Cloud diurnal cycles were analyzed using the Channel 13 Brightness Temperature (BT; infrared wavelength band 10.4 µm) from the Himawari-8 satellite Level 1 atmospheric and cloud property dataset, which covers the Australian region at an hourly temporal and 2 km spatial resolution . The BT data was regridded for the study domain at 2 km spatial resolution, spanning 21.5–14° S and 143–152° E, encompassing a region of north-east Queensland, the central GBR and the Coral Sea, including the coverage areas of three weather radars (Fig. ). Diurnal cloud characteristics were analyzed for the January to March period – including small convective clouds observable with the high resolution Advanced Himawari Imager (AHI). A shortened Coral Bleaching Season (CBS) was analyzed due to data availability (January–March rather than the defined CBS January–April). Previous studies of the same Level-1 dataset have provided detailed insights on diurnal cycles of convection in the tropical Maritime Continent .

Rainfall estimates were obtained from Level-2 historical ground reflectivity retrievals derived from C-band Doppler radar data. This study uses data from the Townsville, Cairns and Willis Island operational weather radars operated by the Australian Bureau of Meteorology, archived on the National Computational Infrastructure . Historical ground reflectivity data has spatial resolution of 1 km, extending horizontally from the radar site up to a range of 150 km in all directions (360 °), with a temporal resolution between 5–10 min (increasing in frequency towards the end of the analysis period). The analysis focuses on the full CBS (January–April), for the years when data from for all three radars are available (2012–2022).

Surface wind data from the Australian Bureau of Meteorology's automated weather stations (AWS) in Townsville and Cairns are used to highlight the complex local-scale influences on surface conditions. AWS data was analyzed for the full CBS for the period of 1990 to 2020, with a temporal resolution of 30 min. The location of both stations is shown in Fig. , with height above sea level of 4.34 m for Townsville and 2 m for Cairns (with wind measurements 10 m above ground level). Hourly averaged wind direction was calculated using the circular mean, which is used for averaging angular data (i.e. taking the average of 350 and 10 ° to result in 0 °).

The Australian Bureau of Meteorology Atmospheric high-resolution Regional Reanalysis for Australia (BARRA-R2) was used to provide meteorological context for the radar rainfall and satellite cloud observations. BARRA-R2 covers the Australian region from 1979 to the present, with a horizontal grid resolution of 12 km and a temporal resolution ranging from hourly to monthly; this study focuses on variables at hourly resolution. Full documentation can be found in . Previous studies have used the BARRA-R2 reanalysis to examine diurnal cycles of temperature and humidity over northern Australia . The higher horizontal resolution of BARRA-R2 compared to global reanalyses enables improved representation of local features such as coastal and topographic circulations, though fine-scale processes (such as convection) remain parametrized.

In this study, BARRA-R2 was used primarily for defining the background wind and local environmental conditions through composite analysis. Large-scale meteorological conditions were characterized by 850 hPa level winds, whose direction and strength relative to the coastline have been shown to influence the diurnal cycle of coastal precipitation . The mean 850 hPa wind direction at 12:00 Local solar Time (LT) from BARRA-R2 was calculated across each radar domain for the full CBS from 2012 to 2022. For each day within the temporal domain the 12:00 LT mean wind direction was computed (by first averaging zonal and meridional components separately before computing wind direction), and each hour of that day (00:00–23:00 LT) assigned to the resultant wind regime. In this way composite results are reflective of a full diurnal cycle. Wind regimes at 850 hPa were categorized as north-easterly (0–90 °), south-easterly (90–180 °), south-westerly (180–270 °) and north-westerly (270–360 °). This classification was designed to establish a clear northerly-southerly distinction, effectively separating tropical air masses (typically warmer and more humid) from those of mid-latitude origin (typically cooler and drier). Due to the variation in coastline orientation across the radar domains, a simple easterly-westerly division was also adopted to ensure consistency and interpretability in the analysis.

