Spatiotemporal Heterogeneity of Long-Term Irrigation Effects on Drought in China’s Arid and Humid Regions

Abstract

Analyzing the spatiotemporal characteristics of meteorological droughts (MD) and agricultural droughts (AD) and their propagation in different climate zones is important for effective drought management, climate adaptation, and food security. This study takes a unique approach by comparing irrigated and rainfed croplands. A comprehensive framework is developed using drought indices, statistical analysis, trend tests, and wavelet transforms. The spatiotemporal evolution patterns, trends, and correlations of MD and AD in Xinjiang and the Middle-lower Yangtze Plain (MYP) are investigated. The main results showed that severe MD events (e.g., Xinjiang 2005–2009 and MYP 2004–2009) significantly impacted rainfed agricultural systems, leading to a decline in vegetation condition. Long-term irrigation can substantially alleviate AD under MD conditions. From 2000 to 2019, AD on irrigated croplands in Xinjiang continuously improved, while rainfed croplands deteriorated significantly during MD events. In contrast, although overall AD in MYP was mitigated, the benefits of irrigation were only evident during severe AD periods and weakened after 2013. Correlation and wavelet analyses revealed different drought propagation mechanisms between irrigated and rainfed croplands, highlighting the key role of local climate conditions and spatial heterogeneity in determining irrigation efficiency. The findings provide important guidance for optimizing drought management strategies, agricultural planning, and sustainable water resource management.

1. Introduction

In recent decades, the intensity and frequency of drought events have increased with global climate change, posing a serious threat to agricultural production and food security [1,2,3,4,5,6,7,8]. Drought is generally defined as a prolonged period of insufficient precipitation or a clear shortage of rainfall, characterized by long duration and wide impact [9,10,11,12,13]. Meteorological droughts (MD), the most common form, refer to an imbalance between water gain and loss caused by mismatches between evaporation and precipitation, and it often leads other types of droughts [14,15,16,17,18]. Agricultural droughts (AD) occur when soil moisture fails to meet the needs of plant growth, leading to water stress in crops [19,20,21]. In China, a major agricultural country, losses from AD account for over 50% of natural disaster losses in agriculture and have increased over the past 20 years, seriously affecting food security [22,23,24].

Current MD monitoring methods rely mainly on data from weather and hydrological stations to build drought indices such as the Standardized Precipitation Index (SPI) [25,26,27], Standardized Precipitation Evapotranspiration Index (SPEI) [28,29,30], and Palmer Drought Severity Index (PDSI) [31,32,33,34]. AD is typically monitored by assessing vegetation growth, coverage, and soil moisture using remote sensing indices like the Normalized Difference Vegetation Index (NDVI) [35], Vegetation Condition Index (VCI) [36,37], Enhanced Vegetation Index (EVI) [38], and Crop Water Stress Index (CWSI) [39]. Thus, using both meteorological and remote sensing data to monitor drought and analyze its spatiotemporal patterns is an important part of research in meteorology and agriculture [22,24,40,41,42,43,44].

MD is the precursor to AD, which represents the development of MD, and has a more direct impact on crop yields and food security [45,46,47,48]. However, whether AD occurs depends on many factors, with irrigation being one of the most important [5,49]. Irrigation supplies water to crops artificially, reducing the impacts of climate change and AD, and preventing crop yield loss [50,51,52]. Nevertheless, irrigated agriculture consumes large amounts of available freshwater, and issues such as low water use efficiency and uneven management make agricultural systems highly vulnerable to extreme droughts under water constraints [53,54,55]. For instance, Zaveri et al. [56] showed, with 40 years of data in India, that while irrigation can reduce the harmful effects of extreme drought, its overall effectiveness has decreased over time. In Gujarat, irrigation contributed the least to grain yield compared to other regions in India and could no longer buffer the negative impacts of extreme heat and drought. Similarly, Song et al. [57] found that in China’s arid and semi-arid regions, agricultural irrigation accounted for over 80% of total water use from 2000 to 2021. Combined with the intensification of extreme drought events, the risk to water resource security may outweigh the benefits of irrigated agriculture. Therefore, although irrigation plays a positive role in countering AD, its long-term effectiveness shows clear spatiotemporal variability due to geographical environment, climate regions, and crop types [53,58].

The dynamic monitoring of crop responses to MD has become a research focus to ensure agricultural production. Global scholars have conducted some research on the response of croplands to MD [59,60,61]. However, most studies have concentrated on drought formation and spread over entire regions, or have compared the monitoring effects of different drought indices [62,63,64]. Studies that analyze drought spatiotemporal patterns and the response mechanisms of AD to MD from the perspectives of climate zones and cropland classification remain limited. China is the largest country in terms of irrigated area, with about 50% of irrigated croplands producing 75% of the nation’s grain and over 90% of economic crops [65,66,67,68]. Xinjiang and the Middle-lower Yangtze Plain are typical representatives of China’s arid and humid regions [69,70]. Combining MD and AD monitoring to study the spatiotemporal patterns in these regions and the impact of irrigation on drought propagation can help achieve food security and efficient water resource use.

