Dynamics of the Net Precipitation in China from 2001 to 2020

Abstract

Net precipitation (NP) is the primary source of soil water essential for the functioning of vegetated ecosystems. By quantifying NP as the difference between gross precipitation and canopy interception evaporation, this study examined the dynamics of NP in China from 2001 to 2020 and the contribution of environmental factors to NP variations was investigated. The findings revealed a multiyear mean NP of 674.62 mm, showcasing a 2.93 mm/yr increase. The spatiotemporal variations in NP were mainly attributed to a remarkable increase in precipitation rather than canopy interception. Notably, climate (temperature, wind speed, surface solar radiation downward and vapor pressure deficit) and vegetation factors (leaf area index and net primary productivity) played a dominant role in NP in 61.53% and 15.39% of China, respectively. The dominant factors contributing to NP changes were vapor pressure deficit (mean contribution rate: −43.68%), temperature (mean contribution rate: 11.69%), and leaf area index (mean contribution rate: 2.13%). The vapor pressure deficit negatively exerts a negative influence on the southern and eastern regions. Temperature and leaf area index have the greatest effect on the northeastern and southwestern regions, respectively. The results provide valuable insights into the pivotal role of climatic and vegetation factors in ecohydrological cycles.

1. Introduction

When precipitation falls on a vegetated surface, a portion is intercepted by the canopy and evaporates directly back into the atmosphere (interception) [1,2,3]. The remaining water reaches the ground via two distinct pathways: throughfall (Tf), which occurs when water passes through the canopy gaps and drips off, and stemflow (Sf), which refers to the flow of water drained to an individual plant’s stem base [4,5]. The effective amount of water that ultimately reaches the ground is termed net precipitation (NP), representing the sum of Tf and Sf [5,6]. The redistribution of rainfall at the ground surface gives rise to variability in the amount and spatiotemporal pattern of NP observations at the surface when compared to precipitation that falls on the top of the canopy. Generally, NP is less than gross precipitation above the canopy, except during cloud or fog events [7], due to water storage and evaporation [5]. The redistribution process enhances the spatiotemporal variation in NP by altering the horizontal and vertical distribution of water [4,8,9]. It affects soil moisture patterns [10], nutrient budgets [11], and localized preferential flow [12] within vegetated ecosystems. Therefore, comprehending the characteristics of NP is crucial for advancing our understanding of hydrologic and biogeochemical processes [6].

The measurement of NP has typically been conducted by directly measuring Tf and Sf [2,5]. These measurements are typically taken at specific points during both liquid and frozen precipitation events [13,14,15], using capture devices designed for this purpose [16,17]. The overwhelming majority of NP studies have been conducted at the field plot scale and focused on individual plant species or mixed species at specific sites [18,19,20]. These studies have greatly improved our understanding of the fine or local scale of rainfall partitioning processes and contribute a wealth of data that can be used for comprehensive synthesis at larger scales. Particularly in recent years, some studies have synthesized and analyzed observations of Tf and/or Sf from independent publications in order to quantify the patterns of NP by vegetation at regional or global scales [4,5,9,18,21]. Previous studies have indicated that the median levels of relative Tf and Sf globally are approximately 75% and 3% of total rainfall, respectively [5,9]. Although these syntheses provided information about the general pattern at a large scale, the number of observations at certain sites is insufficient, resulting in a gap in data for NP in low vegetation and mixed precipitation events [22,23,24].

A multitude of studies have demonstrated that NP is significantly influenced by meteorological conditions and vegetation characteristics. In particular, precipitation exerts a direct influence on NP, as storm magnitude determines the canopy storage capacity, while storm intensity can influence the time to saturation [2,19]. The stand-scale NP shows marked variation in relation to the characteristics of the dominant and understory vegetation [25]. NP was found to decrease in leafless conditions [26] and to be strongly correlated with leaf area index (LAI) for herbaceous plants [27,28]. Consequently, alterations in the states of leaves throughout the seasons and associated fluctuations in precipitation conditions can also exert a considerable influence on NP [29]. Furthermore, other factors, such as temperature, sunshine duration, humidity, and wind speed, also exert some influence on the variability of NP to some extent. The majority of current studies focus on the fine-scale drivers that shape highly localized patterns [5]. Nevertheless, the aforementioned studies did not perform a quantitative analysis of the impact and relative importance of potential drivers, which limits the accuracy of large-scale predictions under changing climate conditions.

