Significant Attribution of Urbanization to Triggering Extreme Rainfall in the Urban Core—A Case of Dallas–Fort Worth in North Texas

Abstract

While rainfall occurs for several reasons, climate change and urbanization influence its frequency and geographical disparities. Although recent research suggests that urbanization may lead to increased rainfall, insights into how urbanization can trigger rainfall remain limited. We selected the Dallas–Fort Worth (DFW) metroplex, which has minimal orographic and coastal influences, to analyze the urban impact on rainfall. DFW was divided into 256 equal grids (10 km × 10 km) and grouped into four clusters using K-means clustering based on the urbanization ratio. Using Multi-Sensor Precipitation Estimator data (with a spatial resolution of 4 km), we examined rainfall exceeding the 95th percentile (i.e., extreme rainfall) on low synoptic days to highlight localized effects. The urban heat island (UHI) effect was estimated based on the average temperature difference between the urban core and the other three non-urban clusters. Multiple rainfall events were monitored on an hourly basis. Potential linkages between urbanization, the UHI, extreme rainfall, wind speed, wind direction, convective inhibition, and convective available potential energy were evaluated. An intense UHI within the DFW area triggered a tornado, resulting in maximum rainfall in the urban core area under high wind speeds and a dominant wind direction. Our findings further clarify the role of urbanization in generating extreme rainfall events, which is essential for developing better policies for urban planning in response to intensifying extreme events due to climate change.

1. Introduction

Spatial–temporal rainfall pattern evaluation is imperative to important and useful references for decision-making in the context of climate adaptation [1,2]. Recently, extreme rainfall events have become more frequent and intense globally, and urban development has gradually been recognized as a key aspect influencing this trend. Urban surfaces—dominated by impervious materials (i.e., asphalt and concrete)—modify the surface energy budget by increasing sensible heat flux and reducing evapotranspiration [3]. Such urbanization-led modifications give rise to the urban heat island (UHI) effect, wherein urban core areas with higher built-up density are significantly warmer than their rural neighborhoods. This localized warming in specific urban pockets can amplify atmospheric instability and initiate or, in some cases, could intensify convective storms, especially during the warm season [4]. It is important to note that surface characteristics such as albedo may also trigger urban heat island intensity and, in turn, influence convective activity and rainfall anomalies [5]. While current research suggests that regional-scale UHI effects have been widely studied, the influence of micro-scale UHIs—localized hotspots within cities—on extreme rainfall remains underexplored, largely due to the limitations of spatial resolution in the existing studies.

The interlinkages between urbanization and storm enrichment are intricate and spatially heterogeneous. As cities grow/expand both outward and upward, their internal thermal structure becomes increasingly fragmented, creating a mosaic of micro-UHIs. Such micro-UHIs, embedded within the broader city-scale heat island, differ due to transformations in land use, variations in building materials, an increase/decrease in surface albedo, and differences in vegetation cover [6,7,8]. Recent studies indicate that these thermal pockets can potentially serve as localized convective triggers, which can create uplift zones that favor storm cell initiation or intensification. However, such findings are often based on modeled or satellite data with coarse spatial resolution [7,9]. Hence, the need for high-resolution granular observational data to validate these dynamics has become increasingly imperative in urban areas, particularly in the context of flood-prone cities.

In connection with this, the current study selected several notable extreme rainfall events and used high spatial and temporal resolution radar rainfall data to investigate how variations in micro-UHI intensities within an urban area may influence localized rainfall peaks. Such an evaluation enables us to capture storm behavior at scales relevant to neighborhoods and infrastructure planning, which is usually overlooked in other UHI and rainfall studies. Integrating the rainfall data from radar with urban land use and land cover intelligence provided by the National Land Cover Database (NLCD) could help us evaluate and spatially align observed rainfall patterns with fine-scale urban features. The findings from this study would indicate if storm intensities often peak over zones with higher surface temperatures, which would correspond to the built-up density in the urbanized localities and most impervious urban areas. This would further support the hypothesis that micro-UHIs enhance rainfall at a sub-city scale.

