1 Introduction

Vegetation is a crucial resource that influences ecosystem services, the global carbon cycle, and the environmental impact of climate change (Reddy et al., 2020; Weng et al., 2023). The study of temporal vegetation dynamics can serve as an environmental health indicator, linking the pedosphere, hydrosphere, and atmosphere, providing regulatory interlinkages (Chang et al., 2023). Consequently, vegetation trend analysis has been often combined with climatic data to evaluate terrain degradation and predict its future ecological health. Recent studies reveal the severe impact of climate change on vegetation growth, posing threats to environmental sustainability, biodiversity, and human well-being (Nguyen et al., 2023; Zhang & Ye, 2020). Numerous studies debated the variation in vegetation sensitivity to climate change and how the gradient of sensitivity deviates across middle and high latitudes as climate change impacts worsen in the last two scores (Afuye et al., 2021b; Zhang & Li, 2023). Climate change impacts vegetation by altering precipitation and temperature seasonality, affecting soil moisture content, photosynthesis, and effective organic carbon sequestration (Anav et al., 2015). Nevertheless, the trade-offs found in satellite sensors will advance our knowledge of degraded areas of vegetation cover and their potential impacts on induced changes in water-energy balance.

Studies have assessed the relationship between precipitation, temperature, and evapotranspiration (ETo) with the Vegetation Condition Index (VCI) and the Normalize Difference Vegetation Index (NDVI) in arid and semi-arid regions including South Africa (Mahasa et al., 2023; Mishra et al., 2021; Mpanyaro et al., 2024; Murungweni et al., 2020; Orimoloye et al., 2022). This provides a crucial insight into the mechanisms underlying the response to carbon storage in terrestrial ecosystems in response to climate change. Meanwhile, stronger climatic severity is expected in the future at varying scales (IPCC, 2014). For instance, Svenning and Sandel (2013) explored uncertainties in vegetation dynamics due to near-future climatic patterns and revealed a notable declivity, resulting in moderate loss of ecosystem structural components at continental and landscape scales Zoran et al. (2016) assessed the feedback mechanisms between climate and forest vegetation in a highly vulnerable urban forest unit using a short time gap (2000–2016). Their assessment reported a significant impact of climate change, exacerbated by atmospheric pollution, air mass dynamics, deforestation, and epidemics, on biomass production, phenological patterns, and species composition (Zoran et al., 2016). A study found a stronger correlation between NDVI and rainfall in Iran, indicating a stronger agreement between NDVI and VCI in monitoring vegetation drought dynamics by 3-month rainfall (Rahimzadeh-Bajgiran et al., 2012). In South Africa, a discovery was reported on the dissimilar variability in stream flow behaviour within quasi-local catchments, influenced by topographic complexity, and their response to declining precipitation trends (Owolabi et al., 2022). Hence, the vegetation response in complex topographies in arid and semi-arid environments is worthy of scientific attention, especially at the reduction scale. The large impact of atmospheric drive is minimal due to other site-specific conditions that mitigate climate change impact (Afuye et al., 2021a; Liu et al., 2022). Previous studies have linked variations in rainfall, rising temperatures, and higher radiation to reduced cloud cover and CO2 fertilisation to increased vegetation growth during different seasons (Hong et al., 2022; Kalisa et al., 2019; Muir et al., 2021). Few studies have assessed vegetation dynamics and their correlation with climate in South Africa due to disparities in data sources and classification algorithms (van Rooyen et al., 2018; Cho & Ramoelo, 2019; Phiri, 2020; Misra et al., 2020; Vukeya et al., 2023). Dyosi et al. (2021) found weak to moderate connections between NDVI, precipitation, and land surface temperature in the Amathole District Municipality (ADM) of Eastern Cape Province. NDVI3g and ERA5-Land precipitation show a low correlation in ADM during spring–autumn (winter) seasons, with a positive (negative) correlation across various vegetation types and climate variables (Afuye et al., 2022). Previous research has provided insights into the mechanisms by which terrestrial vegetation responds to climate change and variability (Ali et al., 2023; Zhang et al., 2018). The understanding of South Africa’s aridity and decreasing precipitation trends, particularly in the ADM region, is still limited. The understanding of bimodal climatic zones and complex topographical features controlling vegetation change rates at the landscape scale is still limited. There is a gap in understanding the bimodal climatic zones and complex topographical features controlling the rate of vegetation change at the landscape scale.

Recent climate change and anthropogenic disturbances negatively impact hydrological variability in ADM, Eastern Cape Province, leading to water level decline and high threats to nature’s services (Amathole District Municipality, 2017). These include food and water security, vegetated ecosystems, and sustainable human well-being as drought severity continues to present multiple threats to plant growth and ecosystem functioning (Graw et al., 2020). Satellite-based remote sensing has been widely used to provide valuable multi-source data for assessing long-term vegetation cover dynamics and land cover change using MODIS and Landsat observations (Afuye et al., 2024; Mu et al., 2020; Timm Hoffman et al., 2018). VCI and NDVI are sensitive vegetation indicators used for monitoring vegetation dynamics in response to climate change, providing reliable data for climate adaptation and decision-making processes (Didan et al., 2015; Pei et al., 2018). The MODIS-specific compositing method was used to assess the spatial variation of vegetation cover characteristics in seasonal VCI trends, quantifying vegetation greenness under stress (Senapati & Das, 2022). The Mann–Kendall trend test and Pearson’s correlation coefficient were utilised to establish the correlation between VCI and NDVI, temperature, precipitation, and reference evapotranspiration (ETo) series (Jiang et al., 2021; Vicente-Serrano et al., 2015). Hence, the study utilised multi-source remote sensing imagery to analyse the spatiotemporal variations and dynamics of vegetation in Amathole District Municipality (ADM), South Africa. The main objectives of the study are to (i) analyse seasonal variability in vegetation cover characteristics in green VCI quality from 2001 to 2022, and (ii) assess the extent of interannual variations and decadal trends of NDVI, considering the transitional attribute of the region. The deductions of this study are crucial for enhancing environmental planning, policy decision-making and management of degraded vegetation and forest resources.

