Located within the tropical and subtropical monsoon zone of southern China (Figure 1), Guangdong Province experiences high temperatures and humidity year-round, with hot, humid summers and warm, humid winters, which is a typical humid tropical region (Ye et al., 2022; Zhang et al., 2017). It is one of China’s most rapidly urbanizing provinces (Cai and Deng, 2020), with the Pearl River Delta-centered Guangdong-Hong Kong-Macao Greater Bay Area (GBA) being one of the world’s largest bay economies (Hui et al., 2020). Guangdong experienced more frequent and intense heatwaves and extreme rainfall in recent years due to urbanization and global climate change. For example, the frequency of severe heatwaves rose by 0.141 events per year between 1960 and 2015 (Jiang et al., 2015; Wu and Du, 2011; Zhang et al., 2018). Therefore, Guangdong Province encompassing GBA is a particularly suitable region for investigating the impacts of urbanization and extreme weather on wet-bulb temperature in humid regions.

Long-term meteorological observations, including daily precipitation, air temperature, and relative humidity, were collected from 86 stations in Guangdong (Figure 2; Hong Kong and Macao excluded due to inaccessible data) via the China Meteorological Data Network (CMDN) for 1951–2018. Land use data were obtained from the China Land Cover Dataset (CLCD) on Google Earth Engine, covering 1985–2019. Population size data were obtained from the 2010 Sixth National Census, while population density data were sourced from the Gridded Population of the World Version 4 (GPWv4).

To investigate the impact of urbanization on WBT, we first classified 86 meteorological stations in Guangdong Province as either urban or rural. This classification was based on a scoring system using three indicators: population size, population density, and land cover (Zhong et al., 2024).

Population size was retrieved from the 2010 Sixth National Census at the county level. Stations in counties with populations <500,000 were scored as rural (score = 0), those with populations >1,000,000 as urban (score = 1), and those with populations between 500,000 and 1,000,000 received a score of 0.5.

After aggregating the three scores for each station, stations with a cumulative score of ≥2 were designated as urban, while those scoring below this threshold were categorized as rural (Zhong et al., 2024). Urban–rural station pairs were selected based on strict criteria for spatial and climatic comparability. The pairing process adhered to the following conditions: (1) an elevation difference of less than 100 m, (2) a spatial distance of less than 80 km, and (3) a latitudinal distance of less than 50 km. To account for elevation-related temperature differences, a vertical lapse rate of 0.0065° C m⁻¹ was applied (Zhang et al., 2023). For each urban station, if only one rural station met the criteria, it was selected as the pair. In cases where multiple rural stations qualified, the geographically nearest station was chosen (Jiang et al., 2020).

The wet-bulb temperature (WBT, ° C) was calculated using elevation-corrected air temperatures and relative humidity (Equation 1, Stull, 2011): (1) WBT = T ⋅ atan [ 0.151977 ⋅ ( RH + 8.33659 ) 1 / 2 ] + atan ( T + RH ) − atan ( RH − 1.676331 ) + 0.00391838 ⋅ ( RH ) 3 / 2 ⋅ atan ( 0.023101 ⋅ RH ) − 4.68035 where T is air temperature (° C) and RH is relative humidity (%). The inverse tangent function (atan) is used as a nonlinear correction term, enabling a more accurate representation of WBT variations under diverse meteorological conditions.

The urban–rural WBT difference ( ΔWBT , °C) was then calculated for each station pair over the period 1967–2018. Although meteorological records at some observation sites extend back to 1951, the 1967–2018 period was selected to ensure data from a more consistent and comprehensive network of stations, thereby maintaining temporal consistency and data reliability. The ΔWBT for each station pair was then computed in Equation 2: (2) ΔWBT = WBT u − WBT r where WBT u refers to the WBT at the urban station, and WBT r refers to the WBT at the rural station.

Figure 2 illustrates the spatial distribution of final urban–rural meteorological station pairs. Initial attempts at pairing revealed challenges, namely: the issue of one-to-many associations (i.e., a single site having multiple potential matches) and the failure to find pairs for all sites, a problem most common in regions with dense station distributions. To address these challenges, a refined two-step pairing strategy was adopted. First, all urban stations for which at least one potential rural counterpart could be identified were retained. Second, for each of these urban stations, the geographically closest rural station was selected as its definitive pair. This process resulted in 38 distinct urban–rural station pairs, providing coverage for the majority (90.9%) of cities in Guangdong Province. The four subregions (the Greater Bay Area, Eastern Coastal, Northern Guangdong, and Leizhou Peninsula) were used in subsequent analysis and highlighted to emphasize their spatial distribution (Figure 2).