Unlike broader studies that treat cities as uniform thermal bodies and provide partial insights, we here focus on intra-urban variability to understand how subtle variances in surface temperature and land use configuration translate into quantifiable differences in rainfall outcomes. For instance, it could be synthesized if compact high-density zones consistently exhibited more intense rainfall than surrounding low-intensity developed areas during the same storm events. This situation suggests that not only the existence of urban development but also its spatial configuration and thermal fingerprint could potentially play key roles in shaping rainfall storm response [10]. Affirmation of such findings highlights the importance of micro-scale analysis in climate adaptation planning, particularly for urban flooding and drainage system design.

While many previous investigations have used satellite-based or modeled precipitation data, such approaches often average out local extremes or exclude short-lived convective bursts. One of the potential reasons for this is the unavailability of high-resolution (spatial and temporal) data products. By leveraging event-level radar data, we were able to trace storm progression on an hourly basis and document precisely where rainfall amplification occurred. Our study builds on prior work demonstrating that consistent synoptic patterns, such as enhanced moisture and low-level convergence, can trigger urban rainfall anomalies linked to urban heat island effects [11]. In nearly every event that was analyzed in this study, the spatial correspondence between rainfall maxima and micro-UHI zones was specifically analyzed to provide evidence on their interlinkages. Such investigation emphasizes that fine-scale urban heterogeneity is a critical but underrepresented driver of hydrometeorological extremes [12].

Thus, our study contributes towards new insights into the urban rainfall nexus by providing high-resolution, empirical evidence of micro-UHI-driven enhancement of extreme rainfall events. In this regard, the specific objective of this study is to put forth answers to several critical questions. First, how do rainfall storm dynamics vary within an intra-urban built-in environment? Secondly, what is the influence of identified micro-UHIs on storm dynamics on finer temporal scales under different scenarios (after the maximum and minimum UHI effect, and under the maximum and minimum UHI effect)? It further demonstrates whether localized thermal anomalies within urban environments can significantly affect the intensity and distribution of rainfall during storms. Findings from such evaluations could have the potential to support the incorporation of micro-UHI dynamics in urban flood models and infrastructure resilience planning. As urbanization continues and cities densify, the potential for micro-scale thermal influences on weather extremes will likely increase, making this line of research both timely and essential.

2. Materials and Methods

2.1. Study Area

Dallas–Fort Worth (DFW) is located in North Texas and has an average annual rainfall ranging from 500 to 1200 mm. The city’s climate is characterized as humid subtropical. Notably, DFW is free from orographic effects because it has terrain. Additionally, it is distant from the nearest coast, thereby reducing the impact of coastal events on rainfall. Consequently, DFW represents the potential to investigate and understand the influence of urbanization on extreme rainfall variation, without the influence of external physiographic features. Based on area coverage, DFW is the largest landlocked metropolitan area in the USA. The population in DWF is increasing rapidly. The US Census Bureau’s 2020 report stated a population increase of 1.2 million over the last decade, which is projected to rise from the current 7.6 million to 10.5 million by 2040 [13]. For this study, the entire DFW area was divided into 256 equal-sized grids (Figure 1) in order to capture the spatial heterogeneity in urbanization. Additionally, the influence of urban heterogeneity on extreme rainfall at a relatively micro-scale is investigated using these 256 grids. Each grid has a spatial resolution of approximately 10 × 10 km. These grids completely cover the seven counties of the DFW area, including Dallas, Tarrant, Collin, Ellis, Johnson, Rockwall, and Denton. The center of these 256 grids is approximately located in the middle of the cities of Dallas and the Fort Worth. Based on the available radar rainfall data, this study utilized data from 2000 to 2016. Although it would have been ideal to conduct the study using data from recent years, the unavailability of observational radar data hindered us from doing so. Hence, this study can also be viewed as a potential baseline for the upcoming assessments once the new data becomes available.

2.2. Land Use Data

The National Land Cover Database (NLCD), maintained by the Multi-Resolution Land Characteristics Consortium (MRLC) and the U.S. Geological Survey (USGS), was utilized to process urbanization-related information (percentage of the total area in each grid). The NLCD has consistently allocated substantial resources to engage in production-focused research, and an accuracy assessment is provided after the release of each NLCD product (2001 onwards). While the overall accuracies of land use land cover classification have been increasing with each new product, the NLCD product available for 2016 has an accuracy of 91%. Derived from earth observation data (i.e., Landsat), NLCD products have a spatial resolution of 30 × 30 m, and it has been used in various studies on urban issues [14,15]. Therefore, we are of the view that this dataset is appropriate for urban-related studies, such as this work.

