Introduction
The Yunnan-Guizhou Plateau plays a vital role in protecting river basins such as the Yangtze and Pearl Rivers [1, 2]. It is also a major global karst region, with karst areas comprising 53.4% of China’s total [3]. The cold climate and karstification render the region’s ecology extremely fragile [4].
Vegetation is crucial to ecosystems, global material cycling, energy balance, and water and carbon cycles [5-8]. It is especially important in preventing desertification in karst regions [9, 10]. Recent decades have seen the Yunnan-Guizhou Plateau affected by significant climate changes and human activities, resulting in a highly vulnerable ecology [10-12]. Considering the uncertain stability of vegetation changes, in-depth research on the region’s vegetation dynamics and their response to climate change is essential for understanding its impact on terrestrial ecosystems.
The Vegetation Index, particularly the Normalized Difference Vegetation Index (NDVI), is vital for evaluating vegetation conditions and ecological changes. NDVI applications accurately reflect surface vegetation conditions and monitor long-term, widespread changes [13-18].
Current research on the Yunnan-Guizhou Plateau’s vegetation focuses on the extensive Southwest and smaller scales, such as provincial and county levels, with few studies addressing overall NDVI changes and climate responses [3]. Tian et al. (2017) studied the spatiotemporal variation characteristics of vegetation cover in Guizhou Province and believed that the vegetation cover in Guizhou Province showed a continuous growth trend, and the growth rate of vegetation cover in karst areas was higher than that in non-karst areas [19]. Qiao et al. (2021) analyzed the spatiotemporal changes in vegetation cover in southwestern China from 2000 to 2016 and found that under the action of karst ecological restoration projects, the risk of rocky desertification can be effectively reduced [20]. Jiang et al. (2021) analyzed the spatiotemporal changes and potential influencing factors of NDVI under different altitudes and land use patterns in the southwestern region. The results showed that the critical elevation of NDVI was 3400 meters, and the responses of NDVI to climate change and human activities were different [21]. Xu et al. (2019) analyzed the influencing factors of vegetation mutation in the southwest region around 2001 and believed that on a 2.5-year time scale, vegetation mutation was influenced by climate change, especially temperature rise [22]. Xu et al. (2023) discovered a regional trend of vegetation improvement in southwest China from 2000 to 2020, with the change in NDVI being primarily influenced by meteorological factors such as minimum temperature, sunshine duration, and precipitation [2].
This study examines the vegetation status and changes in the Yunnan-Guizhou Plateau from 1981 to 2020, using trend and anomaly analysis with rainfall and temperature data. It aims to understand the past four decades’ vegetation changes and their climate responses, providing scientific support for regional ecological protection and desertification management.
Research Area and Methodology
Overview of the Research Area
The Yunnan-Guizhou Plateau, situated in southwestern China (97.5°N-109.6°N, 21.0°E-29.3°E), encompasses the provinces of Yunnan and Guizhou. As depicted in Fig. 1, the region predominantly features Broadleaf Forests (BDF), Needleleaf Forests (NDF), Tropical Monsoon Forests (TMF), Farmlands (FL), Grasslands (GL), and Shrubs (SHR), according to the China Vegetation Map (1:1,000,000) dataset. The plateau has a subtropical monsoon climate and is susceptible to severe desertification, as shown in Fig. 1. The dry season runs from November to March and the wet season from April to October, with temperature and precipitation peaking in July.
Data Sources and Processing
This study uses two NDVI products for analysis. The first, GIMMS NDVI, comes from the AVHRR GIMMS NDVI 3g.v1 dataset, covering 1982 to 34 years later. This dataset has a 15-day temporal resolution and a 1/12° spatial resolution. Monthly NDVI values are computed using the Maximum Value Composite method to mitigate the influence of cloud cover and aerosols.