2 Methods and materials

2.1 Study area

Amathole District Municipality (ADM) in South Africa, located in the Eastern Cape Province, between 32° 34′ 29.99° S and 27° 12′ 17.40° E. (Fig. 1). The district has a population of 892,637 and covers an area of 21,595 km2 (8,338 sq. mi) with an elevation of 1,963 m (6440 ft) above sea level (StatSSA Statistics South Africa Community Survey, 2016). The municipality is composed of six local municipalities, including Amahlathi, Ngqushwa, Great Kei, Mnquma, Mbhashe, and Raymond Mhlaba. The district is the second-richest biodiversity characterized by its subtropical east coast and the temperate south coast biogeographical areas (Nel et al., 2011). The land is diverse, encompassing well-watered coastal and semi-arid Karoo, succulent, thornveld, moist mountainous, and thicket areas (Amathole District Municipality, 2017). ADM experiences a subtropical climate with summer rainfall peaking at 400 mm/year, with fluctuations between 700 and 1000 mm/year in some areas. Monthly temperatures range from 1.5 to 2.5 °C, with winter temperatures reaching 21 °C and summer temperatures reaching 28 °C (Amathole District Municipality, 2020). ADM consists of diverse eco-topographic undulating grasslands and the Amatole Mountain range, resulting in various habitats due to shifting climate and influencing factors like soil, topography, and vegetation types.

Fig. 1
figure 1

Map of Amathole District showing the local municipalities in the Eastern Cape Province, South Africa

Full size image

2.1.1 MODIS data acquisition and pre-processing

Vegetation dynamics and variation were assessed in the study area using Terra Moderate Resolution Imaging Spectroradiometer (MODIS). Terra MODIS provides high radiometric sensitivity (12bits) at a spatial resolution of 250 m with 36 spectral bands between 0.405 and 14 μm. The Adaptive Processing System for MODIS creates levels 1A and 1B geo-location with cloud masks and high-resolution MODIS land surface-atmosphere products separated into Distributed Active Archive Centres (DAAC) for conveyance to the end-users (Sundaresan et al., 2014). To obtain the MOD13Q1 data, the area and layers of interest were mined using an area sample. The Enhanced Vegetation Index (EVI) and pixel reliability, delineating GeoTiff file format with geographic projection were convolved and analysed based on product quality assessment. The MODIS-specific compositing technique that relied on product quality assessment was used to produce the VIs at 16-day intervals and remove low-quality pixels. The pixel reliability Quality Assurance Layer of MOD13Q1 was used to mask the pixels affected by atmospheric disturbances such as cloud, orbital drift, sensor, and degradation. The layers are classified from −1 to + 5 based on the capability of the vegetation condition index ranging from healthy to unhealthy vegetation status between 0 and 1 for this assessment (Orimoloye et al., 2021). Marginal and poor data are suitably recognised as accurate in the pixel reliability bands considered for the study. The evaluation of the Terra MODIS based on the VCImax value of vegetation status experienced in each time or season represents a key factor in the quality of vegetation vigour, and it is broadly used to analyse the spatial distribution of vegetation dynamics (Ali et al., 2023; Wang et al., 2022). Information in Table 1 shows the categorical interval of vegetation status in determining vegetation canopy structure, interpretation, and classification validation.

Table 1 The description of the vegetation condition index categorical scales used in vegetation status (UN-SPIDER recommended index)
Full size table

2.1.2 Meteorological dataset

The long-term averaged meteorological parameters such as precipitation (mm), temperature 0C maximum (Tmax) and minimum (Tmin), relative humidity (%), wind Speed (m/s) maximum (WSmax) and minimum (WSmin), and solar radiation (W/m2) were employed for this study. The variables starting from 2001 to 2022 were obtained from the Application for Extracting and Exploring Analysis Ready Samples (AppEEARS) and retrieved from the National Aeronautics and Space Administration (NASA): Prediction of Worldwide Energy Resource database (https://power.larc.nasa.gov/cgi-bin/v1/DataAccess). Furthermore, the average temperature was obtained from the maximum and minimum temperatures. Related studies have demonstrated that these satellite and model-based products can provide accurate data on solar and meteorological resources in areas where surface measurements are sparse or missing (Bai et al., 2011; Tecuapetla-Gómez et al., 2021). Hence, the study methodology navigates the overall indices and climatic parameters ranging from formulas to software used for this study in Table 2.

Table 2 Overall data information and software
Full size table

The Application for Extracting and Exploring Analysis Ready Samples provided the MODIS data, while NASA’s Prediction of Worldwide Energy Resource databases provided the climate variables (AppEEARS Team, 2020). The Vegetation Condition Index (VCI) as proposed by Kogan in 1995 was utilised for this study. The VCI is based on relative NDVI adjustment with reference to the NDVI minimum historical value. As a result, the VCI compares the values inside a specific pixel in a similar period in previous years to the current Vegetation Indices (VI), like the NDVI or the Enhanced Vegetation Index (EVI), an improved version of the NDVI. Therefore, the VCI was derived using Eq. 1:

$$ VCI_{ijk} = \frac{{VI_{ijk} - VI_{i,min} }}{{ VI_{i,max} - VI_{i,min} }} \times 100 $$
(1)

