Rainfall Trends and Multi-scale Variability in the Water-receiving Area of the Zhejiang East Water Diversion Project.

Authors

ZENG Tian-li¹, ZUO Xiao-xia¹, YANG Yu¹, DAI Huan¹, WU Mu-hong², ZHONG Lü-bin², CHEN Shu-yang³

1. Zhejiang Design Institute of Water Conservancy and Hydroelectric Power Co., LTD., Hangzhou 310002, China
2. Qingyuan County Water Conservancy Bureau, Lishui, Zhejiang 323800, China
3. Zhejiang Yansi Information Technology Co., Ltd., Hangzhou, Zhejiang 310051, China

Abstract

This study systematically investigates the spatiotemporal distribution characteristics, long-term trends, multi-scale variability, and regional correlation of rainfall in 15 typical sub-regions of Eastern Zhejiang based on long-term daily rainfall observation data from 1961 to 2022 (62 years). Employing the Mann-Kendall (MK) trend test, Sen’s slope estimation, Hurst exponent analysis, and multi-scale sliding window analysis, this research provides comprehensive insights into rainfall dynamics within the influence area of the Eastern Zhejiang Water Diversion Project.

Regarding spatial distribution, daily average rainfall ranges from 3.47 mm to 5.68 mm, exhibiting a distinct “high-in-center, low-in-west, medium-in-east” spatial pattern. The central high-value area (Yuyao Plain Mazhu Midstream Area) reaches 5.68 mm, while the western low-value area (Nansha Plain Area) records only 3.47 mm.

In terms of long-term trends, Mann-Kendall tests reveal that all 15 sub-regions exhibit statistically significant upward trends (p<0.05). Sen’s slope estimates demonstrate spatial differentiation in trend magnitude following a “high-in-coastal, medium-in-river, low-in-hilly” pattern. Hurst exponent analysis (H > 0.5) confirms that the observed upward trends possess long-term persistence.

The multi-scale variability analysis through sliding window reveals pronounced scale dependence. At the 3-month scale, precipitation exhibits high instability with large trend fluctuations (±16 mm) and low correlation. At the 12-month scale, trends become more stable (±1.7 mm) with improved correlation, indicating that long-term patterns are controlled by large-scale climate factors.

With respect to regional correlation, inter-regional correlations strengthen with increasing time scale, rising from 0.83 (3-month) to 0.87 (12-month). Notably, the Yuyao Plain Mazhu Midstream Area displays a unique “increase-then-decrease” correlation pattern, reflecting its special hydro-geographic conditions.

For engineering implications, the post-2010 intensification of rainfall fluctuations necessitates optimized scheduling strategies for the Eastern Zhejiang Water Diversion Project to enhance flood control and supply security under changing climate conditions.

Keywords

Eastern Zhejiang Water Diversion Project; rainfall spatiotemporal distribution; Mann-Kendall test; Sen’s slope estimation; Hurst exponent; multi-scale analysis; scale dependence

1. Introduction

1.1 Regional Hydrological and Climate Context

The Eastern Zhejiang region, situated on the highly developed eastern coast of Zhejiang Province, China, serves as a critical economic engine for the Yangtze River Delta. Characterized by a typical subtropical monsoon climate, the region receives abundant annual precipitation exceeding 1400 mm. However, this apparent abundance masks a severe structural challenge: the spatiotemporal distribution of water resources is extremely uneven. Temporally, over 60% of the annual rainfall is concentrated in the Meiyu season (plum rains, typically June-July) and the Typhoon season (August-September), creating a pendulum swing between flood risks in summer and drought risks in winter and spring. Spatially, water resources decrease from the mountainous southwest to the coastal northeast, inverse to the gradient of economic development and population density.

The Eastern Zhejiang Water Diversion Project, which commenced full operation in June 2014, represents the largest cross-basin water transfer infrastructure in Zhejiang Province. It serves 19 counties (districts) across the cities of Hangzhou, Shaoxing, Ningbo, and Zhoushan, delivering an average annual diversion volume of 890 million m³. This project acts as a lifeline for regional water security. However, in the context of accelerating global climate change, recent hydro-meteorological observations indicate that precipitation patterns in this region are undergoing significant and non-stationary shifts. The frequency and intensity of extreme precipitation events—both varying from prolonged droughts to localized torrential rains—have increased markedly. These shifts pose unprecedented challenges to the project's operational logic: water source stability is threatened by shifting seasonal baselines, while conveyance channel safety is jeopardized by flash floods. Therefore, a systematic, multi-scale analysis of rainfall trends and variability specifically within the project's receiving areas is not merely an academic exercise but an urgent engineering necessity for adaptive management.

1.2 Research Progress on Precipitation Variability

1.2.1 Global Precipitation Change and the "Wet-Gets-Wetter" Paradigm

The Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) confirms unequivocally that human influence has intensified the global water cycle [1]. Since the 1950s, precipitation variability over mid-latitude land areas has increased, a phenomenon consistent with the thermodynamic Clausius-Clapeyron relationship, which predicts a ~7% increase in atmospheric water-holding capacity per degree Celsius of warming [2]. Recent high-impact studies have further solidified the "wet-gets-wetter" paradigm. Madakumbura et al. (2021) provided robust evidence that anthropogenic climate change is driving an intensification of extreme precipitation events globally, particularly in monsoon regions [R1]. Furthermore, Kotz et al. (2022) demonstrated that the variability of precipitation is increasing at a rate faster than mean precipitation itself, leading to more volatile hydro-climatic regimes that challenge traditional water resource management assumptions [R2].