The second NDVI product, MODIS NDVI, is derived from the MOD13A3 dataset and covers the period from 2001 to 2019. It offers a temporal resolution of one month and a spatial resolution of 1 km. Pixels averaging below 0.1 NDVI are considered non-vegetative. This study uses high-resolution meteorological data covering mainland China from 1982 to 2018, spanning 40 years. It has a 0.1° spatial resolution and uses monthly data for analysis. To align the datasets, vegetation and climate data were resampled to a 1/12° spatial resolution using nearest-neighbor interpolation. Land cover classifications are obtained from the China Vegetation Map (1:1,000,000) dataset. This dataset categorizes vegetation into seven distinct types: Broadleaf Forests (BDF), Needleleaf Forests (NDF), Tropical Monsoon Forests (TMF), Farmlands (FL), Grasslands (GL), and Shrubs (SHR). All other land cover is classified as ‘other’.
Methodology
Ordinary Least Squares for Vegetation Dynamics Analysis
This study uses Ordinary Least Squares (OLS) with univariate linear regression to analyze pixel-level NDVI trends. The least squares method is extensively employed in grid-based trend analysis, enabling a clear reflection of vegetation change rates [23, 24]. The NDVI trend is calculated as follows:
Here, n represents the years monitored by remote sensing, Ni is the NDVI value of the i-th growing season, and S is the regression slope. A positive value of S indicates an increasing trend in NDVI, and conversely, a negative value indicates a decreasing trend.
Anomaly Calculation Method
An anomaly is the deviation of a meteorological variable (like temperature or precipitation) from its long-term average in a specific area. In meteorology, anomalies indicate if data for a period is high or low compared to a long-term average (e.g., 30-year average) [25, 26]. The anomaly calculation formula is:
In this equation, X represents the variable at a specific time or period, M is the long-term average of that variable, and S is the standard deviation of the long-term variable.
Results and Analysis
Temporal Changes in Vegetation Growth and Hydrothermal Characteristics
Interannual Variability of NDVI
Fig. 2 shows the interannual variation trends of two NDVI products throughout the year and during the dry and rainy seasons. Vegetation cover on the Yunnan-Guizhou Plateau consistently increased during the study period. The annual GIMMS NDVI values varied from 0.62 to 0.67, showing a significant growth rate of 0.0009/a (P<0.001). During the datasets overlapping period, MODIS NDVI generally fell below GIMMS NDVI, with annual averages between 0.58 and 0.67. Importantly, MODIS NDVI’s growth rates in the annual (0.0040/a, P<0.001), dry season (0.0044/a, P<0.001), and wet season (0.0037/a, P<0.001) were significantly higher than GIMMS NDVI’s. Both NDVI datasets showed increasing trends during both the dry (November-March) and wet (April-October) seasons. GIMMS NDVI’s growth rate remained consistent across seasons, with no significant increase during the dry season (p = 0.002); however, MODIS NDVI grew faster in the dry season (0.0044/a) than in the wet season (0.0037/a).
Anomaly Analysis of NDVI and Hydrothermal Parameters
Fig. 3 displays the standardized anomalies for NDVI and climatic parameters. Throughout the study, GIMMS NDVI showed a significant upward trend. From 1982 to 1994, GIMMS NDVI mostly showed negative anomalies annually and during both dry and wet seasons. The year 1990 stood out with a significantly higher positive NDVI anomaly compared to other years. Starting in 1995, GIMMS NDVI anomalies were predominantly positive, with notable declines in 2000, 2005, and 2012 (Fig. 3a).
Conversely, the upward trend of MODIS NDVI from 2001 to 2019 was more distinct. The year 2013 was a turning point, shifting from mainly negative to mostly positive anomalies. From 2013 onwards, the anomalies were positive, reaching a peak in 2017 (Fig. 3b).
Temperature anomalies exhibited a similar upward trend from 1982 to 2018. The year 1998 was a turning point, from mainly negative to mostly positive anomalies (Fig. 3c).
Precipitation showed no significant interannual pattern. However, periods of above-average precipitation occurred between 1994-2002, with 1987-1989 and 2009-2013 being drier. The driest years for precipitation from 1982 to 2018 were 1989, 2009, and 2011 (Fig. 3d).
Fig. 3.
Standardized anomalies of NDVI and climatic parameters: a) GIMMS NDVI, b) MODIS NDVI, c) temperature, d) precipitation.