The main objectives of this study include the following: (1) Select Xinjiang and MYP as study areas, extract continuously irrigated and continuously rainfed croplands from 2000 to 2019, and explore their spatial distribution. (2) Use PDSI and VCI to build time series of MD and AD in the study areas and analyze the spatiotemporal distribution patterns in irrigated and rainfed croplands. (3) Apply the Mann–Kendall trend test to explore the development trends of MD and AD in different cropland types. (4) Employ correlation analysis and wavelet transform analysis to examine the correlation and significance between MD and AD in the two cropland types. The results of this study can support the sustainable development of irrigated agriculture and provide decision support for scientific drought mitigation and disaster reduction.

2. Materials and Methods

2.1. Study Area

To investigate the propagation of MD and AD under continuous irrigation in both arid and humid regions, two study areas were selected: the Xinjiang Uyghur Autonomous Region (Xinjiang) and the Middle-lower Yangtze Plain (MYP) in China. Xinjiang (73°E–96°E, 34°N–49°N) is located in northwest China and is the largest provincial-level region by land area, covering approximately 1.66 million km².

Xinjiang has complex terrain: the Altai mountains lie in the north, the Kunlun mountains in the south. The Tianshan mountains run through central Xinjiang, dividing it into northern and southern parts, resulting in mountain–oasis–desert systems with distinctly different hydro-climatological conditions [71]. The Junggar Basin and the Tarim Basin are located in north Xinjiang and south Xinjiang, respectively. Deeply inland with scarce water resources, Xinjiang receives only about 150 mm of precipitation annually and features a typical temperate continental arid climate. Xinjiang is an important agricultural base, producing cotton, wheat, melons, and other food and economic crops [72,73]. MYP (108°E–123°E, 25°N–35°N) covers six provinces and one municipality (Hunan, Hubei, Jiangxi, Anhui, Zhejiang, Jiangsu, and Shanghai). As one of China’s three major plains, MYP has a subtropical monsoon climate with an annual precipitation of 1000–1500 mm and annual mean temperature of 14–18 °C. MYP is a key region for commercial grain production as well as for important economic crops such as cotton, hemp, and oil crops [74,75,76].

Xinjiang and MYP represent typical arid and humid regions in China, where agricultural practices differ significantly. Most croplands in Xinjiang rely on irrigation, while MYP, despite abundant precipitation, experiences uneven distribution and frequent droughts and floods, requiring irrigation to alleviate water stress [77,78]. The geographic locations and croplands distributions are shown in Figure 1.

2.2. Data

2.2.1. Irrigated and Rainfed Croplands Dataset

This study used croplands data as a mask for drought analysis. The annual irrigated cropland maps across China (IrriMap-CN) [79] dataset was selected as the base map for irrigated croplands. IrriMap-CN is derived from the National Land Cover Dataset (NLCD) [68,80], multiple feature variables (vegetation indices, meteorological, topographic and environmental variables) and random forest classifiers. It provides the annual spatial distribution of China’s irrigated croplands at 500 m resolution for the 20-year period from 2000 to 2019, with an average accuracy of about 80%. IrriMap-CN is obtained using a supervised classification method based on remote sensing, which is not constrained by statistical data and can provide important basic data for fields such as climate change, ecohydrology, and agricultural management. To ensure data integrity and continuity, the annual IrriMap-CN data were subtracted from the NLCD data to remove irrigated croplands, yielding the annual rainfed croplands dataset (RainMap-CN) from 2000 to 2019 at 500 m resolution. Finally, the time series of irrigated and rainfed croplands datasets were accumulated separately, retaining only those croplands pixels identified as irrigated or rainfed in all 20 years. This resulted in the continuously irrigated and continuously rainfed croplands from 2000 to 2019 for subsequent drought analysis.

2.2.2. Meteorological and Agricultural Drought Indices Dataset

PDSI provides a comprehensive measure of drought severity by using precipitation and temperature to estimate the soil moisture supply and demand balance and is widely used for monitoring long-term MD conditions at provincial, basin, or national scales. In this study, monthly PDSI data at 4 km resolution from 2000 to 2019 were downloaded from the global TerraClimate [81] database to represent MD. TerraClimate calculates evapotranspiration, precipitation, and temperature based on the Thornthwaite–Mather water balance model to provide a monthly high-resolution global climate dataset and drought index. The PDSI has been standardized by Z score, where positive values indicate water surplus and negative values indicate water deficit. VCI can assess the relative health of vegetation by comparing current vegetation conditions to historical minimum and maximum values. It is particularly effective for monitoring AD because it directly captures the impact of water deficit on vegetation growth. Therefore, based on satellite observation data, monthly VCI at 500 m resolution for the study areas were calculated from 2000 to 2019 to represent AD.

