Along with climate change and rapid urbanization, recent urban rainstorm flood events have been presented as compound disasters characterized by rainfall-flood-urban inundation coupling, featuring rapid risk transmission, difficult forecasting, severe disaster situations, and a lack of defense standards. This study identified key scientific issues that need to be addressed in the study of urban rainstorm-flood compound disasters, including coupled integrated forecasting and rapid early warning methods based on the formation mechanism of urban rainstorm-flood compound disasters, the combination probability of urban rainstorm floods encountering each other, the return period and defense standard of compound disasters, the principle of risk transmission superposition, and the rapid diagnosis and suppression mechanism of urban rainstorm-flood compound disasters. Research content and technical routes oriented towards scientific problems were proposed. It is believed that solving the challenges of early warning for urban rainstorm-flood compound disasters, rapid risk diagnosis, and rapid suppression responses are crucial for the prevention and control of urban compound water disasters.

With the intensification of global climate change and rapid urbanization, extreme rainfall events are occurring with increasing frequency, substantially altering the hydrological characteristics of urban surfaces and weakening their natural regulation and storage capacity. Consequently, the traditional urban flood management paradigm dominated by rapid drainage has become insufficient to cope with the growing complexity of urban flood risks. This study systematically investigates the mechanisms by which urban spatial patterns and infrastructure in urban flooding and points out that impervious surfaces, building layouts, topography, street networks, urban sprawl patterns, green space distribution, and water body distribution directly affect the risk of urban flooding by altering surface runoff paths, drainage efficiency, and rainwater storage capacity. By analyzing the synergistic mechanisms and quantitative research outcomes of gray infrastructure (e.g., drainage systems) and green-blue infrastructure (e.g., rain gardens and river-lake systems), this study reveals that the limitations of gray infrastructure in handling extreme rainfall events have driven the development of blue-green infrastructure. This forming a“grey-green-blue”collaborative system that enhances urban flood prevention, drainage, and stormwater storage capacity. Furthermore, this study systematically summarizes the research progress on urban flood simulation methods and monitoring technologies that account for urban spatial morphology and infrastructure impacts. The results indicate that physical mechanism models (e.g., SWMM, MIKE URBAN, and TELEMAC-2D), statistical empirical models, and machine learning models (e.g., CNN-LSTM coupled models) each possess distinct advantages in urban flood simulation. The integration of diverse monitoring technologies, including conventional sensors, remote sensing, GIS, social media data, the Internet of Things, and digital twin systems, is accelerating the transition of urban flood management toward intelligent and data-driven frameworks. In light of the current research gaps, this study suggests that future research should adopt multi-scale, multi-factor, and multi-objective perspectives to further advance the coordinated optimization of urban spatial morphology and infrastructure systems. Particular attention should be given to exploring the dynamic adjustment and effective integration of flood control and drainage design standards under changing environmental conditions, promoting the coordinated operation and benefit quantification of “gray-green-blue” infrastructure systems, and leveraging multi-source big data and digital twin technologies to enable intelligent urban flood emergency response, thereby facilitating a shift toward more refined and resilient urban flood management. This paper provides a systematic summary of flood simulation methods, monitoring techniques, and flood risk mitigation measures that consider the influence of urban spatial patterns and infrastructure, discusses the advantages and shortcomings of various aspects of flood risk prevention and control, and highlights that future research should focus on elucidating the coupling mechanisms between spatial patterns and infrastructure and promoting the intelligence, refinement, and integration of urban flood management.