GIMMS NDVI anomaly trends closely matched those of temperature anomalies. Both showed mainly negative anomalies before 1995-1998, turning positive afterward. The notable NDVI spike in 1990 could be due to simultaneous increases in temperature and precipitation. The sharp decline in GIMMS NDVI in 2012 may be associated with the extended drought of 2009-2012, marking a 38-year precipitation low. Spatial Patterns of Vegetation and Hydrothermal Changes.
Spatial Distribution of Vegetation Growth Rate
Fig. 4 illustrates the spatial distribution of NDVI growth trends. From 1982 to 2015, the GIMMS NDVI trends across the Yunnan-Guizhou Plateau showed significant spatial heterogeneity. GIMMS NDVI growth rates ranged from -0.0057 to 0.0045 per annum (Fig. 4a). About 56.88% of the study area showed a statistically significant NDVI increase (p<0.05), indicating notable improvements in vegetation growth. Zhaotong and Qujing in Yunnan experienced the fastest vegetation growth. Conversely, a small portion (4.8%) showed a significant decline in NDVI (p<0.05), mainly in central and northwestern Yunnan and Bijie and Guiyang in Guizhou. Yuxi and Kunming in Yunnan faced the sharpest declines. For 2001-2019, MODIS NDVI growth rates ranged from -0.3588 to 0.3243 per annum (Fig. 4b). MODIS NDVI trends closely followed GIMMS NDVI, with 69.18% of pixels indicating a significant increase and 2.45% a significant decrease. The main difference between the two NDVI datasets appeared in western Guizhou, such as Bijie, where MODIS NDVI showed a significantly sharper increase compared to the more moderate trend of GIMMS NDVI.
Spatial Distribution of Hydrothermal Parameter Change Rates
Fig. 5 depicts the spatial distribution of temperature and precipitation change trends. Temperature change rates ranged from -0.022 to 0.131°C per annum (Fig. 5a). From 1982 to 2018, about 92.5% of pixels showed a statistically significant temperature increase (p<0.05), especially in central and western Yunnan (Baoshan, Dali, Kunming, Yuxi) and northeastern areas (Zhaotong). Only 1.9% of pixels showed a decreasing trend, mainly in central Guiyang and northeastern Tongren, Guizhou. Precipitation change rates ranged from -4.387 to 5.533 mm per annum (Fig. 5b). Regarding precipitation, approximately 37.6% of pixels had a positive slope, yet only 3.2% exhibited a significant increase (p<0.05), primarily in southern Yunnan and parts of central and eastern regions. Conversely, roughly 5.7% of pixels showed a significant decrease (p<0.05), scattered across areas like Dali Bai Autonomous Prefecture, Lincang, western Pu’er, and Honghe Hani and Yi Autonomous Prefecture in Yunnan. Compared to temperature, precipitation trends were generally insignificant across most of the Yunnan-Guizhou Plateau over the last four decades.
Correlation Analysis Between NDVI and Hydrothermal Parameters
Fig. 6 shows the spatial distribution of correlations between NDVI and climate parameters. From a time-series perspective, GIMMS NDVI and temperature correlation coefficients ranged from -0.74 to 0.83 (Fig. 6a). About 50.78% of pixels showed a statistically significant positive correlation (p<0.05), mainly in most Guizhou and Yunnan regions. Of these, 38.6% had a highly significant correlation (p<0.01), especially in Guizhou, southwestern Yunnan (Lincang, western Pu’er, Honghe), and parts of eastern Yunnan (southern Zhaotong, northern Qujing). Only 4.15% of pixels exhibited a statistically significant negative correlation, with 2.31% showing a highly significant negative correlation, mostly in Yunnan.
GIMMS NDVI and precipitation correlation coefficients varied from -0.52 to 0.66 (Fig. 6d). Only approximately 6.74% showed a significant positive correlation and 0.95% a significant negative correlation, mainly in northwestern Yunnan. Most regions did not exhibit a significant correlation between NDVI and precipitation.
For MODIS NDVI, temperature correlations ranged from -0.82 to 0.87 (Fig. 6g). About 17.61% of pixels showed a significant positive correlation, primarily in southern and northeastern Guizhou and sporadically in northeastern and southern Yunnan. Conversely, approximately 4.33% of pixels exhibited a significant negative correlation, mainly in central and western Yunnan.
