Landscape vulnerability assessment driven by drought and precipitation anomalies in sub-Saharan Africa

Global climate extremes are increasingly frequent and intense, especially in Africa, which is most vulnerable to climate change (de Sherbinin Clim. Change 123 23–37). However, the vulnerability of the landscapes composed of diverse ecosystems to climate extremes is far from being clearly understood. This study constructed a set of index systems based on the ‘exposure-sensitivity-adaptive capacity’ framework to assess landscape vulnerability driven by abnormal drought and precipitation in sub-Saharan Africa. In addition, correlation analysis was used to discover factors affecting landscape vulnerability. The results showed that a high level of landscape vulnerability was determined by high exposure and high sensitivity, as adaptive capacity exhibited little difference. The drought and wet events occurred in 80.9% and 51.3% of the climate change-dominated areas during 2001–2020, respectively. In areas where drought anomalies occur, about 8% of the landscapes, primarily formed by sparse vegetation and grasslands, were susceptible to drought. Moreover, in areas with abnormal precipitation, high vulnerability occurred only in about 0.6% of landscapes mostly covered by grasslands and shrubs. In addition, the intensity of landscape vulnerability driven by drought was higher than that driven by precipitation anomalies in the areas that experienced both dry and wet anomalies. Furthermore, the greater the deviation of landscape richness, diversity, and evenness from the normal climate state, the stronger the landscape vulnerability. The results add new evidence for landscape instabilities—an obvious contrast driven by drought and wetness—from the perspective of landscape vulnerability. The methodology of assessing landscape vulnerability established in this study can provide a new way to guide the regulation of landscape composition in response to frequent climate extremes on a macro level.


Introduction
Global climate change, characterized by climate warming, has become one of the most prominent risk issues. Africa is amongst the world's most vulnerable to climate change impacts, although African countries contribute the least to global greenhouse gas emissions (de Sherbinin 2013). Moreover, the spatial distribution of precipitation under global climate change will become more uneven (O'Gorman 2015), with extreme drought and wet events likely to become more common in most of Africa during this century and continue to increase under medium-tohigh emission scenarios (Han et al 2019). Frequent climate extremes inevitably have a huge impact on different levels of ecological organization, including species growth and distribution (Zimmermann et al 2009, Xu et al 2019, ecosystem structure and function (Hoover et al 2014, Frank et al 2015, as well as landscape pattern and process (Okin et al 2018). Thus far, much attention on the impact mechanism of climate change have received from the species and ecosystem levels, but less from the landscape level, which was mainly aimed at human landscapes (Perdinan and Winkler 2014) and geomorphic landscapes (Temme et al 2009, Mezösi et al 2012. But it is worth noting that, unlike the study of climate stress on discontinuous populations or ecosystems, the spatial relationship between adjacent ecosystems can be taken into account at the landscape dimension to explore the threat posed by climate change on a whole composed of different ecosystems in a given spatial and temporal scale (Opdam et al 2009). As a level from a macro perspective, heterogeneous landscapes formed by interacting ecosystems are conducive to maintain a wide range of ecosystem services and ecosystem stability in response to climate change (Turner et al 2012, Hengeveld et al 2014. Meanwhile, effective functions are simultaneously supported and constrained by sustainable landscape structures (Zang et al 2017). Landscape vulnerability driven by climate change puts the emphasis on detecting the relationship between the spatiotemporal evolution of landscape patterns and climate change, in an attempt to construct a stable and sustainable landscape pattern to prevent the potential consequences given rise to climate change. Hence, comprehensively understanding adverse impacts exerted by climate change from a landscape perspective is necessary, which is an extension and supplement to explore the ecosystem response mechanism at the macro dimension as an effective way to mitigate the risks of climate change (Mayer et al 2016, Inkoom et al 2018.
Landscape vulnerability, as an objective attribute, reflects the instability of the landscape system caused by man-made or natural factors and the sensitivity to external disturbances (Forman et al 1993). Currently, human-induced land use changes (Zhou et al 2020) and fires disturbing forest landscapes (Nitschke et al, 2008, Virah-Sawmy et al 2015 have been mostly studied as drivers of landscape vulnerability composed of natural ecosystems, while climate extremes as a possible cause remain poorly understood. In addition, methods for assessing landscape vulnerability are not yet normative. Multi-index comprehensive evaluation (Vyskupova et al 2016), and index systems based on landscape pattern (Zhou et al 2020) and landscape function (Mezösi et al 2012, Cook et al 2019 were pervasively carried out, which embodied climatic variables, as well as potential impacts and variations of landscape composition. However, the focus is attached to the internal landscape attributes without specifying detailed climate stress, and the correspondence between extrinsic climate stress and intrinsic landscape change is not clarified. It has some limitations that apply only one of these methods to quantitatively analyze the vulnerability resulted from climate change. The concept of vulnerability, the degree to which exposure, sensitivity, and adaptive capacity affect its characteristics and extent, has been extensively used for assessment (IPCC 2007). Exposure characterizes the likelihood of a system being threatened as a result of its location (Turner et al 2003a). Sensitivity is defined as system instability that can lead to rapid and irreversible changes following the disturbance of the natural environment (Thomas 2001). Adaptive capacity refers to a system's ability to maintain certain structures and functions after a disturbance (Deangelis 1980). External exposure and internal resistance jointly constituted the vulnerability (Turner et al 2003a). To adequately extend the impact of climate change to the landscape level, this study proposes a new approach under the framework of 'landscape exposure-landscape sensitivity-landscape adaptability' , specifically, to reveal the exposure risk by identifying climate abnormal events, and to evaluate the sensitivity and adaptability driven by different climate events based on the landscape pattern index. This can offer a sort of possibility to obtain scientific information on the effects of climate change at the landscape dimension.
This study made an effort to investigate the landscape vulnerability driven by increasingly frequent drought and wet events in sub-Saharan Africa, which was peculiarly prone to severe climate impacts. The research aims are: (1) to estimate the vulnerability of landscapes experiencing either category of climatic anomalies in climate-dominated regions excluding human pressures, as well as distinguish the resulting differences; and (2) to explore which factors contribute to landscape vulnerability. This research raises new insights into landscape vulnerability assessment from the perspective of ecosystems complex in areas likely to be affected by unforeseen climate change.