The relationship between prehistoric human activity and the natural environment has long been a central focus of archaeology. Uncovering the dynamic adaptation mechanisms of prehistoric agricultural societies under different topographical and hydrological conditions is of great scientific and practical importance. It deepens our understanding of past human-land interactions and provides insights for addressing rapid future climatic and environmental changes. As a key agricultural center in southwestern China, the Chengdu Plain, with its rich archaeological record spanning the Neolithic Guiyuanqiao Culture to the Qin-Han period, serves as an excellent case study for this research. This study investigated the spatiotemporal evolution of agricultural settlement patterns and their adaptation to topographical and hydrological conditions on the Chengdu Plain from the Guiyuanqiao Culture to the Qin-Han period. By integrating data from 223 archaeological sites with high-resolution digital elevation models (DEMs) and river vector data, we employed GIS-based spatial analysis and statistical methods. The key topographical (slope and elevation) and hydrological variables (distance to the nearest river and difference in elevation from the river) were extracted for each site. Statistical analyses, including probability cumulative distribution and period-wise comparative assessments, were conducted to analyze human preferences for settlement locations. Furthermore, to infer demographic trends, a regional radiocarbon (¹⁴C) probability density curve was used as a proxy for relative population changes and correlated with settlement patterns and agricultural information revealed by archaeobotanical evidence. The study revealed a phased settlement distribution pattern: initial occupation of peripheral uplands during the Guiyuanqiao period, southward expansion into the low-lying core plain during the Baodun period, northward contraction to the Yazi River Basin during the Sanxingdui period, renewed southward expansion during the Shierqiao period, and a more dispersed layout during the Qin-Han period. These shifts are closely associated with population pressure, flood risks, and migration events. Statistical results indicated that > 80% of the sites were located within 1,000 m of a river, and 74.7% had slopes < 3°, reflecting an overall preference for flat, river-proximate environments conducive to rice agriculture. However, significant variations were observed across these periods. Sites from the Guiyuanqiao Culture, linked to dryland farming populations migrating from the arid northwest, were predominantly located on higher-elevation slopes (average slope: 3.3°; average elevation: 516.3 m) and at greater distances from rivers (average distance: 1,221.7 m), indicating a strategy to avoid overly humid conditions. The Baodun period witnessed a surge in site numbers and expansion onto flatter terrain (average slope: 2.5°) alongside the dominance of rice cultivation to support population growth. However, this aggregation in low-lying areas increased vulnerability to floods from the Min River tributaries, likely prompting a northward relocation to the better-drained Yazi River Basin during the subsequent Sanxingdui period. Correspondingly, sites during this period showed a slight increase in the average slope and an increase in the proportion of dry crops (22.4%). The Shierqiao period saw renewed southward expansion and increased reliance on rice (82.7%) and renewed flood threats. During the Qin-Han period, the influx of northern migrants revitalized dry farming traditions, leading to settlement expansion on steeper slopes (average slope: 3.4°) and an increase in dry crop cultivation (22.3%). Rice agriculture persisted in the lowland plains and was supported by mature pond-field irrigation systems.

This study investigated the history and technical experience of adaptive water management in Suzhou under flood conditions using historical documents and hydrogeological and archaeological data. The Suzhou area developed agricultural water conservancy by constructing an integrated water network of large-scale polders, known as diked-canal polders or tangpu weitian. Higher dikes were constructed while developing low-lying polder areas, and the clear water from Taihu Lake was diverted to the coastal areas with higher elevations for irrigation. The tangpu weitian system supported the water conservancy foundation for a stable and high-quality irrigation and drainage environment between the low- and high-lying areas of Suzhou. Floods in Suzhou intensified after the destruction of the tangpu weitian system during the early Northern Song Dynasty. The Northern Song government emphasized water management in the low-lying plains centered on Suzhou via the implementation of numerous drainage projects, excavation of river channels, construction of sluice gates, and reclamation projects in a waterlogged environment. Land reclamation in the waterlogged lake and marshland areas centered on Suzhou intensified during the Southern Song Dynasty, when the government was responsible for dredging and digging waterways; therefore, the canal system centered on Suzhou was constructed. The water flow of Taihu Lake and surrounding lakes is regulated by the Suzhou Canal system and discharged into the Wusong River. An analysis of the water conservancy management model of Suzhou Ancient City from the late Tang Dynasty to the Song Dynasty revealed that the water network of Suzhou Ancient City played a central role in flood regulation of the Taihu Lake plain area. The high groundwater level, effective lake regulation system, and vast aquatic environment influenced by floods were the environmental foundation for Suzhou Ancient City, which was a water conservancy hub of the eastern Taihu Lake plain during the Northern Song Dynasty. The main models for flood and water resource management in Suzhou Ancient City during the Song Dynasty were to draw better water resources from Taihu Lake into the city and create freshwater flow. The river network of Suzhou Ancient City integrated functions such as water diversion, drainage, and transportation, fully utilizing rainwater and flood resources. These experiences and models provide good reference values for future water conservancy.

