Characterizing Drought Patterns and Vegetation Responses in Northeast China: A Multi-Temporal-Scale Analysis Using the SPI and NDVI

Abstract

Drought significantly reduces global agricultural productivity and destabilizes ecosystems. As the primary grain-producing region and a key ecological buffer zone in China, Northeast China is experiencing intensifying drought stress. However, the regional-scale characteristics of refined drought and the impact mechanisms on different types of vegetation in the Northeast are rarely investigated. In this study, we analyzed the spatial and temporal characteristics of drought over 30-, 60-, 90-, 180-, 270-, and 360-day time scales in Northeast China using the Standardized Precipitation Index (SPI) based on high-precision daily precipitation data simulated by CLM3.5 from 2008 to 2023. Additionally, we used the MODIS Normalized Difference Vegetation Index (NDVI) to elucidate the response of vegetation to drought across different land use types. The results showed that SPI-30 was the most sensitive for drought detection, and there was a clear trend of drought aggravation in the northern part of the Northeast region. The strongest correlation between vegetation and drought was found in September. A significant lag in the response of vegetation to drought was observed in May, June, July, and August, with the best correlation observed at a one-month lag. In addition, the degree of response to drought varies among different types of vegetation. Grasslands are the most sensitive to drought, while woodlands and wetlands have a weaker response. This study provides a reference for assessing the dynamics of refined climates at different spatial and temporal scales and offers actionable insights for ecosystem management in climate-sensitive agricultural regions.

1. Introduction

Droughts, characterized by their complex internal structure, uncertain frequency, and extensive impacts, significantly impede a range of social activities, including agricultural production, economic development, and ecological protection [1,2,3,4,5]. Statistical data indicate that economic losses from meteorological disasters account for over 70% of the total losses caused by natural disasters, with droughts alone contributing approximately 50% of these losses [6]. As a country highly susceptible to droughts, in China, an average of over 200,000 km² of cropland was affected by drought annually from 1984 to 2018 [7]. In the context of climate change, the unpredictability of the duration and intensity of global extreme events is increasing [8]. The Sixth Assessment Report of the IPCC projects that future drought events will have broader and more intense impacts [9]. Additionally, the intervals between droughts and other extreme weather events, such as heat waves, are expected to shorten, posing greater challenges to regional resilience [10]. Therefore, identifying drought events and investigating changes in their characteristics is crucial in mitigating drought losses and providing a scientific basis for drought management and decision-making.

Droughts are typically classified into four primary types: meteorological, agricultural, hydrological, and socio-economic [11]. Meteorological drought, the most common type, serves as the driving force for other drought types [12]. Identifying and characterizing meteorological drought is a crucial component of drought research. Several widely used drought indices include the Standardized Precipitation Index (SPI), the Standardized Precipitation Evapotranspiration Index (SPEI), and the Palmer Drought Severity Index (PDSI) [13]. Despite their widespread use, these indices have limitations. The PDSI, which integrates temperature, precipitation, and soil factors, is widely used for agricultural drought assessment but is highly sensitive to soil information and has fixed time scales [14,15,16]. The SPEI, which combines precipitation and potential evapotranspiration, effectively accounts for moisture deficits but requires complete data, lacks sufficient temporal continuity for drought monitoring, and is calculated on a monthly scale, limiting its ability to accurately identify drought onset and termination [17]. In contrast, the SPI, which requires only precipitation data, quantifies precipitation deficits or surpluses on any time scale, providing a more accurate assessment of drought conditions [18,19]. However, most previous drought monitoring studies have focused on monthly scales, with fewer addressing daily scales. Wang et al. [20] developed a daily SPI to address the limitations of temporal resolution in existing monthly drought assessments. The daily drought index overcomes the limitations of the monthly index, providing more detailed information on drought conditions and enabling the monitoring of drought characteristics across multiple time scales, including 30-day, 60-day, 90-day, and longer periods. Therefore, enhancing the temporal resolution of drought indices is essential in accurately analyzing drought characteristics.

Drought assessment must be data-driven, and traditional methods of drought monitoring primarily rely on field observations and meteorological assessments. However, obtaining sufficient drought monitoring sampling sites on a regional scale is labor-intensive, and field sampling faces methodological challenges [21]. Additionally, the use of other variables, such as temperature and precipitation, in meteorological assessments is often limited by the lack of sites data [22,23]. With the continuous advancement of science and technology, the emergence of exogenous data has provided new technical means of drought monitoring. Satellite remote sensing data have emerged as an effective and reliable tool in this context [24,25,26]. Over the last several decades, scholars have developed precipitation products with varying spatial and temporal resolutions by integrating international meteorological satellite data, machine learning algorithms, and data assimilation techniques [27]. Despite the availability of numerous precipitation datasets, their accuracy often falls short of the requirements for current drought event assessments. For example, ERA5, one of the most widely used precipitation datasets, has a spatial resolution of only 0.1°, which may not be sufficient for detailed drought analysis [28]. There have been many studies using precipitation data developed from pavement models with high spatial resolution. However, the land surface model requires multiple models to drive the generation of precipitation data, and this data generation has been taking place since 2008 [29].