To better understand the convective environments associated with each wind regime, the thermodynamic structure of the atmosphere was examined using vertical profiles of equivalent potential temperature (θe). Profiles were constructed at the BARRA-R2 grid point nearest to each radar site (Townsville, Cairns, and Willis Island), using temperature and specific humidity fields across all pressure levels, calculated using the open-source MetPy Python package. θe is a combined measure of temperature and moisture, and it provides a useful indicator of an air parcel’s convective potential. Higher values of θe suggest warmer or more moisture-rich conditions, whereas lower values are associated with colder or drier conditions. The vertical structure of θe also reveals information about boundary layer depth and stability, taking into account latent heat release from condensation of the moisture in the air. Analysis of θe anomalies associated with each wind regime was done by constructing a smoothed rolling mean climatology with a 31 d centered window (for the period 1979–2024 January–April) to remove seasonal signals.

Instantaneous rain rate (R mmh-1) was derived from the historical ground radar reflectivity factor (Z mm6m-3) using the Z-R relationship, Z=aRb. Constants a and b used in this study (see Appendix Table for values) are estimated by using rain gauge observations near each radar site. The exception to this is the Willis Island radar, which does not have a co-located in situ rain gauge for calibration , and also exhibits the highest proportion of missing temporal data (6.65 % of times analyzed), and is therefore associated with larger uncertainty. Prior to the diurnal analysis, the rain rate data required pre-processing to remove regions with erroneous data caused by the partial radar beam blockage from complex topography around the Townsville and Cairns radar sites. To identify regions with erroneous rain rate data, the climatological mean rain intensity over 2012–2022 was calculated, and a lower threshold was applied to exclude affected data from the diurnal analysis. No lower threshold was applied to the Willis Island radar data, as the highest elevation on the island is 20 m above sea level and topographic blockage is absent.

Diurnal cloud and rainfall analyses focused on three distinct regions: coastal land, coastal ocean and open ocean, as indicated in Fig. . Coastal land is defined as land areas within the Cairns and Townsville radar domains, while coastal ocean encompasses the oceanic portions of these radar domains. These regions were delineated using land-sea masks derived from the National Oceanic and Atmospheric Administration (NOAA) 2-min Gridded Global Relief Data (ETOPO2) v2 dataset, which provides high-resolution topography, bathymetry and shoreline information . In this classification, coastal ocean areas are defined by surface elevation below 0 m (z<0) and coastal land by elevation above 0 m (z≥0). The open ocean is defined as the area within the Willis Island radar domain. Given that Willis Island lies about 450 km from the Queensland coastline, and measures only about 500 m in length and 150 m in width, it is considered to have little to no land-surface influences on local convection and rainfall processes.

The diurnal mean hourly values of cloud BT, rain intensity and rainfall frequency across the CBS were computed for each of the three regions – coastal land, coastal ocean and open ocean – under the defined wind regimes. Rain intensity and frequency were calculated using a ±30-min window around each hour to match the hourly temporal resolution of the cloud BT dataset. Rain intensity calculations include all time periods (i.e. raining and non-raining) to maintain consistency with BT diurnal cycles. Rain frequency was defined as the percentage of points in a given time step where rain rate was greater than 0.1 mmh-1. An example of satellite-derived channel 13 BT and corresponding radar rain rate retrievals at a single time step over the Cairns radar domain is provided in Fig. .

The diurnal evolution of cloud and rainfall was examined through a Hovmöller analysis. A zonal transect was chosen which spans the coastal land, coastal ocean and open ocean regions (shown in pink in Fig. ). This transect crosses both the Cairns and Willis Island radar domains, offering simultaneous data on cloud and rainfall characteristics for coastal land, coastal ocean and open ocean regions. Mean hourly values of BT and rain intensity were calculated for each longitude point along the transect using a narrow latitude range (16.6–17° S), to minimize radar-induced noise. These data were used to generate Hovmöller diagrams that represent the mean diurnal cycle of each variable and reveal patterns in the spatiotemporal propagation of cloud and rainfall features across the domain.

The large-scale wind regimes during the CBS period play a crucial role in shaping local thermodynamic environments and convective behavior. This section examines the characteristics of the four wind regimes (850 hPa winds) defined in Sect. . Figure a shows histograms of the 850 hPa winds for each regime at the three study sites, while the monthly frequency distribution is presented in Fig. b–d. Additional statistical summaries are provided in Appendix Table .