Coastal cities are increasingly exposed to compound pluvial-coastal flooding under climate change and sea-level rise. When intense rainfall coincides with elevated downstream water levels, tide-induced backwaters suppress gravity drainage and can rapidly escalate sewer surcharges and surface inundation. This study quantifies the amplification effect of tidal backwater on urban waterlogging in a tidal river-network system and evaluates the mitigation potential of graded sluice operation under extreme conditions. The Jinfeng-Cuiping area in Zhuhai, southern China, was selected as a representative low-gradient coastal catchment. A designed tide frequency curve was derived using a maximum entropy framework and fitted with multiple candidate probability distributions. The goodness-of-fit was evaluated using the Kolmogorov-Smirnov (K-S) test, and the selected distribution was used to obtain the design high-tide levels for typical return periods. Because long-term observations at the target downstream boundary (Shijiaozui sluice) are limited, design tide levels were transferred from the long-record Denglongshan gauge to Shijiaozui through an empirically established relationship based on overlapping water-level observations, thereby enabling long-series-based boundary design while maintaining local representativeness. A loosely coupled 1D-2D urban flood model was then built by linking the U.S. EPA Storm Water Management Model (SWMM) for drainage hydraulics with LISFLOOD-FP for surface inundation. The SWMM-simulated node overflow hydrographs were converted into boundary-condition files and imposed on the corresponding grid cells in the 2D model to reproduce the spatiotemporal evolution of surface flooding. The coupled workflow was calibrated and validated against multiple historical waterlogging events using both inundation extent and water-level processes, ensuring that the model can be used for mechanism identification and scenario comparison. Scenario simulations were conducted for the heavy rainfall event of 4 May 2024 under three classes of downstream boundary conditions: (i) free outflow (no tidal constraint), (ii) rainfall encountering 1-, 10-, 50-, 100-, and 200-year design tides with direct imposition of tide levels at the downstream boundary, and (iii) the same tide scenarios with graded sluice operation driven by the head difference between the inside and outside water levels. Key response metrics included overflow-node count and proportion, overflow duration and depth indicators, sewer surcharge classification, and inundation depth/area statistics. Results indicate that tidal backwater substantially increases system overload and surface waterlogging. Under the 4 May 2024 rainfall, the overflow-node proportion reached 43.46% in the free-outflow case and increased to 46.35% under the 200-year tide, with the proportion of highly surcharged pipes increasing to 52.41%, indicating severe drainage stress. The tide backwater also amplified surface flooding, with the maximum inundation depth increasing to 2.85 m (approximately 40% higher than free outflow) and the total inundation area expanding to 3.62 km² (129% larger than free outflow). Graded sluice operation provided measurable peak-reduction benefits in the 200-year tide scenario: compared with directly imposing tide levels as the downstream boundary, the maximum inundation depth decreased by 20.7%-26.2%, the total inundation area was reduced from 3.62 km² to 3.0 km² (a 17.1% reduction), and the high-risk zone area decreased by 13.5%, although improvements in overflow duration and system-wide surcharge conditions remained limited. Overall, downstream tide levels are confirmed as a dominant external amplifier of pluvial waterlogging in coastal river-network cities, and tide-aware sluice operation provides a practical avenue for peak reduction and disaster mitigation under extreme compound conditions. The proposed design-tide-to-model workflow supports tide-aware drainage assessments, sluice operation designs, and compound flood risk management in similar coastal urban settings.