where, VIijk is the weekly/monthly/DOYs VI value for pixel i in week/month/DOYj for year k, and VIimax and VIjmax are the maximum and minimum values of many years of maximum and minimum VI for pixel i respectively. Therefore, the pixel i for the VCI value is VCIijk between the day/week/month of the years (DOYj) for year k. The higher and lower values indicate healthy and unhealthy vegetative status hence, the resultant rate of the observed value of VI in the previous years was set in the range of extreme values (i.e., maximum, and minimum) as shown in Table 1. The MODIS compositing technique was used to assess the spatiotemporal vegetation variation and dynamics of VCI based on EVI, which has a specific advantage over other vegetation indices including NDVI (Orimoloye et al., 2021; Yadav & Borana, 2019). MODIS data can assess and compare dry conditions and vegetated months, and the reflected light is not affected by ground cover plants or airborne particles due to the temporal history. The time-series Landsat NDVI data were also acquired between 2001 and 2022 using Landsat-5 TM, Landsat-7 ETM + , and Landsat-8 OLI sensors. These surface reflectance imageries which cover the entire seasons were projected using the World Geodetic System (WGS84), Zone 29N, Universal Transverse Mercator (UTM) coordinate system after being resampled to 30 m resolution. They were obtained descending from the very top of the atmosphere from USGS products (https://earthexplorer.usgs.gov). In all imageries, the pixels with cloud or snow cover were removed by the mask created by (Zhu & Woodcock, 2012). For each pixel, the land-surface radiance with the highest NDVI values was mosaicked and applied to the annual imageries, and the resulting imageries were used to create a maximum value composite (MVC) (Brown et al., 2015). The variation and trends in vegetation greenness can be better represented by vegetation indices deduced from MVC values, and less likely to be affected by atmospheric effects (Walther et al., 2016). The homogenised regression-based method was used to calculate surface reflectance from Landsat 5 TM, Landsat 7 ETM + and Landsat 8 OLI sensors (Su et al., 2017). Regression equations were computed using the mean values of the MVC of surface-reflectance imagery within each sample. The entire NDVI and MVC imagery from Landsat 5TM, Landsat 7 ETM + , and Landsat 8 OLI for the corresponding years were calibrated using the linear regression equations (Erasmi et al., 2021). The NDVI was integrated with climate datasets to assess the role of climate variability and change in response to spatiotemporal vegetation variation and dynamics in the study area. Hence, the NDVI is calculated as:

$$ {\text{NDVI }} = \frac{{\left( {{\text{NIR}}{-}{\text{RED}}} \right)}}{{\left( {{\text{NIR}} + {\text{RED}}} \right)}} $$
(2)

where, the NIR and RED are the averaged reflectance values of NIR within waveband ranges chosen to match the data in the near-infrared and red channels.

2.2 Study methods

In this study, the case study research design used multi-source remote sensing imagery based on comparative analysis to assess vegetation variation and dynamics in the study area. All analyses were carried out at seasonal and annual scales. This study categorises summer months as December–February (DJF), the autumnal months as March–May (MAM), the winter months as June–August (JJA), and the spring months as September–November (SON). Terra MODIS VCI was employed because of its spatiotemporal capabilities in detecting vegetation variation and dynamics (Chere et al., 2022; Reddy et al., 2020). The MODIS-specific compositing method was used to evaluate the spatial variation of vegetation cover characteristics of the seasonal VCI trends in quantifying the variation of vegetation greenness under stress (Senapati & Das, 2022). The Landsat NDVI time series was utilised to assess the extent of interannual variations and decadal trends of NDVI during this period (Erasmi et al., 2021), considering the transitional attribute of the region. Figure 2 shows modified planning, pre-processing, and data-processing techniques adapted from the United Nations Platform for Space-based Information for Disaster Management and Emergency Response (UN-SPIDER) proposed procedures for drought monitoring using vegetation (https://www.un-spider.org/advisory-support/recommended-practices/recommended-practice-drought-monitoring-using-vegetation)

Fig. 2
figure 2

Chain processing flowchart for data selection criteria

Full size image

2.2.1 Pearson’s correlation analysis

The study conducted a Pearson correlation analysis between NDVI, temperature, precipitation, and evapotranspiration as dependent variables and independent variables from 2001 to 2022. This was carried out to examine the relationship between vegetation and meteorological variables (Dhanapriya et al., 2022). The correlation coefficient analysis (rxy) is computed as:

$$ r_{xy} = \frac{{\mathop \sum \nolimits_{i} \left( {x_{{i - \overline{x}}} } \right)\left( {y_{{i - \overline{y}}} } \right)}}{{\sqrt {\mathop \sum \nolimits_{i} \left( {x_{{i - \overline{x}}} } \right)^{2} - \sum \left( {y_{{i - \overline{y}}} } \right)^{2} } }} $$
(3)

where xi and yi are independent and dependent variables, and x and y are the sample mean values with a value ranging from −1 to 1 respectively.

2.2.2 Penman–Monteith method to estimate the potential evapotranspiration

The daily climatic data including the minimum and maximum temperature, precipitation, sunshine, wind speed, solar radiation, and relative humidity (RH) were assessed for missing data. The usability of the data was ascertained as missing data was less than 3% annually hence, the nearest neighbourhood method of interpolation based on 7-day trends was computed for the missing links. Consequently, where missing data is co-occurring for days in a month, the missing days were calculated from concomitant regimes in other years in accordance with the Owolabi et al., 2021 approach. The minimum and maximum temperature, precipitation, sunshine, wind speed, solar radiation, and RH were collated to simulate the reference evapotranspiration (ETo) series from 2001 to 2022. In doing so, the mean temperature was obtained from the maximum and minimum temperatures. The selected climatic datasets were summed up for their monthly average to improve the accuracy considering the impact of the interpolated values on the standard error margin. The Penman–Monteith method, recommended by the World Meteorological Organization (WMO) and Food and Agriculture Organization (FAO) was used to calculate potential evapotranspiration (PET) (Lang et al., 2017; Dimitriadou and Nikolakopoulos 2021). The computation was performed in CROPWAT software v8.0.1.1 based on Eq. (4).

$$ ET_{0} = \frac{{0.408 \times S_{VP} \times \left( {R_{n} - G} \right) + \frac{{900 \times K_{P} }}{{T_{ave} - 273}} \times v_{2} \times \left( {e_{s} - e_{a} } \right)}}{{S_{VP} + K_{P} \times \left( {1 + 0.34 \times {\upomega }_{2} } \right)}} $$
(4)

where, SVP denotes the absorption vapour pressure slope function (kPa/°C), Rn represents solar radiation (106W/m2), G is the Soil heat flux density (106W/m2) and KP is the psychometric constant (kPa/°C). While Tave represents the daily average temperature (°C), ω2 represents the wind speed at 2 m height (m/s), es denotes the saturated vapour pressure (kPa), and ea represents the exact vapour pressure (kPa).