The south-easterly wind regime, representing the trade winds, dominated across all three sites, accounting for 57 %–67 % of the CBS period. At Townsville, south-easterly winds were least frequent in January (about 50 %) but increased steadily to 75 % by April. Cairns and Willis Island show relatively similar frequencies in January–February followed by a marked increase in March and April. This north-south contrast may reflect stronger tropical influences at the lower-latitude sites early in the season, and a progressive strengthening of the trade wind belt as the Inter-Tropical Convergence Zone (ITCZ) shifts northward.

The north-easterly wind regime was the second most common regime across all three sites during the CBS, occurring 15 %–23 % of the time. It was most commonly observed in January and February over Townsville (31.5 % and 27.6 % respectively). Cairns and Willis Island exhibit less month-to-month variation in north-easterly frequencies (19.1 %–20.3 %) during the same period of time, with a similar drop-off from March onward (Fig. c, d). This regime’s seasonal decline coincides with the strengthening south-easterlies, indicating weakening of tropical influences.

North-westerly and south-westerly regimes were relatively infrequent, each occurring in 6 %–13 % of CBS periods, with peak frequencies during January-February and a marked decline by April. South-westerlies became especially rare at Willis Island in April (3.3 %). While the northwesterly regime is likely associated with transient tropical lows, monsoonal flow, or monsoon burst events , the south-westerly regime is likely more closely linked to the passage of cold fronts driven by deep mid-latitude troughs along the east coast . The enhanced westerly flow and deep convection typical of north-westerlies are also consistent with active MJO phases influencing northern Australia .

In summary, south-easterly trades dominate the CBS period, with a clear seasonal transition replacing early-season tropical and cyclonic regimes – which drive the higher frequency of northerlies . Latitudinal differences suggest stronger tropical influence at Cairns and Willis Island, particularly early in the season, likely modulated by MJO activity, monsoon flows and ITCZ positioning.

The CBS mean and wind-regime-averaged θe profiles for the three sites are shown in Fig. a, c, e, with anomalies relative to a smoothed rolling mean climatology shown in Fig. b, d, f. The four large-scale wind regimes can be interpreted in terms of two key controls on convective potential: (i) the moisture and temperature characteristics of the air mass (θe magnitude), and (ii) the vertical stability structure (θe gradient and inversion strength). To complement the θe analysis, corresponding Convective Available Potential Energy (CAPE) and vertical velocity cross-sections from the BARRA-R2 reanalysis are presented in Appendix (Figs. –), providing additional insight into the spatial patterns of instability and large-scale vertical motion associated with each regime.

The four wind regimes represent fundamentally different thermodynamic and dynamical environments for convection. The south-easterly regime represents a cool, dry, and highly stable environment, with persistent negative θe anomalies through the column. Stronger warming aloft relative to the boundary layer enhances stability above the 600 hPa inversion. This suppression is reflected in uniformly low CAPE, generally below 1000 Jkg-1 over land and below 1500 Jkg-1 over the adjacent ocean (Fig. d, h). The absence of large-scale ascent over the marine sector further limits opportunities for convective development (Fig. d). In contrast, the north-easterly regime is characterized by a warm and moist lower and mid-troposphere, but weak vertical θe gradients hence only moderate instability. Despite the moderate CAPE (2000–2500 Jkg-1) over the ocean, a pronounced band of deep ascent along the coastal ranges highlights the importance of orographic lifting in triggering convection (Fig. b). The interaction of moist north-easterly flow with the steep coastal terrain appears to locally release the available instability, focusing convection along the windward slopes.

The south-westerly regime exhibits a more heterogeneous structure, with weak positive θe anomalies over land suggesting some surface-driven destabilization, while open ocean conditions remain close to climatology. CAPE shows a marked land-sea contrast, exceeding 3000 Jkg-1 over the coastal ocean but remaining substantially lower further inland (Fig. c, g). Localized ascent along the windward coastal ranges highlights the importance of orographic forcing (Fig. c). However, the ascent is weaker than in the north-easterly regime, likely because the south-westerly flow encounters less abrupt upwind terrain. As a result, convection is likely influenced more by land surface heating and boundary-layer growth than by orographic lifting alone. By contrast, the north-westerly regime stands out as the most convectively favorable, with strong positive θe anomalies in the lower and mid-levels and enhanced vertical gradients indicative of an unstable air mass. CAPE exceeds 2000–3000 Jkg-1 across much of the domain and is accompanied by deep, widespread ascent (>0.02 ms-1), demonstrating that both thermodynamic instability and large-scale dynamical forcing support convection. This structure is consistent with monsoonal or tropical intrusions that inject warm and moist air masses, providing both moisture supply and reduced stability . Together, these characteristics highlight how large-scale flow direction modulates both the source characteristics of air masses and the vertical thermodynamic structure that governs convective potential across the study region.