2.2.3. Remote Sensing and Other Auxiliary Dataset

Using Moderate-resolution Imaging Spectroradiometer (MODIS) satellite data and Google Earth Engine (GEE) platform, this study employed NDVI data from the MOD13A1 product at 500 m resolution for all clear-sky scenes over the study areas from 2000 to 2019, which were used to calculate the VCI. Additional auxiliary data includes basic geographic data such as administrative boundaries and river vectors.

Table 1. The datasets in this study. NLCD and PDSI were resampled to 500 m resolution.

Dataset	Spatial Resolution (m)	Time Resolution (Year)	Usages in This Study	Source/Reference
IrriMap-CN	500	2000–2019	Irrigated croplands mask	[79]
NLCD	100	2010	Rainfed croplands extraction	[80]
PDSI	4000	2000–2019	Meteorological drought	[81]
MOD13A1	500	2000–2019	Calculation of VCI	(NASA)
Shapefile	/	/	Study area extraction	/

summarizes all the datasets. For the differences in spatial resolution of multi-source data, we used the nearest neighbor method to resample the NLCD and PDSI data to 500 m to ensure spatial consistency.

2.3. Methods

2.3.1. Meteorological Drought Disaster Extraction

PDSI is based on the water balance equation and was first proposed by Palmer in 1965 [34]. It considers the influence of previous moisture conditions when assessing the regional water balance. PDSI mainly uses temperature and precipitation to monitor wet and dry states and is widely used to indicate the severity of MD. The calculation formula of PDSI is as follows:

$$\widehat{P}=\widehat{ET}+\widehat{R}+\widehat{RO}-\widehat{L},$$

(1)

$$d=P-\widehat{P},$$

(2)

$$Z=K_j·d,$$

(3)

$${PDSI}_i=a·{PDSI}_{i-1}+b·Z_i,$$

(4)

where $\widehat{P}$ is the monthly precipitation under suitable climate conditions; $\widehat{ET}$ is the evapotranspiration; $\widehat{R}$ is the soil moisture recharge; $\widehat{RO}$ is runoff; $\widehat{L}$ is the soil moisture loss; d is the water deficiency, $P$ is the actual precipitation, $Z$ is the water deficit index, and $K_j$ is the climate correction coefficient for month $j$ of the year. The $a=0.897$ and $b=1/3$ are the empirical coefficients. PDSI values range between −4 and 4; the lower the value, the more severe the MD. Based on the widely used standard thresholds, the specific classification of PDSI is shown in

Table 2. Different levels of the PDSI.

PDSI Values	Dryness/Wetness Levels
PDSI ≤ −4.0	Extreme drought
−4.0 ≤ PDSI < −3.0	Severe drought
−3.0 ≤ PDSI < −2.0	Moderate drought
−2.0 ≤ PDSI < −1.0	Mild drought
−1.0 ≤ PDSI < 1.0	Normal or wet
1.0 ≤ PDSI < 2.0	Mild wet
2.0 ≤ PDSI < 3.0	Moderate wet
3.0 ≤ PDSI < 4.0	Severe wet
PDSI ≥ 4.0	Extreme wet

[31].

Since this study mainly focuses on the propagation mechanisms of MD and AD over the past 20 years in croplands areas and considering the phenological differences in crops, we only selected PDSI data during the crop growth period and used the average value synthesis method to obtain the annual PDSI. Specifically, for a given year, only the monthly PDSI data from April to October for Xinjiang croplands were selected and averaged on a per-pixel basis to represent that year’s MD. Similarly, for MYP, the monthly PDSI data from March to November were selected and averaged.

2.3.2. Agricultural Drought Disaster Extraction

The basic principle of VCI is to compare the current NDVI with its historical minimum and maximum values, thereby indicating the current status of vegetation relative to its historical range. By removing the influence of location and ecosystem on NDVI, VCI shows a high sensitivity to AD and is ideal for large-scale remote sensing AD monitoring.