For this study, land use data for the years 2001, 2004, 2006, 2008, 2011, 2013, and 2016 were acquired from the NLCD and preprocessed for further evaluations. For example, the land cover in NLCD products is divided into 15 different categories. However, in this study, we are only concerned with the developed land type. In the NLCD categorization scheme, developed land type categories include the following: Developed, Open Space; Developed, Low Intensity; Developed, Medium Intensity; and Developed, High Intensity. Hence, the developed land type categories were grouped to estimate urbanization (%) following the procedure provided in [16].

2.3. Rainfall Data

To analyze rainfall events, this study used the Multi-Sensor Precipitation Estimates (MPE) data. This product was introduced by the National Weather Service to correct radar quantitative precipitation estimation (QPE) and to blend data from various sensors for the sake of comprehensiveness and clarification [17]. The MPE data include radar data from various instruments, including Multi Radar Multi-Sensor, and undergo a comprehensive quality check, incorporating local gauge bias correction [18]. Until the year 2003 (i.e., the early phase of MPE), the data were only based on the Stage-III radar gauge product, and MPE became available online after that. This data product provides hourly rainfall information at a 4 × 4 km spatial scale. For this study, MPE data in the Hydrologic Rainfall Analysis Project (HRAP) coordinate system were converted to the WGS 1984 Geographic Coordinate System (GCS). Different sophisticated Python scripts (using Python version 3.9) were developed to acquire MPE rainfall values at the required temporal scales for the study area. As noted earlier, the MPE data used in this study were from 2000 to 2016, as these were the only available data at the time this research was conducted.

2.4. Sounding Data

Given the fact that wind is also an important influencer in the distribution of rainfall, its role was examined in integration with land cover. The distribution of extreme rainfall was observed for each wind regime. It is noted that the National Weather Service station at Fort Worth is the only station in the DFW area having daily upper-air sounding data (Fort Worth, KFWD, ID: 72249). Hence, sounding data at the Fort Worth station were acquired to find the wind direction for each day at 700 hPa. These data were acquired because wind direction and speed at 700 hPa are most suited for urban studies [4,19] and better represent convective storms, which were analyzed in this study. Moreover, the sounding data are available at 00z (7 p.m. local time) and 12z (7 a.m. local time) at the Fort Worth station. The wind direction at 700 hPa was averaged to represent daily storms (covering day and night). The data were acquired from the University of Wyoming online archives maintained by the Department of Atmospheric Sciences [20].

2.5. Temperature Data for UHI Analysis

The Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) provides the global land surface conditions at 0.5 (approximately 50 km) spatial resolution from 1980 to the present [21]. In this study, temperature (°C) at 2 m from the MERRA-2 product was used to perform UHI analysis. Hourly data at night from six locations/stations in and around the DFW area were downloaded. Three stations are located in the urban core, and three stations are in the rural areas of the DFW. Temperature was averaged for three urban and three rural stations to calculate the UHI, which is the difference in temperature between urban and rural areas. The UHI is usually prominent at night under clear skies and calm wind conditions [22]. Hence, the UHI in our study was estimated on the night preceding the storm only for those days satisfying these conditions (low cloud cover and low wind speeds). Four scenarios were developed to comprehensively evaluate the situation: (1) low UHI and low wind speed (LULW), (2) low UHI and high wind speed (LUHW), (3) high UHI and low wind speed (HULW), and (4) high UHI and high wind speed (HUHW). Here, wind speed means the speed observed from the sounding data during the event of extreme rainfall, which is the day after the associated UHI was estimated. High and low refer to the upper and lower 25th percentile data, respectively. The distribution of extreme rainfall in the study area under these scenarios was examined.

2.6. Extreme Rainfall, Spatial–Temporal Patterns, and Influencing Factors

2.6.1. Identification of Extreme Rainfall Events

In the literature, various approaches are employed to classify extreme rainfall, with no consensus on a particular method. For instance, fixed daily rainfall values have been used to classify extreme rainfall [19,23,24]. Otherwise, daily rainfall exceeding a certain threshold (e.g., 95th percentile) is also considered extreme rainfall [25]. For this study, the rainfall amount exceeding the 95th percentile threshold at the DFW International Airport rain gauge is considered an extreme rainfall day. The DFW airport lies in Grid 152. To capture urban signatures in extreme rainfall, the Spatial Synoptic Classification (SSC) technique was adopted [26,27]. This air mass classification scheme contains seven significant weather type classifications: Dry Polar (DP), Dry Moderate (DM), Dry Tropical (DT), Moist Polar (MP), Moist Moderate (MM), Moist Tropical (MT), and Transitional (T). The SSC weather classification data for the Dallas–Fort Worth station have been acquired from the Kent State University archive [26]. It is noted that SSC weather classifications are available for the DFW station from 1953 to 2023.