As a pivotal agricultural production base in China, the Northeast region is crucial in national food security [30]. However, the Northeast region experiences significant spatial and temporal fluctuations in precipitation and is vulnerable to drought. Over the past 20 years, 25% of the total cultivated area has been affected by drought, posing a serious threat to the economic development of the region [31,32]. Studies have indicated that the Northeast region has experienced a significant aridification trend since 1961 [33]. The frequency and intensity of drought events in the Northeast have increased notably after 2000 compared to the pre-2000 period [34]. Climate model projections under the RCP8.5 scenario indicate that the summer PDSI in the region is expected to decline by 2.5–3.0 units by the end of the 21st century, suggesting a continued escalation in drought risk [35,36]. Additionally, numerous global-scale studies have confirmed that drought is a dominant factor in the decline of vegetation productivity [37]. With the increasing frequency of drought events in Northeast China, their impact on regional vegetation growth and ecosystem stability has become increasingly pronounced [38,39]. The effects of drought on vegetation growth in Northeast China exhibit significant regional differences [39]. As the primary commercial grain base in China, the Northeast region accounts for approximately 25% of the country’s grain production, and drought poses a significant threat to national food security [40]. Research on drought in this region not only provides a scientific basis for enhancing China’s food security and water resource management but also serves as a reference for the sustainable development of other regions in similar climatic zones. Despite there having been previous studies on the characteristics of drought in Northeast China and its relationship with vegetation response, few have examined the characteristics of drought changes at higher resolutions and over finer time scales or the effects of drought on different vegetation types. Therefore, this study is essential in ensuring food production and ecosystem management in the Northeast region [41,42,43].

In this study, we aimed to better understand the regional-scale characteristics of fine-scale drought and its impacts on different types of vegetation. The specific objectives were as follows: (i) to explore the spatiotemporal dynamics, including drought frequency, duration, intensity, and severity, of fine-scale drought at multiple time scales; (ii) to analyze the trends in drought under different land use types, such as dry cropland, woodland, grassland, and wetland; (iii) to examine the effects of fine-scale drought on vegetation for different land use types in terms of correlation and lag. The study results can contribute to the understanding, prediction, and mitigation of fine-scale drought and can be applied to other drought hotspots worldwide.

2. Materials and Methods

2.1. Study Regionfour Distinct Seasons: Warm and Rainy Summers and Cold and Dry Winters

The Northeast region covers a total area of approximately 1,267,000 km² and is situated between 38° N and 53° N latitude and 115° E and 135° E longitude (Figure 1). It comprises Liaoning Province, Jilin Province, Heilongjiang Province, and the eastern part of the Inner Mongolia Autonomous Region. Spanning the mesothermal and cold temperate zones from south to north, the region experiences a temperate monsoon climate characterized by. The average annual precipitation ranges from 300 to 1000 mm, transitioning from humid and semi-humid zones in the southeast to semi-arid zones in the northwest.

The Northeast region is characterized by a complex topography, featuring the Changbai Mountains in the east, the Lesser Khingan Mountains in the north, and the Greater Khingan Mountains in the west. The average elevation of the Northeast Plains is approximately 200 m above sea level. As China’s most important grain-producing area, the region devotes 45.46% of its total area to crops, with maize, rice, soybeans, and wheat being the primary crops. Additionally, the region boasts substantial forest resources, with forest cover accounting for 25.81% of the study area.

2.2. Data Source

2.2.1. Precipitation Data

This study employed daily-scale precipitation data from the European Centre for Medium-Range Weather Forecasts (ECMWF) and GLDAS (Global Land Data Assimilation System). Community Land Model version 3.5 (CLM3.5), developed by the National Center for Atmospheric Research (NCAR), was utilized to simulate the exchanges of energy, moisture, and carbon cycles between the land surface and the atmosphere. Released in 2007, this model incorporates recent observational data and theoretical advancements from the fields of ecology and hydrology. It features a modular design that includes components for vegetation dynamics, soil stratification, and snowpack processes. The driving data for the CLM3.5 and the assimilation system require meteorological variables such as the near-surface air temperature, barometric pressure, specific humidity, wind speed, downward shortwave surface radiation, downward longwave surface radiation, and precipitation rate. Additionally, the spatial resolution of the CLM3.5 data was enhanced using the Delta downscaling method. This method involves interpolating low-resolution simulated data to a high-resolution grid and then calculating the relationship between the simulated and observed data at the raster scale for the historical period’s multi-year average.

This study employed CLM3.5 precipitation data for the period from January 2008 to December 2023, with a spatial resolution of 2 km × 2 km, obtained via the Delta downscaling method. Additionally, ERA5 daily precipitation data were selected for the period from May to September 2008–2014, with a spatial resolution of 0.1°. To accurately analyze the precipitation remote sensing products, data from 90 meteorological sites in the study area were selected to verify the accuracy with CLM3.5 and ERA5 precipitation data, respectively.