1.2.2 Regional Responses in Eastern China

In China, the response to global warming exhibits significant regional heterogeneity. While northern China has experienced complex shifts between drying and wetting, eastern and southern China generally show a trend of increasing precipitation intensity [15,16]. For the Yangtze River Basin and the southeast coastal zone, recent studies have identified a "dual-intensification" pattern: both the frequency of extreme heavy rainfall and the duration of dry spells are increasing. Zhou et al. (2023) analyzed the changing characteristics of the East Asian Summer Monsoon, revealing that the northward shift of the rain belt has led to more concentrated and intense rainfall events in the Yangtze River Delta [R3]. Li et al. (2022) focused on the spatiotemporal evolution of precipitation extremes in Zhejiang, identifying a significant correlation between rising sea surface temperatures (SST) in the East China Sea and the amplification of coastal typhoon rainfall [R4]. Zhang et al. (2024) further highlighted that urbanization in this megalopolis region may be exacerbating these trends through the urban heat island effect, creating localized "rain islands" [R5].

1.2.3 Methodological Advances in Trend Detection

The detection of trends in non-stationary hydrological series relies on robust statistical methods. The Mann-Kendall (MK) test [6,7] and Sen's slope estimator [11] remain the gold standards for assessing trend significance and magnitude due to their resilience against outliers [8]. However, recent methodological advancements emphasize the need for integrated frameworks. The Hurst exponent [12] has seen renewed application in evaluating the "long-term memory" or persistence of detected trends, offering a predictive dimension often missing in traditional trend analysis [R6]. Moreover, the recognition of scale dependence has driven the adoption of multi-scale approaches. Wavelet analysis and sliding window techniques are increasingly used to decompose hydrological time series, revealing how statistical characteristics (such as variance and correlation) evolve from short-term weather noise (days to months) to long-term climate signals (years to decades) [13,14]. Wu et al. (2023) demonstrated the utility of multi-scale entropy in characterizing the changing complexity of rainfall regimes in changing environments [R7].

1.2.4 Engineering Implications for Water Diversion

The intersection of climate change and large-scale water diversion projects is a growing field of inquiry. Research on the South-to-North Water Diversion Project has highlighted the vulnerability of inter-basin transfer systems to simultaneous droughts in source and receiving areas [R8]. For the Eastern Zhejiang Water Diversion Project, the challenge is acute: the system must balance flood control during the typhoon season with water supply during dry spells. However, most existing studies focus on the physical infrastructure or hydraulic optimization [R9], with fewer studies systematically linking long-term climatic trends to operational resilience. Recent work by Liu et al. (2023) calls for "climate-adaptive scheduling" that dynamically integrates trend detection into real-time decision-making frameworks [R10].

1.3 Research Gaps and Objectives

1.3.1 Research Gaps

Despite substantial progress, critical gaps remain regarding the specific context of the Eastern Zhejiang Water Diversion Project: Lack of Fine-Scale Analysis: Most existing studies operate at provincial or basin scales (e.g., Qiantang River Basin). There is a notable lack of high-resolution analysis targeting the 15 specific sub-regional receiving areas. This "granularity gap" means that local variations—such as the difference between coastal plains and inland hills—are often smoothed out, leading to one-size-fits-all management strategies that are inefficient. Limited Integrated Assessment: While individual methods (MK, Sen's) are widely used, few studies systematically combine trend significance, magnitude, and persistence (Hurst exponent) into a unified framework. Understanding whether a trend is a temporary fluctuation or a persistent shift is crucial for long-term infrastructure planning. Insufficient Multi-scale Analysis: The scale dependence of rainfall variability and inter-regional correlations in this coastal transition zone remains under-investigated. It is unclear how the spatial synchronization of rainfall events changes from seasonal to annual scales, which is vital for coordinating multi-reservoir operations. Engineering Gap: There is a disconnect between climatic research and engineering application. Most studies stop at trend detection without translating findings into operational recommendations for source-receiver coordination and risk management under the specific constraints of the water diversion project.

1.3.2 Research Objectives

Addressing these gaps, this study focuses on the 15 sub-regions of the Eastern Zhejiang Water Diversion Project. Based on a continuous, quality-controlled 62-year dataset (1961-2022), it aims to achieve four objectives: Spatial Analysis: To quantify the multi-year average daily rainfall and identify spatial clustering patterns (high/medium/low-value areas) to provide a precise baseline for regional water balance assessments. Trend Assessment: To detect trend significance, quantify magnitudes, and evaluate persistence using an integrated MK-Sen-Hurst framework, specifically identifying "hotspots" of rapid climate response that require prioritized adaptation measures. Multi-scale Analysis: To characterize rainfall variability and the evolution of inter-regional correlations at 3-, 6-, and 12-month scales, distinguishing between local weather noise and regional climate signals to inform multi-time-scale scheduling. Engineering Application: To synthesize these findings into scientific recommendations for differentiated water supply strategies, source-receiver coordinated optimization, and adaptive management, directly enhancing the project's resilience to future climate uncertainty.

2. Study Area and Data

2.1 Overview of the Study Area

2.1.1 Geographic Location and Topography

The study area is located in the eastern part of Zhejiang Province, China, covering 15 typical sub-regions distributed across five administrative zones: Xiaoshan, Shaoyu, Yuyao, Cixi, and Ningbo (Figure 1). This region serves as the primary water receiving area for the Eastern Zhejiang Water Diversion Project.

The topography features a complex transition from coastal plains in the east to inland hills in the west, which significantly influences regional hydrological processes. The characteristics of the three major topographic zones are summarized in Table 1a.

Table 1a. Topographic Zones and Characteristics

Table 1b. The 15 Sub-regions of the Study Area

2.1.2 Hydro-climate Characteristics

The region is characterized by a subtropical monsoon climate with substantial precipitation variability. Annual precipitation multi-year averages range between 1273.5 mm and 1454.8 mm, with high inter-annual variability as historical maximums reach 2050.1 mm and minimums drop to 682.2 mm. Regarding seasonal distribution, rainfall is highly concentrated in the Meiyu season (June-July) and Typhoon season (August-September), which together account for over 60% of the annual total. This concentration often leads to alternating flood and drought risks.

