Drylands are hyper-arid, arid, semi-arid and dry-sub-humid climate zones (Figure 1.3.9) where rainfall is less than 65 per cent of the ‘potential evapotranspiration’ (i.e. the amount of evaporation that would occur if enough water were available) (Middleton and Thomas, 1992). They occupy over 46 per cent of the Earth’s surface and host 38 per cent of the world’s human population (more than 2 billion people) (Cherlet et al., 2018). Vegetation types include deserts, grasslands, shrublands, woodlands, savannas, Mediterranean forests and tropical dry forests (see and for tropical dry forests and savannas). Due to their extent and the chronic water deficit, these areas are of particular concern in the face of global changes, and so we assess them separately here, despite some overlap with the tropical forest and savanna and grassland biomes above. 

Figure: 1.3.9
Figure 1.3.9: Global distribution of dryland subtypes based on the aridity index. Source: WAD3-JRC (Cherlet et al., 2018).

Recent estimates suggest that one-fifth of drylands are degraded as a result of climatic variations and human activities (Burrel et al., 2020; about 9 per cent in IPCC SRCCL, 2019). Major pressures on drylands (Cherlet et al., 2018) include:

  • Climate change – for example, changes in precipitation, temperature, seasonal and interannual variability and frequency of extreme events. Projections indicate that some drylands might become more humid, whereas others may become drier (Huang et al., 2016; Pravalie et al., 2019). These expectations are uncertain though (Lian et al., 2021). 
  • Land use intensification – for example, grazing (the main use of drylands, at 62 per cent) (Cherlet et al., 2018), water extraction, deforestation, agriculture and urbanisation. 
  • Perturbations – for example, fires, insect outbreaks and biotic invasions.

The dynamics of drylands depend strongly on the interaction between these pressures, such as climate change and local perturbations (Rilig et al., 2023). 

Evidence for tipping dynamics 

Different lines of evidence point toward the existence of tipping dynamics in drylands, including past and current ecosystem transitions, bistability of dryland states at the global scale, thresholds along environmental gradients, and feedback mechanisms maintaining persistent dryland states.  

Abrupt transitions have historically occurred in several dryland systems. Palaeo evidence reveals abrupt shifts into and out of African Humid Periods (Pausata et al., 2020), including a notable greening of the Sahara during the early to mid-Holocene, followed by its abrupt desertification around 5,500 years ago (Shanahan et al., 2015; Claussen et al., 2017; Hopcroft and Valdes, 2021, Claussen et al., 1999). Positive/amplifying feedback mechanisms between vegetation and the monsoon in North Africa are thought to be important (Charney et al., 1975). Climate change projections suggest ‘Sahel Greening’ might partially occur again in the future (Erfanian et al., 2016; IPCC SR1.5, 2018; Dosio et al., 2021). In dune systems, stratigraphic records covering 12,000 years have found coexistence of a vegetated, stabilised state and a bare active state in dune systems in northern China, with occasional sharp shifts in time between those contrasting states and hysteresis (Xu et al., 2020). 

Shrub encroachment may also reflect tipping dynamics. Long-term data from Jornada Experimental Range (northern Chihuahuan Desert, New Mexico, USA) showed abrupt transitions from grasslands to shrublands triggered by a combination of climatic and human (i.e. overgrazing) factors during the last 150 years (Beltelmeyer et al., 2011; D’Odorico et al., 2012). Transitions from Mediterranean forests to shrublands have been reported under a combination of dry conditions, wildfires (Baudena et al., 2020; Acacio et al., 2009; Mayor et al., 2016) and herbivory (van der Wouw et al., 2011). 

Figure: 1.3.10
Figure 1.3.10: Map of drylands vulnerability to predicted changes in aridity for 2100 based on the IPCC RCP8.5 scenario (i.e. under the assumption of sustained increase in CO2 emissions). Abrupt decays in plant productivity, soil fertility and plant cover were identified beyond aridity threshold values of respectively 0.54, 0.7, and 0.8 (Berdugo et al., 2020). The map displays areas that are expected to cross one (or several) of those thresholds in aridity level. Light-grey areas are areas that are not drylands today. Figure from (Berdugo et al., 2020).

Satellite observations indicate that 5 per cent of drylands have experienced an abrupt loss of vegetation cover over the last 20 years, while 18 per cent underwent an abrupt increase in vegetation (Berdugo et al., 2022). 

Evidence suggests that, in drylands, sequential abrupt shifts in plant productivity, soil fertility and plant cover occur at increasing aridity thresholds, respectively corresponding to aridity values of 0.54, 0.7, 0.8 (Berdugo et al., 2020) (Figure 1.3.10). A higher dependence of vegetation on water has been reported at aridity values of around 0.6 (Nemani et al., 2003), producing a decline in productivity with increasing aridity (Berdugo et al., 2020) and an increase in tree mortality events with hotter droughts (Hammond et al., 2022). 

At aridity levels around 0.7, abrupt declines in vegetation are related with losses of soil fertility (Delgado-Baquerizo et al., 2013; Berdugo et al., 2020), changes in vegetation spatial structure, (Kéfi et al., 2007, 2011; Berdugo et al., 2017; Berdugo et al., 2019) which influences soil hydrological connectivity and resource loss at the landscape scale (Mayor et al., 2013; Rodriguez et al., 2018; Mayor et al., 2019), increases in the dominance of shrublands (Berdugo et al., 2020), and rapid shifts in the composition of soil microbial communities and soil functionality (Maestre, 2015; Lu, 2019; Delgado-Baquerizo, 2020; Zhang et al., 2023). 