Higher VCI values indicate higher vegetation water content, while lower values indicate that the vegetation is under drought stress. VCI is calculated based on the relative position of the current NDVI value within its historical range (maximum and minimum values) for a given location, and it essentially reflects the relative vegetation health in a specific area. Therefore, the fixed classification thresholds may not accurately reflect the actual drought status in different regions. The calculation formula for VCI is as follows:

$$NDVI={\displaystyle \frac{Nir-Red}{Nir+Red}},$$

(5)

$$VCI={\displaystyle \frac{{NDVI}_i-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}},$$

(6)

where Nir and Red denote the reflectivity of the near-infrared and red bands, respectively. ${NDVI}_i$ is the NDVI value of the $i$ month for a given year; ${NDVI}_{max}$ and ${NDVI}_{min}$ are the maximum and minimum NDVI values of the $i$ month from 2000 to 2019, respectively.

Due to the influence of cloud cover, MOD13A1 may contain null values or noise pixels, especially during the rainy season in China. Therefore, we used the Savitzky–Golay (SG) filter [82] to reduce outliers and fill in missing values. The SG filter is a digital smoothing technique that can effectively correct data anomalies and preserve the original shape and features of time series data by fitting a polynomial to a moving window of data points. To obtain the monthly composite data of VCI, we calculated 16-day VCI data from MOD13A1 NDVI. Then, the VCI time series from 2000 to 2019 was filtered using SG filter, and the filtered VCI in each month was averaged to obtain the VCI monthly data. We compared the results with those obtained from the MOD13A3 (1 km) monthly product and confirmed that the higher spatial resolution of MOD13A1 (500 m) can monitor drought more accurately. Finally, similar to PDSI, the average VCI for each year was computed using the data from April to October for Xinjiang and March to November for MYP to monitor AD.

2.3.3. Determination of Drought Trends

The Mann–Kendall (MK) test is a time series analysis and prediction technique used to detect trends and abrupt changes in a set of time series data [83,84]. As a nonparametric statistical test, the MK test does not require the data to follow any specific distribution and is not affected by extreme values. It works well with non-normally distributed data and is commonly applied in research on climate change, hydrology, and environmental science [85,86].

For one time series $X$ with sample size $n$, the calculation formula for the statistic $S$ of the MK test is as follows:

$$\mathrm{S}={\sum }_{k=1}^{n-1}{\sum }_{j=k+1}^{n}sgn(x_j-x_k).$$

(7)

Here, $sgn$ is a sign function and can be computed as:

$$sgn\left(x_j-x_k\right)=\left\{\begin{array}{c}1,x_j-x_k>0\\ 0,x_j-x_k=0\\ -1,x_j-x_k<0\end{array}\right..$$

(8)

In this study, the MK test was applied to the PDSI and VCI time series separately to analyze the trends in MD and AD over the past 20 years for both irrigated and rainfed croplands in Xinjiang and MYP.

2.3.4. Construction and Analysis of Correlation Model

Correlation analysis is a statistical method used to measure the degree of association between two or more variables, with the correlation coefficient quantifying the strength of this association. The Pearson correlation coefficient (PCC), also known as the product-moment correlation coefficient, expresses both the degree and direction of the correlation between two variables [87,88,89]. It is usually denoted by the symbol r, which ranges from −1 (perfect negative correlation) to 1 (perfect positive correlation), with 0 indicating no correlation. The PCC is calculated based on the covariance and standard deviation and is suitable for continuous data that are assumed to follow a normal distribution.

In this study, both PDSI and VCI are continuous time series data, and they approximately meet the normal distribution assumption. Therefore, the PCC was calculated to analyze the correlation between MD and AD. The specific formula for PCC is as follows:

$$R_{xy}={\displaystyle \frac{\sum _{i=1}^n\left[\right(x_i-\overline{x}\left)\right(y_i-\overline{y}\left)\right]}{\sqrt{\sum _i^n{(x_i-\overline{x})}^2}\sqrt{\sum _i^n{(y_i-\overline{y})}^2}}},$$

(9)

where $R_{xy}$ represents the correlation coefficient between variables $x$ and $y$, and $x_i$ and $y_i$ are the PDSI and VCI values of the ith year, respectively. $\overline{x}$ and $\overline{y}$ represent the average of PDSI and VCI in 20 years, respectively.

2.3.5. Construction of Wavelet Analysis Model

Cross-wavelet transform (XWT) and wavelet coherence (WTC) analysis are powerful techniques that combine wavelet transformation with statistical analysis [90]. They are widely used in signal processing and time series data analysis to study the time–frequency characteristics of signals and to explore the correlations and resonant patterns between time series data.