Study area
The study area is a staple of the African continent, which lies south of the Sahara Desert. The climate varies from equatorial and tropical to arid and subtropical climates (figure 1), with an annual average temperature between 0 • C and 30 • C (WMO 2022b). Rainfall varies considerably by region, ranging from near zero in dry areas such as the Kalahari Desert to over 3000 mm per year in the Congo-Guinean rainforests (WMO 2022a). Diverse climatic conditions have contributed to a significant richness of ecosystems, where the most widespread land cover are forests, shrubs, grasslands, and croplands accounting for approximately 40%, 20%, 16%, and 15% of the total area (Zanaga et al 2021). The plateaus lie to the east and south, while altitudes progressively diminish towards the west and north, which mostly consists of ancient mass originating from Precambrian crystalline and metamorphic rocks (Shahin 2002). Ethiopian Highlands form the largest continuous area of altitude, and the lowest point is Danakil Depression (Jones et al 2013). The major drainage basins are Figure 1. Location of the study area and climate zones in sub-Saharan Africa. Note: The climate zone layer was derived from the Köppen-Geiger climate map (https://koeppen-geiger.vu-wien.ac.at/), and areal percentages of different climate zones were calculated based on the high resolution (5 m) raster map.
the Nile, Congo, Zambezi, and Niger systems (UNEP 2010). Water is held in these rivers, large dams, lakes, and wetlands. Additionally, there is a great soil diversity, with Ferralsols dominating in humid and sub-humid zones, Lixisols in semi-arid zones, and Leptosols in desert regions (Kihara et al 2012).

Data sources and preparation
The data used in this study are listed in table 1. To assess landscape vulnerability on each grid, all downloaded datasets were firstly processed into the same coordinate system. Climate data with low resolution were downscaled to 0.041 • spatial grid based on available high-resolution products, and standardized precipitation evapotranspiration index (SPEI) and normalized difference vegetation index (NDVI) datasets were pre-processed with the same resolution using resample method, which was implemented in ArcGIS 10.2 (Esri, Redlands, CA). The land cover dataset maintained the original resolution to capture landscape patches on each grid.