2.2.2. Land Use Type

The land use data were obtained from the China Multi-Period Land Use Remote Sensing Monitoring Database, released by Xu Xinliang’s team at the Institute of Geographic Sciences and Resources, Chinese Academy of Sciences at http://www.resdc.cn (accessed on 12 August 2024). The data are generated on the basis of Landsat satellite image interpretation and are suitable for land use/cover change analysis on the regional and national scales, with a spatial resolution of 30 m × 30 m. In this study, four land use types were extracted from this product for the year 2020: dry croplands, forests, grasslands, and wetlands. These types were used as the baseline data for land use classification. Historical data comparison and verification revealed that the primary land use types in the study area had remained stable over time, with no large-scale transformations occurring. To account for minor local changes, spatial buffer zones were established to ensure that the potential impact of data timeliness on the study results remained within an acceptable range.

2.2.3. Normalized Difference Vegetation Index (NDVI)

Landsat remote sensing image data were selected for vegetation index analysis, and the data were obtained from the Google Earth Engine (GEE) Cloud Computing Platform at https://code.earthengine.google.com/ (accessed on 22 August 2024). In this study, Landsat 5 and 8 satellite data were used; they provide medium resolution multispectral imagery suitable for vegetation monitoring. All image preprocessing and NDVI calculations were performed using the GEE platform. And the GEE platform directly invoked the preprocessed Landsat surface reflectance dataset to ensure the consistency of data quality. NDVI data for the period of May to September 2008–2023 in Northeast China were selected. The NDVI data used in this study are 16-day composite products generated using the Maximum Value Composite (MVC) method, which integrates multi-view imagery. This method selects the highest NDVI value for each raster cell within the 16-day period as the output. Given that vegetation indices typically peak under clear-sky conditions and are significantly depressed by cloud cover, the MVC approach effectively filters out cloud interference. The 16-day integration period ensures that even with temporary cloud cover, each raster cell has a high probability of being observed under clear skies, thereby producing a vegetation index that closely approximates ground truth. Consequently, this NDVI product offers high data integrity and reliability, effectively overcoming the common issue of cloud interference in single-time-phase imagery.

2.3. Methods

2.3.1. Standardized Precipitation Index (SPI)

Drought indices are critical for drought assessment because they facilitate a clear understanding of the complex interrelationships across multiple study areas. The SPI is a parameter solely based on precipitation data. The SPI is not directly normalized by the arithmetic mean and standard deviation in the calculation process. Instead, it is firstly fitted to the probability distribution suitable for describing the characteristics of precipitation, such as gamma distribution, in order to accurately portray its non-negativity, right skewedness, and zero-value characteristics [44]. Subsequently, it is standardized by transforming the probability values to the quantile of standard normal distribution through the cumulative distribution function [45]. This method fully considers the non-normal characteristics of precipitation data, and its theoretical basis lies in eliminating the influence of the original distribution through the probability distribution transformation. The process of enabling precipitation data at different spatial and temporal scales can be presented in a standardized and comparable form. This ensures the statistical reasonableness and reliability of drought assessment results. The SPI compares the precipitation for a particular period with the total precipitation for the same period over all recorded years, facilitating the temporal analysis of drought phenomena. The SPI is calculated as follows:

Let the amount of precipitation at a given time (t) be denoted as x₀:

$$\mathrm{SPI} =\mathrm{S}\left\{\mathrm{t} -{\displaystyle \frac{\left({\mathrm{c}}_2\mathrm{t} - {\mathrm{c}}_1\right)\mathrm{t} + {\mathrm{c}}_0}{ \left[\left({\mathrm{d}}_3{\mathrm{t}}^2+{ \mathrm{d}}_2\mathrm{t} + {\mathrm{d}}_1\right)\mathrm{t}+1.0\right]}}\right\}$$

(1)

$$(1.0 \ge \mathrm{F} > 0.5, \mathrm{S}=1; \mathrm{F} \le 0.5, \mathrm{S}=-1)$$

(2)

$$\mathrm{t}=\sqrt{\ln {\displaystyle \frac{1}{{\mathrm{F}}^2}}}$$

(3)

where F is the probability that the random variable x < x₀; c₀, c₁, c₂, d₁, d₂, and d₃ are constants with values of c₀ = 2.515517, c₁ = 0.802853, c₂ = 0.010328, d₁ = 1.432788, d₂ = 0.189269, and d₃ = 0.001308, respectively. These values are derived from the classical polynomial approximation method found in mathematical manuals, which converts cumulative probabilities into normalized indices by simplifying the formula. The drought classifications for SPI are shown in

Table 1. SPI drought levels.

Drought Level	SPI
No Drought	SPI > −0.5
Light Drought	−1.0 < SPI ≤ −0.5
Moderate Drought	−1.5 < SPI ≤ −1.0
Severe Drought	−2.0 < SPI ≤ −1.5
Extreme Drought	SPI ≤ −2.0

.