Convective rainfall, rather than frontal or typhoon rainfall, is significantly affected by urban forms [28]. MT days, characterized by warm and humid air, lead to convective rain, particularly in the summer season. MT days are synoptically benign, reducing the influence of large-scale weather patterns. A study conducted in Oklahoma City utilized the SSC technique and focused on MT days, exploring the roles of urbanization and wind on rainfall [29]. However, both MT and MM days are included here because MM can have similar characteristics to those of MT on cloudy days in the summer season and can thus give rise to convective events [26]. Furthermore, this investigation focuses on exploring the variability of extreme rainfall, and there are more MM days with rainfall above the 95th percentile threshold. Therefore, MM days meeting extreme rainfall criteria occurring before or after MT days have been included alongside the MT days. Thus, days with MM or MT characteristics and exceeding the 95th percentile threshold have been selected (Figure 2). This study encompasses the period from May to September, which is considered the summer season in the DFW area. The number of MT and MM (low synoptic days) for May, June, July, August, and September is 28, 31, 16, 18, and 21, respectively. MPE data were used to explore the variability of extreme rainfall during these 114 days.

2.6.2. Spatial Clustering of Land Use to Evaluate Extreme Rainfall

To highlight patterns of extreme rainfall among grids with similar urbanization characteristics, K-means clustering is applied to group the 256 grids into four different clusters based on urbanization (%) in 2001. Similar results were obtained if the clustering criteria used land use data from 2016 instead of 2001. To classify the cells into different clusters, we employed the K-means method and used the urbanization ratio (i.e., the percentage of urbanization for each cell) as a parameter. The elbow method of selected optimal number of K-means clusters indicated that the grids in the study area could be divided into four clusters based on the intensity of urbanization in the area: Urban Core (UC, ~12% grids), Urban Periphery (UP, ~15% grids), DFW North (DFW-N, ~30% grids), and DFW South (DFW-S, ~43% grids), as shown in Figure 3. Most of the grids in the UC cluster have a high percentage of urban development, hence named Urban Core. Grids in the UP have a high rate of urbanization, whereas DFW-N and DFW-S have similar levels of urban land cover and were named owing to their geographic location.

2.7. Storm Tracking

Storm tracking was conducted on an hourly basis for selected days out of the total 114 events (the number of extreme rainfall days). These days include low-wind-speed and high-wind-speed days for the prevailing window (i.e., Southwesterly (SW) winds). Low-wind-speed days were when the wind speed was less than the 25th percentile, while high-wind-speed days covered days when the wind speed was more than the 75th percentile.

2.8. Influence of UHI on Extreme Rainfall

In addition to extreme rainfall variation, the UHI distribution was evaluated. This analysis was conducted for the selected days only. The UHI was estimated for each hour of the selected day in the study area. The magnitude and time of the maximum UHI of the day were calculated for days when the peak rainfall in the study area occurred immediately after the maximum UHI. In such cases, the maximum UHI’s effect on extreme rainfall in each cluster was further analyzed. In addition, days when peak rainfall occurred before the maximum UHI were also analyzed.

2.9. Role of Other Environmental Thermodynamic Parameters

In this study, boundary layer height (BLH) and convective inhibition (CIN) were analyzed in conjunction with urban heat island (UHI) intensity to assess their combined influence on convective storm development. BLH refers to the depth of the atmospheric layer near the surface, where turbulent mixing occurs, affecting moisture and heat exchange that are critical for storm formation. CIN represents the amount of energy suppressing convection; higher CIN values indicate a stronger cap preventing storm initiation, while lower values favor convective development.