Methods
The methodology is schematically illustrated in figure 2. Firstly, residual trend analysis (Evans and Geerken 2004) was used to separate the effects of climate change and human activities on vegetation dynamics, and the region dominated by climate change was selected as a basis to identify climatic anomalies for subsequent vulnerability assessment. Secondly, an assessment indicator system was constructed based on the coupling framework of 'exposure-sensitivity-adaptive capacity' to explore the impact of climatic anomalies on landscape vulnerability. Specifically, the climate change velocity was used for indicating the degree of landscape exposure to climate change, and indices composed of landscape metrics under different climate disturbance states were used to evaluate the landscape sensitivity and adaptive capacity to climate change. In addition, correlation analysis was used to identify the factors affecting landscape vulnerability. To improve the evaluation accuracy, landscape vulnerability was calculated on each 4.5 km × 4.5 km grid in ArcGIS 10.2 (Esri, Redlands, CA).

Identifying climatic anomalies in
where NDVI pre , NDVI obs and NDVI res are predicted NDVI, observed NDVI and NDVI residual, PRE and  TEM refer to precipitation (mm) and temperature ( • C), a and b are the regression coefficients, and c is a constant.
Then the trends of NDVI pre , NDVI obs and NDVI res were calculated using monadic linear regression trend analysis (Stow et al 2007): where n is the total years; NDVI i is the NDVI in the year of i. Slope > 0 indicates an increasing trend. Climate-dominated vegetation variation was identified based on Slope NDVIobs , Slope NDVIpre and Slope NDVIres shown in table 2 (Sun et al 2015).

Identifying climatic anomalies
The SPEI is a drought index that describes phases of anomalous dry and wet conditions by normalizing the difference between precipitation and potential evapotranspiration (Vicente-Serrano et al 2010). SPEI indicator is often used to identify drought and wet events due to the advantage of combining the capacity to include the effects of evaporative demand on drought and wetness assessment with multi-scalar nature of identifying impacts on diverse systems (Isbell et al 2015). An SPEI threshold less than −1.0 was discriminated as a dry event during 2001-2020 (Ayugi et al 2020), while an SPEI greater than 1.0 represented a wet event.

Landscape vulnerability assessment
Landscape vulnerability is jointly assessed by the exposure of the landscape system to climate change, as well as the sensitivity and adaptive capacity in the face of external disturbances. Generally, higher exposure and sensitivity increase vulnerability, while greater adaptive capacity is the contrary (Koley and Jeganathan 2022). Exposure, sensitivity and adaptive capacity can be combined into a landscape vulnerability index (LVI) (Ippolito et al 2010): where LEI is landscape exposure index; LSI cc is landscape sensitivity driven by climate change; LACI cc is landscape adaptive capacity to climate change. The equation not only captures immediate response of a landscape determined by exposure and sensitivity (Du et al 2022), but also takes into account that mitigation effects of adaptive capacity are only effective over longer timescales (Duan et al 2022).

Landscape exposure assessment
Exposure, which reflects only extrinsic climate conditions, is widely estimated using climate change velocity (Sachan et al 2022). As a climatic landscape metric, climate change velocity describes the degree to which a system is exposed to climate variations over time or space (Loarie et al 2009, Hamann et al 2015. It takes into account present and future climate conditions, which is conducive to identifying climate change hotspots based on climate dynamics (Lai et al 2022). The velocity is normally calculated as the ratio of temporal gradient (derived by linearly regressing the annual time series, • C yr −1 or mm yr −1 ) to spatial gradient (described as the differential of adjacent grid cells, • C km −1 or mm km −1 ) to reveal the speed of change in km per year (Loarie et al 2009). Taking multiple components into account, the velocity can be defined as the absolute sum of the velocities for annual total precipitation and mean annual temperature (Asamoah et al 2021), which is calculated using R codes (Hamann et al 2015). LEI is assessed by climate change velocity, with higher velocity suggesting greater exposure:

Landscape sensitivity assessment
Landscape sensitivity, which indicates landscape instability to disturbance, can be measured jointly by the magnitude of disturbance on landscapes and the resulting probability of landscapes suffering enormous damage (Zeng et al 2022). Disturbances generally alter landscape patterns, making patches more complex, discontinuous and heterogeneous (Sun et al 2022). Thus, the disturbance is typically quantified using landscape fragmentation, dispersion, and dominance metrics that condense the information on these landscape variations (Nematollahi et al 2022). Driven by a common feedback mechanism of external disturbances and internal dynamics, landscape sensitivity can be characterized by an index (LSI) (Xie et al 2013, Zhou et al 2020: where N is the total number of landscape type i; n i is the number of patches of landscape type i; n is the total number of patches; A i is the area of the ith landscape; A is the total area of landscape; F i is landscape fragmentation index, with high values indicating a decrease in mean patch size and an increase in the number of patches; S i is landscape division index referring to the dispersion of patch distribution in the same landscape; landscape dominance index (D i ) measures the degree to which one or few ecosystems predominate the landscape in terms of proportion; a, b and c represent weights of F i , S i , D i , set to 0.5, 0.3 and 0.2; d and e are the weights of the two components of D i , assigned as 0.4 and 0.6; landscape depletion index (H i ) represents the vulnerability of internal structure and characterizes the ability to resist external disturbances, with 0.28, 0.23, 0.19, 0.14, 0.09 and 0.04 assigned to the values of unutilized land, water area, farmland, grassland, forest and construction land (Zhou et al 2020, Xu et al 2021. Each grid's indices were obtained based on the land cover dataset in ArcGIS 10.2 (Esri, Redlands, CA). Generally speaking, the greater the deviation of landscape conditions from the original normal level during the climatic anomaly, the weaker the ability of the landscape formed by ecosystems to resist climatic disturbance (Hossain and Li 2021). Landscape sensitivity driven by climate change (LSI cc ) can be obtained by the following formula: where LSI n and LSI e are LSI values during normal years and the year climate extremes occurred.
where P i is the proportion of the landscape occupied by patch type i; m indicates the total number of patch types; SHDI, Shannon's diversity index, implies diversities of patch types and mosaic patterns; SHEI, Shannon's evenness index, means even distribution of area among patch types (Shannon and Weaverc 1964); PRD is Patch richness density, reflecting a number of different patch types present within landscape boundary (Wu 2007). These indices were calculated in ArcGIS 10.2 (Esri, Redlands, CA). Similar to the adaptive capacity of the ecosystem described by Isbell et al (2015), LACI cc is quantified as: LACI n , LACI e , and LACI e+1 represent adaptive capacity during normal years, during a climate event, and the year after a climate extreme, respectively.

Analysis of factors affecting landscape vulnerability
To exclude additional effects from other factors, the partial correlation was used to measure the net correlation between landscape vulnerability and exposure, sensitivity, and adaptive capacity. Since the landscape metrics related to sensitivity and adaptive capacity were irrelevant, the relationship between landscape vulnerability and landscape factors was analyzed using Pearson's correlation (Pearson 1895).

Drought and precipitation anomalies in climate-dominated areas
The relative contribution rate of climate change to vegetation dynamics is shown in figure 3(a). Overall, climate change is not the dominant driver of vegetation variation in the vast majority of sub-Saharan Africa. The total areas in which climate change alone contributed to vegetation variability accounted for 33.8%, especially in the 27.1% of areas with increased NDVI. The areas of vegetation change driven by human and climate synergistic factors covered 18.5%, of which the average contribution rate of climate change was 46.9%. Considering that this study aims to probe into the landscape vulnerability driven by climate change, the areas where climate change alone explained ecosystem variation served as a basis for identifying climatic anomalies to assess vulnerability. Furthermore, based on the SPEI indicator, the area proportion with drought and wet events were 80.9% and 51.3% in climate change-dominated areas, respectively (figures 3(b) and (c)), and the areas with both kinds of abnormal weather accounted for 42.7%, mainly in the southeast.