For two discrete time series $x_n$ and $y_n$, the XWT is expressed as $W^{XY}$=$W^X{W}^{Y\mathrm{*}}$, where * denotes the complex conjugate; $\left|W^{XY}\right|$ represents the cross-wavelet power; arg ($W^{XY}$) indicates the local relative phase between $X_n$ and $Y_n$ in the time–frequency domain. $P_k^X$ and $P_k^Y$ denote the cross-wavelet power and the background power spectrum of the two time series, respectively. The formula is as follows:

$$D({\displaystyle \frac{\left|W_n^X(s)W_n^{Y*}(s)\right|}{{\sigma }_X{\sigma }_Y}}<P)={\displaystyle \frac{Z_v(p)}{v}}\sqrt{P_k^X{P}_k^Y},$$

(10)

where $Z_v\left(p\right)$ is the confidence level corresponding to probability $p$, derived from the square root of two ${\chi }^2$ distributions. The phase angle of the cross-wavelet is defined as:

$$\alpha =\mathit{arg}\left(X,Y\right)=arg\left[{\sum }_{i=1}^n\mathit{cos}\left({\alpha }_i\right),{\sum }_{i=1}^n\mathit{sin}\left({\alpha }_i\right)\right].$$

(11)

WTC measures the degree of coherence in the time–frequency space of the XWT. It is defined as the square of the absolute value of the smoothed cross-wavelet spectrum normalized by the smoothed wavelet power:

$$R_n^2(s)={\displaystyle \frac{{\left|S(s^{-1}{W}_n^{XY}\left(s\right))\right|}^2}{S(s^{-1}{\left|W_n^X\left(s\right)\right|}^2)\cdot S(s^{-1}{\left|W_n^Y\left(s\right)\right|}^2)}}.$$

(12)

Here, $S$ is a smoothing operator:

$$S\left(W\right)=S_{scale}\left(S_{time}\left(W_n\left(s\right)\right)\right),$$

(13)

and $S_{scale}$ and $S_{time}$ represent the smoothing along the wavelet scale axis and time axis, respectively. The calculation method is as follows for the Morlet wavelet:

$${\left.S_{scale}\left(W\right)\right|}_n={\left.(W_n\left(s\right)\times c_2\mathsf{\Pi }\left(0.6s\right))\right|}_n,$$

(14)

$${\left.S_{time}\left(W\right)\right|}_s={\left.(W_n\left(s\right)\times c_1^{{\displaystyle \frac{-t^2}{2s^2}}})\right|}_s,$$

(15)

within which $c_1$ and $c_2$ are the determined normalization coefficients, and $\Pi$ is the rectangle function.

3. Results

3.1. Continuous Irrigated and Rainfed Croplands

According to Section 2.2.1, we extracted the pixels that were consistently identified as irrigated or rainfed croplands from the IrriMap-CN and RainMap-CN datasets for the years 2000–2019. This yielded the continuous irrigated and rainfed croplands for Xinjiang and MYP. Based on the spatial distribution maps (Figure 2), continuous irrigated croplands in Xinjiang cover 31,757 km² (approximately 69%), with most located in southern Xinjiang. In contrast, continuous rainfed croplands, covering 14,288 km² (about 31%), are mainly located in northern Xinjiang. In MYP, continuous irrigated croplands cover 134,466 km² (approximately 59%), mainly in the northern part (including eastern Hubei, northern Anhui, and most of Jiangsu), while continuous rainfed croplands cover 92,671 km² (about 41%), primarily concentrated in Hunan, Jiangxi, and Zhejiang and appearing as fragmented and scattered cropland pixels.

In subsequent analyses, these continuously irrigated and rainfed croplands in this section are used as the main research objects.

3.2. Spatiotemporal Patterns of Drought

3.2.1. Spatiotemporal Distribution of Drought in Xinjiang

The PDSI time series curve (Figure 3a) shows no obvious trend in the change in MD. Both irrigated and rainfed croplands exhibit similar PDSI trends, with values mostly above −1, indicating that the MD of Xinjiang’s croplands experienced normal or wet conditions from 2000 to 2019. In 2003 and 2017, the PDSI reached its maximum values, representing extremely wet years over the past 20 years, while the minimum values occurred in 2009, indicating severe MD. Figure 4a,b display the spatial distribution of PDSI during these extreme years, and Figure 4d shows the 20-year average PDSI. Overall, Xinjiang croplands experienced normal wet and dry conditions from 2000 to 2019, and northern Xinjiang was more humid than southern Xinjiang.

For VCI, the maximum and minimum values from 2000 to 2019 indicate that high VCI values are mainly in western northern Xinjiang (Figure 5a), while low VCI values are primarily in eastern northern Xinjiang and central southern Xinjiang (Figure 5b). Most croplands show maximum VCI values after 2010 (over 85%, Figure 5c), and most minimum VCI values occurred before 2010 (over 86%, Figure 5d), suggesting that AD has eased in recent years.