The nature of drought is the cumulative effect of precipitation shortages. Because the SPI calculation is based on precipitation sequences in different cumulative time windows, it can analyze multiple time scales. In drought studies, selecting the multi-time-scale SPI to calculate drought characteristics can provide a more comprehensive assessment of the drought situation in the study area. The shorter-time-scale SPI can quickly respond to precipitation changes and capture short-term drought events. But it only reflects changes over the last 30 days. In contrast, the longer-time-scale SPI can reflect the cumulative impacts of drought on water resources and assess long-term drought risks [46,47]. Additionally, the multi-time-scale SPI can smooth short-term fluctuations, reduce misjudgments, integrate multiple factors, enhance the spatial and temporal comparability of drought characteristics, and facilitate the comparison and reflection of the spatial and temporal evolution of drought patterns across different regions. Using only a single time scale would fail to capture the immediate, persistent, and cumulative characteristics of meteorological droughts simultaneously, leading to a one-sided understanding of the drought formation mechanism and evolutionary process. Therefore, this study calculated SPIs for 30-, 60-, 90-, 180-, 270-, and 360-day time scales.

2.3.2. Run Theory

Run Theory is a method used to analyze time series data, drought characteristics, and related concepts. A “run” refers to a phenomenon where events with the same properties continue to occur, and the length of the run is defined by the duration of these events. Drought duration is defined as the total time between the beginning and end of consecutive drought events. Drought severity is defined as the cumulative underachievement of successive drought indices below the thresholds. Drought intensity is defined as the average of the values from the beginning to the end of the drought. The principle of Run Theory is illustrated in Figure S1. A drought event is deemed to have begun when the drought index x₀ falls below −0.5, and it is considered to have ended when the drought index x₀ rises to or exceeds −0.5.

2.3.3. Trend Analysis

Linear regression is a statistical method used to analyze the relationship between two or more variables. The slope is a key parameter in linear regression analysis, indicating the change in the dependent variable for each unit change in the independent variable. On a temporal scale, trends in drought characteristics can be analyzed using linear regression and subsequently fitted to simulate trends for each raster grid in the study area, facilitating the assessment of drought evolution. Therefore, the calculated slope values can reflect the actual change in drought characteristics over a specific period. The specific formula for the slope is as follows:

$$\mathrm{Slope}={\displaystyle \frac{\mathrm{n}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\left(\mathrm{i} \times { \mathrm{DC}}_{\mathrm{i}}\right) - {\sum }_{\mathrm{i}=1}^{\mathrm{n}}\mathrm{i} \times {\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{DC}}_{\mathrm{i}}}{\mathrm{n}{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{\mathrm{i}}^2 - {\left({\sum }_{\mathrm{i}=1}^{\mathrm{n}}\mathrm{i}\right)}^2}}$$

(4)

where Slope is the slope of the raster regression equation indicating the rate of change of drought characteristics over time, DC_i represents the average DC value in year i, and n is the length of the study period.

When Slope > 0, it indicates that the DC of the raster point is increasing; when Slope = 0, it indicates that the DC of the raster point remains relatively stable during the study period. When Slope < 0, it indicates that the DC of the raster point is decreasing.

2.3.4. Correlation Analysis

The Pearson correlation coefficient method quantifies the linear correlation between drought indicators and vegetation indices, thereby revealing the relationship between drought and vegetation growth. The correlation coefficients indicate the degree to which the SPI reflects the relationship between water changes and vegetation; a positive correlation indicates that the drought index is strongly correlated with the vegetation condition. The larger the correlation coefficient, the closer the relationship between precipitation and vegetation. The Pearson correlation coefficient can not only analyze the correlation between drought and vegetation but also reveal the lagged phenomenon between drought and vegetation [39,48]. Currently, the Pearson correlation coefficient is widely used in drought research. It helps to comprehensively assess the relationship between drought and vegetation growth and provides a scientific basis for ecosystem protection and management. The formula for the correlation coefficient is as follows:

$${\mathrm{R}}_{\mathrm{xy}}={\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}\left[\right({\mathrm{x}}_{\mathrm{i}} - \stackrel{\mathrm{-}}{\mathrm{x}}\left)\right({\mathrm{y}}_{\mathrm{i}} - \stackrel{\mathrm{-}}{\mathrm{y}})]}{\sqrt{{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{x}}_{\mathrm{i}} - \stackrel{\mathrm{-}}{\mathrm{x}})}^2{\sum }_{\mathrm{i}=1}^{\mathrm{n}}{({\mathrm{y}}_{\mathrm{i}} - \stackrel{\mathrm{-}}{\mathrm{y}})}^2}}}$$

(5)

where R_xy is the correlation coefficient of x and y, x_i and y_i are the SPI and NDVI values in year i, respectively, while $\overline{\mathrm{x}}$ and $\overline{\mathrm{y}}$ are the mean values of SPI and NDVI, respectively, and n is the length of the time series.

To ensure spatial and temporal consistency between the daily SPI and the 16-day synthetic NDVI dataset, the data were spatially and temporally processed. Spatially, the SPI data were interpolated to match the raster resolution of the NDVI dataset. Temporally, the daily SPI data were aggregated into 16-day averages to correspond with the NDVI period. This processing unified the spatial and temporal scales of the two datasets, enabling Pearson correlation analysis. Following the calculation of the Pearson Correlation Coefficients, a significance test was conducted to assess the significance of these coefficients (p < 0.05).