Figure 3 presents the temporal variations of annual mean WBT for all stations and for the urban and rural subgroups. During the study period, the overall mean WBT across all stations was 19.39° C. Urban stations exhibited a higher average WBT of 19.48° C, compared to 19.04° C for rural stations. This observed difference (0.44° C) implies an urban impact on moist heat conditions. Overall, the mean WBT increased slightly at a rate of 0.136° C decade⁻¹ (p value = 0.001). The rate of increase was 0.132° C decade⁻¹ (p value = 0.001) for urban stations, compared to a slightly higher rate of 0.151° C decade⁻¹ (p value <0.001) for rural stations. This difference in warming rates resulted in a slight narrowing of the urban–rural WBT difference over the study period.

To assess the spatial distribution of urbanization’s impact on annual mean WBT changes, the UE and UC metrics were calculated for each station pair (Figure 4). On average, urban WBT increased 0.0027° C decade⁻¹ (p < 0.001) faster than rural WBT. This positive UE demonstrates that urbanization has exacerbated the urban–rural WBT difference. Furthermore, the UC analysis revealed that within Guangdong Province, urbanization accounted for an average of 3.13% of the total observed WBT increase in urban areas, highlighting its contribution to the warming of these urban environments.

Spatially, stations exhibiting statistically significant upward WBT trends were widespread, particularly in the GBA and in western and eastern Guangdong. Within the GBA, urban stations exhibiting UE values less than 0 are predominantly found in the Pearl River Delta (PRD), particularly in cities like Guangzhou, Foshan, and Jiangmen. This suggests that in these mature, highly urbanized centers, the net impact of ongoing urban development on WBT trends is less pronounced compared to their rural counterparts, potentially leading to a stabilization or even a narrowing of the urban–rural WBT difference (Liang et al., 2022). Conversely, stations in regions such as eastern Guangdong, characterized by rapid recent expansion, showed higher rates of WBT warming and larger urbanization contributions (Zhong et al., 2024).

To better understand urban–rural WBT differences across Guangdong, we analyzed WBT trends in four primary subregions, namely the northern Guangdong, the GBA, the eastern coast, and the Leizhou Peninsula (delineated in Figure 2; trends are presented in Figure 5). Distinct regional patterns emerged. In the rapidly urbanizing Leizhou Peninsula, urban areas exhibited a faster rise in WBT (0.175° C decade⁻¹, p < 0.001) than rural areas (0.145° C decade⁻¹, p = 0.004), suggesting a pronounced urban heat island (UHI) effect on air temperature is observed, coexisting with a relatively weak urban dry island effect. In contrast, the GBA, characterized by more mature and slowing urbanization, showed slower urban warming (0.123° C decade⁻¹, p = 0.005) compared to rural areas (0.174° C decade⁻¹, p < 0.001), indicating that a weakening urban dry island effect (a relative increase or lesser suppression of urban humidity over time) may exert a more dominant influence on WBT trends than the sensible heat component of the UHI.

Urban–rural WBT trends and their resulting differences are also influenced by local climate factors across Guangdong’s diverse subregions. For instance, along the eastern coast, there are more comparable WBT warming trends between urban sites (0.201° C decade⁻¹, p < 0.001) and their adjacent rural counterparts (0.182° C decade⁻¹, p < 0.001), which may be explained by that sea breeze circulation may moderate local temperatures. In northern Guangdong, a mountainous region with extensive forest cover, the rural WBT warming again outpaces that observed in urban areas, with averaged urban WBT (18.20° C) being higher than the rural WBT (17.80° C).

In the GBA, characterized by advanced and mature urbanization, the average UE was −0.072° C decade⁻¹ (p < 0.001), with the UC being the lowest among all subregions (Figure 6). This negative trend suggests that as urban centers reach advanced stages of development, urbanization ceases to be the primary driver, exacerbating the WBT difference between the city and its rural surroundings. Conversely, subregions such as the northern mountainous region and the eastern coastal region, which are still undergoing active (but not yet saturated) urbanization, generally exhibited positive or near-zero average UE and UC values. This indicates that in these developing areas, urban development continues to contribute to the observed warming of urban WBT, either by causing faster warming than in rural areas (positive UE) or by ensuring urban areas warm at least in line with regional trends (near-zero UE). The Leizhou Peninsula exhibited the highest average UE (+0.044° C decade⁻¹, p < 0.001) and UC (21.90%), reflecting an impact of recent urban expansion that both elevates urban WBT and accounts for a substantial portion of its overall warming trend.