3. Results

3.1. Storm Tracking at an Hourly Scale

How the Maximum UHI Varied for Both Wind Speed Scenarios

Under the dominant window (SW), there were 16 and 17 low- and high-speed-wind days, respectively (see

Table 1. The table shows the number of days on which peak rainfall occurred in a cluster after the maximum UHI (post-max UHI peak) and before the maximum UHI (pre-max UHI peak) for both wind conditions.

Condition	Low Wind				High Wind
	UC	UP	DFW-N	DFW-S	UC	UP	DFW-N	DFW-S
Post UHI peak	9	0	1	0	4	0	1	0
Pre UHI peak	5	0	0	1	11	0	1	0

). For 10 out of 16 days under low wind speed, peak rainfall in the study area occurred after the maximum UHI. In nine of these ten days, peak rainfall occurred in the UC cluster (Figure 4), and only one day of peak rainfall happened in the DFW-N cluster. There were six days under low wind speed when the peak rainfall occurred in the study area before the maximum UHI value. Even on these days, rainfall fell in the UC cluster that exceeded the peak level by five. This shows that for both pre- and post-maximum UHI scenarios in low-wind-speed days, the majority of the time, peak rainfall was noted in the UC cluster.

Under high wind speed, peak rainfall in the study area occurred before the maximum UHI in 11 out of 17 days (Table 1). In only five days, peak rainfall occurred after the maximum UHI, and on four out of these five days, peak rainfall happened in the UC cluster (Figure 5). For pre- and post-maximum UHI scenarios, peak rainfall was primarily observed in the UC cluster, consistent with the low-wind-speed days.

For both wind speeds and post-max UHI scenarios, the maximum UHI occurred in the afternoon. The peak rainfall in the UC cluster was primarily noted after 1 to 3 h of the maximum UHI.

3.2. Peak Rainfall Difference Between the UC and Other Clusters

On days with low wind speeds, the peak rainfall difference between the UC and all the clusters was higher in the pre-max UHI conditions than in the post-max UHI (Figure 6). This implies that the rainfall occurring after the maximum UHI was widely distributed, resulting in higher rainfall in other clusters. The maximum difference was observed in the DFW-N cluster, where peak rainfall occurred in pre-max UHI conditions.

Low wind speeds could not advect rainfall downwind of the UC cluster. However, the maximum difference in peak rainfall was observed in the DFW-S cluster (upwind) for post-max UHI days, highlighting the role of the UHI in influencing rainfall distribution, resulting in less rainfall upwind during low wind speeds.

Like low-wind-speed days, the average peak rainfall difference for high-wind-speed days was higher in the pre-max conditions than the post-max UHI (Figure 6). However, the average peak rainfall difference on high-wind-speed days was lower than on low-wind-speed days under pre-max conditions for UP and DFW-N clusters. Compared to the pre-max UHI situation, the average peak rainfall difference was reduced in the post-max UHI, similar to low-wind-speed days. High wind speed in combination with the UHI (for the post-max UHI condition) influenced rainfall and caused less intense rain in the upwind areas (DFW-S) compared to the UP and DFW-N clusters.

Two events have been analyzed for the maximum UHI-initiating rainfall in the study area under low and high wind speeds, respectively.

3.3. Low Wind Speed and Maximum UHI (Example of Rainfall Occurring After Maximum UHI)

On 31 May 2016, a maximum UHI of 3.4 °C occurred around 14:00, which, after a span of 4 h, generated peak rainfall in the UC cluster (Figure 7 and Figure 8). The peak rainfall in the UC was almost double that of the other three clusters. This rainfall event started at 15:00, immediately one hour after the maximum UHI, and lasted for almost 5 h.

Considering the peak rainfall occurring in the UC cluster, the convective inhibition (CIN) and boundary layer height (BLH) were tracked in it (Figure 9). Immediately after the maximum UHI, the CIN began to decrease in the UC cluster, while BLH increased, and its peak coincided with the extreme rainfall peak. While examining the UHI-induced rainfall variation, it was noted that the BLH was consistent with the rainfall in the city area. However, in our study, we observed that this happened under low wind speeds.