3. Results

3.1. Evaluation of Precision of Precipitation Data

The CLM3.5 model daily precipitation data were aggregated to obtain monthly precipitation data for the Northeast region from 2008 to 2014. In this study, both daily and monthly precipitation data were used for accuracy validation within the study area. As shown in Figure S2, the results indicate that the Satellite daily precipitation data correlation coefficient with site data R = 0.5313 (p < 0.05). The overall coefficient of determination R² = 0.8532, slope K = 1.0798, and correlation coefficient R = 0.9237 between the monthly satellite precipitation data and the measured monthly precipitation at the sites passed the significance test at p = 0.05. Satellite monthly precipitation tends to overestimate the actual precipitation in areas with low values and underestimate the actual precipitation in areas with high values. The correlation coefficient between ERA5 and the daily precipitation data at the meteorological sites R = 0.3605 (p < 0.05). The overall coefficient of determination R² = 0.6294, slope K = 4071.4, and correlation coefficient R = 0.7933 for the monthly precipitation data of ERA5 and the measured rainfall data at the sites indicate that the accuracy of the ERA5 precipitation data is relatively low compared to the measured rainfall data. Overall, the monthly precipitation data from the CLM3.5 model exhibit a more pronounced correlation and consistency with site-measured precipitation in the study area, making them suitable for use as a data source for precipitation-related studies.

3.2. Drought Evolution Trends and Response to Precipitation

Drought conditions in the study area have changed over the past decade and a half. This increased climate change could significantly trigger changes in regional attributes, which could lead to greater uncertainty regarding economic development in the region. Different time scales can characterize different moisture features. SPI-30 can reflect detailed changes in moisture deficit, SPI-90 can reflect seasonal changes in moisture deficit, and SPI-180, SPI-270, and SPI-360 can reflect mid-term and long-term changes in moisture deficit. As a crucial grain production base in China, monitoring the precipitation situation during the growing period of crops in the Northeast region is of significant importance. The primary crop-growing areas in the study region are concentrated in the Songnen Plain, the Sanjiang Plain, and the Liaohe Plain (Figure 1).

In this study, we selected daily CLM3.5 precipitation data with a 2 km spatial resolution in the watershed to calculate the SPI and analyze the evolutionary trends and patterns of drought in the Northeast region (Figure 2). Overall, the SPI at all time scales has shown a consistent wetting trend since 2008, aligning with the results of existing studies [49]. The region has predominantly been free of severe drought, with most events classified as light droughts. Moderate drought conditions were observed in 2011, after which there was a gradual shift towards wetter conditions. As the time scale increases, the fluctuations in the SPI curve tend to flatten out. The frequency of alternating drought and non-drought periods gradually decreases, the durations of different drought events lengthen, and the trend towards wetting becomes more pronounced. This is because small droughts occurring in shorter periods cannot be recognized as the time scale increases.

The primary cause of drought phenomena is the variation in precipitation. Based on precipitation statistics, we identified 2011 as a low-water year and 2020 as a high-water year. We analyzed the changes in daily precipitation and the SPI during the crop-growing period (May to September) in the three major plains during these high- and low-water years (Figure 3). Overall, the fluctuating trend in the SPI at different time scales is relatively consistent across the plains, both in low-water and high-water years. However, there are differences in the fluctuation amplitude of the SPI. The smaller the time scale, the more pronounced the fluctuations in the SPI, resulting in a higher identification sensitivity to drought and a more sensitive response to drought conditions. Conversely, the larger the time scale, the gentler the fluctuations and the lower the identification sensitivity to drought. At the Liaohe Plain, peak precipitation occurs in late July during low-water years and in late August during high-water years, while peak SPI values are observed in mid-August and mid-September, respectively. In the Sanjiang Plain, precipitation is higher in early June during low-water years, peaking in late August during high-water years, with SPI peaks occurring in mid-June and mid-September. In the Songnen Plain, precipitation is mostly concentrated in mid-July during low-water years and peaks in early September during high-water years, while SPI peaks in early August and mid-September. A lag exists between SPI and precipitation, with this lag being more pronounced in high-water years than in low-water years.

3.3. Characteristics of Spatial and Temporal Changes in Drought

3.3.1. Characteristics of Spatial and Temporal Changes in Drought Frequency

The increased frequency of droughts is widely recognized as a direct consequence of global warming. Understanding this increase is crucial in comprehending and responding to climate change, developing water resource management plans, and assessing the agricultural, ecological, and socio-economic impacts of droughts. In this study, we utilized SPI data from 2008 to 2023 to analyze the spatial and temporal evolution trends and patterns in drought frequency in Northeast China across different time scales (Figure 4 and Figure 5).