China, endowed with extensive and diverse topography and climate types, harbors a rich and intricately complex variety of vegetation characterized by uneven precipitation distribution [30]. The phenomena observed in China, including greening [31,32], increasing precipitation variability [33,34], and shifts in land use [24,35,36], significantly influence the stages of precipitation redistribution. Although many field experiments are conducted at the site level or in typical ecosystems [18,37,38,39], there is a lack of large-scale maps of NP in China and qualifications of the contribution of climate and vegetation factor influence on NP over past two decades. Our study leveraged large-scale evaporation of canopy interception (EI) products and models, enabling us to indirectly map NP as the difference between gross precipitation and EI. This approach is made possible by the Penman–Monteith–Leuning (PML) model [40,41,42], which has been further refined by Zhang et al. [43]. By employing this model, we aim to explore NP patterns in China and investigate the macroscale role of precipitation–vegetation interactions on NP over a long-term period. The main objectives of this study are to (1) characterize the spatiotemporal dynamics across China from 2001 to 2020; (2) quantify the contributions of climate and vegetation factors to determine the most influential factors on spatiotemporal changes in NP. Our study offers a new perspective and a tool that can be used alongside field measurements to enhance our knowledge of water balance dynamics, contributing valuable insights into the effects of China’s greening and climate change on water and energy exchange.

2. Materials and Methods

2.1. Data Collection

EI (mm) data used to calculate NP were sub datasets from the global PML-V2 Evapotranspiration (ET) datasets, which had an 8-day temporal resolution and a spatial resolution of 500 m. Zhang et al. [44] included EI to estimate the three components of ET across the globe and Zhang et al. [43] extensively calibrated the PML-V2 model using data from 95 flux towers for 10 plant functional types worldwide. The PML-V2 EI product for China from 2001 to 2020 was obtained from the Google Earth Engine (GEE) platform (https://developers.google.com/earth-engine/datasets/catalog/CAS_IGSNRR_PML_V2_v017, accessed on 17 June 2023).

The meteorological data, including gross precipitation (PRE, mm), air temperature (TEM, °C), wind speed (WIN, m/s) and relative humidity (RH, %), were obtained from the China National Ground Meteorological Station Basic Meteorological Elements Daily Values Dataset (V3.0) (http://data.cma.cn/, accessed on 21 July 2023). Annual cumulative values were calculated for PRE, and annual mean values were calculated for TEM and WIN. These annual values were then interpolated into gridded data using the ANUSPLIN 4.37 software for improved spatial analysis capabilities. Yearly surface solar radiation downward (SSRD, W/m²) data were derived from the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5-Land reanalysis dataset (https://www.ecmwf.int/en/era5-land, accessed on 23 July 2023). Vapor Pressure Deficit (VPD, kPa) was calculated using a specific equation.

$$\mathrm{VPD}=0.61078\times {\mathrm{e}}^{\frac{17.27\times \mathrm{TEM}}{\mathrm{TEM}+237.3}}\times \left(1-\mathrm{RH}\right)$$

(1)

Vegetation factors including LAI and net primary productivity (NPP) data were obtained from MODIS products (MOD15A2H and MOD17A3) with 500 m and 8-day resolution, processed into annual datasets using the GEE platform.