3.4. High Wind Speed and the Maximum UHI (Example of Rainfall Occurring After Maximum UHI)

A maximum UHI of 6.11 °C occurred around 14:00, which, after a span of 3 h, caused peak rainfall in the UC cluster on 2 May 2009 (Figure 10). The rainfall diminished after reaching a peak value, and then another peak arrived at 20:00, i.e., 6 h after the maximum UHI. During both peaks, the rainfall was clustered around the UC and UP clusters (shown in Figure 11a and Figure 11b for 15:00 and 20:00, respectively). Later on, the rainfall shifted downwind in the DFW-N cluster rainfall. The strongest UHI observed in this study was 6.11 °C. The intense UHI, combined with high wind speeds, fostered rainfall by supplying moisture, resulting in a more extended rainfall period compared to the low-wind-speed event, which diminished early.

Moreover, the CIN and BLH in the UC cluster for this rainfall event showed a low CIN and the highest BLH at the time of the second peak (20:00) in the UC cluster (Figure 12). This implies the possible role of BLH, which may have fostered instability, leading to another peak in rainfall in the UC under high wind speeds following an intense UHI in the study area.

Lastly, the boundary layer height distribution is just an indication of the atmospheric instability that was prominent in the Urban Core (UC) cluster. Among several other reasons, this instability could have been caused by the highest UHI observed in the study area.

4. Discussion

The interplay between the urbanization process, the UHI, and intensifying extreme events is a complex phenomenon. We here provide empirical evidence on the interconnections between these different processes in an intra-urban context—which is less explored in the existing literature. Through this, the current study provided valuable insights into unlocking the nexus of urbanization, the UHI, wind, and extreme rainfall. Our findings provide significant insights into the potential role that urbanization plays in triggering extreme rainfall events, particularly in the urban core cluster of the study area. Our results ascertain the interaction between UHI and wind speed in influencing the timing and spatial distribution of peak rainfall in an extreme event.

The results presented in this study align with existing research that demonstrates the influence of the UHI on localized weather patterns, particularly in dense urban cores. For example, Han et al., (2014) [30] highlighted that the UHI-induced convective rainfall is more likely to be experienced in urban areas that are densely built up. This happens due to localized heating. Our study further adds to this by demonstrating that wind speeds might influence the timing of peak rainfall relative to the maximum UHI; peak rainfall tends to occur post-maximum UHI, while under high wind speeds, it often occurs pre-maximum UHI. Such a nuanced understanding of the temporal dynamics of UHI-induced rainfall is a significant addition to the existing body of literature. Similarly, Bornstein and Lin (2000) [31] demonstrated that urban regions typically experience increased rainfall due to enhanced surface roughness and thermal gradients. We also observe that peak rainfall in DFW predominantly occurs in the UC cluster (i.e., highest density of built-up area), both pre- and post-maximum UHI, which is also in line with Bornstein and Lin (2000) [31].

Furthermore, the role of wind speed in modulating UHI formation and distribution has also been highlighted in several existing studies. For instance, Oke (1982) [3] and Arnfield (2003) [32] highlighted that wind speed plays a crucial role in determining the intensity and spatial extent of the UHI. Our results substantiate this, showing that high wind speeds can advect rainfall downwind, reducing its intensity in the UC cluster. In contrast, low wind speeds allow rainfall to concentrate in urban cores. Our utilization of convective inhibition and boundary layer height (BLH) delivers a mechanistic understanding of UHI–rainfall coupling. During both low- and high-wind-speed scenarios, a decrease in CIN and a peak in BLH followed the maximum UHI. This situation indicates that surface heating could play a noteworthy role in weakening the stability and development of convection. These results are in line with the findings of Zhang et al. (2017) [33], who demonstrated that UHI-enhanced BLH can modulate rainfall intensity and timing in Beijing. Similarly, the importance of CIN and BLH evolution was also stressed by Diffenbaugh et al. (2017) [14], who related thermodynamic variations under warming conditions to extreme weather events. This further demonstrates the potential for intensified urban rainfall events in a warming climate, particularly in rapidly expanding urban regions.

From the modeling perspective, precise rainfall estimations in urban settings necessitate the integration of high-resolution data. In this context, the incorporation of satellite-based approximations into multi-sensor precipitation frameworks has demonstrated promise in capturing urban rainfall heterogeneity [18]. These approximations, when combined with retrospective gauge corrections [17], improve hydrologic simulations that are otherwise limited by spatial rainfall misrepresentation. Furthermore, dynamic rainfall runoff models (i.e., those developed by Lin & Chen (2004) [23]) have represented the importance of non-linear feedbacks between rainfall and surface conditions. These models could be further improved by explicitly coupling UHI indices and synoptic meteorological features to predict urban flood risk more effectively.