Furthermore, Figure 3b shows the VCI changing trends in the two cropland types from 2000 to 2019. Among them, irrigated croplands generally exhibit an increasing VCI trend, indicating continuous improvement in AD conditions. In contrast, the VCI for rainfed croplands decreased significantly from 2005 to 2009, reaching a minimum in 2009, which indicates worsening AD (Figure 6). This corresponds to the significant decline in PDSI during the same period, suggesting that the occurrence of MD led to increased AD in rainfed croplands in Xinjiang, while irrigated croplands were largely unaffected. In addition, the VCI value of irrigated croplands has always been greater than that of rainfed croplands over the past 20 years, with the multi-year average values being 0.57 and 0.53, further confirming the mitigating effect of irrigation on AD.

3.2.2. Spatiotemporal Distribution of Drought in Middle-Lower Yangtze Plain

In MYP, both irrigated and rainfed croplands exhibit similar PDSI trends with no obvious pattern, but the drought severity is higher, particularly during 2004–2009 and 2011–2014 (Figure 7a). The driest years were 2004 and 2011, and the wettest was 2016 (Figure 8). Over the past 20 years, the region generally experienced normal wet and dry conditions, with the northern part slightly drier than the southern part.

The spatial distribution of the maximum and minimum VCI values from 2000 to 2019 shows that high VCI values are mainly found in central Jiangsu and eastern Hubei (Figure 9a), while low values are mainly in central Anhui and northeastern Jiangsu (Figure 9b). Additionally, over 72% of the croplands reached maximum VCI values during 2010–2019 (Figure 9c), and over 77% reached minimum values during 2000–2009 (Figure 9d).

In general, the VCI for both cropland types in MYP increased from 2000 to 2019, indicating a gradual alleviation of AD (Figure 7b). However, unlike Xinjiang, the average VCI in rainfed croplands is consistently higher (0.64) than that in irrigated croplands (0.61) over the past 20 years. In addition, from 2004 to 2009, rainfed croplands experienced a significant drop in VCI, while irrigated croplands remained stable, suggesting that irrigation can counteract MD to prevent crop water stress. After 2013, however, AD in irrigated croplands increased significantly compared to rainfed croplands.

3.3. Trend Test of Drought

To clarify the spatiotemporal trends of drought in Xinjiang and MYP from 2000 to 2019, a MK test was applied to the PDSI time series (p < 0.05) (Figure 10). The results show that there is no significant trend in MD in Xinjiang and MYP over the past 20 years, with non-significant areas accounting for 95.5% in Xinjiang and 99.5% in MYP, further proving that there is no obvious pattern in the change in MD.

For AD, the MK test was applied to the VCI time series for irrigated and rainfed croplands in Xinjiang (p < 0.05) (Figure 11). Results indicate that 73.56% of irrigated croplands show a significant increasing trend in VCI, with only 1.8% showing a significant decrease, suggesting that most irrigated croplands in Xinjiang experienced a marked alleviation of drought. In contrast, approximately 23.4% of rainfed croplands experienced worsening drought, and 27.8% showed no significant trend.

For MYP (Figure 12), 50.3% of irrigated croplands and 63.1% of rainfed croplands show a significant increase in VCI, indicating alleviation of drought (p < 0.05). Unlike Xinjiang, the areas with significant VCI decreases in MYP are small (6.6% for irrigated croplands and 5.6% for rainfed croplands), with the remaining areas showing no significant trend (43.1% and 31.3%, respectively). The proportion of alleviated drought is higher in rainfed croplands than in irrigated croplands, and the area with worsening drought is very small. This suggests that while irrigation in MYP has a clear anti-drought effect, it is not as effective as in Xinjiang.

3.4. Correlation and Significance Analysis

Correlation analysis between PDSI and VCI shows that in Xinjiang most rainfed croplands exhibit positive correlations (about 87.9%), mainly in the western part of northern Xinjiang (Figure 13b), indicating that AD has an obvious positive feedback on MD. In contrast, more than half of the irrigated croplands show negative correlations, mainly in northern Xinjiang, while most areas in southern Xinjiang show positive correlations (Figure 13a), suggesting that MD is not the decisive factor for AD in irrigated croplands. Significance tests (p < 0.05) further show that 86.1% of irrigated croplands in Xinjiang have no significant correlation between PDSI and VCI (Figure 13c), while 44.6% of rainfed croplands show significant correlations (Figure 13d), which was significantly larger than that of irrigated croplands.

In MYP, the distribution of positive and negative correlations in rainfed croplands is nearly equal (50.5% and 49.5%, respectively, Figure 14b). In irrigated croplands, 40.2% show positive correlations and 59.8% show negative correlations (Figure 14a). Among them, regions with positive correlations are mainly in central-northern Anhui and central Zhejiang, while negative correlations are concentrated in central Hubei and central-northern Hunan. In addition, only 9.7% of irrigated croplands in MYP show significant correlations (Figure 14c), compared to 43.3% in C (Figure 14d). These results indicate that irrigation plays a limited role in the propagation of MD to AD in MYP.