The physical geographic division data of China, including the vegetation regionalization map, moisture regions, and land-use types, were obtained from the Resource and Environment Science and Data Center (https://www.resdc.cn, accessed on 21 June 2023). Overall, China was divided into the following eight vegetation regions: I (cold temperate coniferous forest), II (temperate mixed coniferous and broad-leaved forest), III (warm temperate deciduous broad-leaved forest), IV (sub-tropical evergreen broad-leaved forest), V (tropical monsoon forest), VI (temperate grassland), VII (temperate desert), and VIII (alpine vegetation). The physical geographic division is presented in Figure 1a. Furthermore, the land use in China was classified into six categories, as illustrated in Figure 1b.

The time series for all datasets spanned from 2001 to 2020, covering China. For subsequent analysis, all data were processed using a spatial resolution of approximately 1 km and projected to the same coordinate system (GCS_WGS_1984). This spatial resolution was determined based on the coarsest spatial resolution among all datasets.

2.2. Methodology

2.2.1. Calculations of NP

In accordance with the process of rainfall partitioning by vegetation, NP was calculated as the difference between annual PRE and annual EI [1,38,45]. The estimation of NP at large scales was calculated as follows:

$$\mathrm{NP}=\mathrm{PRE}-\mathrm{EI}=\mathrm{Tf}+\mathrm{Sf}$$

(2)

2.2.2. Trend Analysis

The temporal trends of NP were determined using linear regression at the national, regional, and pixel levels. The trends were calculated as follows:

$${\mathrm{TR}}_{\mathrm{NP}}=\frac{\mathrm{n}\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{ix}}_{\mathrm{i}}-\sum _{\mathrm{i}=1}^{\mathrm{n}}\mathrm{i}\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{x}}_{\mathrm{i}}}{\mathrm{n}\sum _{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{i}}^2-{\left(\sum _{\mathrm{i}=1}^{\mathrm{n}}\mathrm{i}\right)}^2}$$

(3)

where n is the length of research years, x_i is the ith year of NP, and TR_NP is the long-term NP trend in China from 2001 to 2020. TR_NP > 0 indicates an increasing trend, while TR_NP < 0 indicates a decreasing trend. Moreover, the F-test was conducted to test the significance of slope in this study, where 0.01 ≤ p < 0.05 stands for statistically significant, p < 0.01 stands for highly significant and p ≥ 0.05 stands for not significant.

2.2.3. Contribution Analysis

The contributions of PRE and EI to the NP trends were quantified using a differential equation [46,47]. PRE represented the influence of the precipitation regime, whereas EI signified the impact of water interception by plants on NP. As NP varied with time, the long-time trend of NP was calculated as follows:

$$\frac{\mathrm{dNP}}{\mathrm{dt}}=\frac{\partial \mathrm{NP}}{\partial {\mathrm{PRE}}_{\mathrm{i}}}\frac{\mathrm{d}{\mathrm{PRE}}_{\mathrm{i}}}{\mathrm{dt}}-\frac{\partial \mathrm{NP}}{\partial {\mathrm{EI}}_{\mathrm{i}}}\frac{\mathrm{d}{\mathrm{EI}}_{\mathrm{i}}}{\mathrm{dt}}+\mathsf{\epsilon }$$

(4)

Equation (4) could also be rewritten as follows:

$${\mathrm{TR}}_{\mathrm{NP}}=\mathrm{C}\left(\mathrm{PRE}\right)+\mathrm{C}\left(\mathrm{EI}\right)+\mathsf{\epsilon }$$

(5)

where C(PRE) and C(EI) are the contributions of PRE and EI to the long-time trend of NP, respectively. ε is the system error.

The relative contribution rate of PRE (CR(PRE)) and EI (CR(EI)) were calculated as follows:

$$\mathrm{CR}\left(\mathrm{PRE}\right)=\frac{\mathrm{C}\left(\mathrm{PRE}\right)}{\mathrm{C}\left(\mathrm{PRE}\right)+\mathrm{C}\left(\mathrm{EI}\right)}\times 100\%$$

(6)

$$\mathrm{CR}\left(\mathrm{EI}\right)=\frac{\mathrm{C}\left(\mathrm{EI}\right)}{\mathrm{C}\left(\mathrm{PRE}\right)+\mathrm{C}\left(\mathrm{EI}\right)}\times 100\%$$