Global climate change and rapid urbanization have intensified the occurrence of extreme rainstorm events, and urban waterlogging has emerged as a critical disaster risk constraining the sustainable development of high-density cities, posing serious threats to life, property, and urban resilience. To characterize urban waterlogging risk in Shenzhen under different design rainstorm scenarios, this study constructed three representative scenarios with distinct return periods and rainfall characteristics—Zhengzhou “7·20”, Shenzhen “9·07” and Shenzhen “8·05” —using the two-dimensional hydrodynamic model LISFLOOD-FP and conducted systematic waterlogging simulations and exposure risk assessments of urban systems. Model validation using 183 historical waterlogging points demonstrated high reliability: within 50-m buffers of these points, maximum simulated water depths generally exceeded 0.3 m (with extreme values exceeding 10 m), effectively reproducing the spatial distribution and severity of severe waterlogging. The results indicate that waterlogging in Shenzhen exhibits a distinct spatial pattern characterized as “deeper in the west, shallower in the east; deeper in urban cores, shallower in suburbs,” driven by the “higher southeast, lower northwest” topography, high impervious surface coverage in western districts, and uneven drainage system loading. Among the three scenarios, the “7·20” design scenario poses the highest risk due to its high rainfall intensity, pronounced peak discharge, and extended duration, with areas experiencing water depths > 30 cm accounting for 26.55% of the study area. Specifically, more than 74,000 buildings and approximately 4.68 million people were exposed to water depth exceeding 1 m, and 46.76% of the total road network (8,966.25 km) was inundated. The “9·07” scenario is characterized by nocturnally concentrated short-duration heavy rainfall, resulting in localized water accumulation in low-lying areas such as Longgang. The “8·05” design scenario exhibits a multi-peak pattern with a pronounced surge and a mid-event rainfall lull that temporarily alleviates accumulation, producing an intermediate risk level relative to the other two scenarios. Critical infrastructure elements exhibit high sensitivity to water depth, with significant differences in risk response. Under the “7·20” design scenario, 1,028 medical institutions, 823 elderly and childcare facilities and 106 emergency shelters, were exposed to high risk, potentially compromising emergency medical services and vulnerable populations; more than eight hazardous chemical enterprises faced potential secondary disasters at water depths exceeding 0.5 m. Spatially, risk to critical infrastructure exhibits a pattern of “western concentration and eastern dispersion.” High-risk clusters are concentrated in Luohu and Longhua (medical and elderly-care facilities), Bao'an (moderate risk), and Nanshan, Luohu, and Guangming (hazardous chemical enterprises). Eastern districts exhibit generally low risk, with localized high-risk pockets confined to elderly-care facilities in Dapeng and Yantian. Futian District demonstrates the strongest protective performance, likely attributable to higher construction standards and more scientifically informed site selection. This study advances the literature by constructing cross-regional and locally representative design rainstorm scenarios and elucidating the coupling mechanism between rainfall characteristics and waterlogging risk in high-density urban environments. The findings provide a scientific basis for hierarchical disaster prevention planning and offer a transferable framework for waterlogging risk management in similar high-density cities nationwide.

Accurate assessment of typhoon-induced disaster losses is often hindered by limited historical data and severe class imbalances, especially in regions with infrequent but high-impact events. These challenges reduce the robustness and generalizability of predictive models, leading to unreliable assessments of potential disaster severity. To address these issues, this study proposes an integrated evaluation method that combines data augmentation using a CTGAN with multiple machine learning algorithms. The objective was to enhance sample diversity, alleviate class imbalance, and improve the accuracy and stability of disaster loss predictions. Wenzhou City, located in Zhejiang Province, China, was selected as the study area because of its frequent exposure to typhoon-related hazards. Twenty typhoon cases from 1994 to 2020 were collected, and a structured dataset was constructed using 13 key indicators. These indicators cover three dimensions: (1) hazard-inducing factors such as maximum wind speed and accumulated rainfall; (2) environmental background conditions, including elevation, river network density, and landform; and (3) socioeconomic exposure and vulnerability, reflected by variables such as population density, GDP per capita, and infrastructure indicators, such as road length and hospital bed count. To represent the level of disaster impact for each event quantitatively, a disaster loss index was calculated and used as the input for k-means clustering. This unsupervised learning approach classified 20 typhoon events into four distinct loss severity levels, forming the basis for subsequent supervised classification tasks. To overcome the limitations of class imbalance, the CTGAN model was employed to generate synthetic samples under specific class-conditional constraints. The generated samples were incorporated into the training set to enrich underrepresented classes and improve the representativeness of the dataset. Five widely used machine-learning models were trained and evaluated: GBDT, XGBoost, LightGBM, CatBoost, and Random Forest. The experimental results demonstrated that the GBDT model outperformed the others in terms of both classification accuracy and generalization performance. This model showed the most consistent results across multiple metrics, including mAP, precision, recall, F1-score, and accuracy. Additionally, a comparative analysis was conducted to explore the influence of synthetic data volume on model performance. The findings revealed that simply increasing the number of synthetic samples does not guarantee continuous improvement; rather, an optimal range of sample sizes exists beyond which model stability may plateau or even decline. This study provides a practical and scalable methodological framework for typhoon disaster loss assessments in data-constrained environments. By leveraging generative modeling and ensemble learning techniques, this study offers insights into the effective application of data-driven methods to support disaster preparedness, emergency response planning, and resilience analysis in other hazard-prone regions.