2.2.3 Mann–Kendall trent test

Analysis of the trend pattern and significance of NDVI inter-annual and seasonal variables, and that of the mean monthly climatic data set such as rainfall, average temperature, and PET were performed. The study used the non-parametric Mann–Kendall (MK) statistical test to evaluate the amplitude of wavelets and their relationship with vegetation dynamics trends over a long-term climate change (Kendall, 1975; Mann, 1945). The MK test has been widely used to assess trends of different vegetation indices including NDVI, VCI and climate variables due to its robustness and the ability to handle outliners (Adeyeri et al., 2022; Fentaw et al., 2023). Trend analysis based on MK statistics; S computed using Eq. (5).

$$ {\text{S}} = \mathop \sum \limits_{a = 1}^{m - 1} \mathop \sum \limits_{b = a + 1}^{m} sgn\left( {p_{b} - p_{a} } \right) $$
(5)
$$ sgn \left( p \right) = \left\{ {\begin{array}{*{20}c} 1 & {if} & {p > 0,} \\ 0 & {if} & {p = 0} \\ { - 1} & {if} & {p < 0} \\ \end{array} } \right., $$
(6)

where, m is the temporal data length provided as b1, b2, b3, …., bm; p is the NDVI and climatic data value whose trends are assessed in the year a and b, and b > 1. The absolute value of S is deduced by summing the positive or negative signs in the dataset. The variance of S is obtained in relation to the structure of the independent samples (Kendall, 1975). The variable variance Var(s) is calculated using Eq. (7).

$$ Var\left( s \right) = \left[ {\left( {2m^{3} + 3m^{2} - 5m} \right) - \mathop \sum \limits_{a = 1}^{k} \left( {2t_{a}^{3} + 3t_{a}^{2} - 5t_{a} } \right)} \right] \times 18^{ - 1} $$
(7)

where, t is the number of ties for ath values, while k also symbolizes the tied groups. If ta is zero, Var(s) is equated as the expected or average value of S, E(s). Z-transformation is computed using Eq. 8 to approximate the normal distribution of S:

$$ Z = \left\{ {\begin{array}{*{20}c} {\frac{S - 1}{{\sqrt {{\text{Var}}\left( {\text{S}} \right)} }}} & {if} & {S > 0} \\ 0 & {if} & {S = 0} \\ {\frac{S + 1}{{\sqrt {{\text{Var}}\left( {\text{S}} \right)} }}} & {if} & {S < 0} \\ \end{array} } \right. $$
(8)

where, the significance level for MK computation is determined by Z statistics, and the null hypothesis suggests that there is no trend in the time series when τ is zero (null hypothesis). Alternatively, MK statistics can be interpreted as exhibiting a trend, depending on Z-transformation and its displacement to the α significance level for a two-sided test. A significant trend can be negative or positive depending on the value of Z (Owolabi et al., 2022). The consistency of trends of variables is determined using Eq. 9:

$$ - Z_{{1 - \frac{\alpha }{2}}} \le Z \le Z_{{1 - \frac{\alpha }{2}}} $$
(9)

The MK assessments utilised the XLSTAT engine in Microsoft Excel to compute the inter-annual NDVI for summer, autumn, winter and spring, revealing climatic patterns in response to climate change and variability.

3 Results

3.1 Summer (DJF)—vegetation dynamics based on the vegetation condition index (2001–2022)

The summer month spatial distribution of vegetation dynamics for the years 2005–2007, 2009, and 2018 are shown in Fig. 3a–b. During the study period, moderate to unhealthy vegetation conditions ranging from 10 to 49% were observed, with a browning trend across different areas. Vegetation cover trends from 2001–2002, 2008, and 2011–2014 show a high greening trend between 50 and 100%, indicating non-conversion of vegetation cover influenced by climate, topography, or vegetation types, particularly in thicket areas. While the years 2003–2004, 2010, 2016–2017, and 2019 depict low to sparse vegetation and reveal dry conditions between 0 and 10% vegetation index. Vegetation stress due to dryness patterns may be influenced by the severity of the summer temperature variation, ranging from warm to hot, which is primarily linked to sub-tropical conditions (Nel et al., 2011). Nevertheless, the summer months (DJF) of 2016 and 2019 were also significantly impacted by extremely dry conditions.

Fig. 3
figure 3

a–b Spatial distribution of vegetation dynamics based on vegetation condition index (VCI) between 2001 and 2022 (Summer). NB: 0–10 ≈ sparse vegetation; 10–49 ≈ unhealthy vegetation; 50–100 ≈ healthy vegetation

Full size image

In particular, the year 2020 witnessed low to sparse vegetation conditions because of the subsequent precipitation decline observed in 2019, suggesting that the area might have received a low precipitation amount. The years 2021–2022 witnessed moderate to unhealthy vegetation conditions between 10 and 49% vegetation index (Table 1). The moderate vegetation cover coincides with increasing precipitation during La Niña episodes as shown in Fig. 3a–b. The development is likely linked to the austral summer period, which typically sees most annual rainfall during Southern Africa’s peak rainy season and is likely influenced by ENSO dynamics (Manatsa et al., 2017). For instance, the Southern part of Ngqushwa local municipality was affected in the years 2004, 2010, 2015–2017, 2019, and 2020 thus, revealing unhealthy vegetation cover compared to other years during the summer months. Understanding the large-scale environment of a study area, its influence on the ocean, and other factors like topography, relief, and oceans, is crucial for comprehending the region’s climate. It is worth noting that the Kei Catchment ranges from Queenstown to the Indian Ocean within the ADM. Consequently, the upper Great Kei River's deeply incised valley, flowing into the Indian Ocean north of East London, may have influenced vegetation cover change in the region.