Landscape exposure, sensitivity, and adaptive capacity
As shown in figures 4(a)-(c), there was a heterogeneity in the exposure to climate change and the sensitivity of various landscapes to drought and wet disturbances, but a homogeneity in the adaptation and self-recovery ability, either. Overall, climate change velocity is 0.2 km yr −1 on average, with 35% of the regions experiencing above-average rates. The coastal regions of Guinea and Sierra Leone will exhibit the highest climate velocities with a maximum value of 0.98 km yr −1 . More than 99% of the areas had LACI cc values between 0 and 0.3, as indicated by the mean values of 0.01 and 0.02 for resilience to drought and wet events, respectively. When drought events have occurred, a higher LSI cc value, with an average of 0.54, occurred mostly in about 7% of climate change-dominated areas, including Mali, Niger, Guinee, Somalia, Namibia, Guinea-Bissau, Senega, and the junction of Angola and the Democratic Republic of the Congo. During wet events, landscape sensitivity was higher in about 9% of the areas located in southwestern Niger, southern Ethiopia, northeastern Tanzania, and south-central Namibia, with a mean LSI cc of 0.61. SHEI ranged from 0 to 1, which can be used to represent landscape dominance. Evergreen broadleaved forests, grasslands, and shrubs with relatively low SHEI values indicated that these ecosystems dominated the landscape, while sparse vegetation was mostly mixed with other ecosystems. Meanwhile, landscapes composed of sparse vegetation, grassland, and shrubland ecosystems were susceptible to drought and precipitation anomalies, suggesting that the constituted landscape could shift from a continuous, homogeneous whole to a fragmented and heterogeneous mosaic. In addition, the same ecosystem is more likely to be exposed to drought than abnormal precipitation. Yet, the sensitivity and adaptive capacity driven by abnormal precipitation is greater than those under drought (figures 4(d)-(f)).

Landscape vulnerability to climate change 3.3.1. Spatial pattern of landscape vulnerability
According to the grading of LEI, LSI cc , and LACI cc indicators, the LVI index was divided into four categories, which were low vulnerability (LVI < 0.1), moderate vulnerability (0.1 ⩽ LVI < 0.2), high vulnerability (0.2 ⩽ LVI < 0.4), and very high vulnerability (LVI ⩾ 0.4). Figure 5 shows the spatial pattern of landscape vulnerability to drought and precipitation anomalies. The majority of the landscape-covered 77.8% of the area-had a low vulnerability to drought impacts. Drought-driven landscapes were more vulnerable in Mauritania, Mali, Niger, Chad in the Sahel, Guinee, and Guinee-Bissau in West Africa, the borders between the Republic of Congo and Angola, and between Zambia and Botswana, accounting for about  8% of the area, while 0.8% of the scattered areas had the highest landscape vulnerability. Meanwhile, the landscape vulnerability in 77.8% and 21.6% of the regions affected by wet events showed low and moderate, respectively. Only 0.6% of the landscapes in south-central Niger, west Ethiopia, and the southwest Democratic Republic of the Congo were relatively vulnerable to precipitation anomalies, which were also drought-prone.
Overall, the intensity of landscape vulnerability driven by drought was higher than that driven by precipitation anomalies. For regions with co-occurring drought and wet events (figure 6(a)), the landscape index with high and very high vulnerability was about 0.02 higher on average under drought-driven than under extreme-wet conditions. The landscapes had higher vulnerability due to drought, accounting for 58%. In comparison, 35% were more vulnerable to extreme wetness ( figure 6(b)). There were no significant differences between the two anomalous events for the vulnerability of the remaining landscapes.

Factors affecting landscape vulnerability
Quantifying the correlation on the dimension of all ecosystems and the specific one showed that landscape vulnerability had a very strong positive correlation with sensitivity and a strong correlation with exposure but was weakly negatively correlated with adaptive capacity (table 3). In addition, to explore the change of landscape patterns deviating from the normal state, landscape metrics reflect differences in multi-year mean values during climate extremes and normal conditions. The result demonstrated a strong correlation between landscape vulnerability and changes in patch richness density, a moderate correlation with changes in Shannon's evenness index and Shannon's diversity index, and a weak correlation with changes in landscape fragmentation and  landscape division. Among the three factors affecting landscape vulnerability, Shannon's diversity index was relatively high and its deviation from normal steady state was larger (figure 7).