To investigate the impact of storm rainfall patterns on urbanized village flooding, seven storm rainfall patterns were designed for a village in Futian District, Shenzhen City, based on the formula for storm rainfall intensity in Shenzhen City The Infoworks ICM model was used to simulate the inundation process under different rainfall patterns with different recurrence periods (10, 50, and 100 years) and to analyze the characteristics of inundation in terms of inundated area and inundated water depth under different storm rainfall patterns. The results show that the single-peak rain type is more serious than the double-peak and uniform rain types in terms of inundation. In particular, the single-peak backward rain type has the largest inundation area, and the inundation area of the rain peak in the late rainfall was larger than that of the rain peak at the front and the center of the rainfall. The uniform type has the smallest degree of inundation among all rain types. Inundation depths vary among different rainfall types. The inundation area and proportion associated with the single-peak center rainfall are obviously higher than those of the other rain types, indicating more severe flooding and greater potential damage. The single-peak type is particularly hazardous in terms of inundation depth under different rain types. The single-peak type, especially the single-peak backward rain type, is easy to cause waterlogging; in the 10-year return period, the 0.15-0.28 m inundation depth is dominant, and with the increase in the return period, inundation depths greater than 0.40 m account for a gradually increasing proportion. The early warning and forecasting program should focus on the consideration of the impact of the single-peak backward rain type on urban waterlogging in Shenzhen City. In this paper, seven different rainstorms were selected for flooding analysis, but rainstorms in the context of climate change have great uncertainty and spatial heterogeneity, and the flooding characteristics of different spatially distributed rainstorms under unfavorable future conditions still need to be further investigated. The results of this study provide a theoretical basis for flood prevention, mitigation, and the sustainable development of urban areas.

Rapid urbanization has changed the watershed subsurface, and the land-use pattern of urban villages is particularly significant, changing their hydrological processes and thus affecting the risk of waterlogging. Coupled with the recent increase in extreme precipitation, it is necessary to explore the role of the land-use layout of rapidly urbanizing villages in the flooding process under extreme precipitation. In this study, taking a village in Shenzhen with drastic changes in land-use patterns as an example, we simulated the flooding situation of different land-use layouts in four phases with different rainstorm recurrence periods (10-, 50-, and 100- years) based on the Infoworks ICM model and analyzed the characteristics of flooded areas and flooded water depths under the land-use layouts in four phases in a comparative way. The results show that, compared with the current site layout, the inundation area of the planned layout after removing some buildings is obviously reduced, and the reduction rates are 91.89%, 73.34%, and 46.32% for the 10-year, 50-year, and 100-year return periods, respectively, indicating that the planned site layout is able to reduce inundation in heavy rainstorms effectively. However, the reduction rate gradually decreases with the increase in the return period, indicating that inundation under low rainstorm return periods can be reduced to a lower level than the current one. However, as the recurrence period increased, the reduction rate gradually decreased, indicating that the effect of reducing inundation in low rainstorm recurrence period is especially obvious and that the layout of the planning site can reduce the inundation water depth, especially for inundated areas larger than 0.50 m. The reduction rates of the inundated areas with inundation water depths larger than 0.50 m in the planning site layout for the 1-in-10-year, 1-in-50-year, and 1-in-100-year periods were 92.74%, 97.24%, and 81.15%, respectively, and were all higher than those of the other three inundation water depth ranges. There are differences in the inundation area and spatial distribution characteristics of different planning periods. In addition, inundation areas vary substantially among regions under different rainstorm recurrence periods. These variations are closely related to the land-use layout of each planning period, and the spatial structure of the urban land-use, layout, and landscape pattern directly affect urban waterlogging. The average percentage of the four inundation water depth ranges under the different recurrence periods of the four periods are 61.87%, 19.12%, 5.42%, and 13.59%, respectively, indicating that the region is dominated by 0.15-0.28 m waterlogging. The results of this study provide a theoretical basis for flood prevention, mitigation, and the sustainable development of urban areas.