3.2 Autumn (MAM)—vegetation dynamics based on the vegetation condition index (2001–2022)

The spatiotemporal patterns of seasonal variation in vegetation dynamics for the autumn month in the year 2002, 2009–2010, and 2017 are shown in Fig. 4a–b. Dry spells (< 10%) led to low to sparse vegetation conditions in the quality of green vegetation cover and distribution during these years. The severity of degradation was unevenly distributed across the region, for example, the Western area of Amahlathi, Raymond Mhlaba local municipalities in 2001, 2006, 2008, 2011, and 2013–2014 were not severely affected in the autumn month compared to other districts (Fig. 4a–b). The years 2006, 2012, 2018, 2020, and 2022 depict a high greening trend between 50 and 100% vegetation index characterised by inter-annual variations in response to climate variabilities in spatial distribution.

Fig. 4
figure 4

a–b Spatial distribution of vegetation dynamics based on vegetation condition index (VCI) between 2001 and 2022 (Autumn). NB: 0–10 ≈ sparse vegetation; 10–49 ≈ unhealthy vegetation; 50–100 ≈ healthy vegetation

Full size image

The spatial distribution of vegetation dynamics reveals moderate to unhealthy vegetation conditions for the years 2004–2005, 2007–2008, 2011, 2013–2016, and 2019 between 10 and 49% vegetation index (Table 1). The years have seen a decline in vegetation and the loss of native vegetation due to recurring dry episodes. The alteration of various human activities may have severe impacts on vegetated ecosystems and the environment. In 2021, vegetation growth in Ngqushwa local municipality decreased slightly due to high precipitation variability and drier conditions in the Southern part. In 2020 and 2022, autumn vegetation conditions were found to be healthy, with a 50–100% vegetation index, possibly linked to prolonged La Niña events (Lioubimtseva, 2015).

3.3 Winter (JJA)—vegetation dynamics based on the vegetation condition index (2001–2022)

The spatial distribution of variation in vegetation dynamics for the winter month was observed in the Eastern part of Mnquma and Mbhashe, and the central areas between 2002, 2004–2005, 2007, 2009–2010, 2015, 2016, and 2017 as shown in Fig. 5a–b. During the study period, moderate to high extreme drier conditions (< 10%) in green vegetation cover quality varied across different parts of the region. During the drier 2015–2017 period, water scarcity severely impacted inhabitants and agricultural activities in the region, potentially leading to food shortages (Amathole Community Newsletter, 2017). The results from this study further revealed that the years 2002, 2004, and 2009 were extremely affected by drier conditions compared to the years 2001, 2010, 2015, and 2016. This explains the temporal changes in terrestrial vegetation in response to climate variability and change, particularly in the region. The vegetation conditions in 2006, 2008, 2014, and 2018 were moderate, with some areas being more affected than others. Conversely, from 2011 to 2013, vegetation greening was high due to reduced grazing pressure on vegetation cover, varying from 50 to 100% vegetation index.

Fig. 5
figure 5

a–b Spatial distribution of vegetation dynamics based on vegetation condition index (VCI) between 2001 and 2022 (winter). NB: 0–10 ≈ sparse vegetation; 10–49 ≈ unhealthy vegetation; 50–100 ≈ healthy vegetation

Full size image

The winter months of 2004, 2009, and 2017 in the Southern parts of Ngqushwa local municipality were more affected than the Northern parts of the municipalities (Fig. 5a–b). On account of the resilient drought conditions that have continued to grip the larger parts of the area, the central areas were more affected in 2002–2005, 2007, 2009–2010 and 2017 in the same winter month. The study area in the winter months of 2001, 2003, 2005, 2007–2008, 2010, 2014–2015, and 2018–2019 witnessed moderate to unhealthy vegetation conditions between 10 and 49% vegetation index. These years depict a spatial pattern of browning trend in natural vegetation areas. The years 2020–2022 witnessed a browning trend of vegetation from moderate to unhealthy vegetation index between 10 and 49%, suggesting that different classes of degradation were observed in the region from non-conversion of vegetation cover to land cover change. This study identified patterns of seasonal vegetation changes for the winter month ranging from multi-seasonal and multi-decadal changes of climatic patterns of vegetation dynamics.

3.4 Spring (SON)—vegetation dynamics based on the vegetation condition index (2001–2022)

The spring month also revealed the spatial distribution of variation in vegetation dynamics in the research area for the years 2005, 2009, 2014, 2017–2018, and 2021 as shown in Fig. 6a–b. These years witnessed moderate to unhealthy vegetation conditions between 10 and 49% vegetation index and depict a browning trend of vegetation distribution, identified with some years more impacted than others. The central region experienced greater impacts in the spring months of 2003, 2007–2008, and 2013 compared to 2010 and 2019 due to similar patterns of land use to land cover change. The spatiotemporal patterns of seasonal variation in vegetation dynamics in 2003, 2007, 2010, 2013, 2016, 2019 and 2021 depict low to sparse vegetation conditions. The study area experienced a low vegetation condition due to dry spells ranging from 0 to 10%.

Fig. 6
figure 6

a–b Spatial distribution of vegetation dynamics based on vegetation condition index (VCI) between 2001 and 2022 (Spring). NB: 0–10 ≈ sparse vegetation; 10–49 ≈ unhealthy vegetation; 50–100 ≈ healthy vegetation

Full size image

During the years 2001–2002, 2004, 2006, 2011–2012, and 2015, the area experienced a high greening trend between 50 and 100% vegetation index, suggesting less vegetation damage during this period. The vegetation conditions in the years 2001, 2012, 2015, 2020 and 2022 witnessed fewer dry conditions, which connotes that the area might have received a large amount of precipitation during the period. The study reveals that the district experiences varied patterns of vegetation dynamics, with low green vegetation cover (< 10%), for most seasons, affecting seasonal VCI trends. The conditions were witnessed especially in the spring months (2010 and 2019) followed by summer (2016 and 2019), autumn (2009–2010 and 2017) and winter (2002 and 2009) as shown in Figs. 3a–b, 4, 5, 6a–b. Drier conditions or droughts significantly impact vegetation growth in semi-arid environments, reducing evapotranspiration and thereby decreasing photosynthesis in plants (Walther et al., 2016).