Figure 1. Location of the study area and distribution of 15 sub-regions

2.2 Data Sources

This study utilizes daily precipitation observation records from 1961 to 2022 (62 years) for 15 representative meteorological/hydrological stations within the study area. Data were obtained from the Zhejiang Provincial Meteorological Bureau and local hydrological monitoring networks. The dataset includes 22,630 daily values per station, capturing multiple decadal climate cycles and providing a robust foundation for long-term trend analysis.

2.3 Data Quality Control

To ensure reliability, a rigorous four-stage data quality control system was applied. The first stage involved continuity assessment, which systematically identified time series gaps; missing data accounted for less than 2% for any station. The second stage focused on outlier detection, applying the 3σ criterion (Pauta criterion) to flag statistical anomalies; flagged values were cross-checked against adjacent stations and synoptic records to distinguish measurement errors from genuine extreme events. The third stage addressed interpolation, where small gaps were filled using linear regression with the most highly correlated neighboring stations, validated against seasonal climatology. The fourth stage comprised homogeneity testing, employing the Standard Normal Homogeneity Test (SNHT) and sliding t-test methods to detect and correct breakpoints caused by station relocations or instrument changes.

The final quality-controlled dataset exceeds 98% completeness and has been verified for homogeneity.

3. Research Methods

This study employs four complementary analytical methods to comprehensively characterize rainfall spatiotemporal patterns: (1) Mann-Kendall trend test for significance assessment, (2) Sen's slope estimation for trend magnitude quantification, (3) Sliding window analysis for multi-scale variability characterization, and (4) Hurst exponent analysis for trend persistence evaluation.

3.1 Mann-Kendall (MK) Trend Test

The Mann-Kendall test is a rank-based non-parametric method widely used to detect monotonic trends in hydro-climatic time series. Its primary advantages are its robustness against outliers and its capability to handle non-normally distributed data [6,7].

For a time series $\{x\_{1},x\_{2},...,x\_{n}\}$, the test statistic $S$ is calculated as:

$S=\sum\_{i=1}^{n-1}\sum\_{j=i+1}^{n}sgn(x\_{j}-x\_{i}) (Equation 1)$

where $sgn(θ)$ is the sign function:

$sgn(θ)=\left\{\begin{matrix}+1,&if θ>0\\0,&if θ=0\\-1,&if θ<0\end{matrix}\right.$

For sample sizes $n\geq 8$, the statistic $S$ is approximately normally distributed with mean $E(S)=0$ and variance:$Var(S)=\frac{n(n-1)(2n+5)-\sum\_{p=1}^{g}t\_{p}(t\_{p}-1)(2t\_{p}+5)}{18}$

where $g$ is the number of tied groups and $t\_{p}$ is the number of observations in the $p$-th tied group.

The standardized test statistic $Z$ is computed to assess significance:

$Z=\left\{\begin{matrix}\frac{S-1}{\sqrt{Var(S)}},&if S>0\\0,&if S=0\\\frac{S+1}{\sqrt{Var(S)}},&if S<0\end{matrix}\right. (Equation 2)$

A positive $Z$ indicates an upward trend, while a negative $Z$ indicates a downward trend. The null hypothesis of no trend is rejected at the significance level $α$ if $|Z|>Z\_{1-α/2}$. In this study, significance levels of $α=0.05$ ($|Z|>1.96$) and $α=0.01$ ($|Z|>2.576$) are used.

3.2 Sen's Slope Estimation

Sen's slope estimator [11] is used to estimate the true magnitude of the linear trend. Unlike linear regression, it is insensitive to outliers. For all data pairs $(x\_{i},x\_{j})$ where $j>i$, the slope $Q\_{k}$ is calculated:

$Q\_{k}=\frac{x\_{j}-x\_{i}}{j-i} (Equation 3)$

The Sen's slope $β$ is defined as the median of all $N=n(n-1)/2$ calculated slopes. A positive $β$ signifies an increasing trend (mm/year). The 95% confidence interval for the slope is estimated using a non-parametric procedure based on the normal distribution of the rank statistic.

3.3 Sliding Window Analysis

Table 3a. Sliding Window Configurations

For a window of size $w$, four statistical moments are calculated at each time step $k$. The arithmetic mean ($μ\\_k$) is computed as $μ\\_k=\frac{1}{w}∑\\_i=k^{k+w-1}x\\_i$. The standard deviation ($σ\\_k$) is calculated as $σ\\_k=\sqrt{\frac{1}{w-1}∑\\_i=k^{k+w-1}(x\\_i-μ\\_k)^{2}}$. The coefficient of variation ($CV\\_k$) is derived as $CV\\_k=\frac{σ\\_k}{μ\\_k}$, providing a normalized measure of variability. The trend slope within window ($a\\_k$) represents the instantaneous rate of change within the window.

3.4 Hurst Exponent Analysis

The Hurst exponent ($H$) quantifies the long-term memory or persistence of a time series [12], estimated via Rescaled Range (R/S) analysis. The procedure involves three steps: first, dividing the series into sub-series of length $n$; second, calculating the range of cumulative deviations ($R$) and standard deviation ($S$) for each sub-series; and third, fitting the power law relationship $(R/S)\\_n∝n^{H}$.

The value of $H$ ranges from 0 to 1, with interpretations summarized in Table 3b.

Table 3b. Hurst Exponent Interpretation

3.5 Inter-Regional Correlation Analysis

To assess spatial coherence, Pearson correlation coefficients ($r$) are calculated between all pairs of the 15 sub-regions. Correlation matrices are computed for data aggregated at 3-month, 6-month, and 12-month scales. This scale-dependent correlation analysis reveals how spatial synchronization evolves from local weather scales to regional climate scales.