MODIS NDVI and precipitation correlations ranged from -0.62 to 0.82 (Fig. 6j). Around 10.78% of pixels displayed a significant positive correlation, mainly in Zhaotong in Yunnan and Bijie in Guizhou, and sporadically in central Yunnan and eastern Guizhou.
Fig. 6.
Spatial distribution of correlation between NDVI and climate parameters (G refers to GIMMS NDVI, M refers to MODIS NDVI, T refers to temperature, P refers to precipitation).
Overall, significantly more pixels showed a positive correlation between NDVI and temperature than those showing a negative one. Most of the negatively correlated pixels were localized in central Yunnan. The majority of regions showed a negative but statistically insignificant correlation between NDVI and precipitation. Areas with a significant positive correlation between NDVI and precipitation were relatively few and primarily located in Yunnan Province.
Both GIMMS and MODIS NDVI showed significant positive correlations with temperature year-round and in both dry and wet seasons. Specifically, during the dry season, GIMMS NDVI had a stronger positive correlation with temperature, marked by a 0.69 correlation coefficient (Fig. 7). The correlation between MODIS NDVI and temperature in the dry season was not statistically significant. Neither NDVI dataset exhibited a significant correlation with precipitation (p<0.05). The analysis indicates that temperature has a more significant impact on vegetation changes than precipitation.
Correlation Analysis of Different Vegetation Types with Hydrothermal Parameters
Fig. 8 displays the correlation between NDVI and climate parameters across various vegetation types. The correlation between NDVI and temperature differed among various vegetation types. Broadleaf Forests (BDF), Needleleaf Forests (NDF), and Tropical Monsoon Forests (TMF) all exhibited a statistically significant positive correlation with temperature. The correlation coefficients for these vegetation types ranged from 0.46 to 0.7, with Broadleaf Forests showing the strongest correlation with temperature.
In contrast, Farmlands (FL), Grasslands (GL), and Shrubs (SHR) demonstrated diverse relationships with temperature across the two NDVI datasets. The correlation patterns were not uniform, indicating that these vegetation types are influenced by a complex set of factors, possibly including human activities.
Interestingly, neither NDVI dataset showed a significant correlation with precipitation across any of the vegetation types. This suggests that precipitation may not be a primary driver of vegetation changes for these specific vegetation categories.
Discussion
Trends in Vegetation Change
Numerous studies have consistently reported an upward trend in vegetation NDVI across various spatial scales [27-32]. Globally, between 34% and 67% of the land area has experienced significant improvements in vegetation growth [33, 34]. Within China, between 53.8% and 77.7% of the land demonstrates an increasing NDVI trend, with a statistically significant increase in approximately 29.3% of the area [35-37]. In southwestern China, some studies have indicated that between 43.7% to 60.2% of the area exhibits an increasing NDVI trend [31, 38-41]. These findings align well with the results of the present study, which covers the period from 1982 to 2019. Specifically, 56.68% of areas show significant increases in GIMMS NDVI, and 69.18% in MODIS NDVI, respectively. These percentages are significantly higher than the average levels reported for the whole of China and its southwestern region.
Fig. 7.
Correlation between NDVI and climate parameters (G refers to GIMMS NDVI, M refers to MODIS NDVI, T refers to temperature, P refers to precipitation).
Fig. 8.
Correlation between NDVI and climate parameters in different vegetation types (G refers to GIMMS NDVI, M refers to MODIS NDVI, T refers to temperature, P refers to precipitation).
This suggests that the Yunnan-Guizhou Plateau has undergone more substantial improvements in vegetation compared to other regions within the study timeframe. The estimated decadal increase rate of NDVI in the southwestern region ranges from 0.015 to 0.035 per decade [39, 42-45]. The trend analysis conducted for the Yunnan-Guizhou Plateau in this study revealed increase rates of 0.0009/a and 0.0040/a for GIMMS NDVI and MODIS NDVI, respectively, aligning with the findings of previous research.