3.5. Wavelet Analysis of Meteorological Drought and Agricultural Drought

Figure 15 and Figure 16 present the XWT and WTC analysis of the PDSI and VCI time series for Xinjiang and MYP. In these figures, thick contour lines indicate significance at the 95% confidence level against red noise, and the lighter black lines represent the cone of influence (COI) to reduce edge effects. The color bar on the right shows the wavelet energy. Arrows indicate the relative phase: an arrow pointing to the right (0° phase angle) indicates that PDSI and VCI are in phase (positive correlation), while an arrow pointing to the left (180° phase angle) indicates that they are out of phase (negative correlation).

In Xinjiang, both irrigated and rainfed croplands show a significant resonant period between PDSI and VCI on a time scale corresponding to 6–8 months (Figure 15a,b). From 2005 to 2019, rainfed croplands consistently show a clear positive correlation between MD and AD (Figure 15d). However, irrigated croplands exhibit a negative correlation during 2005–2010, and after 2010 the correlation becomes positive or non-significant (Figure 15c). This difference is related to the continuous decline in PDSI from 2005 to 2009 in Xinjiang, which reflects severe MD. Overall, the correlation between PDSI and VCI in rainfed croplands in Xinjiang is stronger than in irrigated croplands.

In MYP, multiple resonant periods exist between PDSI and VCI in irrigated croplands, with time scales mainly concentrated at 4 or 8 months. Notably, a clear negative correlation is observed during 2004–2006, while other periods show no clear correlation (Figure 16a,c). In rainfed croplands, there are two significant resonant periods at 6–8 months, and unlike irrigated croplands, the time series of PDSI and VCI during 2004–2009 show a strong positive correlation (Figure 16b,d), indicating that severe MD during this period led to AD.

4. Discussion

4.1. Differences in Irrigation’s Drought Mitigation Effects

Under global climate change, rising temperatures and increased evaporation have led to more frequent and severe drought events over recent decades. The drought-mitigating effect of irrigation on agriculture remains unclear. China is one of the countries most affected by drought, and drought conditions vary significantly across regions. Our results show that the effect of irrigation in alleviating AD differs by climate, croplands spatial distribution and time.

In Xinjiang, a typical arid region, permanent irrigated croplands have shown a continuous improvement in AD over the past 20 years, indicating that irrigation effectively reduces AD. In contrast, rainfed croplands in Xinjiang experienced a significant drop in VCI during 2005–2009, a period that coincided with MD events. This suggests that, without irrigation, MD has a strong impact on AD. Spatially, most irrigated croplands in Xinjiang are located in the drier southern area due to uneven precipitation distribution (more rain in the north, less in the south). With northern Xinjiang receiving about 200 mm of annual precipitation compared to less than 100 mm in the south, the need for irrigation is much higher in the south, highlighting its benefit. In addition, short-term heavy rains in northern Xinjiang during the summer, especially during critical crop growth stages, have a significant positive effect.

In the humid MYP, during severe MD from 2004 to 2009, the VCI of rainfed croplands decreased significantly, while irrigated areas remained more stable. This indicates that there is indeed a clear response relationship between MD and AD, but irrigation plays a mitigating role in extreme drought events, resulting in a non-linear relationship between the two. Notably, after 2013, the drought-mitigating effect of irrigation appears to have weakened, possibly due to an increase in extreme weather events caused by global climate change. These spatial and temporal patterns emphasize that the effectiveness of irrigation largely depends on local climate conditions and the severity of MD. For example, Elliott et al. [91] found that, with global climate change and the shortage of freshwater resources, about 20–60 million hectares of early irrigated croplands worldwide may not meet the water supply and growth needs of crops by the end of this century, potentially shifting to rainfed agricultural systems and causing significant grain production losses in regions such as the western United States, China, West Asia, and South Asia.

Overall, the irrigation effect in MYP is less pronounced than in Xinjiang. This difference may stem from several factors. First, the differing irrigation methods between Xinjiang and MYP could influence the water use efficiency. In water-scarce Xinjiang, water-saving techniques—such as drip irrigation—are widely employed, directly supplying water to the crop root zones and thus achieving higher water use efficiency [92]. Second, irrigation frequency plays a critical role, especially during the key growth stages of crops. For instance, Zhang et al. found that high-frequency irrigation (six times per season) can significantly reduce crop water stress, while low-frequency irrigation (three times per season) may not fully alleviate water deficits [93]. In addition, crop type and planting structure must also be considered. Deng et al. investigated cotton, a high water-demand crop in Xinjiang, and found that water consumption is lowest during the early growth stage and peaks in the mid-stage, which has increasingly pressured the region’s irrigation water supply over recent decades [94].