The diurnal evolution of θe under each wind regime is shown in Fig. as anomalies relative to a smoothed rolling mean climatology over the CBS. Consistent with Fig. , the wind regimes primarily define the background thermodynamic state, setting the moist static energy and stability structure. In particular, north-westerly flow is associated with the highest θe, and weakest stability, north-easterly flow shows intermediate moistenining, and south-easterly flow is characterized by persistently low θe and enhanced stability. Figure further shows how this background state is modulated by the diurnal cycle.

A clear contrast emerges between the coastal land sites (Townsville and Cairns) and the marine site (Willis Island). Over land, θe profiles exhibit strong diurnal modulation of the lower troposphere, especially under north-westerly regimes (Fig. a, f). Positive θe anomalies larger than 6 K occur during the afternoon and early evening, reflecting daytime surface heating and boundary-layer deepening, which enhance near-surface moist static energy and convective potential. Townsville shows the strongest response under north-westerly flows, consistent with more intense land-surface forcing under this regime. Cairns exhibits a similar but slightly more muted behavior except under south-easterly conditions, where the diurnal signal is substantially suppressed (Fig. f, g, h). In contrast, Willis Island shows little systematic diurnal variability across all regimes, indicating that the absence of strong land surface forcing limits sub-daily modulation of θe in the marine boundary layer (Fig. k, l, m, n). This sharp land-sea contrast highlights the dominant role of surface heating in driving diurnal thermodynamic variability. Overall, these features underscore the interplay between large-scale flow direction and surface type in shaping diurnal thermodynamic evolution and the likelihood of convective activity.

Next we examine how these thermodynamic conditions influence the diurnal evolution of cloud and rainfall patterns. In general, differences in vertical stability and moisture content across wind regimes are expected to influence not only the likelihood of convection but also its timing and intensity throughout the day. The mean diurnal cycles of satellite-derived Brightness Temperature (BT) and radar-derived rainfall frequency and intensity under different wind regimes are presented in Fig. , with the seasonal mean conditions shown in Fig. m, n, o. BT is used as a proxy for cloud-top height, with colder BTs indicating deeper convection, and warmer BTs suggesting shallow cloud or clearer-sky conditions.

A first-order feature across all regimes is again a strong land-sea contrast in the diurnal cycle. Over land, BT typically increases through the morning and decreases sharply in the afternoon to evening, consistent with boundary-layer growth and the development of deep convection, accompanied by a corresponding increase in rainfall frequency and intensity. Over the coastal ocean, this signal is phase-shifted, with rainfall and colder BTs generally occurring from evening to early morning. In contrast, Willis Island exhibits minimal diurnal variability in both BT and rainfall, reflecting weak surface forcing in a more homogeneous marine environment.

Superimposed on this common structure are regime-dependent modulations that primarily affect amplitude, timing, and convective characteristics. The north-easterly regime is characterized by relatively colder BTs and enhanced rainfall compared to the seasonal mean, especially over land. Cairns experiences the strongest diurnal signals among all three sites, with BTs over land warming through the morning (reaching 277 K at 11:00 LT) and cooling rapidly into evening (253 K by 18:00 LT), indicative of vigorous deep convection. Townsville shows a similar warming trend (approximately 1 K h⁻¹) but with more gradual cooling, about half of the rate at Cairns, likely reflecting weaker orographic enhancement. Rainfall frequency and intensity at both land sites closely follow this evolution, peaking in the afternoon-evening, especially at Cairns, with moderate amplitudes of 5 % and 0.61 mmh-1. A notable feature is that peak rainfall at 16:00 LT precedes the coldest BTs by 3 h, suggesting that the minimum BTs likely reflect the later anvil-dominated stage of convection rather than the peak rainfall phase.