Landscape vulnerability with various compositions
Different ecosystems formed unique landscape patterns. The ecosystems in the drought-prone areas, in descending order of landscape vulnerability indices, were sparse vegetation, grasslands, shrubs, deciduous broad-leaved forest, rainfed croplands, and evergreen broad-leaved forests. In particular, 60% of the landscapes with high and very high vulnerability were composed of deciduous broad-leaved forests, grasslands, shrubs, and sparse vegetation, where the landscape structure was mostly a mixture of sparse vegetation and grasslands and a combination of deciduous broad-leaved forests and shrubs. Additionally, in the regions prone to abnormal wet events, grasslands, shrubs, deciduous broad-leaved forests, and rainfed croplands accounted for more than 60% of the areas with high and very high vulnerability, which were dominated by individual grasslands, mixed deciduous broad-leaved forests, and shrubs, combined rainfed croplands and mosaics croplands.
The greater diverse and more evenly distributed intra-landscape ecosystems related to deciduous broad-leaved forests, evergreen broad-leaved forests, shrubs, grasslands, and rainfed croplands, as well as the lower diversity and more even distribution in the landscape composed of sparse vegetation, resulted in a relatively high vulnerability (figure 8). Generally, the vulnerability of landscapes formed by deciduous broad-leaved forests and grasslands was greater under drought-driven than the one under precipitation anomalies, but as the diversity within the constituent landscapes increased, the vulnerability to abnormal precipitation became relatively higher. The vulnerability of the landscape composed of evergreen broad-leaved forests, shrubs, and rainfed croplands caused by abnormal precipitation was greater than that caused by drought. In contrast, the vulnerability of shrubs and rainfed croplands was converted to greater under drought when the species were evenly distributed and greater diversity, respectively.

Landscape vulnerability in different climate zones
Drought mainly occurred in savannah, tropical semiarid, and tropical desert climates, with an area proportion of 41.4%, 21.1%, and 20.8%, while 47.8% and 18.4% of precipitation anomalies occurred in savanna and tropical semi-arid climate zones. Figure 9 revealed that landscape vulnerability was highest in tropical desert and savannah climates, whether in the case of drought or wet events.

Discussion
Landscape vulnerability in this study added new evidence to the negative impact of drought and precipitation anomalies on African landscape. Unlike studies on specific responses of landscape to climatic disturbances, the result of this study revealed inherent properties of landscape systems. Previous studies have verified climate change's negative or positive effects on landscape level, which varies by climate variables and regions. Drought and heavy precipitation have been reported to reduce landscape diversity and the  only indicated correlations between climatic factors and landscape pattern indices. The assessment in this study was applied to explore vulnerability attributes of the landscape, consisting of the stress from climate, landscape shift, and recovery after disturbance. It reflected the difference in landscape instabilities driven by drought and wetness. Generally speaking, a high level of vulnerability can result from high exposure, high sensitivity and low adaptive capacity (Cook et al 2019). This study revealed that landscape vulnerability driven by climate change in sub-Saharan Africa was primarily determined by exposure and sensitivity to climate change (table 3) rather than adaptive capacity. It suggested that reduced exposure or sensitivity could reduce overall vulnerability. The impact of climate change on the landscape depends in part on the degree to which a system is exposed to climate change (Garcia et al 2014). To quantitatively assess this exposure, it is necessary to understand how the climate has changed over space or time (Loarie et al 2009). Climate change velocity is a vector description used to evaluate the rate of climate displacement, with longer length indicating greater exposure (Dobrowski and Parks 2016). Additionally, velocity is a simple function of changes in climatic conditions in a particular landscape and contains no biological information, which reflects the intrinsic characteristics of climatic conditions (Hamann et al 2015, Cook et al 2019).
Research on global velocity has demonstrated that Sahel Africa, southwestern Africa, and most areas in central Africa are likely to suffer greater climatic pressures (Li et al 2018, Shi et al 2021. There has been an increasing trend in the frequency and intensity of drought and wet events in Africa during this century (Ahmadalipour et al 2019). The current persistent drought in North Africa will experience the sharpest rise in the future, and the drought's severity is expected to continue to increase in West Africa and South Africa (Liu et al 2021). Meanwhile, exposed to precipitation extreme, West Africa and East Africa are projected to become high-exposure centers in the future (Chen and Sun 2021). In this study, the distribution of landscapes with high vulnerability to drought and wet events overlaps with these high exposure areas (figures 4(a), (b) and 5). Currently, the distribution of precipitation in Africa varies in space owing to complex topography and circulation, with a consequent unevenness of dry and wet conditions (CRED 2019, WMO 2021). Long-term drought persists in arid and semi-arid regions, accounting for 56% of sub-Saharan Africa (Otte and Chilonda 2002). This is particularly true for the drought-prone areas of the Horn of Africa, West Africa and the Sahel (IPBES 2018), where drought events are more widespread than wet events (figures 4(a) and (b)).
In the face of external climate change threats, the inherent anti-interference ability of the system is the main factor affecting vulnerability. Numerous studies have shed light on the negative impacts of climate change on species and ecosystem levels in Africa. On the one hand, climate change is shifting species ranges by altering niches (Broennimann et al 2006). On the other hand, intensive and frequent climatic anomalous events have different impacts on fragile ecosystems (Leal Filho et al 2021). A landscape is a dynamic mosaic of multiple patches of interacting flora or ecosystem (Şenik and Kaya 2021), in which changes are unavoidable due to patch modification. Furthermore, landscape vulnerability is related to landscape richness, diversity, and evenness, which can reflect landscape heterogeneity explaining spatiotemporal inhomogeneity and complexity of components. Landscape heterogeneity is an important feature of landscape structure and controls landscape functions and dynamic processes, which determines the disturbance resistance and resilience of landscapes (Turner et al 2003b). Landscape heterogeneity composed of sparse vegetation was weak owing to the low diversity and high evenness, while the one formed by forest was relatively strong (figure 7). Previous studies have elucidated that ecosystems dominated by herbaceous cover (sparse grasses, shrubs and savannas) are most sensitive to climate change (Hawinkel et al 2016, Ceccarelli et al 2022, whereas the expected impacts of climate change are lower in African humid forests (Asefi-Najafabady and Saatchi 2013). Likewise, this study concluded that the vulnerability was higher in landscapes composed of savannahs and shrubs and lower in tropical rainforest climates (figure 9).
Water availability (precipitation and soil moisture) has been identified as the major natural driver of landscape heterogeneity and landscape change (Musakwa and Wang 2018). The occurrence of extreme drought and wet events may further alter the characteristics of heterogeneity by changing the species richness within the mosaic and at the edges (Gwitira et al 2014). For example, related research proclaimed that increased precipitation could trigger a transition from grasslands and shrublands to savanna in the Sahel with intense climate change (Yu et al 2015, Aleman et al 2017. In Southern Africa, vegetation cover may convert to a less biologically productive and lower photosynthetic type during the dry season (Bunting et al 2018). In the face of precipitation variability, the landscape diversity and evenness comprised of sparse vegetation and grassland under drought deviated significantly from the normal steady state, and the deviation of the one was larger when grassland and shrub responded to abnormal precipitation (figure 7). The widely fluctuating heterogeneity indicates that unstable landscape systems are more susceptible to climatic anomalies. In regions with co-occurring drought and wet events, landscapes with high evenness (0.01 higher) and low diversity (0.01 lower) under drought events resulted in less heterogeneity compared to abnormal precipitation events, which contributed to higher landscape vulnerability towards drought. Meanwhile, because sufficient soil moisture enhanced vegetation activity during the growing season, precipitation anomalies were always accompanied by a high resistance (Hossain and Li 2021). Droughts could lead to a higher landscape vulnerability. In addition, unlike relatively short-duration of extreme precipitation events, droughts tend to last longer in Africa, sometimes over months or years, triggering more intense negative impacts.