3.5 Inter-annual and decadal trends of vegetation series and meteorological variables

Figure 7 shows the magnitude of the differences between precipitation and average temperature connotes some form of correlation between the two climatic variables. The study revealed consistent patterns in average temperature and precipitation changes, with some years showing more severe VCI than others (Figs. 3a–b, 4, 5, 6a–b). The decadal precipitation trends witnessed inter-annual variation from 2001 to 2010 indicating that the period received 120 mm of rainfall (Fig. 7). Therefore, the variations in rainfall might have influenced vegetation dynamics during the study period. The increasing trend of rainfall between 2001 and 2010 witnessed the highest peak in 2006 with 140 mm, while the other years fell between 100 and 60 mm with variations in vegetation dynamics. The study area experienced a decreasing trend in precipitation from 2011 to 2019, particularly in 2018 and 2019, with a rainfall range between 120 and 0 mm. In 2020, there was a significant increase of 52 mm in rainfall, while in 2021, there was a slight decrease of 48 mm. In 2022, the rainfall trend reached 91 mm, possibly due to prolonged La Niña events and climate patterns, indicating a significant environmental change (Maneja et al., 2020).

Fig. 7
figure 7

Inter-annual variability in precipitation and average temperature from 2001 to 2022

Full size image

Consequently, 2003, 2007, and 2009 were noted to have received more than 80 mm of rainfall, while the year 2016 received around 5 mm. The results reveal a significant decrease (< 10%) in the quality of green vegetation conditions during spring months 2003, 2005, 2007–2010, 2013, 2016, 2019 and 2021. The vegetation conditions during autumn periods from 2002–2003, 2009–2010, 2016–2017, and 2019 showed similar patterns. As a result, all these years witnessed a decline in precipitation amount between 75 mm and less than 50 mm. The winter months revealed a decline in vegetation trend between 2002, 2004–2005, 2007, 2009–2010, and 2017, while the summer months revealed the least changes, particularly between 2004, 2016–2017, and 2019–2020 respectively. Nevertheless, the VCI in these periods were among the most affected years as shown in (Figs. 3a–b, 4, 5, 6a–b, and Table 1). For instance, the precipitation in 2011 was higher compared to what was observed in 2014, 2020 and 2022 (Fig. 7). The winter months of 2007, 2009 and 2017 recorded a low amount of rainfall between 65 and 40 mm and low vegetation cover, in 2017, confirming the results of the VCI (Fig. 5a–b). The study region experienced low vegetation cover dynamics (Fig. 6a–b), as identified in the vegetation condition index results (Table 1). The study indicates that 2019, 2016, 2010, and 2004 all recorded the highest average temperatures of 18.35 °C, 18.05 °C, and 18.0 °C respectively. The results revealed that temperatures between 2006 and 2011 were the lowest at 16.6 °C and 16.45 °C, while other years witnessed temperatures between 16.65 °C and 17.8 °C (Fig. 7). The results show that the 2010 spring months were severely impacted by extreme dry events, possibly due to millennium climate change in a semi-arid region.

Tables 3 and 4 show the trend pattern of the meteorological data and vegetation series from 2001 to 2022 during the study period (Tables 3 and 4). The Mann–Kendall trends test revealed a decreasing annual precipitation time series in Amathole District Municipality (Ʈ = −0.833 and ꞵP = −12.124), with a significant increase in 2011 (p-value < 0.001) (Table 3). The trend pattern of average temperature in 2011 and 2012 shows that these years were identified to be affected by a sharp decrease in temperature as shown in Fig. 7. This might have interrupted the water balance and induced rainfall events between 130 and 120 mm respectively. Vegetation growth and trends in the region were influenced by the average temperature of 18 °C between 2004, 2010, 2016 and 2022, possibly due to surface energy increasing and decreasing with temperature (Wang & Dickinson, 2012).

Table 3 Series of Mann–Kendall trend test
Full size table
Table 4 Seasonal variation of Mann–Kendall trend test
Full size table

The Mann–Kendall trend test revealed a significant decrease in the average temperature time series in ADM (Ʈ = 0.267 and ꞵT =  + 0.023), with a significant increase in 2001 and an increasing trend in 2010 (p-value < 0.018) (Table 3). The development during these periods might have contributed to the temperature trends and is most likely influenced by the sensitivity effect of vegetation response to changes in water availability. This might have enhanced productivity and distribution. In addition, the MK trend test revealed a decreasing trend in the average temperature time series between 2011 and 2022 (Ʈ = 0.778 and ꞵT = 0.244) (p-value = 0.002) (Table 3). The trend pattern in the year 2011–2022 recorded the highest average temperatures between 18.5 °C, and 18 °C in 2016, 2019, and 2022 respectively. During these years, vegetation experienced moderate to extremely dry conditions due to increased temperature, indicating a significant change in the ecosystem.

The results show the correlation between seasonal NDVI trends and simulated evapotranspiration (ETo) using the MK test and Sen’s slope from 2001 to 2022. The analysis of inter-annual NDVI for all seasons, including summer, autumn, winter, and spring, revealed climatic patterns and ETo changes in response to climate change and variability (Tables 3 and 4). The Mann–Kendall trend test revealed that > 50% of the region experienced negative ETo anomalies between 2008 and 2010, with 2009 being the driest year, as shown in Fig. 8. Between 2002–2005 and 2014–2017, ETo anomalies exhibited a positive trend of > 40%, with 2016 being the wettest year, significantly affecting water budget, plant growth, and ecosystem functioning. Similarly, between 2021 and 2022, ETo anomalies increased by up to 91 mm, despite inter-annual variations in the region’s rainfall pattern from 2001 to 2022. Besides, the decadal trends witnessed variation in the amount of ETo with approximately 4.5 MJ/m2y−1 observed between 2001 and 2022 (Figs. 8 and 9). Drier conditions negatively impact vegetation growth, reducing evapotranspiration and affecting rainfall variation due to their negative effects on vegetation growth across seasons.