In contrast, the south-easterly regime is characterized by the warmest BTs and the weakest rainfall across all land regions, consistent with shallow, low-topped cumulus clouds and suppressed deep convection. Unlike the north-easterly regime, BT and rainfall diurnal cycles in the south-easterly regime over Cairns and Townsville are more synchronized and subdued, though rainfall intensity over Cairns is slightly enhanced in the afternoon – again likely linked to localized orographic lifting and elevated terrain heating, which can partially overcome the otherwise stable background conditions. Over the coastal and open oceans, both BT and rainfall variability remain weak throughout the day, consistent with stable marine boundary layers dominated by shallow trade-wind cumulus and limited diurnal destabilization.

South-westerly regime exhibits very strong diurnal modulation in both BT and rainfall, suggesting enhanced surface-driven convection. Over land, BTs are largest around midday (296 K at Cairns and 298 K at Townsville), indicating minimal cloud cover, before dropping sharply in the afternoon. Correspondingly, rainfall frequency and intensity are minimal around midday but increase rapidly in the early afternoon, reaching maxima in the late afternoon over land and shifting to early morning over the coastal ocean. Cairns generally experiences more frequent and intense rainfall than Townsville, particularly in the afternoon-evening period. Oceanic regions show comparable rainfall magnitude near the coast, while Willis Island again shows little diurnal variability, reflecting its open-ocean character. These patterns align with the south-westerly regime fostering deeper, more organized convection through localized surface heating and atmospheric destabilization.

The north-westerly regime stands out as the most convectively active, with the coldest BTs and highest rainfall frequency and intensity. Over land, convection deepens through the afternoon, with progressively colder BTs and increasing rainfall. The most intense rainfall over land peaks at 17:00 LT, whereas peak rainfall frequency and cold BTs lag by 3 h, again reflecting a transition from active convection to mature and stratiform/anvil-dominated cloud. Over the coastal ocean, strong diurnal structure is evident, with daytime (11:00–14:00 LT) BT warming and suppressed rainfall followed by nocturnal intensification. The large amplitudes of cloud and rainfall signals over the coastal ocean highlight the strong diurnal forcing under this regime. In contrast, Willis Island shows only weak signals in both fields, consistent with the open-ocean environment. These features are consistent with enhanced instability associated with elevated θe and strong vertical gradients fueled by moist north-westerly inflow. Diurnal patterns over land and coastal ocean reflect surface heating and boundary-layer dynamics.

Overall, the diurnal evolution of brightness temperature and rainfall reflects a consistent interaction between large-scale wind regimes and surface forcing. The wind regimes control the background thermodynamic environment and convective potential, while land-sea contrasts regulate the timing, propagation, and amplitude of the diurnal cycle, with the strongest organization occurring under south-westerly and north-westerly regimes and the weakest under south-easterly trade conditions.

Building on the discussion of diurnal BT and rainfall cycles, Hovmöller diagrams were produced to examine the spatial evolution and propagation of convection through time. Using the method described in Sect. , Fig. shows BT and rain intensity (averaged across both raining and non-raining periods) under each wind regime defined by the Cairns domain with terrain elevation in Fig. e, f (for reference).

A consistent feature across all regimes is the strong influence of coastal geometry and topography on convective initiation and propagation. Convection typically develops over or near the coastal ranges during the afternoon before propagating either inland or offshore depending on the prevailing background flow. The extent, speed, timing, and organization of this propagation vary substantially between regimes, indicating strong modulation by the large-scale thermodynamic and kinematic environment identified in the previous sections.

The clearest offshore propagation occurs under the north-westerly regime. In this regime, BTs below 253 K develop by 16:00 LT over coastal land, particularly near regions of elevated terrain (Fig. a, e). Nighttime propagation of convection is evident over the coastal and open oceans, with BTs below 263 K reaching 200–300 km offshore, and rainfall extending to the eastern edge of the Cairns radar domain. Inland propagation is comparatively weak in this regime. The coldest BTs are largely confined to near-coastal waters (within 100 km of the coastline) and are associated with relatively slow-moving rainfall features (2–5 ms-1). In contrast, warmer BTs propagate farther offshore more rapidly, at speeds of 12–15 ms-1. In this regime, the coldest BTs correspond closely with the most intense rainfall.