(7)

Furthermore, to separate the individual contributions of climatic factors (TEM, WIN, SSRD and VPD) and vegetation factors (LAI and NPP), the Elastic Net Regression model was employed at the pixel grid. The model utilized the annual average values of these variables from 2001 to 2020 for each raster cell, thereby ensuring the precision and reliability of the outcomes. The model exhibited robustness against multicollinearity and facilitated variable selection [48,49]. The model can be expressed as follows:

$$\underset{\left\{{\mathsf{\beta }}_0,{\mathsf{\beta }}_1,{\mathsf{\beta }}_2,\dots ,{\mathsf{\beta }}_{\mathrm{p}}\right\}}{\min }\left[{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\left({\mathrm{Y}}_{\mathrm{i}}-{\mathsf{\beta }}_0-{\sum }_{\mathrm{j}=1}^{\mathrm{p}}{\mathsf{\beta }}_{\mathrm{j}}{\mathrm{X}}_{\mathrm{i}\mathrm{j}}\right)}^2-\mathsf{\lambda }\left(\mathsf{\alpha }{\sum }_{\mathrm{j}=1}^{\mathrm{p}}\left|{\mathsf{\beta }}_{\mathrm{j}}\right|+\frac{1-\mathsf{\alpha }}{2}{\sum }_{\mathrm{j}=1}^{\mathrm{p}}{\mathsf{\beta }}_{\mathrm{j}}^2\right)\right]$$

(8)

where i = 1, 2, … n (a total of n grids); j = 1, 2 … p (p = 6, means six independent variables); X was the standardized value of independent variables; Y was equal to the standardized value of NP; β was the regression coefficient and β0 was the intercept coefficient; α was the mixing parameter, which determined the balance between L1(α = 0) and L2(α = 1) regularization. In this study, α = 0.5 was chosen to ensure effective contributor selection and strong robustness. λ was the hyperparameter that controlled the strength of regularization and the optimal value of λ depended on the performance of cross-validation.

The coefficients obtained (β₁, β₂, …, β_p) were used to build the regression model to assess the contributions of independent variables to the dependent variable. The individual driver’s relative contribution rate was expressed accordingly, with the positive and negative signs determined by the original coefficients. The R package glmnet was used to complete the build the regression model.

$$\mathrm{CR}\left({\mathrm{X}}_{\mathrm{i}}\right)=\frac{{\mathsf{\beta }}_{\mathrm{j}}}{\sum _{\mathrm{j}=1}^{\mathrm{p}}{|\mathsf{\beta }}_{\mathrm{j}}|}$$

(9)

Path analysis was employed in this study since it could effectively decompose the correlation between the independent and dependent variables in cases where there were complex interactions between the independent variables or the independent variable indirectly affected the independent variable through other independent variables [35]. The magnitude of the path coefficient indicated the impact of the independent variable on the dependent variable, with its sign denoting positivity or negativity. The R package lavaan conducted the path analysis in this study.

3. Results

3.1. Dynamics of NP in China from 2001 to 2020

3.1.1. Spatiotemporal Variation in Mean NP

The multiyear mean NP (Figure 2a) from 2001 to 2020 exhibited a significant gradient, decreasing from the southeast to the northwest in China. The national mean value of multiyear annual NP was 674.62 mm, with a range of 3.23 mm to 4694.25 mm (Figure 2a). Regions with high NP values (>1000 mm) were concentrated in the southeastern parts of China, particularly along the coastal regions. In contrast, arid and lowland terrains in northeastern China had low NP values (<200 mm). At the regional level, region V had the highest annual NP with a mean value of 1376.47 mm, followed by region IV with a mean value of 1115.30 mm (Figure 2c). On the other hand, region VII had the lowest annual NP with only 227.95 mm (Figure 2c). The remaining regions had annual mean NP values ranging from 344.82 mm to 603.56 mm (Figure 2c).