To further investigate the spatial patterns underlying the aforementioned regional variations, we examined the spatial distribution of the urban–rural WBT difference under different extreme-event conditions, including non-extreme events (i.e., normal), extreme precipitation events, high-temperature events, and compound heat-drought events (Figure 7). During periods without extreme events (Figure 7a), the urban–rural WBT difference across most of the province was modest, typically ranging between −1° C and +1° C, indicating generally weak, albeit spatially heterogeneous, urban–rural WBT differences under normal conditions. Notably, the GBA area exhibited small median WBT differences (≤1° C). In contrast, this core was encircled by a distinct ring of significantly higher WBT differences, where values frequently exceeded 2° C, with localized peaks reaching up to 4° C.

Under extreme precipitation events (Figure 7b), the spatial pattern of the urban–rural WBT difference closely resembled that of non-event periods, with most areas maintaining a similar range of values. In contrast, both high-temperature events (Figure 7c) and compound heat-drought events (Figure 7d) generally reduced in the WBT difference across most regions. Specifically, the WBT difference often decreased, and the overall range of values narrowed from approximately 6° C during non-event periods to about 4° C. Despite the general reduction, weak positive urban–rural WBT differences (i.e., urban WBT slightly higher than rural WBT) persisted in many areas. Notable exceptions were the GBA and northern mountainous regions, where a wet-bulb cold-island effect emerged, meaning urban WBT became lower than their rural counterparts. Nevertheless, the spatial pattern remained largely consistent with that of non-event periods. The Leizhou Peninsula exhibited an increase in its urban–rural WBT difference (by approximately 0.5–1.0° C) during both high-temperature and compound heat-drought events. This might be due to localized topographic and climatic factors on the peninsula partially counteracting the broader suppressive impact that high ambient temperatures typically exert on the urban–rural WBT differences.

Our findings reveal that urbanization’s impact on the urban–rural WBT difference is dependent on the stage of urban development, leading to significant spatial heterogeneity. In regions undergoing active urbanization, such as eastern Guangdong, a positive UE was observed, indicating that urban WBT is warming faster than in the surrounding rural areas. This is consistent with the urban heat island effect, where urbanization alters the hydrothermal properties of the surface by increasing temperature (Wu et al., 2019).

In contrast, in mature, highly urbanized centers like the GBA core, we found a negative UE, indicating that urban WBT is warming more slowly than in the countryside. This reflects a transition from a heat-island-dominated system to one where the urban dry island effect becomes more influential. In these advanced stages, the vast expanses of impervious surfaces and loss of vegetation lead to significant reductions in urban relative humidity and suppressed evapotranspiration (Luo and Lau, 2018; Pyrgou et al., 2020). This finding is corroborated by the observed long-term decrease in relative humidity across the GBA (Lin et al., 2020), where the urban dry island effect has been strengthening (Luo et al., 2022). The cooling influence of this reduced humidity on WBT can eventually outweigh the warming influence of the UHI (Du et al., 2019), causing the urban–rural WBT difference to narrow. In contrast, rural areas maintain higher humidity levels through sustained evapotranspiration, thereby experiencing a more pronounced rise in WBT as background temperatures increase (Du et al., 2016; Jiang et al., 2019).

High-temperature and compound heat-drought events generally reduce urban–rural WBT differences in GBA and northern mountainous areas, sometimes even leading to the emergence of wet-bulb cold island. Conversely, the Leizhou Peninsula and eastern coastal areas show enhanced urban–rural WBT differences, while extreme precipitation exhibits opposite spatial patterns. Moreover, the consistent effects of high-temperature and heat-drought events indicate that drought’s regulatory role in humid thermal environments is relatively limited.