Fig. 8
figure 8

Long-term precipitation and evapotranspiration time series from 2001 to 2022

Full size image
Fig. 9
figure 9

Long-term evapotranspiration and inter-annual NDVI trends from 2001 to 2022

Full size image

The results further support the findings based on the vegetation condition index (VCI) (Figs. 3a–b, 4, 5, 6a–b) thus, revealing a notable trend at 95% confidence level for almost the entire study period. The Mann–Kendall trend test showed a significant decrease in time series ETo in the region (Ʈ = −0.556 and ꞵ = −0.365), with a notable increase from 2002 to 2005, and a corresponding decrease in 2009 and 2010 (p-value = 0.029) respectively (Table 3). The long-term trend analysis revealed a significant correlation between precipitation and ETo in the order of (P < 0.001) and (p < 0.029), with vegetation dynamics compared to the inter-annual NDVI trends (P < 0.817) (Tables 3 and 4). The results reveal a strong correlation between the simulated ETo and inter-annual NDVI trends, highlighting their interrelationship in terms of seasonal and inter-annual variations (Fig. 9 and Table 4). The relationship between the seasonal NDVI trends and simulated ETo anomalies and precipitation time series connotes that rainfall drives ETo variability (Figs. 8, 9, 10). Therefore, the assessments revealed a significant temporal relationship between rainfall anomalies and the inter-annual variation in NDVI trends compared to the ETo anomalies. The positive correlation between ETo and rainfall anomalies aligns with Pohlert (2016) and Botai et al. (2020) findings, which strongly indicated a strong correlation between ETo and rainfall variation.

Fig. 10
figure 10

Long-term precipitation and inter-annual NDVI trends from 2001 to 2022

Full size image

4 Discussion

4.1 Analysis of VCI trend dynamics

The study assessed the seasonal variation in vegetation cover characteristics of VCI and NDVI across Amathole District Municipality (ADM) using various meteorological variables from 2001 to 2022. The study reveals that temperature significantly influences the shift in green vegetation during the summer and spring months (Figs. 3a–b, 4, 5, 6a–b). The study reveals that vegetation cover characteristics are sensitive to climatic changes, indicating shifts in eco-topographic undulating grassland areas and the Amatole Mountain range (Mucina et al., 2014). Sen’s slope estimates for temperature change with respect to brownness were higher in the western region, explaining the significant drought sensitivity effects in the east. Similarly, the greenness latitudinal gradients were slightly lower in the central parts of ADM. The study reveals that extreme environmental conditions have caused northward shifts in vegetation brownness in a large area and southward movement of greenness in a small portion during autumn and winter (Figs. 4a–b and 5a–b). Temperature can improve vegetation greenness in small areas due to higher evapotranspiration rates and their impact on rainfall variation. The last decadal period experienced a significant decrease in precipitation, possibly due to a negative correlation between temperature and precipitation in a small area with negative greenness gradients. The negative correlation between vegetation indices and meteorological variables in the region could impact local and regional vegetation dynamics, impacting water-energy balance, human health, and ecosystem functioning.

The results indicate that the decrease in precipitation and rise in temperature between 2001 and 2022 pose significant threats to vegetation and ecosystems across ADM. Consequently, the recurring dry episodes may have negatively impacted water availability, food security, and plant productivity. Overall, the study found a significant decline (< 10%) in the quality of green vegetation condition index from moderate to low due to prolonged dryness in autumn and winter, affecting seasonal VCI trends (Table 1). The year 2016 and 2019 were marked by extremely dry conditions across the seasons. During these periods, the situation may have significantly impacted numerous inhabitants and water-dependent services, particularly farmers (Amathole Community Newsletter, 2017). The results align with previous research on seasonal vegetation indices during the spring and summer (Afuye et al., 2022; Cho & Ramoelo, 2019; Dyosi et al., 2021; Graw et al., 2017; Murungweni et al., 2020; Timm Hoffman et al., 2018). The 2010 VCI experienced extreme dry events in spring, while the autumn and summer months showed moderate to sparse vegetation cover (Figs. 3a–b and 6a–b and 7). This connotes that the rising temperatures may have influenced climate and vegetation changes in the region. The decline in soil moisture content during spring, autumn, and summer months may be exacerbated by edaphic factors and steppe soil conditions (Amiri & Gocić, 2021). Moreover, the 2010 VCI revealed moderate to low winter vegetation cover, indicating temporal changes in terrestrial vegetation due to climate variability and changes across ADM.