On a seasonal scale, NP exceeds 200 mm in summer in most semi-humid and humid regions, with some southwestern humid regions also reaching this threshold in spring and autumn (Figure 2b). In contrast, arid and semi-arid regions have NP values consistently below 100 mm per year (Figure 2b). The multi-year seasonal mean NP in China peaks in summer (346.56 mm), followed by spring (154.18 mm), autumn (134.24 mm) and is lowest in winter (42.67 mm) (Figure 2c).

Figure 3a shows a contrasting spatial pattern in the percentage of NP/PRE. The NP/PRE ratio was low in humid and semi-humid regions and reached its lowest point in the northeast region and differed across seasons. The percentage of NP/PRE across China fluctuated between 30.85% and 100%, with an average value of 94.89% (Figure 3c). Region I had the lowest percentage of NP/PRE at 85.58%, while the remaining seven regions had higher values, ranging from 92.54% to 97.93%. At the seasonal scale (Figure 3b), the region with the greatest canopy interpretation differed from the other regions. In spring, the Yunan and Hainan provinces exhibited the lowest NP/PRE. In summer, region I and the majority of region IV exhibited low NP/PRE values. In autumn, the southern regions of region IV exhibited low NP/PRE values. In winter, the lowest NP/PRE values were observed in regions IV and V, particularly in Yunnan and the southern coastal cities.

3.1.2. Spatiotemporal Trend of NP

The annual mean NP in China exhibited a significant upward trend of 2.93 mm/yr (p < 0.05) with regional variations (Figure 4). Notably, only region V was characterized by a decreasing trend (−3.91 mm/yr), while seven other regions displayed an increasing trend in annual mean NP. These upward trends ranged from 0.48 (region VIII) to 7.35 mm/yr (region II). In seasonal terms (Figure 4), summer and autumn exhibited a highly significant (p < 0.01) and significant (p < 0.05) increasing trend in mean NP values, respectively, at 2.24 mm/yr and 1.43 mm/yr. Conversely, spring and winter displayed a non-significant downward trend in seasonal mean NP. The change in annual NP indicated that 75.96% of China had experienced an increase, while 24.04% had witnessed a decrease (Figure 5). Within this context, 3.63% of the whole area of China demonstrated a highly significant increase (p < 0.01), particularly in the Sichuan basin and the mountain range in the northeastern section. Conversely, the decreasing trend was mainly concentrated in China’s southern border, spanning from southeastern Tibet and southern Yunnan to southwestern Guangxi, predominantly south of 22°30′N latitude.

3.2. Potential Drivers of Spatiotemporal Variation in NP

3.2.1. Contribution of PRE and EI in Controlling NP Trend

From 2001 to 2020, both PRE and EI exhibited rising trends at rates of 3.39 mm/yr and 1.63 mm/yr, respectively (Figure 6). Analysis of their contributions to the overall NP trends revealed that PRE provides significant and positive contributions, while EI played a secondary role, contributing negatively at 32.45% (Figure 6). On a national scale (Figure 7), NP trends in the majority of regions (72.95%) were dominated by PRE. Only a small percentage of regions were primarily influenced by EI, concentrated in the Hanzhong Basin in region IV, the North China Plain in region III, and the Dzungarian Basin in region VII. The NP trend in these regions was dominated by EI, resulting in patterns resembling meandering curves and circles along mountain ranges due to the influence of elevation and climate. At the regional level (Figure 6), PRE accounted for more than 60% of the contributions. Specifically, values in regions VIII and V were over 70%, with the highest value reaching 82.52% in region I. Similar contributions of PRE and EI to the NP trend were only observed in region VII.

3.2.2. Climate and Vegetation Potential Drivers in NP Variation

The Elastic Net Regression analysis revealed substantial spatial variability in the contributions of various factors to NP at the national scale (Figure 8). TEM (Figure 8c) was identified as the principal positive driver, contributing to NP changes in China with a mean contribution rate of 11.69% (Figure 9), and exerted a predominant role in 15.04% of the country’s area (Figure 10). It is noteworthy that in regions III and VI, its contribution rate exceeded 20% (Figure 9).