Under high-temperature and heat-drought conditions, both urban and rural areas experience elevated air temperatures. However, changes in the urban–rural WBT differences primarily stem from the differing evaporative capacities of urban versus natural surfaces. In rural regions, the abundance of natural surfaces allows for gradual soil moisture release that helps maintain relatively higher air humidity (Luo et al., 2021). This leads to a more pronounced increase in WBT compared to urban areas, where high proportions of impervious surfaces result in lower baseline humidity (Du et al., 2019; Hao et al., 2018; Hao et al., 2023; Ward et al., 2016) and thus smaller WBT rises. This contrast is particularly evident in regions such as the GBA, where high impervious surface coverage makes urban areas significantly drier than their surroundings, and in the northern mountainous areas, where extensive forests maintain near-surface humidity through sustained transpiration during heat events. In both cases, these contrasting land surface conditions lead to a greater increase in rural WBTs.

The effect of precipitation on WBT is complex. For instance, recent studies using global gridded products and General Circulation Models found that light rain can exacerbate humid heat by increasing near-surface humidity with only a minor reduction in solar radiation, especially over arid and semi-arid regions (Zhang et al., 2024). Our study investigates extreme precipitation in a humid region, where the balance between radiative cooling and humidity enhancement varies, a dynamic that appears to be mediated by the land surface. During extreme precipitation events, as both urban and rural areas approach saturation, this land surface dependence becomes critical, particularly the difference in drainage efficiency. This reveals a paradoxical relationship in our humid region where heavy rainfall, despite its cooling effect on air temperature, may exacerbate moist-heat stress. In urban areas, the prevalence of impervious surfaces leads to rapid runoff (Miao and Tang, 1998), yet the retained surface water keeps humidity levels high (Ramamurthy and Bou-Zeid, 2014). This massive spike in local humidity may overwhelm the drop in air temperature, increasing the WBT and perceived thermal stress, to which human thermoregulation is highly sensitive (Raymond et al., 2020). In contrast, rural areas benefit from vegetation interception and swift soil infiltration (Dai et al., 1999; Zhou et al., 2009), which limits the proportion of water available for evaporation (Hao et al., 2018) and results in a relatively smaller humidity increase. Consequently, this divergence amplifies the urban–rural WBT difference, particularly in the GBA and northern regions, where urban WBTs rise more sharply.

Interactions between coastal land-sea breezes and urban systems further influence urban–rural WBT difference, with studies showing that urban-coastal microclimates can be confounded by complex interactions that favor hot-wet compound occurrences (Ganguli and Merz, 2025). For instance, during heat-drought events, sea breeze advection along the Leizhou Peninsula and the eastern coast can partially mitigate drought-induced humidity losses (Du et al., 2004) and reduce the urban–rural humidity gradient. Nevertheless, enhanced urban heat island effects in these coastal zones tend to increase the WBT differences. Under extreme precipitation, as relative humidity approaches saturation in both urban and rural areas, the modulating influence of the sea breeze diminishes. Additionally, turbulent mixing during typhoons or severe convective storms accelerates air exchange between urban and rural areas, reducing both temperature and humidity gradients and thereby narrowing the WBT differences.

The interaction between urbanization and extreme climate events is fundamentally complex and nonlinear (Zhao et al., 2021; Chen H. P. et al., 2024; Chen Y. et al., 2024). The UE and UC provide first-order estimates of the urbanization signal in station-based records (Zhong et al., 2024), while the EUCΔR offers a normalized measure of how extreme events alter the urban–rural WBT difference. It should be noted that these metrics are designed as diagnostic tools to capture the net outcome of such complex interactions, rather than to unravel the underlying nonlinear physics. This inherent limitation underscores that our approach is not a substitute for mechanistic modeling studies (e.g., Zhao et al., 2021), but rather complements them by providing an essential observational benchmark that quantifies the real-world magnitude of these interactions and reveals their spatial heterogeneity.

The spatial representativeness of this study is somewhat constrained by the uneven distribution of urban and rural stations. While urban stations are predominantly concentrated in the GBA, rural stations are mainly situated in northern Guangdong. As a result, some rural stations could not be paired with urban counterparts, and certain urban stations lacked nearby rural stations for comparison. This spatial imbalance may impact the representativeness of the UE and UC indicators, potentially affecting the generalizability of the regional analysis.

Furthermore, our station-based approach cannot resolve the fine-scale spatial heterogeneity of WBT. Future research should leverage high-resolution meteorological grid data and atmospheric models (e.g., Weather Research and Forecasting model with an urban canopy scheme) to overcome the above limitations. Such an approach would allow for a more detailed analysis of the thermodynamic and dynamic processes at the sub-city scale, enabling a clearer attribution of the mechanisms driving WBT changes during and after extreme precipitation events.