In terms of spatial distribution, the area with a significant increase in MODIS NDVI is larger than that of GIMMS NDVI. The areas experiencing significant increases were primarily distributed in Bijie, Zhaotong, Liupanshui, southwest Guizhou, the northeast of Zunyi and Tongren, Wenshan, as well as parts of Lincang and Pu’er in the southwest. One contributing factor is the favorable water and heat conditions that promote vegetation growth in places like Zhaotong [46, 47]. Another reason is the implementation of measures to combat stony desertification through initiatives such as returning farmland to forest, mountain closure, and afforestation projects, as well as soil and water conservation efforts [48]. These actions have significantly improved the status of vegetation growth in central and western parts of Bijie, Liupanshui, and southwest Guizhou where stony desertification is concentrated within karst areas.
The areas exhibiting no noticeable changes are mainly located in southeast Guizhou, Xishuangbanna, Dali, Nujiang, Diqing, and central Yunnan. These areas have high vegetation coverage and have not been seriously damaged, such as tropical rainforests in Xishuangbanna and primary forests in southeast Guizhou.
On the other hand, there has been a significant decrease observed mainly in Kunming, Dali Qujing, Guiyang, and Anshun. The urbanization expansion in these cities has led to the reduction of vegetation area, particularly evident in ecologically fragile karst areas.
Impact of Hydrothermal Conditions on Vegetation Change
Role of Temperature and Precipitation
Global warming and regional precipitation shifts have been well-documented. Studies have revealed a notable warming trend since 1960 in Guizhou Province [40, 49]. This study, covering the Yunnan-Guizhou Plateau from 1982 to 2019, reveals that 92.5% of the area experienced a significant temperature increase, though precipitation trends are inconclusive.
Precipitation and temperature are considered the two most crucial climatic factors affecting vegetation dynamics [50], as they influence plant growth and distribution through control of photosynthesis and respiration [30, 51, 52]. Within suitable ranges, temperature and precipitation promote physiological and biochemical plant reactions [53]. The correlation between NDVI and climate factors varies across different times and spaces. A previous study found that in temperate and cold Eurasia, a decline in summer NDVI was related to decreased precipitation, increased spring temperatures enhanced vegetation growth, and summer and seasonal vegetation growth were driven by precipitation [17, 54]. Chen et al (2021). demonstrated that in western Xinjiang, temperature significantly affects NDVI in high-precipitation areas, whereas precipitation is a key factor in low-precipitation areas [43]. In the humid climate of the southwestern region, previous studies highlight that temperature has a more dominant influence on vegetation growth than precipitation does [31, 39, 41, 42, 55-57]. Significantly, in this research, 50.78% of GIMMS and 17.61% of MODIS NDVI pixels show a positive correlation with temperature. This aligns with previous findings that temperature impacts key physiological processes such as photosynthesis, respiration, and nutrient utilization [58-60].
Regional Variability and Anomalies
Interestingly, 4.15% of GIMMS NDVI pixels and 4.33% of MODIS NDVI pixels exhibit a significant negative correlation with temperature. These areas, often susceptible to drought, include Lijiang, Kunming, Chuxiong, Dali, Yuxi, northern Honghe, and Dehong [61, 62]. Elevated temperatures in these regions can exacerbate water evaporation, reducing water available for vegetation growth and thereby stressing vegetation [17]. Zhe et al. (2021) found that rising temperatures could inhibit vegetation growth in South Tibet [61], mainly due to enhanced evapotranspiration and permafrost degradation from climate warming, potentially creating soil temperature, moisture, and nutrient imbalances [63-65]. Previous research identified a significant negative correlation between NDVI and temperature in the middle and southern AHRB, areas with drought-prone grasslands [17]. These previous researches suggest that under conditions of low precipitation, rising temperatures may inhibit plant growth.
Our study identifies a positive correlation between NDVI and precipitation in 6.74% and 10.78% of areas, respectively, primarily in Kunming, Dali, Yuxi, northern Honghe, Wenshan, Diqing, and Zhaotong. The area in Yunnan where NDVI exhibits a negative correlation with temperature overlaps with the area where NDVI shows a positive correlation with precipitation. These findings imply that in arid and hot climates, even minimal precipitation can significantly benefit vegetation growth.