4. Results and Analysis

4.1 Rainfall Spatial Distribution Characteristics

4.1.1 Overview of Spatial Pattern

Analysis of the 62-year (1961-2022) daily rainfall records reveals that the study area exhibits a distinct spatial distribution pattern characterized as "high-in-center, low-in-west, medium-in-east" (Figure 2). The daily average rainfall across the 15 sub-regions ranges from 3.47 mm to 5.68 mm, with an overall regional average of 3.98 mm.

This spatial pattern reflects the combined influences of maritime moisture availability from the East China Sea, topographic modification of regional airflow, local water body distributions and evapotranspiration, and confluence effects in river valley systems.

Figure 2. Spatial distribution of daily average rainfall (mm) across 15 sub-regions in Eastern Zhejiang (1961-2022)

4.1.2 Western Low-Value Area

The western zone, encompassing three sub-regions, records the lowest daily average rainfall with a zonal mean of 3.80 mm.

The lower rainfall in the western zone can be attributed to several factors. First, airflow subsidence effects occur because the western portions lie in the lee side of regional topographic features, potentially experiencing subsidence of air masses that reduces precipitation efficiency. Second, the transitional hills between the western plains and adjacent highlands act as terrain barriers that may intercept moisture-bearing air masses, creating a partial rain shadow effect. Third, compared to coastal areas, the western zone is farther from the primary maritime moisture source (East China Sea), resulting in reduced moisture availability.

4.1.3 Central High-Value Area

The central zone represents the rainfall maximum of the study area, with six sub-regions recording a zonal mean of 4.26 mm—12% higher than the western zone.

The central high-value area, particularly the Yuyao Plain Mazhu Midstream Area with its exceptional 5.68 mm daily average, owes its elevated rainfall to:

River Confluence Effects: The Mazhu Midstream Area is located at a major confluence zone where multiple tributaries of the Yao River system converge. This creates enhanced local water vapor availability and potential convergence of low-level airflow along river valleys.

Dense Water Network: The presence of numerous rivers, lakes, and irrigation channels significantly augments local moisture cycling through increased evaporation and transpiration.

Topographic Convergence: The central plains represent a natural collection zone for airflow from surrounding elevated terrain, enhancing low-level moisture convergence and orographic triggering of convection.

Notable Observation: The Yuyao Plain Mazhu Midstream Area (5.68 mm) receives 64% more daily rainfall than the Nansha Plain Area (3.47 mm), despite their relatively close proximity (~30 km). This remarkable gradient underscores the importance of local geographic factors in modulating regional precipitation patterns.

4.1.4 Eastern Medium-Value Area

The eastern zone, comprising six sub-regions along the coast and in the Yao River downstream area, records an intermediate zonal mean of 3.88 mm.

Despite direct exposure to maritime air masses from the East China Sea, the eastern coastal zone does not record the highest rainfall. This apparent "Coastal Paradox" is explained by:

Lack of Lifting Mechanisms: The flat coastal terrain provides minimal orographic lift to incoming maritime air masses.

Marine Boundary Layer Stability: The stable marine boundary layer over the cool East China Sea can suppress convective development.

Coastal Wind Effects: Strong onshore winds can transport moisture inland before significant precipitation occurs.

Internal Gradient: An interesting pattern emerges within the Cixi Plain series, showing a westward decrease from Midstream (3.90 mm) to West River (3.63 mm), potentially reflecting local moisture dynamics.

4.1.5 Summary of Spatial Distribution

Table 2. Daily Average Rainfall Statistics by Zone

The higher coefficient of variation (CV = 0.172) in the central zone reflects the exceptional status of the Yuyao Plain Mazhu Midstream Area, while the low CV (0.042) in the eastern zone indicates more homogeneous conditions among coastal sub-regions.

4.2 Long-term Trend Evolution Characteristics

This section presents a comprehensive analysis of rainfall trends across the 15 sub-regions of Eastern Zhejiang from 1961 to 2022, including Mann-Kendall test results, Sen's slope estimates, decadal variations, spatial patterns of trend intensity, uncertainty assessment, and Hurst exponent persistence analysis.

4.2.1 Trend Significance Analysis (Mann-Kendall Test)

The Mann-Kendall test was applied to the annual rainfall series of all 15 sub-regions. Results demonstrate that all regions exhibit statistically significant upward trends at the 0.05 significance level (Table 1).

Table 3. Mann-Kendall Test Results for Annual Rainfall (1961-2022)

Spatial Pattern: The Z-values reveal clear spatial differentiation. The strongest trends (Z > 4.0) are concentrated in the western Nansha area and the central/eastern coastal plains (Yubei Midstream, Cixi East/Midstream). In contrast, the hilly areas (Shaoyu) and the highly variable Mazhu Midstream area show relatively weaker signals (Z < 2.3), though still statistically significant.

4.2.2 Decadal Variations and Inter-annual Patterns

Figure 3 presents the annual trend change chart showing simultaneous variations across all 15 sub-regions from 1961 to 2022. The region exhibits pronounced decadal-scale oscillations superimposed on the long-term upward trend.

Figure 3. Annual trend change chart of daily average rainfall (1961-2022)

The key temporal features are summarized in Table 3c.

Table 3c. Key Temporal Features of Rainfall Variation (1961-2022)

A critical observation is the post-2010 intensification, characterized by increased frequency and amplitude of fluctuations. The Yuyao Plain Mazhu Midstream Area reached a record high of approximately 9 mm/day in 2021. This intensification suggests a shift toward more extreme precipitation regimes under climate change.

4.2.3 Individual Sub-region Trend Analysis

Detailed trend analysis was performed for each sub-region using Sen’s slope estimation. The 15 sub-regions are categorized into three groups based on their geographic characteristics and trend magnitudes.