Overall, Figure 4 indicates that the annual precipitation in the three plains has shown an increasing trend since 2008, followed by a decreasing trend, which aligns more closely with the trend in drought frequency across different time scales. However, there are differences in drought frequency identification across various time scales. The larger the time scale, the coarser the identification of drought, meaning it can only detect drought occurrences over longer periods. Consequently, as the time scale increases, the frequency of identified drought occurrences decreases. Drought frequency also varies across the plains, with the Liaohe Plain experiencing a relatively high frequency of droughts and the Sanjiang Plain and Songnen Plain experiencing lower frequencies. For the Liaohe Plain, the highest drought frequency was recorded in 2011, with 10 times, mostly short-term droughts. In 2009, the region saw eight drought events, with a relatively higher proportion of medium-term droughts. In contrast, the Liaohe Plain had the lowest frequency of drought in 2021, with most droughts occurring in the short term and fewer medium-to-long-term droughts. The Sanjiang Plain had the highest frequency of drought in 2010, with more droughts in the medium and long term compared to other years. The Sanjiang Plain experienced the lowest frequency of drought in 2023, with fewer droughts occurring in the medium and long term. The Songnen Plain had the highest drought frequency in 2009, with more droughts occurring in the medium to long term. In contrast, the Songnen Plain had a lower drought frequency in 2021, with more droughts occurring in the short term.

On the spatial scale, the trend of drought change in the Northeast region tends to flatten out as the time scale increases. At different time scales, SPI-30 and SPI-60, similar to the seasonal time scale SPI-90, exhibit a relatively clear trend of decreasing drought frequency, primarily concentrated in the Songnen Plain and the Sanjiang Plain. In contrast, the SPI-180, SPI-270, and SPI-360 scales show relative stability, with only a slight increase in drought frequency in the northern part of Northeast China and the western part of the Sanjiang Plain. Thus, while short-term and seasonal droughts in the Northeast have decreased in frequency, changes in medium- and long-term droughts are less pronounced.

3.3.2. Characteristics of Spatial and Temporal Changes in Drought Duration

Drought duration is not only a significant meteorological indicator but also has profound impacts on ecology, agriculture, the economy, and social development. Figures S3 and S4 illustrate the spatial and temporal distribution of drought duration at different time scales in the Northeast region. Slope trend analysis was conducted at six time scales to characterize the trends in drought duration changes.

Across the time scales, the trends in drought duration and drought frequency were consistent, both showing an overall decreasing trend. The identification of drought duration varied by time scale. In the three plains, drought durations were longer in 2009 and 2011 and increased with increasing time scales. After 2018, none of the droughts lasted long. The Liaohe Plain had the shortest drought duration in 2021, with drought duration reaching 0 days on the interannual scale. The Sanjiang Plain had the shortest drought duration in 2023, with most droughts occurring in the short term and decreasing in duration with increasing time scales. The Songnen Plain had a shorter duration of drought in 2021, with durations varying considerably on the monthly and interannual scales.

From the spatial change trend, it is evident that the change trend in drought duration in the Northeast region is relatively distinct. The overall trend of increasing drought duration is concentrated in the northern part of the Northeast, while the trend of decreasing drought duration is concentrated in the middle region of Northeast China. As the time scale increases, the trend of drought duration change becomes increasingly significant. The upward trend gradually extends to the south, while the downward trend gradually extends from the eastern Songnen Plain and the western Sanjiang Plain to the middle region of the entire Northeast.

3.3.3. Characteristics of Spatial and Temporal Changes in Drought Intensity

Drought intensity is a key indicator of the extent of drought, reflecting the degree to which a region experiences insufficient water supply at a given time. Figures S5 and S6 show the temporal and spatial trends in drought intensity at different time scales in the Northeast, respectively, and are used to analyze the degree of change and distribution of drought intensity from 2008 to 2023.

As can be seen from Figure S5, drought intensity exhibits a consistent decreasing trend. Overall, drought intensity is relatively weak in the Sanjiang Plain, while it is higher in the Liaohe Plain. The intensity of drought in the three plains was stronger in 2009 and 2011, peaking at the SPI-180 and SPI-270 time scales, indicating that a long-term period of high-intensity drought occurred during this period. In contrast, the Liaohe Plain, Sanjiang Plain, and Songnen Plain experienced weaker drought intensity in 2021, 2023, and 2021, respectively, with drought intensity being more intense only for shorter periods.

Figure S6 shows that the trend of change in drought intensity is consistent with the trend of change in drought duration, and the trend of change in drought intensity becomes more pronounced as the time scale increases. The trend of increase in drought intensity gradually spreads in the northern part of the Northeast, but the area occupied by the trend of increase in drought intensity is relatively small. Meanwhile, the trend of apparent decrease in drought intensity gradually extends from the Sanjiang Plain area to the central and southern parts of the entire Northeast.

3.3.4. Characteristics of Spatial and Temporal Changes in Drought Severity

The severity of drought reflects the extent of its impact on agricultural production, ecological environment, and other aspects. The assessment of drought severity is crucial in understanding the causes, impacts, and development trends of drought, thereby enabling the implementation of more effective measures. Figures S7 and S8 illustrate the temporal and spatial evolution of drought severity at different time scales.