Implications for Policy and Planning

The results from this study have several policy- and planning-related implications, particularly in the context of urban planning and disaster management. From a scientific point of view, our work underscores the need to account for urbanization and the induced UHI in integration with wind speed as the critical factors in understanding the dynamic rainfall patterns in cities. This is specifically important as well as relevant for climate simulations and weather-related modeling, where precise representations of urban environment and different effects are critical for advanced and reliable forecasts [34]. For policymakers and other stakeholders (i.e., urban planners and practitioners), the findings from this study highlight the need to consider the UHI effect in the designing and management of urban infrastructure. Future urban development plans must account for the updating of drainage systems and flood mitigation strategies through informed UHI-influenced rainfall occurrence in cities to avoid any maladaptation practices. Similarly, urban regions, especially with higher UHI intensity, may require enhanced stormwater management systems to handle the increased rainfall intensity and volume—as demonstrated in this study [7,8].

Additionally, the results of this study suggest that the spatial distribution of rainfall has implications for land use planning. For example, the concentration of peak rainfall in the UC cluster (high-density core urban areas) advocates that urban development should be carefully managed to minimize the risk of flooding. This could encompass tactics such as growing green spaces, which can help mitigate the UHI effect, as well as water runoff [35], and implementing zoning regulations that restrict high-density development in areas prone to extreme rainfall [36]. Similarly, there is a need for cohesive urban planning that takes into account the potential impact of the UHI on local weather patterns. One potential way out could be the development of urban heat island mitigation strategies, such as the use of reflective materials in building construction and the promotion of urban forests/vegetation [37]. Such measures can provide multiple benefits (i.e., help reduce the intensity of the UHI and, consequently, the risk of extreme rainfall events).

Conclusively, given the role of the UHI in contributing to extreme rainfall in the UC cluster, future urban resilience planning must integrate UHI intensity as a variable in flood models, include green infrastructure to mitigate urban heating, and adopt zoning practices that account for thermally vulnerable regions. The need for high-resolution data, multi-sensor integration, and continued urban climate monitoring is critical for mitigating the intensifying effects of UHI-induced rainfall and associated risks.

5. Conclusions

There is a pressing need for ongoing research into the effects of urbanization on local weather patterns. As cities continue to grow, understanding the multifaceted interlinkages between the UHI, wind speed, and rainfall will be indispensable for devising effective and efficient strategies to reduce and adapt to the risks associated with extreme weather events. In this context, we deliver valuable insights into the role of urbanization in influencing extreme rainfall events in cities, given the heterogeneous distribution of development levels within urban areas. The results from this study underscore the importance of considering wind speed in understanding UHI-induced rainfall patterns and have important implications for urban planning, disaster management, and policy development. Through integrating these insights into urban planning and policy, cities can better manage the risks associated with extreme rainfall and build more resilient urban environments.

Using radar observational data, land use and land cover information, sounding data, and temperature, we conducted a multifaceted investigation into how rainstorm dynamics change within the intra-urban environment, given the heterogeneous development levels. Furthermore, we investigated how the UHI effect might influence the onset and distribution of peak rainfall in different developed areas of the city. The hourly tracking of extreme rainfall reveals that the peak hourly rain occurred in the urban core cluster and mainly after the maximum UHI in the study area. Under low wind speeds, the rainfall event did not last long; however, under high wind speeds, the rainfall event caused two high peaks (hourly) in the UC cluster for the intense UHI recorded (i.e., 6.5 °C). Our results suggest that, for both pre- and post-maximum UHI scenarios on low-wind-speed days, the majority of the time, peak rainfall was observed in the urban core cluster (i.e., where built-up density is the highest). It is further ascertained that for both wind speeds and post-max UHI scenarios, the maximum UHI occurred in the afternoon. The peak rainfall in the UC cluster was primarily noted after 1 to 3 h of the maximum UHI. Hence, the segments of cities that are prone to the UHI could also face extreme rainfall events in a warming climate.

Our findings reinforce the notion that urban centers, particularly those with pronounced UHI effects, act as “heat islands” that locally modify mesoscale weather progressions. The situation as such becomes increasingly relevant under climate change, where enhanced UHI effects may exacerbate already intensifying precipitation patterns due to global warming.