In the process of rapid urbanization, roads undergo drastic changes that provide convenience for travel and have a profound impact on waterlogging. To explore the influence of road layout on the flooding process in rapidly urbanizing areas, taking a village in Shenzhen as an example, the characteristics of road density and road connectivity under four phases of road planning layout were first analyzed. Then, the influence of different road layouts of the four phases on the flooding process of rainstorms with different recurrence periods (10-, 50-, and 100-year periods) was simulated based on the Infoworks ICM model to compare the impacts of different road layouts of the four phases on the flooding characteristics, such as inundated area and inundated water depth, under different road densities and connectivity. The results show that the road densities and connectivity of the planned roads in the four phases increase gradually, from the current situation of 0.0051 m/m² to 0.0091 m/m², and from 1.86 to 2.47, respectively, which represent increases by 78.30% and 32.82%, respectively, indicating that the regional road connectivity increases significantly after the construction of different roads in different periods. There are differences in the inundation area under different road layouts for the same return period, especially the largest difference at the 10-year return period, indicating that storm water pipe network drainage plays an obvious role when the storm magnitude is small. The ratios of the inundation area under the largest one-phase road layout to that under the smallest one-phase road layout at the 10-, 50-, and 100-year return periods are 1.24, 1.15, and 1.07, respectively, indicating that the inundation area under the largest one-phase road layout is 1.24, 1.15, and 1.07 times larger than that under the smallest one-phase road layout; with the increase of rainstorm return period, the difference of inundation area under different road layouts of the four phases decreases gradually. The spatial distribution of the inundation area under different road layouts for the same return period differs, and regional topography and road connectivity are the main reasons affecting the spatial distribution of the inundation area. The depth of inundation of the four phases of roadways under different return periods is mainly concentrated in the range of 0.15-0.28 m. The results of this study provide theoretical basis for the prevention and control of flooding and the sustainable development of urbanized areas.

Using observed monthly precipitation and atmospheric circulation data from 1980 to 2020, this study examines the interdecadal variability in main rainy season precipitation in the Two Lakes Basin and its associated atmospheric drivers. On average, precipitation during the main rainy season accounts for 41.4% of the annual total, with a maximum contribution of 49.9% in 1993 and a minimum of 31.3% in 1991. Spatially, four wet and three dry subregions were identified. Wet areas include northeastern Jiangxi, the Mufu–Jiuling Mountains along the Hunan–Jiangxi border, the northern end of Xuefeng–Wuling Mountains in northwestern Hunan, and Nanling Mountain in the southern basin. Dry areas include the Hengshao Basin, the Dongting Lake region in northern Hunan, and southern Jiangxi. From 1980 to 2000, main-season precipitation exhibited a significant upward trend, characterized by increased rainfall in the central basin and decreases in the northwestern basin and the Hengshao Basin. Temporally, four distinct phases emerged: 1980–1991 (below average), 1992–1999 (above average), 2000–2013 (below average), and 2014–2020 (above average). In Phase 1, most of the basin experienced below-average precipitation, except northwestern Hunan. Phase 2 saw basin-wide above-average precipitation. In Phase 3, only the southern basin received above-average rainfall, while other areas were drier. In Phase 4, precipitation was again above average for most of the basin. The transitions between precipitation phases are strongly linked to variations in the position and intensity of the western Pacific subtropical high and coordinated changes in mid- to high-latitude circulation. During Phase 1, Eurasian mid- to high-latitude circulation exhibited weak meridional disturbances, leading to southerly anomalies and water-vapor divergence over southern China. Cold and warm air masses converged farther north, resulting in reduced rainfall over the Two Lakes Basin. In Phase 2, an anomalously strong Baikal Ridge promoted southward intrusions of cold air, which converged with warm, moist air over southern China, enhancing water-vapor convergence and precipitation. In Phase 3, a relatively 500 hPa flow over Eurasia inhibited southward cold-air penetration and produced widespread water-vapor divergence, suppressing rainfall. During Phase 4, a strengthened trough-ridge pattern associated with the westerly jet favored southward cold-air movement and its convergence with warm, moist air in the basin, leading to enhanced precipitation.