4.2.3.1. Coastal High-Change-Rate Regions

The coastal zone exhibits the highest trend magnitudes among all regions, reflecting enhanced sensitivity to oceanic climate changes. The Cixi Plain Midstream Area records the highest trend in the entire study area, with a Sen’s slope of 7.99 mm/year; its time series shows a consistent upward trajectory from approximately 800 mm in the late 1960s to over 2000 mm in the 2020s, with strong positive anomalies frequently observed after 2010 (Figure 4).

The Nansha Plain Area, despite being the driest region with the lowest baseline rainfall, shows the second-highest trend magnitude of 7.95 mm/year, with rainfall increasing by approximately 50% over the study period; a major cluster of positive anomalies is evident from 2010 to 2022, indicating a rapid wetting trend in this historically drier western zone (Figure 5). The Cixi Plain East River Area displays a Sen’s slope of 7.15 mm/year with a clear increasing pattern particularly accelerated after 2000, exhibiting temporal evolution similar to the neighboring Cixi Midstream Area but with slightly lower magnitude (Figure 6). The Cixi Plain West River Area records a Sen’s slope of 6.42 mm/year, indicating a westward decrease in trend magnitude within the Cixi system, likely reflecting increasing distance from the primary moisture source of the East China Sea (Figure 7).

The trend characteristics of coastal regions are summarized in Table 4a.

Table 4a. Trend Characteristics of Coastal High-Change-Rate Regions

4.2.3.2. River and Inland Plain Regions

The river and inland plain regions exhibit moderate trend magnitudes, characterized by strong hydrological connectivity and pronounced decadal oscillations. The Yuyao Plain Yaojiang Downstream Area shows the highest trend within this group at 6.66 mm/year, featuring a large variability range of approximately 800-2300 mm, clear decadal oscillations, and strong precipitation increases after 2010 (Figure 8). The Yaojiang Upstream Area exhibits a Sen’s slope of 6.46 mm/year with a trend pattern closely mirroring the downstream area, reflecting strong hydrological connectivity and shared climatic influences within the Yaojiang river system (Figure 9).

Both the Yuyao Plain Upstream and Downstream areas show an identical Sen’s slope of 6.30 mm/year, indicating synchronized behavior and common climate forcing across the inland plains with a moderate growth rate typical of the central zone (Figures 10-11). The Mazhu Midstream Area is unique in this group, exhibiting the largest variability (range 1250-3250 mm) and highest baseline rainfall in the region; although its Sen’s slope is a moderate 6.20 mm/year, the absolute increases are substantial with a historic high approaching 3250 mm in 2021, and the extreme inter-annual variability partially masks the linear trend signal, resulting in a lower Z-value of 2.28 despite large absolute gains (Figure 12).

The Yubei Plain Midstream Area shows a Sen’s slope of 5.20 mm/year with a very consistent upward trend; despite the moderate slope magnitude, it has a high Z-value of 4.20, indicating a steady and significant increase over the study period (Figure 13). The Fenghui Plain Area records a Sen’s slope of 5.08 mm/year, representing a below-average trend for the central zone, characterized by relatively stable inter-annual variability compared to its neighbors (Figure 14).

The trend characteristics of river and inland plain regions are summarized in Table 4b.

Table 4b. Trend Characteristics of River and Inland Plain Regions

4.2.3.3. Hilly and Mountainous Regions

The hilly and mountainous regions exhibit the lowest trend magnitudes, likely due to terrain blocking effects and stable microclimates that buffer large-scale climate signals. The Shushan Plain Area, located in the western region, shows a Sen’s slope of 5.20 mm/year; while moderate, its variability is higher than that of coastal zones, reflecting its transitional geographic position (Figure 15). The Yubei Plain Upstream Area, influenced by hilly terrain, shows an identical Sen’s slope of 5.20 mm/year with a moderate but consistent trend typical of inland transition zones (Figure 16).

The Jiangbei Zhenhai Plain Area presents an interesting case: despite its coastal location, it records a Sen’s slope of only 4.91 mm/year, which is lower than other coastal regions, suggesting different synoptic influences or local microclimatic effects (Figure 17). The Shaoyu Plain Area exhibits the lowest trend in the entire study region at 4.00 mm/year; the surrounding hilly terrain likely creates a stable microclimate and terrain blocking effects that limit moisture advection, effectively buffering the large-scale warming signal (Figure 18).

The trend characteristics of hilly and mountainous regions are summarized in Table 4c.

Table 4c. Trend Characteristics of Hilly and Mountainous Regions

4.2.4 Spatial Pattern of Trend Magnitude

Figure 19 synthesizes the Sen's slope estimates, revealing a "Coastal > River Basin > Hilly" spatial pattern. Coastal areas show ~50% stronger trends (avg 7.45 mm/yr) than hilly areas (avg 4.91 mm/yr), indicating greater sensitivity to marine climate warming.

Figure 19. Spatial pattern of rainfall trend magnitudes

4.2.5 Trend Uncertainty Analysis (95% Confidence Intervals)

Figure 20 presents the 95% confidence intervals for Sen's slope estimates. A critical finding is that the 95% confidence intervals for ALL 15 sub-regions include zero. This indicates substantial statistical uncertainty in the magnitude of the trend, primarily due to the high inter-annual variability of the monsoon climate. While the MK test confirms the existence of trends, their precise rate of change varies significantly. The Mazhu Midstream Area displays the widest confidence interval (~400 mm/year), consistent with its extreme fluctuations.

Figure 20. Rainfall variation trends and 95% confidence intervals

4.2.6 Trend Persistence Analysis (Hurst Exponent)

To evaluate trend reliability, Hurst exponent ($H$) analysis was performed.