Dataset Discrepancies and Human Influence
The GIMMS and MODIS NDVI datasets display discrepancies. While GIMMS NDVI values are typically higher, MODIS NDVI shows a faster growth rate, broader growth area, and more pronounced trend than GIMMS NDVI. The proportion of MODIS NDVI pixels exhibiting significant changes is 71.63%, among which, the proportion of NDVI pixels showing a significant correlation with temperature and precipitation is 33.42%. Other regions experiencing significant changes may be attributed to anthropogenic influences. Yang et al. (2022) identified human activities as the dominant factor in vegetation cover improvement and degradation from 2001 to 2019, with average influences of 62% and 59%, respectively [55]. Since the year 2000, China has launched various ecological restoration projects [44, 57, 66], possibly explaining the rapid increase in MODIS NDVI and its weaker correlation with temperature. The correlation between MODIS NDVI and temperature for the entire year and dry season is comparatively weaker compared to that observed between GIMMS NDVI and temperature. GIMMS NDVI encompasses a broader temporal range than MODIS NDVI. From a statistical standpoint, the impact of hydrothermal conditions on NDVI mitigates the impact of anthropogenic activities.
Vegetation Type-Specific Correlations
Vegetation types exhibit diverse relationships between NDVI and temperature. Broadleaf forests, needleleaf forests, and tropical monsoon forests all show a significant positive correlation with temperature across both NDVIs. Conversely, farmlands, grasslands, and shrubs show varied temperature relationships in the two NDVIs, possibly reflecting their greater sensitivity to human activities. Compared to forests, farmlands, grasslands, and shrubs are more susceptible to human activities. Forest ecosystems are more stable, with hydrothermal conditions predominantly influencing changes [37, 43].
The primary focus of this study is the impact of hydrothermal conditions on vegetation growth. Future research will quantitatively evaluate the effects of human activities and extreme climatic events to clarify the ecological changes’ driving mechanisms in the Yunnan-Guizhou region.
Conclusions
This study investigates vegetation dynamics on the Yunnan-Guizhou Plateau from 1982 to 2019. By employing trend and anomaly analyses, and juxtaposing these with temperature and precipitation datasets, we have illuminated the intricate interplay between vegetation changes and hydrothermal shifts over the past four decades. Key findings include:
1. NDVI Seasonal Trends: Both GIMMS (1982-2015, 0.0009/a) and MODIS (2001-2019, 0.0040/a) datasets reveal a significant upward trend. Importantly, MODIS NDVI’s growth rate exceeds that of GIMMS, indicating improved vegetation vitality in the Yunnan-Guizhou region after 2001. Since 2000, China has implemented a series of ecological restoration projects, resulting in an intensified impact of anthropogenic activities on vegetation improvement, which can explain the accelerated growth of vegetation after 2001.
2. Spatial Dynamics: NDVI trends across the Yunnan-Guizhou Plateau show significant spatial heterogeneity. Notably, 56.88% (GIMMS NDVI) and 69.18% (MODIS NDVI) of the area show significant vegetation growth. Notably, the trend of vegetation growth in Guizhou surpassed that observed in Yunnan. Impressively, 92.5% of the study area experiences a temperature increase, while precipitation trends remain stable.
3. Hydrothermal Interactions: NDVI fluctuations reflect temperature anomalies, emphasizing temperature’s critical role. Certain years suggest drought-induced NDVI decreases, such as in 2012. Both NDVI datasets correlate with temperature annually and seasonally but show little correlation with precipitation. Approximately 50.78% (GIMMS) and 17.61% (MODIS) of pixels show a positive correlation with temperature. Meanwhile, 4.15% (GIMMS) and 4.33% (MODIS) of pixels, especially in Yunnan’s drought-prone areas, exhibit a negative correlation with temperature, indicating increased evaporation and subsequent vegetation stress. The areas with Positive correlations between NDVI and precipitation count for 6.74% and 10.78%, mainly in central Yunnan and the Bijie region, where the climate is dry and hot, precipitation benefits vegetation growth.
4. Vegetation Variability: Broadleaf, needleleaf, and tropical monsoon forests maintain a consistent positive temperature correlation in both NDVI datasets. Conversely, farmlands, grasslands, and shrubs show diverse temperature relationships, likely influenced by increased human activity.