Overall, the drought severity in the Liaohe Plain, Sanjiang Plain, and Songnen Plain shows a decreasing trend, with the Sanjiang Plain exhibiting the lowest drought severity. In 2009 and 2011, the severity of drought in the three plains was particularly strong, especially in the Sanjiang Plain. Conversely, the Liaohe Plain, Sanjiang Plain, and Songnen Plain exhibited weaker drought severity in 2021, 2008, and 2021, respectively. It can be observed that the Sanjiang Plain and Songnen Plain display more prominent drought characteristics in dry years, while in non-drought years, they exhibit wetter conditions compared to other regions.

As can be seen from Figure S8, the trend of change in drought severity is consistent with the trend of change in other drought characteristics, both of which become more pronounced with the increase in time scale. The trend of increasing drought severity is primarily distributed in the northern part of the Northeast, with a slight extension as the time scale increases. Meanwhile, the trend of decreasing drought severity extends from the middle region of the Northeast, particularly in the Songnen Plain and the Sanjiang Plain, towards the south.

4. Discussion

4.1. Response of Different Land Use Types to Drought Characteristics

In recent years, the detrimental effects of drought on natural ecosystems have garnered increasing attention. Various land use types exhibit distinct physiological adaptations under drought conditions [50]. Since the SPI-30 time scale is the most sensitive in detecting drought, this study investigates the trends in drought characteristics across different land use types (dry cropland, forests, grasslands, and wetlands) in the Northeast region at the SPI-30 scale (Figure 6).

Overall, the frequency, duration, intensity, and severity of droughts across different land use types exhibited a weakening trend during the period 2008–2023. The drought conditions for the four land use types were most severe in 2011 and lighter in 2018 and 2023. Previous studies have shown that different land use types exhibit varying degrees of sensitivity to drought, thereby eliciting distinct drought responses. These differences arise from the unique structural and physiological characteristics of each vegetation type, such as stomatal conductance and light utilization efficiency [51].

Dry cropland experienced 7.8 times of drought in 2011, with a total of 205 drought days. The intensity and severity values reached -1.13 and -234, respectively. Since 2018, drought conditions in dry cropland have mitigated significantly. However, the frequency, duration, intensity, and severity of droughts occurring in dry cropland have remained at a higher level compared to other land use types. This suggests that dry croplands are more susceptible to droughts [52,53]. Grasslands exhibited the most significant fluctuations in drought characteristics. In 2011, droughts occurred on average 6.7 times, with a total duration of 182 days, an intensity of −1.18, and a severity of −218. By 2023, only four drought events occurred, with a total duration of 47 days, an intensity of −0.81, and a severity of −46. Grasslands showed the highest sensitivity to drought among all land use types [54]. This high sensitivity is likely attributed to the fact that grassland vegetation types possess shallower root systems and are more dependent on the availability of surface water [55,56]. Consequently, the rate of change in grasslands is both faster and more pronounced under drought conditions. The response of woodlands and wetlands to drought has remained relatively low. In 2011, they experienced 6.8 and 6.9 drought events, respectively, with total durations of 192 and 188 days, an intensity of -1.18, and severities of -227 and -223, respectively. This indicates that woodlands and wetlands exhibit a weaker response to drought and are more resistant to its effects [52]. The shorter, milder droughts experienced by woodlands are attributed to their deeper root systems compared to other vegetation types. These deeper root systems enable them to draw water from deeper soil layers and rely less on surface moisture [57,58]. Wetlands not only had lower drought susceptibility but also exhibited more stable and gentle fluctuations. Their resistance to drought was comparable to that of forested land, and they demonstrated a higher degree of overall stability. This can be attributed to the excellent hydrological conditions and ecosystem regulation functions of wetlands. Located in water-rich areas, wetlands possess a strong soil water-holding capacity, enabling them to effectively cope with environmental stresses such as drought [59].

It is worth noting that in 2022, woodlands and grasslands experienced extreme dryness, with the drought intensity reaching the highest level in sixteen years. In contrast, dry cropland and wetlands were less significantly affected by dry conditions in 2022. Although the Northeast experienced a high overall level of precipitation in 2022, its distribution was uneven, particularly in areas with dense vegetation cover such as woodlands and grasslands. Under the influence of the La Niña effect, the Northeast continued to experience high temperatures in 2022. The substantial transpiration from vegetation consumed large amounts of water, thereby exacerbating drought conditions. In contrast, dry cropland has sparse vegetative cover, while wetlands, with their greater water content, are less impacted. This study not only effectively identified the responses of different surface cover types to drought but also achieved significant improvements in precision and accuracy. More accurate results were obtained using daily-scale precipitation data with a spatial resolution of 2 km. This approach not only verified the reliability of previous studies but also provided a more reliable database for subsequent research.

4.2. Trends in Vegetation and Response to Drought Trends in Vegetation and Response to Drought

Precipitation is recognized as a key factor regulating the spatial and temporal dynamics of vegetation. Prolonged precipitation deficits can cause widespread drought, leading to vegetation degradation. Therefore, this study investigated the response of vegetation to the SPI-30 time scale during the May-September period from 2008 to 2023 in the Northeast (Figure S9, Figure 7 and Figure 8).