The Hurst exponent analysis yields consistent results across all 15 sub-regions, with all H values exceeding 0.5 (range: 0.5595 to 0.6257). The Nansha Plain Area exhibits the highest persistence ($H=0.6257$), aligning with its strong trend. In contrast, the Mazhu Midstream Area shows the lowest persistence ($H=0.5595$), where high variability reduces the persistence signal.

The consistent finding of $H>0.5$ across all regions confirms that the observed upward trends possess long-term memory and are likely to continue in the near future, rather than being random fluctuations. This provides robust evidence for long-term water resource planning.

Table 2. Hurst Exponent Values for Annual Rainfall (1961-2022)

4.3 Multi-time Scale Variation Characteristics

Sliding window analysis was employed to examine the scale-dependent behavior of rainfall in Eastern Zhejiang. By analyzing statistical characteristics at 3-month, 6-month, and 12-month scales, this section reveals how rainfall variability, trends, and patterns evolve across different temporal aggregation levels. The Cixi Plain East River Area is used as the representative case for detailed illustration, with regional comparisons provided.

4.3.1 Three-Month Scale Analysis (Seasonal/Intra-seasonal)

Figure 21 presents the 3-month sliding window analysis results, displaying four key statistics: daily average rainfall, standard deviation, coefficient of variation, and trend slope.

Figure 21. 3-month sliding window analysis for Cixi Plain East River Area

4.3.1.1 Detailed Observations

The daily average rainfall exhibits rapid, high-amplitude oscillations ranging from 0 to 10 mm, with clear seasonal signals showing summer maxima and winter minima. Peak values approach 10 mm during strong monsoon periods, while troughs drop near zero. The standard deviation ranges from 0 to 8 mm, tracking rainfall seasonality, with peaks exceeding 6 mm during periods of high variability such as typhoon sequences. The coefficient of variation frequently exceeds 1.0 (range 0 to 1.5+), indicating extreme relative variability where the standard deviation often surpasses the mean. The trend slope shows chaotic oscillations with rapid sign reversals within a range of ±5 mm; at this scale, “trends” are dominated by noise and lack persistence.

4.3.1.2 Regional Statistics at 3-Month Scale

Across all 15 sub-regions, the 3-month scale is characterized by a trend fluctuation range of ±16 mm (-16.44 to +16.60 mm) and a minimum correlation of 0.578. The key finding is that precipitation at this scale exhibits high instability and spatial heterogeneity, driven largely by local convective systems and terrain effects rather than large-scale climate patterns.

4.3.2 Six-Month Scale Analysis (Semi-annual)

Figure 22 presents the 6-month sliding window analysis, revealing how aggregation to half-year scales modifies statistical characteristics.

Figure 22. 6-month sliding window analysis for Cixi Plain East River Area

4.3.2.1 Detailed Observations

The daily average rainfall fluctuations smooth out to a range of 2.5-7.5 mm, with the annual cycle becoming clearer and reduced extremes compared to the 3-month scale. The standard deviation peak values are reduced by approximately 30% (range 1.5-5.5 mm). The coefficient of variation stabilizes around 0.7-0.8, with fewer events exceeding 1.0. The trend slope fluctuation range reduces significantly to ±2 mm, with trends showing longer wavelengths and fewer reversals, indicating emerging persistence.

4.3.2.2 Regional Statistics at 6-Month Scale

At the 6-month scale, the trend fluctuation range narrows to ±5 mm (-5.30 to +5.09 mm), representing a 70% reduction in volatility compared to the 3-month scale. The minimum correlation increases to 0.623. The key finding is that semi-annual aggregation filters out high-frequency weather noise, revealing clearer decadal patterns and improved spatial coherence.

4.3.3 Twelve-Month Scale Analysis (Annual)

Figure 23 presents the 12-month (annual) sliding window analysis, demonstrating the most stable characterization of rainfall behavior.

Figure 23. 12-month sliding window analysis for Cixi Plain East River Area

4.3.3.1 Detailed Observations

The daily average rainfall shows smooth, quasi-sinusoidal oscillations within a constrained range of 2-5 mm, with long-term decadal variations becoming the dominant feature (e.g., early 1990s peak, mid-2010s rise). The coefficient of variation drops to a stable range of 0.4-1.0, indicating predictable relative variability. The trend slope fluctuations are drastically reduced to ±0.5 mm, with annual-scale trends showing high persistence and reversing only with major climate shifts.

4.3.3.2 Regional Statistics at 12-Month Scale

At the 12-month scale, the trend fluctuation range narrows to ±1.7 mm (-1.58 to +1.73 mm), representing approximately 90% reduction from the 3-month scale. The minimum correlation reaches 0.640, the highest among all scales. The key finding is that at the annual scale, rainfall patterns are controlled by large-scale climate modes (e.g., ENSO, Monsoon) rather than local weather, resulting in maximum stability and spatial synchronization.

4.3.4 Scale Dependence Synthesis

The analysis confirms a strong scale dependence in rainfall characteristics. Regarding variability reduction, trend fluctuation amplitude decreases by approximately 90% from 3-month (±16 mm) to 12-month (±1.7 mm) scales. In terms of correlation enhancement, spatial coherence improves as the analysis window widens, with minimum correlation increasing from 0.578 to 0.640. The physical interpretation of this hierarchy reflects the transition from weather-dominated processes (stochastic, local) at short scales to climate-dominated processes (organized, regional) at long scales. This finding suggests that annual statistics provide the most reliable basis for long-term water resource planning, while short-term operations must account for extreme instability.

4.4 Scale Dependence of Regional Rainfall Correlation

This section examines the inter-regional correlation patterns of rainfall across the 15 sub-regions of Eastern Zhejiang. Correlation matrices were computed for 3-month, 6-month, and 12-month aggregated series to reveal how spatial coherence evolves with temporal scale.

4.4.1 Three-Month Scale Correlation Analysis

Figure 24 presents the correlation matrix at the 3-month scale.