Under the intensifying impacts of anthropogenic climate change, the increasing frequency and severity of compound extreme climate events have emerged as critical challenges with cascading consequences for ecosystems, infrastructure, and human livelihoods. This study presents a comprehensive analysis of four categories of compound extremes: Warm-Wet (WW), Warm-Dry (WD), Cold-Dry (CD), and Cold-Wet (CW) events in Fujian Province, a subtropical coastal region of Southeast China characterized by complex topography and monsoon climate variability. By integrating high-resolution observational data from 20 meteorological stations with outputs from 14 CMIP6 climate models, we developed a robust methodological framework to project future changes under different warming scenarios. This study employed a multistage analytical approach. First, we rigorously evaluated four statistical downscaling techniques (Delta, Daily Translation, Quantile Mapping, and Local Intensity Scaling) to correct for systematic biases in precipitation and temperature simulations. Our results demonstrate that Quantile Mapping provided the most accurate representation of precipitation (RMSE = 37.68 mm, R = 0.87) and minimum temperature patterns, while the Delta method excelled in correcting mean (RMSE = 2.25°C, R = 0.989) and maximum temperature (RMSE = 2.28°C, R=0.982). The multi-model ensemble (MME) approach significantly improved the simulation accuracy and reduced precipitation and temperature errors compared with individual models. The projections under the four Shared Socioeconomic Pathways (SSP) scenarios revealed distinct warming trajectories. Fujian Province will reach 1.5℃ and 2℃ warming thresholds between 2035-2048 across all scenarios, with notable divergence at higher warming levels. The 3°C threshold is projected to occur by 2082 under SSP2-4.5 (medium emissions), 2075 under SSP3-7.0 (high emissions), and as early as 2069 under SSP5-8.5 (very high emissions), highlighting the critical importance of emission mitigation. Analysis of compound extremes shows complex, non-linear responses to warming. At moderate warming levels (1.5°C/2°C), the region experiences significant reductions in warm events (WW: −11.13/−10.95 d/a; WD: −8.24/−8.33 d/a) accompanied by increases in cold events (CD: +9.28/+2.59 d/a; CW: +4.87/+2.63 d/a). However, this pattern undergoes a dramatic reversal under 3°C warming, with WD events surging by 14.28 d/a while other event types decline, indicating a potential regime shift toward more arid conditions. Spatial analysis revealed pronounced geographical heterogeneity. The northwestern mountainous regions and central-southern regions emerged as hotspots of compound extreme variability, with WD events increasing by up to 32.1 days/year in coastal areas. This spatial patterning reflects the interactions between topography, monsoon dynamics, and urbanization effects. Methodologically, this study advances regional climate projection techniques through several innovations: (1) the development of a multi-criteria framework for downscaling method selection, (2) the demonstration of MME superiority in regions with complex terrain, and (3) the creation of a transferable protocol for compound event analysis. These findings provide critical insights for climate adaptation planning in subtropical monsoon regions, particularly for disaster risk assessment and infrastructure resilience. Future research directions include incorporating dynamic downscaling, assessing population exposure risks, and evaluating ecosystem impacts under these changing compound extreme regimes.