4.2 Inter-annual and decadal trend of NDVI and reference evapotranspiration series

Figure 8 displays the long-term trends of the inter-annual NDVI and reference evapotranspiration series. The first decadal series of potential evapotranspiration (PET) showed decreasing trends, while the second series from 2011 to 2016 showed positive trends (Figs. 8 and 9). Besides, the first and second quarters of PET trends experienced a significant decline between 2006–2009, and 2017–2020, respectively. The initial PET series showed inconsistent precipitation and temperature trends, indicating haphazard climate variability between 2001 and 2014, influenced by other PET parameters. This includes wind speed, relative humidity, solar radiation, and sunlight hours used to simulate the reference evapotranspiration in this study. The second decal PET trends reveal a positive disparity with negative climatic variability from 2015 to 2017, partly due to El Niño events. In the second decade, the PET trends depict a positive disparity with the negative climatic variability from 2015 to 2017 partly attributed to El Niño events. The United Nations Office for the Coordination of Humanitarian Affairs reported that southern African regions, particularly South Africa including Amathole areas, had experienced historical dry episodes (UNOCHA, 2016; Ibebuchi, 2021). The historical dry episodes resulted in a severe shortage of regional crop yield and food supply affecting approximately 25 million populations in need of relief (Funk et al., 2018). The deductions support the results and are identified as being affected by the dynamics of vegetation based on the seasonal VCI and NDVI assessed in the study region (Figs. 3a–b, 4, 5, 6a–b and Figs. 7, 8, 9, 10). The negative correlation between VCI, NDVI, temperature, and precipitation may be influenced by declining precipitation and rising temperature responses, as well as vegetation parameters (Amiri & Gocić, 2021). Studies show that vegetation variation and dynamic patterns are influenced by various ecological processes such as competition, succession, disturbance, and colonization (Han et al., 2023; Sun et al., 2015). In essence, vegetation variation and dynamic patterns may be influenced by various environmental factors like climate change, soil type, topography, and land use practices.

The study reveals that vegetation in the area varies in complexity due to wetness and dryness regimes, with a decline in ETo anomalies in 2009 and a positive increase in 2016. The significant decrease in ETo anomalies can be attributed to a decrease in wind speed due to high sensitivity and a lesser extent to a decrease in net solar radiation (Gocić & Arab Amiri, 2021). Wind speed and sunshine hours can decrease ETo anomalies in the region, while positive anomalies may result from increased wind speed, temperature trends, decreased precipitation, and high ETo anomalies (Wang et al., 2013). The PET trend from 2021 to 2022 is positive, with steady increases in precipitation, temperature, and inter-annual NDVI (Figs. 7, 8, 9, 10). The development can be attributed to the semi-arid microclimate dynamics, which are influenced by the physical and biological properties of the soil ecosystem (Afuye et al., 2022; Mndela et al., 2023). Overall, the NDVI trend analysis revealed high consistency in the first decade’s climatic series, while the second decadal series showed haphazard variability between 2018 and 2019 (Figs. 9, 10). From 2018 to 2019, PET and precipitation decreased, with inter-annual NDVI significantly influencing the highest average temperature, affecting vegetation growth and performance. Extreme temperatures significantly impact vegetation change, leading to sparse vegetation, while moderate temperatures are conducive to the growth of vegetation (An et al., 2018). The study enhances understanding of the intricate relationship between meteorological factors and vegetation conditions by integrating VCI and NDVI with meteorological variables. Thus, the scale of vegetation stress in microscale regions improves assessment accuracy, necessitating governments to implement early mitigation measures.

4.3 Limitations and future scope

The study’s limitations are due to the coarse resolution of satellite-derived vegetation indices, which may not provide adequate resolution to accurately distinguish land use, urbanisation, warming, and ecological restorations. Future research may use a higher spatial resolution dataset to better evaluate vegetation variation and sensitivity to climate change and distinguish the impacts of soil properties, land surface temperature, and human activity on vegetation cover. The VCI, with its reliability and accuracy surpassing the NDVI, has demonstrated its sensitivity to both short-term and long-term environmental responses like droughts, floods, and wildfires. However, the VCI has limitations in areas with sparse vegetation, such as deserts or urban areas, where the NDVI may not accurately reflect vegetation conditions. The VCI provides a quantitative measure of vegetation variation and dynamics in the area, but interpretation requires contextual knowledge of vegetation type, environmental factors, and land use practices. The insights from these findings can guide sustainable land management strategies through optimised tree planting, restoration policy, and improved research at national and global scales. Further assessment of big datasets like the regional coupled climate-vegetation model is needed to determine a more robust relationship between vegetation indices and meteorological factors. Therefore, predicting future variations in vegetation and climate change is crucial for developing more effective management strategies.

5 Conclusion

The MODIS-specific compositing method and Pearson correlation analysis were employed to evaluate VCI, NDVI, and meteorological variables based on Landsat and MODIS observations. The case study used multi-source remote sensing imagery to assess vegetation variation and dynamics in Amathole District Municipality, South Africa, across different seasons. The results showed that the VCI for 2010 in spring was the most affected by extreme dry events. The conditions were observed especially in the spring months (2010 and 2019), followed by summer (2016 and 2019), autumn (2009–2010 and 2017), and winter (2002 and 2009). Low vegetation conditions in these years indicate a negative spatial distribution of irregularities in precipitation and temperature across seasons. The irregularities in precipitation between 2010 and 2022 may be linked to the reference evapotranspiration series. The negative trend in autumn and winter causes severe implications on crop yields and food security in the region. The years aforementioned may have significantly impacted agricultural production and the economy. On average, the autumn and winter months experienced low vegetation, with some years (0–10%) experiencing a browning trend and much higher in the western region. The spring and summer months experienced high vegetation greenness, with significant drought sensitivity in the east due to atmospheric variations in vegetation change. The government, institutions, and policymakers can effectively address vegetation stress and drought sensitivity in ADM, safeguarding livelihoods, enhancing ecosystem services, and fostering climate resilience.

The study found a shift in inter-annual and decadal NDVI trends, with decreasing vegetation trends causing high drought sensitivity in the region. The study found a possible link between rainfall anomalies and the inter-annual variation in NDVI trends, which contrasts with ETo anomalies. A positive correlation (P < 0.001) was found between precipitation and ETo anomalies with vegetation dynamics, with small areas showing a greening trend due to increased ETo rates and temperature change. This study emphasises the importance of vegetation management on local and national scales for effective climate change adaptation in multilevel environmental governance. To address recurring dryness episodes, adaptive and sustainable interventions are crucial to reduce the impact of climate change and human-induced disturbances on nature’s services. South Africa needs to adopt integrated ecosystem-based adaptation (EbA), information-based model (IBM), and learning-based planning (LBP), including local knowledge bases. The study calls for scaling climate adaptation actions to resilient drought conditions and recommending suitable responses to rapidly changing climate conditions.