Figure 24. Inter-regional correlation matrix (2-month scale)

Statistical Overview and Spatial Patterns

The correlation analysis at the 3-month scale reveals an average correlation coefficient of 0.83 across the region, with values ranging from a minimum of 0.709 to a maximum of 1.000. The lowest correlation (0.709) is observed between the Yuyao Plain Mazhu Midstream Area and the Nansha Plain Area.

Geographically, distinct clusters emerge even at this short time scale. The Coastal Cluster exhibits high internal coherence; for instance, the correlation between Cixi East River and Cixi Midstream reaches 0.905. Similarly, the River Basin Cluster demonstrates strong hydrological connectivity, evidenced by the perfect correlation (1.000) between Yuyao Upstream and Downstream areas. However, the Yuyao Plain Mazhu Midstream Area stands out as an anomalous region, showing systematically lower correlations with most other regions (e.g., 0.709 with Nansha). This decoupling likely reflects its unique hydro-geographic conditions and high local variability, which are more pronounced at seasonal scales.

Interpretation

At the seasonal scale (3-month), local meteorological factors such as convective systems and terrain effects play a dominant role. This dominance results in significant spatial heterogeneity across the study area, leading to relatively lower inter-regional correlations compared to longer time scales.

4.4.2 Six-Month Scale Correlation Analysis

Figure 25 presents the correlation matrix at the 6-month scale.

Figure 25. Inter-regional correlation matrix (6-month scale)

Statistical Changes and Cluster Strengthening

As the aggregation scale extends to 6 months, the average correlation coefficient increases slightly to 0.85, representing a net gain of 0.02 compared to the 3-month scale. The minimum observed correlation also improves to 0.718.

This scale extension leads to a noticeable strengthening of regional clusters. The coastal cluster, in particular, shows significant improvement in internal coherence; for example, the correlation between Cixi Midstream and West River rises from 0.899 to 0.938. This enhancement suggests that semi-annual aggregation effectively filters out a portion of the high-frequency noise associated with short-term weather events, thereby revealing stronger underlying oceanic climate signals, especially in coastal zones.

4.4.3 Twelve-Month Scale Correlation Analysis

Figure 26 presents the correlation matrix at the 12-month (annual) scale, showing the highest overall spatial coherence.

Figure 26. Inter-regional correlation matrix (12-month scale)

Overall Coherence and Anomalous Patterns

The 12-month scale analysis demonstrates the highest level of spatial synchronization, with the average correlation reaching 0.87 (an increase of 0.04 from the 3-month scale). Correlation coefficients across the matrix range from 0.681 to 1.000.

A notable finding is the “Increase-then-Decrease” pattern observed in the Yuyao Mazhu Midstream Area. For several region pairs, such as with Nansha, the correlation first increases from the 3-month scale (0.709) to the 6-month scale (0.718), but then unexpectedly decreases at the 12-month scale (0.681). This anomaly suggests that while seasonal aggregation initially improves coherence, the annual total rainfall in the Mazhu area may be decoupled from regional patterns. This could be due to specific local hydrological processes, such as moisture recycling, which operate on different timescales than the broader regional climate.

In terms of spatial patterns, the Coastal Cluster reaches maximum coherence with an average correlation of approximately 0.93, driven by common large-scale marine climate forcing. In contrast, the Hilly Cluster shows the smallest increase in correlation, indicating that terrain-induced heterogeneity persists even when data are aggregated to annual scales.

4.4.4 Summary of Correlation Evolution

Table 4. Correlation Statistics Across Time Scales

Implications for Management

The scale-dependent correlation patterns have direct implications for water resource management. High-correlation regions, such as the Coastal and River Basin clusters, can be managed as coordinated systems for long-term planning, leveraging their synchronized climate response. Conversely, low-correlation regions like the Mazhu Midstream Area require independent operational strategies and robust risk management plans, as their hydrological behavior may diverge significantly from the regional average, especially during extreme years.

5. Discussion and Conclusions

5.1 Discussion

5.1.1 Mechanisms of Spatial Differentiation

The study reveals a distinct “high-in-center, low-in-west” rainfall pattern. The exceptional rainfall in the Mazhu Midstream Area (5.68 mm/day) is likely driven by confluence zone enhancement. As a major convergence point of the Yao River tributaries, this area experiences channelized airflow and forced uplift. Furthermore, the dense network of rivers and wetlands enhances local evapotranspiration, creating a moisture recycling feedback loop that sustains high precipitation. In contrast, the Nansha Plain Area (3.47 mm/day) in the west likely suffers from lee-side subsidence effects and terrain barriers that intercept moisture transport from the coast.

5.1.2 Climate Change Signatures and Comparison

The universal upward trends (4.00-7.99 mm/year) across all 15 sub-regions align with the global “wet-gets-wetter” paradigm documented in IPCC reports [1] and global analyses [4]. Our finding of a “Coastal Enhancement” pattern (average trend 7.45 mm/year) is consistent with studies in the Yangtze River Delta [22] and Zhejiang Province [25], which link enhanced coastal precipitation to warming sea surface temperatures and intensified moisture flux from the East China Sea.

Significantly, we observed a post-2010 intensification of rainfall variability, with record highs occurring in 2021 (e.g., ~3250 mm in Mazhu). This aligns with findings by Su et al. [21] regarding increased extremes in the Yangtze Basin. The high Hurst exponents ($H>0.5$) for all regions provide novel evidence that these upward trends are not temporary fluctuations but persistent climate shifts.

5.1.3 Implications for Engineering Operations

The findings have direct implications for the Eastern Zhejiang Water Diversion Project. A notable source-receiver mismatch exists: the primary water source area (Mazhu) has the highest rainfall but also the highest variability, while the key recipient areas (Nansha, Shaoyu) have the lowest baseline rainfall but are experiencing the fastest increasing trends. This dynamic requires flexible allocation strategies. Regarding scale-dependent management, the increase in spatial correlation from 0.83 (3-month) to 0.87 (12-month) suggests that strategic planning can be coordinated regionally based on annual climate coherence. However, tactical operations for flood control must remain independent at the sub-regional level due to lower short-term correlations and high local variability.