Under the north-easterly regime (Fig. b), widespread moderate coastal rainfall persists throughout the day, consistent with a moist environment, while colder BTs (below 263 K) develop over land and adjacent coastal waters during the evening to early morning period. From around 12:00 LT, BTs below 273 K propagate both inland and offshore, although the inland signal is more coherent and appears more directly linked to coastal convective development. Inland propagation speeds reach approximately 30 ms-1, extending more than 300 km westward from the coastline. The coldest BTs are typically found on the leeward side of the mountain ranges between 19:00–22:00 LT, traveling westward – suggesting convection enhanced by mountain effects that produce strong thunderstorms propagating westward into the evening. Offshore propagation is also apparent in the BT field, though this signal is less clearly tied to coastal convection and may partly reflect broader open-ocean cloud variability. Compared to the north-westerly regime, convection appears less organized offshore but more effective at penetrating inland.

For south-westerly regimes (Fig. c), coastal rainfall exhibits a pronounced diurnal cycle, with little to no rainfall occurring from early morning through to midday before convection rapidly intensifies during the afternoon. BTs below 283 K form over land during the afternoon before propagating both inland and offshore. The coldest BTs occur between 18:00–00:00 LT over land. Early morning convection moves offshore by approximately 100 km, associated with light rainfall. Rainfall contours generally track the BT evolution closely, although peak rainfall (blue) often occurs prior to the appearance of the coldest BTs – again suggesting these may correspond to high-level anvil clouds decoupled from topographic influence.

Under south-easterly regimes, light-to-moderate coastal rainfall is present (similar to north-easterly regimes) but confined to a narrower coastal region. Cooler BTs (283 K) start propagating inland from the coast by 15 LT, penetrating more than 200 km inland at speeds of 12–15 ms-1, with the coldest BTs emerging about 250 km from the coast in the late evening between 20:00–03:00 LT, suggesting storm enhancement during inland propagation. Rainfall follows a similar westward propagation, though slightly slower (10-15 ms-1), and generally tracks the BT contours. Compared with the other regimes, convection remains weaker and less spatially extensive, consistent with the more stable trade-wind environment discussed earlier. Localized regions of enhanced afternoon rainfall over elevated terrain further indicate an important role for orographic forcing under otherwise suppressed convective conditions. Over the coastal ocean, convection is less coherent and no strong offshore propagation signal is evident.

The spatial patterns and propagation characteristics of the BTs and rainfall suggest that land-sea breeze circulations play a variable but important role in shaping the diurnal development of convection over north-east Queensland, depending on the prevailing wind regime. Figure presents surface wind roses at 07:00 and 15:00 LT, alongside the diurnal evolution of hourly average wind speed and direction at Cairns and Townsville weather stations. We use station-based surface wind observations here rather than the BARRA-R2 reanalysis as the 12 km horizontal resolution of the reanalysis may be too coarse to reliably resolve these coastal circulations. While a classic wind reversal pattern is not observed (likely due to the region’s complex coastal geometry and orographic influences), distinct afternoon landward winds indicative of sea-breeze influence are observed at both stations. Similarly, evening seaward winds consistent with land-breeze activity are also evident, supporting the presence of thermally driven circulations despite terrain-induced variability.

Further detail on the variable influence of land-sea breeze circulations can be seen in the contrasting diurnal surface wind patterns observed across different wind regimes at Cairns (Fig. in the Supplement). Afternoon landward winds are evident across all regimes, though with varying strength, which is a reflection of the modulating effect of background winds and possibly the topographic flows. This variation helps explain the differing speeds and inland reach of the convective propagation seen in Fig. , where convection initiates near the coast in the early afternoon and moves inland at speeds characteristic of sea-breeze dynamics. In contrast, seaward winds are most pronounced under the two westerly regimes, aligning with the offshore-propagating convection that peaks in the evening shown in Fig. . This suggests a potential role for nocturnal land breezes in triggering marine convection, though the extent of offshore propagation also depends on ambient moisture availability and background wind strength, which influence the distances over which colder BTs are observed.