The NDVI in Northeast China showed a significant upward trend over time (k = 0.0015/a, p < 0.05). The interannual increase in NDVI showed a highly stable linear trend, indicating that the vegetation change process is characterized by persistence and regularity. Similarly, Liu et al. [60,61,62] found that the overall trend of the NDVI in Northeast China showed an improving condition. The occurrence of high-NDVI zones reflects the positive effects of ecological restoration measures on vegetation recovery in Northeast China, while the presence of low-NDVI zones may be related to extreme weather conditions during this period [63,64]. As a crucial grain production base in China, the Northeast plays an irreplaceable role in ensuring national food security and maintaining market stability. The crop growing areas in the Northeast are primarily concentrated in the Liaohe Plain, Sanjiang Plain, and Songnen Plain. During the growing period of crops, the vegetation cover in these three plains shows a significant increasing trend. This result indicates a continuous increase in vegetation cover in the region, suggesting that crops are growing well and contributing to national food security. Spatially, there were significant differences in the trends of vegetation cover in Northeast China. The overall rate of change in NDVI ranged from −0.00485 to 0.00355 per year. Areas with an increasing NDVI (slope > 0) accounted for 85.8% of the total study area, while areas with a decreasing NDVI (slope < 0) accounted for 14.2%. NDVI increases are primarily distributed in the central and southwestern parts of the Northeast, as well as the eastern plains. In contrast, NDVI decreases were observed in the Greater Khingan Mountains and Lesser Khingan Mountains in the northern part of the Northeast and higher elevations in the southeastern part. This pattern indicates that the NDVI distribution in the Northeast region exhibits significant spatial and temporal heterogeneity [65]. This may be attributed to the fact that higher-elevation areas experience lower temperatures, poorer water and heat conditions, and infertile soils, all of which are detrimental to vegetation growth. In contrast, lower elevation plains benefit from better light and water conditions, promoting better vegetation growth and a more pronounced increase in vegetation cover.

Many researchers have used the Pearson correlation coefficient to explore the relationship between drought and vegetation. By calculating the correlation coefficients between drought indices (e.g., SPI, PDSI) and vegetation indices (e.g., NDVI, EVI), the degree of vegetation response to drought can be analyzed [66]. This approach provides an important basis for ecosystem and climate research [67,68]. From the perspective of the correlation between vegetation and drought, the correlation between the SPI and NDVI during the growing season of crops in the Northeast increased with increasing months, reaching a maximum correlation of 0.29 in September. This indicates that the vegetation response to drought is most significant in September, during the late growing season [69]. In addition, we found that the correlation between the NDVI and SPI, with a one-month lag, is strongest during May–August. Particularly, the correlation is most significant in August, with a correlation coefficient of 0.39. In contrast, the lag effect is not significant in September. This is because September is a critical period for the formation of yields of major grain crops in the Northeast. During this period, water and heat conditions are better matched, favoring crop growth and development, and are less affected by drought. After September, temperatures in the Northeast region drop significantly, and temperature becomes a dominant factor influencing the NDVI. As a result, the SPI in September showed the strongest correlation with the NDVI, with no significant lag effect. The response of vegetation to drought varies across different land use types due to differences in community structure, biodiversity, and water use patterns in vegetation ecosystems [70]. Overall, grassland correlates more strongly with dry cropland, while woodland correlates less strongly with wetland. This indicates that grasslands and dry croplands are more sensitive to meteorological drought and more vulnerable to its impacts, while woodlands and wetlands are less affected [71]. Grasslands and dry croplands are primarily composed of shallow-rooted plants that can only obtain water from the surface. In contrast, woodlands have a deeper root system that allows them to access and consume more soil moisture, while wetlands are well watered [72,73,74,75].

5. Conclusions

In this study, we analyzed the characteristics of drought in Northeast China and the response of vegetation to drought using the SPI derived from high-precision daily precipitation data. The conclusions drawn are as follows:

• The SPI-30 index is most sensitive to drought identification. In terms of drought characteristics, the Liaohe Plain, Sanjiang Plain, and Songnen Plain experienced the most frequent, severe, and prolonged droughts in 2009 and 2011, with less severe droughts in 2018 and 2021. These findings correlate with the trends in total annual precipitation across the three plains. Spatially, drought trends were more pronounced in the northern part of the Northeast region, while drought conditions eased in the central and southern parts.
• From the perspective of different land use types, grasslands exhibit greater overall changes and are more sensitive to drought, while wetlands show relatively slower changes and weaker responses to drought. In comparison, dry cropland and grasslands collectively experience a slightly more severe state of drought, whereas woodlands and wetlands are less severely affected.
• Since 2008, vegetation in Northeast China has exhibited a significant greening trend. During the crop growth period, the vegetation cover of the three plains (Liaohe Plain, Sanjiang Plain, and Songnen Plain) has shown a significant increasing trend, with R² values of 0.78, 0.62, and 0.59, respectively. The correlation between the SPI and NDVI was stronger during the crop growth period. Specifically, the correlation was most significant between grassland and dry cropland and slightly weaker between woodland and wetland, and the most significant correlation was observed between the SPI and NDVI with a one-month lag.