In contrast, VPD (Figure 8f) was the largest negative driver, contributing to NP with a mean rate of −43.68% (Figure 9). VPD had a dominant influence on NP change in 55.39% of China (Figure 10), particularly in most of southern and eastern China. The mean contribution rate of VPD reached up to −50% in regions I and II (Figure 9).

Furthermore, WIN and SSRD exhibited a negative impact on NP, with a mean contribution rate of −4.27% and −6.67%, respectively (Figure 9). The negative impacts of SSRD (Figure 8e) were concentrated in Yunnan, Hebei, and the Junggar Basin, while those of WIN (Figure 8d) were observed in the Hanzhong Basin and southeastern Tibet. The positive influences were observed in specific regions only. SSRD exhibited a positive mean contribution rate in regions I and VIII, while WIN demonstrated a positive mean contribution rate in region VII (Figure 9).

The vegetation factors LAI and NPP exerted both positive and negative influences on NP, with mean contribution rates of 2.13% and −0.98%, respectively (Figure 9). The positive impacts of LAI (Figure 8a) were concentrated in southwestern China, whereas NPP (Figure 8b) was observed in northeastern China. These vegetation factors exerted a predominant influence on 11.07% and 6.77% of China, respectively (Figure 10).

With regard to the findings of the path analysis (Figure 11), it was found that SSRD had the largest direct negative impact in China, while NPP exhibited the highest direct positive impact. Among the six factors, TEM and WIN had similar direct negative effects on NP, whereas LAI and NPP demonstrated strong positive associations. At the regional scale, NPP exhibited the highest positive direct effect on regions I, III, and VII. The largest negative direct effect was demonstrated by SSRD on regions I, II, III, IV, VI, VII, and VIII. VPD exhibited the highest positive direct impact in region II, while in region V, it had a large negative direct impact. Additionally, LAI demonstrated a high negative direct effect on regions I, VII, and VIII, and a high positive direct effect on region VI. The greatest positive impact of TEM was observed in region V, while the greatest negative impact was observed in regions II and VIII.

In the context of the combined effects of climate and vegetation factors on NP variation (Figure 12), climate changes were identified as the primary driver of changes in NP in approximately 61.53% of China. Within this category, 55.41% of regions exhibited negative changes, while 6.12% showed positive changes. The area percentage changes in NP, induced by vegetation factors and co-factors, were relatively similar. They accounted for 15.39% and 14.32%, respectively, and included both negative and positive impacts. Both climate and vegetation factors contribute to a higher area percentage of negative effects than positive effects on NP. In certain regions, climate factors were the main cause of the decrease in NP, while vegetation factors played a leading role in influencing both the decrease and increase in NP changes. Regions experiencing the joint influence of both climate and vegetation factors were scattered in northeastern parts of China.

4. Discussion

Between 2001 and 2020 in China, we utilized the canopy water balance equation to indirectly measure the difference between PRE and EI, deriving NP. The PRE data were obtained from observational stations in China, while the sub-datasets of EI were sourced from PML ET datasets [43,50]. The performance of the PML model and datasets was validated using data from 95 flux towers globally, resulting in a Root Mean Square Error of 0.69 mm/day [43]. In this study, we decided not to use PRE data in the PML-V2 model due to concerns about the extension of the study area and data accuracy in China. Despite both data sources undergoing quality control [34,50], the inconsistency might still introduce uncertainty into NP estimation.

