocturnal inertial oscillation mechanism that has been invoked to explain nighttime warm-sector rainfall in South China. Nor can single-point profilers rule out contamination of vertical velocity measurements by falling hydrometeors aloft before rain reaches the gauge. What the study does deliver is a resolved, five-year, station-based picture of the pre-onset environment, and a set of candidate kinematic indicators that are continuously observable by operational profiler networks.

The implications for forecasting are significant. Quantitative precipitation forecast skill for warm-sector heavy rainfall remains low in both global and regional models, precisely because convective initiation in this regime is governed by mesoscale and boundary-layer processes that conventional observations and models resolve poorly. The new results suggest that monitoring the late-hour rise in low-level jet occurrence and the intensification of low-level ascent could provide short-range warning signals, complementing the thermodynamic information available from moisture and instability fields. Related work has already shown that profiler mesonets can resolve lower-tropospheric convergence preceding heavy rainfall, and observing system simulation experiments over the Beijing-Tianjin-Hebei region indicate that assimilating profiler data can improve mesoscale wind analyses and short-term rainfall forecasts. The next steps, the authors argue, are detailed case studies combining profiler records with weather-radar reflectivity and dual-polarization fields to test whether back-building or training storm organization underlies the large accumulations, convection-permitting simulations to verify the jet-convergence-ascent pathway, and event-level validation to establish the specificity, useful lead time, and false-alarm characteristics of the newly identified precursor signals.

Subject of Research: Warm-sector heavy rainfall identification and precursor low-level dynamic signals over the eastern China plains

Article Title: Warm-sector heavy rainfall over the eastern China plains: identification, characteristics, and precursor dynamic signals

Article References: Deng, W., Guo, J., Li, N., Zhang, Z., Wang, Y., & Yang, Y. (2026). Warm-sector heavy rainfall over the eastern China plains: identification, characteristics, and precursor dynamic signals. Atmospheric Chemistry and Physics, 26(19), 14259-14275. https://doi.org/10.5194/acp-26-14259-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-14259-2026

Keywords: warm-sector heavy rainfall, eastern China plains, convective initiation, low-level jet, radar wind profiler, moisture flux convergence, nowcasting, vertical wind shear, nocturnal rainfall, rain gauges, ERA5 reanalysis, forecasting