5.2 Conclusions

Based on the systematic analysis of 62 years of daily rainfall data (1961-2022), this study draws five main conclusions. First, regarding spatial pattern, rainfall follows a distinct “high-in-center, low-in-west” distribution, with the Central zone (Mazhu Midstream) receives 64% more rainfall than the Western zone (Nansha), driven by river confluence effects and local topography. Second, in terms of universal increase, all 15 sub-regions show statistically significant upward trends ($p<0.05$), with trend magnitudes exhibiting a clear spatial gradient: Coastal (7.45 mm/yr) > River Basin (6.38 mm/yr) > Hilly (4.91 mm/yr). Third, concerning trend persistence, Hurst exponent analysis yields values between 0.56 and 0.63 for all regions, confirming that the observed upward trends are persistent and likely to continue in the near future. Fourth, with respect to scale dependence, rainfall characteristics are highly scale-dependent; as analysis scales extend from 3 to 12 months, trend fluctuation amplitude decreases by approximately 90% (from ±16 mm to ±1.7 mm), and inter-regional correlation improves by 5-11%, indicating that long-term patterns are controlled by regional climate factors while short-term patterns are dominated by local weather noise. Fifth, regarding the correlation anomaly, the Mazhu Midstream Area exhibits a unique “increase-then-decrease” correlation pattern, reflecting its distinctive hydro-geographic behavior as a semi-independent confluence zone.

5.3 Recommendations

Based on the research findings, the following recommendations are proposed, summarized in Table 5.

Table 5. Recommendations for Water Diversion Project Management

References

[1] IPCC, 2021. Climate Change 2021: The Physical Science Basis.

[2] Seneviratne et al., 2012. Changes in climate extremes and their impacts.

[3] Kundzewicz et al., 2008. Implications of projected climate change for freshwater resources.

[4] Westra et al., 2013. Global increasing trends in annual maximum daily precipitation.

[5] Donat et al., 2016. More extreme precipitation in the world's dry and wet regions.

[6] Mann, H.B., 1945. Nonparametric tests against trend.

[7] Kendall, M.G., 1975. Rank Correlation Methods.

[8] Yue et al., 2002. Power of the Mann–Kendall and Spearman's rho tests.

[9] Hamed, K.H., 2008. Trend detection in hydrologic data.

[10] Burn & Elnur, 2002. Detection of hydrologic trends and variability.

[11] Sen, P.K., 1968. Estimates of the regression coefficient.

[12] Hurst, H.E., 1951. Long-term storage capacity of reservoirs.

[13] Alexander et al., 2006. Global observed changes in daily climate extremes.

[14] Zhang et al., 2011. Indices for monitoring changes in extremes.

[15] Zhai et al., 2005. Trends in total precipitation and frequency of daily precipitation extremes over China.

[16] Zhai et al., 1999. Changes of climate extremes in China.

[17] Qian & Lin, 2005. Regional trends in recent precipitation indices in China.

[18] Wang & Zhou, 2005. Observed trends in extreme precipitation events in China.

[19] You et al., 2011. Changes in daily climate extremes in China.

[20] Jiang et al., 2007. Temporal and spatial trends of precipitation in Yangtze River basin.

[21] Su et al., 2006. Recent trends in observed temperature and precipitation extremes.

[22] Wang et al., 2020. Innovative trend analysis of rainfall in Yangtze River Delta. Water.

[23] Li et al., 2020. Climate variability of rainfall in Yangtze River estuary. Water.

[24] Zhao et al., 2022. Projection of climate change impacts on runoff. Water.

[25] Chen et al., 2015. Spatial and temporal trends of extreme precipitation over Zhejiang.

[26] Xu et al., 2010. Temporal trend of precipitation and runoff in major Chinese rivers.

[27] Gao & Xie, 2014. Trends and physical causes of extreme precipitation in China.

[R1] Madakumbura, G.D., et al. (2021). Anthropogenic influence on extreme precipitation over global land areas seen in records of precipitation intensity. Nature Communications, 12, 3944.

[R2] Kotz, M., et al. (2022). The effect of rainfall changes on economic production. Nature, 601, 223–227.

[R3] Zhou, B., et al. (2023). Projected changes in the East Asian summer monsoon and associated precipitation under the 1.5°C and 2°C warming targets. International Journal of Climatology, 43(2), 987-1002.

[R4] Li, X., et al. (2022). Spatiotemporal variability of extreme precipitation in Zhejiang Province, China, and its teleconnection with SST. Theoretical and Applied Climatology, 148, 123–138.

[R5] Zhang, Y., et al. (2024). Urbanization effects on precipitation extremes in the Yangtze River Delta: A multi-scale analysis. Urban Climate, 53, 101822.

[R6] Wang, F., et al. (2022). Long-term persistence of hydrologic series in the Yellow River Basin using an improved Hurst exponent. Journal of Hydrology, 610, 127956.

[R7] Wu, Z., et al. (2023). Multi-scale entropy analysis of rainfall complexity in changing environments. Water Resources Research, 59(4), e2022WR033456.

[R8] Zhao, T., et al. (2021). Vulnerability of inter-basin water transfer systems to synchronous drought: A case study of the South-to-North Water Diversion Project. Journal of Environmental Management, 293, 112847.

[R9] Chen, H., et al. (2022). Optimization of water diversion scheduling based on uncertainty analysis. Water, 14(3), 452.

[R10] Liu, J., et al. (2023). Climate-adaptive scheduling for large-scale water infrastructure: Framework and application. Engineering, 25, 112-125.