The multiyear annual mean NP was 674.62 mm (annual mean NP/PRE: 94.89%). However, comparative analyses with existing research show disparities. This value was found to be considerably higher than that observed in field experiments that were limited to specific locations and periods following the rainfall event. In China, Zhang et al. [39] found median Tf and Sf values of 74.3% and 2.7%, respectively. On a broader scale, Sadeghi et al. [5], based on 644 global observations, reported median relative Tf and Sf values of 75.8% and 2.2%. The values were derived from a quantitative synthesis of peer-reviewed papers based on site observation. Consequently, this value differs from previous studies, potentially due to scale gaps in time and space [39,45,48]. In our study, the mean NP/PRE ratio was employed as a means of reflecting the actual amount of precipitation that reached the ground over the course of the entire year. Additionally, the tendency for ecosystem-level Tf measurements to exceed those at the single/tree level may contribute to this discrepancy [9,51]. Nevertheless, when the value was compared with the EI interception rates obtained from a model based on remote sensing data over a long-term period at a global scale, the value was within the normal range. For instance, the EI rate obtained from a physics-informed hybrid machine learning was 8.6% [52]. He et al. [53] employed a Revised Gash Model, reporting an average annual ratio of EI to PRE of 22.30% in China. The EI interception rates of different grassland types in the Qinghai-Tibet Plateau are as follows: alpine meadows, 3.11%; mountain meadows, 3.82%; temperate grassland, 2.45%; and temperate desert grassland, 1.13% [54]. Large-scale studies on EI suggest that the annual mean NP/PRE ranges from 70% to 99%.

The spatial pattern of NP was primarily influenced by the spatial distribution of precipitation and vegetation types [5,6]. Under the monsoon climate, precipitation decreased from southeast to northwest [55,56], which aligned with the distribution of NP. The seasonal NP in summer accounts for over 50% of the annual NP as the precipitation levels were significantly elevated. Furthermore, seasonal fluctuations in rainfall characteristics are accompanied by changes in tree canopy characteristics [40]. Consequently, the seasonal variability in precipitation might be as significant as the seasonal variability in leaf state. Vegetation types changed from tropical or subtropical forest to temperate steppe or temperate desert (Figure 1a), characterized by a sparser canopy structure that retained less rainwater [17,39]. In contrast, the spatial pattern of the NP/PRE ratio was found to be relatively low in the humid and semi-humid regions, with the lowest values in the northeast (Figure 3a). The annual NP/PRE ratio in this study reflected the influence of seasonal variability in vegetation and precipitation on the capacity of the canopy to intercept precipitation to some extent. The southern region generally had a higher comprehensive interception capacity of vegetation compared to the north [39]. The northern regions of VI, VII, and VIII in China were arid and semiarid areas with low NP values. These areas were dominated by shrubs, alpine, and desert vegetation types, and exhibited low vegetation coverage [39]. Exceptions in northern China include region I, which was a cold temperate coniferous forest, with low NP and NP/PRE values. The low amount of NP and the high interception ratio were attributed to the higher total plant leaf area of coniferous plants due to their needle leaf form [9,53]. According to Magliano et al. [4], the interception ratio in needle leaf form was higher than in broadleaf or pinnate leaf forms.

At the temporal scale, the annual mean NP, PRE, and EI in China exhibited a slight upward trend of 2.93 mm/yr, 3.39 mm/yr, and 1.63 mm/yr, respectively. This increase was consistent with the findings that interception, Tf, and Sf increase with rainfall event size [57]. The upward trend rate of EI was lower than that reported by He et al. [53] but higher than that reported by Yan et al. [58]. Although studies have reported an increase in precipitation amount and variability [33,34,55], 25% of regions have exhibited a decreasing trend in annual NP, primarily concentrated in southwestern China. At the regional level, only region V showed a decreasing trend in NP, potentially linked to a decrease in monsoon precipitation in southwest China [59], coupled with an increase in evapotranspiration in these regions [60].

In situ studies have shown that the amounts of NP depend on vegetation attributes, rainfall regime characteristics, and meteorological conditions [5,6,52]. Our findings indicate that, in most regions of China, it is PRE, rather than EI, that governs the interannual variation of NP, explaining approximately 68% of the NP trend. It was generally agreed that precipitation characteristics, represented by PRE, predominantly determine NP [57,61]. However, it overlooks other aspects of rainfall characteristics, such as duration, intensity, variation, and frequency [52]. Future research should incorporate these additional characteristics of rainfall events to enhance the understanding of the impact of changes in rainfall regime on NP, especially in regions with distinctive characteristics like dryland and low-vegetated areas in northwestern China [57,62].