Conclusion
This study demonstrated that high exposure and high sensitivity could result in a high level of landscape vulnerability in sub-Saharan Africa. Drought and wet events occurred in 80.9% and 51.3% of climate change-dominated areas during 2001-2020, respectively. About 7% of landscapes were highly sensitive to drought and 9% to precipitation anomalies, accordingly, while adaptive capacity was homogeneous throughout the area. Moreover, landscape vulnerability caused by drought was stronger than that caused by abnormal precipitation in areas that experienced both dry and wet anomalies. About 8% of landscapes, mainly composed of sparse vegetation and grasslands, were vulnerable to drought; only 0.6% for abnormal precipitation were chiefly covered by grasslands and shrubs. In addition, the greater the deviation of landscape richness, diversity and evenness from the normal climate state was, the stronger the landscape vulnerability would be.
This study examined the landscape vulnerability driven by drought and precipitation anomalies, but landscape responses to climatic events of different intensities (moderate, severe, and extreme) were not distinguished. Further quantitative study on multilevel climatic events and landscape responses on a longer time scale is needed to compare the differences in landscape vulnerability under various disturbance levels and identify the vulnerability thresholds corresponding to disturbance transitions. Additionally, it should be noted that vulnerability was assessed based on exposure, sensitivity, and adaptive capacity, but indices for measuring adaptive capacityall about landscape composition-are still insufficient despite this being an exploration. The next challenge is to incorporate landscape configuration metrics into exploring factors affecting vulnerability, as the main controlling factors of adaptive capacity are confirmed.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).