Global warming has intensified the spatiotemporal heterogeneity of precipitation, but the typical grid spacing of land-atmosphere coupling models remains much larger than the spatial scale of real precipitation events. This scale mismatch forces models to distribute rainfall uniformly within a grid cell, which can distort rain-rate estimation and subsequently introduce systematic biases into land-surface hydrological simulations. Vegetation canopy interception is particularly sensitive because it is the evapotranspiration component that responds most rapidly to rainfall and directly controls the partitioning of precipitation into interception, evaporation, throughfall, and stemflow, thereby influencing surface energy and water fluxes. In this study, we quantify the global spatiotemporal patterns of rainfall coverage (μ), defined as the fractional area within a model grid cell where rainfall actually occurs. Using bias-adjusted WFDE5 precipitation as the model forcing field and MSWEP V2.8 as the benchmark “actual” precipitation, we derive a global μ dataset at 0.5° spatial resolution and 3-hourly temporal resolution for 1980-2018 and then aggregate it to monthly means for model application. The climatological mean μ over 1981-2018 is 0.36, exhibiting strong spatial contrasts: high values occur in equatorial and tropical rainforest regions, whereas low values dominate subtropical arid and desert zones. Seasonally, μ follows the order June-August > December-February > March-May > September-November, with the largest seasonal amplitude in mid- to high-latitude regions of the Northern Hemisphere. We incorporate the monthly varying μ into the rainfall interception parameterization of Community Land Model version 5 (CLM5), thereby representing subgrid precipitation heterogeneity in canopy interception calculations. Offline global simulations were conducted at a 1° resolution for 1980-2018 using the CLM5 Satellite Phenology configuration. The performance of the modified model (CLM5_μ) is systematically evaluated against four widely used global evapotranspiration products that provide canopy interception estimates, including GLEAMv3.5, GLEAMv4.2, PML, and ERA5. The results demonstrate that introducing rainfall coverage substantially improves the realism of the simulated global canopy interception. The multi-year mean interception decreases from 56.66 mm in the baseline CLM5 to 49.13 mm in CLM5_μ, bringing simulations closer to the range of benchmark products. Spatially, the most pronounced improvements occurred in mid- to low-latitude humid and semi-humid regions (e.g., parts of North America and northern Eurasia) and in arid-to-semiarid transition zones (e.g., the Sahel and Central Asian grasslands), where the baseline model tended to overestimate interception under the uniform rainfall assumption. Temporally, CLM5_μ shows a higher proportion of grid cells with statistically significant correlations to benchmark products and enhanced spatial continuity of correlated areas, especially during summer months in mid- to low latitudes and during non-growing seasons in high-latitudes. Improvements are relatively limited in primary tropical rainforests, high-latitude cold regions (e.g., Siberia and northern Canada), and extremely arid areas. Overall, this study provides a practical and physically interpretable pathway for incorporating spatial heterogeneity of precipitation into canopy interception parameterization. By accounting for rainfall coverage dynamics, the proposed scheme reduces interception biases induced by scale mismatches and strengthens the capability of land-surface models to represent water and heat flux partitioning on a global scale.

Stable isotopes in precipitation (δ²H, δ¹⁸O) serve as effective tracers for short-term weather processes. The complex structure of typhoons and their interactions with multiscale circulation systems result in significant heterogeneity in precipitation distribution, duration, and intensity, limiting the understanding of isotopic fractionation mechanisms. Typhoon Doksuri (2305), the fifth typhoon of 2023 and the second strongest to strike Fujian Province, produced a distinct multi-stage extreme precipitation process under inner core circulation and peripheral mesoscale convective systems. High-resolution hourly isotope data from Fuzhou, combined with meteorological data and HYSPLIT model simulations, were analyzed to characterize isotopic variation and influencing factors. The precipitation δ¹⁸O values ranged from −20.33‰ to −2.01‰, with a weighted mean of −13.69‰. Three stages were identified: (i) minimal rainfall (2.4 mm) with enriched δ¹⁸O values (mean −2.84‰), influenced by peripheral moisture, weak convection, and sub-cloud re-evaporation; (ii) intensified rainfall (63.8 mm) under spiral rainbands, where δ¹⁸O showed strong negative correlation with upstream cumulative precipitation, reflecting extended fractionation through convection and Rayleigh distillation, yielding depleted values (mean −15.64‰); (iii) extreme rainfall (263.2 mm) triggered by mesoscale convective systems, where δ¹⁸O ranged from −16.27‰ to −7.54‰ (mean −13.27‰), showing slight depletion followed by progressive enrichment due to mixing of depleted residual moisture and enriched marine input. A clear decoupling between isotope and rainfall intensity was observed: the strongest precipitation stage exhibited higher δ¹⁸O than the moderate stage, underscoring that isotopic signatures are governed primarily by moisture fractionation history rather than local rainfall intensity. Typhoon precipitation isotopes are jointly regulated by local convection, upstream transport, source region processes, and cumulative fractionation. These findings provide a basis for isotopic model validation and enhance understanding of typhoon-driven precipitation dynamics.