To delve further into NP changes, we examined six additional climate and vegetation factors. Many previous studies have highlighted the pivotal role of VPD in regulating the trade-off between carbon and water fluxes [63,64]. During rainfall, VPD affected EI by influencing the difference in vapor pressure between wet surfaces and surrounding air. High VPD and dry air conditions tended to accelerate evaporation, whereas low VPD and humid conditions might have slowed down the process [42,65]. TEM influences NP through its regulation of PRE. As global temperatures rise, precipitation amount and variability continue to increase [33,34]. At the national level, TEM had a positive impact on NP in China, particularly in northeastern regions, the Sichuan basin, and Huang-Huai-Hai Plain. These regions are affected by the East Asian monsoon, which is characterized by the influx of moist air from the Pacific Ocean during the summer [66,67]. This leads to increased evaporation and moisture, which in turn increases the likelihood of rainfall [68]. SSRD and WIN influence rainfall and subsequent processes of interception loss and evaporation. However, their contributions to NP in China at the yearly time scale were relatively modest in our study. In contrast to previous studies that emphasized the sensitivity of NP to wind magnitude at fine scales [5,6], the annual average WIN would alleviate the impact of extremely high wind speeds on NP. Additionally, the yearly averaged SSRD might temper the distinct seasonal and daily patterns in the relationship between solar radiation and evaporation [58,60].

To evaluate the influence of vegetation dynamics on NP changes, we identified LAI and NPP as key indicators. The national average contribution rate of NPP and LAI indicated that vegetation dynamics had a weak impact on NP changes in China. Previous studies have demonstrated that the characteristics of the vegetation community exert a significant influence on a small scale [7,27,69]. However, at a larger scale, meteorological factors such as precipitation characteristics, wind speed, and evaporation emerge as the dominant drivers [70]. Excluding the scale gap, the saturation phenomenon of the remote sensing-based vegetation index also leads to inaccuracies in estimating the vegetation growth status in densely vegetated areas [71,72]. Additionally, site-based studies underscore the importance of various canopy structure variables, including leaf shape, canopy thickness, size, and forms, in influencing rainfall redistribution processes [9,73,74]. Neglecting these structural properties of vegetation could potentially limit the accuracy of contributors for regional NP [17,39,73]. Therefore, future studies need to thoughtfully select vegetation variables characterizing biophysical properties, considering both data availability and accuracy.

5. Conclusions

This study investigated the dynamics of NP for different vegetation-type regions across China during 2001–2020 and analyzed the contribution of climate change and vegetation dynamics to the interannual variation in NP. The canopy water balance equation was employed to indirectly ascertain the discrepancy between PRE and EI, thereby deriving NP. The spatial distribution of NP exhibited a decreasing trend along the moisture gradient from southeast to northwest, in a manner analogous to that observed for precipitation across China. The NP/PRE ratio was found to be relatively low in the humid and semi-humid regions, exhibiting seasonal variability. This reflected the influence of seasonal variability in vegetation and precipitation on the capacity of the canopy to intercept precipitation. In general, the trend of NP has been on the rise over the past two decades, with an average change rate of 2.93 mm per year in China. This is due to a significant increase in precipitation during the summer and autumn seasons.

The contribution analysis revealed an increase in the trend of NP, with the spatiotemporal variations in NP being primarily attributed to changes in precipitation rather than EI. Notably, climate factors played a more dominant role in NP variation than vegetation factors in the majority of regions across China. The most significant factors influencing NP changes were VPD, TEM and LAI. The VPD exerted a negative influence on the southern and eastern regions. TEM and LAI exerted the most influence on the northeastern and southwestern regions, respectively. In conclusion, the results provide valuable insights into the pivotal role of climatic and vegetation factors in ecohydrological cycles. Further research should investigate the influence of vegetation characteristics, rainfall patterns and seasonality on the redistribution of rainfall.