This study highlights the importance of tailoring irrigation strategies to local conditions. In arid regions like Xinjiang, long-term irrigation can effectively reduce crop water stress. However, managing water resources remains critical. Optimizing irrigation systems, adjusting irrigation schedules, and modifying cropping patterns are necessary to balance irrigation efficiency and water resource sustainability [95]. In humid regions such as MYP, irrigation should be focused on mitigating extreme drought events. Given the increasing climate risks, maintaining effective irrigation in these areas is a significant challenge [96]. Future research should aim to identify priority regions for irrigation and areas with high irrigation potential, thereby providing a robust basis for decision-making in agricultural water resource management under changing climatic conditions.

4.2. Innovation and Limitations of the Study

Previous studies have demonstrated that drought indices are effective in representing different types of droughts, making the response of AD to MD a research focus. However, most studies have concentrated on short-term drought events or evaluated croplands as a whole. This study started with the “impact of irrigation in both arid and humid regions on AD relief” and conducted a long-term discussion and analysis from 2000 to 2019. By separating irrigated and rainfed croplands, our research reveals different drought propagation mechanisms. For instance, in Xinjiang, irrigated croplands showed a decoupling between MD and AD during severe drought events, while rainfed croplands exhibited a strong positive coupling. Furthermore, our integrated methodology combines drought indices (PDSI and VCI) with advanced statistical and wavelet analysis techniques (MK test, PCC, XWT, and WTC). This multi-method approach enhances our understanding of the spatial and temporal relationships between MD and AD, providing a detailed perspective on how irrigation can alter AD in different climate zones.

However, this study has some limitations. First, it only relies on PDSI and VCI as indicators of MD and AD. Although these indices are widely used in drought assessments, combining additional drought indices according to local conditions and making comprehensive comparisons could provide a more complete evaluation of drought conditions. And the error inherent in the irrigated croplands base map dataset also affects the accuracy of the results of this study. In addition, although we selected data from April to October and March to November to approximate the crop growing season, this does not fully capture the complexity of crop phenology. Different crops and growth seasons may respond differently to water stress during critical development stages. For example, Zhang et al. [97] evaluated the adaptive differences in various crops to drought risk in the central United States from 1981 to 2012, and found that irrigation significantly alleviated drought for corn and soybean but had no effect on wheat. Deb et al. [98] found that while the yield of irrigated croplands increased significantly compared with rainfed croplands, the yield increase differed between corn and soybean. Therefore, future work should consider more detailed time segmentation aligned with regional crop calendars. Finally, due to data limitations, this study only covers 20 years of drought indices and disaster events, lacking historical data before 2000, which may limit the reflection of longer-term drought trends and irrigation effects.

5. Conclusions

This study comprehensively examined the spatiotemporal characteristics, trends, and correlations of MD and AD on continuous irrigated croplands and rainfed croplands in Xinjiang and MYP from 2000 to 2019. The main conclusions are as follows: (1) Severe MD events in Xinjiang from 2005 to 2009 and MYP from 2004 to 2009 led to increased AD in rainfed croplands. In contrast, irrigated croplands experienced more stable agricultural conditions, demonstrating that irrigation effectively mitigates AD under MD conditions. (2) Long-term data show a clear improvement in AD on irrigated croplands in Xinjiang from 2000 to 2019. MK trend test revealed that 73.56% of irrigated croplands had a significant alleviation of drought. In MYP, although AD in both cropland types showed overall improvement, the benefits of irrigation were less evident. In this humid region, irrigation only provided significant benefits during extreme MD (e.g., 2004–2009), and its effectiveness diminished after 2013. (3) Correlation and wavelet analyses revealed different drought propagation mechanisms between irrigated and rainfed croplands. Rainfed croplands generally showed a strong positive correlation between PDSI and VCI (Xinjiang from 2005 to 2019 and MYP from 2004 to 2009), indicating that MD directly leads to AD. In contrast, irrigated croplands exhibited negative correlations during some periods (Xinjiang from 2005 to 2010 and MYP from 2004 to 2009), suggesting that irrigation can offset the impact of MD on vegetation. In addition, although irrigation in MYP also helps to alleviate drought, its overall effect is lower than in Xinjiang. (4) Overall, spatial heterogeneity highlights that the effectiveness of irrigation varies. In arid regions like Xinjiang, irrigation plays a critical role. In contrast, in the generally wet MYP, the benefits of irrigation are limited to periods of MD, and in some cases, irrigated croplands may even experience more severe AD than rainfed croplands. Moreover, the occurrence of AD is not only related to MD and water shortage is only one contributing factor. Factors such as pest infestations, crop types, and topography may also contribute to the development of AD.