At aridity thresholds of 0.8, abrupt decays in plant productivity and vegetation cover occur (Berdugo et al., 2020) and can lead to a nonlinear increase in soil erosion (Mora and  Lázaro, 2013; Elwell and Stocking, 1976; Francis and Thornes, 1990; Mayor et al., 2013).

Figure: 1.3.11
Figure 1.3.11: Schematic showing key feedbacks that could lead to dryland tipping. Coloured disks represent some of the main feedbacks described in the text (vegetation-rainfall in blue, biogeochemical feedback in red and ecohydrological feedback in blue). White arrows represent positive effects (an increase in the variable at the source of the arrow leads to an increase of the variable at the end of the arrow) and red negative effects. SOC stands for Soil Organic Carbon. See Mayor et al. (2019) for a more detailed version of the ecohydrological feedback.

A number of feedback mechanisms are known to occur in drylands, operating across ecosystem elements and at different spatio-temporal scales. Theoretically, such feedbacks can lead to bistability and abrupt transitions between stable states in drylands (Holling, 1973; Noy-meir, 1975; May, 1977; Scheffer et al., 2001; Walker et al., 2004), although such alternative states do not necessarily exist in all regions (Ma et al., 2023). Several key feedbacks can be identified (Figure 1.3.11):

  • Soil microbial communities (biogeochemical feedback; small scale): Microbial biomass and diversity in drylands are intricately linked to variations in water availability and organic matter (which change along the global aridity gradient; (Zhang et al,. 2023). Soil microbes, such as bacteria, fungal decomposers and mycorrhizal fungi, are fundamental for the breakdown of complex litter and organic matter. By decomposing organic matter, microbes are critical in the build-up of soil carbon stocks, which is essential in the maintenance of moisture in dry soils. Soil moisture, in turn, is needed for organic matter decomposition. 
  • Plant-plant interactions: In drylands, plants are known to facilitate the recruitment and growth of other plants, leading to the formation of vegetation patterns. The positive interactions between plants, i.e. facilitative effects, involve effects on microclimate, soil conditions and herbivores impacts:
    • Plant-soil feedback (small to medium scale): Plants enhance local soil conditions through several means, such as nutrient and water retention (‘islands of fertility’), microclimate influence and erosion prevention (Aguiar and Sala 1999; Schlesinger et al., 1990; Rietkerk et al., 2000; D’Odorico et al., 2007). These processes boost vegetation growth and contribute to the formation of spatial patterns. 
    • Ecohydrological feedbacks (medium scale): Plants aggregate and form spatial patterns of vegetation patches interspersed in a matrix of bare soil (Aguiar and Sala, 1999). The spatial connectivity of the bare soil (runoff-source areas) affects the redistribution of water, nutrients and sediments at the patch and landscape scale, which in turn shapes vegetation cover and spatial pattern (Mayor et al., 2013, 2019). These local (patch) and global (landscape) connectivity-mediated feedbacks affect the productivity and resilience of the ecosystem (Mayor et al., 2019).
    • Vegetation-herbivore feedback (medium scale): Herbivores graze/browse on palatable plants, which stimulates regrowth (McNaughton 1983); they then keep eating at the same places because the resprouts are soft and more easily digestible. This in turn allows the recruitment of unpalatable plants in areas without herbivores. An excess of grazing on palatable plants can prevent regrowth and lead to vegetation transitions from diverse, palatable to unpalatable dominated plant communities (Cingolani et al., 2005). 
  • Vegetation-fire feedback (medium to large scale): Fire can facilitate a transition from forest to shrublands. Shrublands recover faster and burn easier, generating a positive/amplifying feedback (e.g. dry Mediterranean regions in Portugal (Acacio et al., 2009), Spain (Baudena et al., 2020)). Replacement of native Mediterranean forests by pine forest plantations or invasion by exotic non-woody plants can contribute to this feedback (e.g. central Chile) (Pauchard et al., 2008; Gomez-Gonzalez et al., 2018) (see also: Tropical [] and Boreal [] forests). 
  • Vegetation-rainfall positive/amplifying feedbacks (large scale): Vegetation is largely controlled by local climate, but modelling studies suggest that it can also influence regional precipitation by modifying the atmospheric energy and water budget (Charney, 1975; Dekker et al., 2007). This large-scale albedo-precipitation and evapo-transpiration–precipitation feedback could have significant implications for ecosystem resilience. (see also: Tropical [] and Boreal [] forests).

Global dryland assessments suggest two different ecosystem states can exist at intermediate aridity levels (‘bistability’). Drylands with aridity levels between between 0.75 and 0.8 (i.e. in the transition zone between semi-arid and arid drylands) may be in one of two different states, with higher and lower vegetation cover, with large contrasts in soil fertility, nutrient capture and nutrient cycling (Berdugo et al., 2017). Observing different ecosystem states across an area with similar conditions does not in itself prove those ecosystems are bistable. However, the global tendency for these two states to emerge, combined with our understanding of feedbacks in these ecosystems and observations of threshold responses, suggests that these could represent alternative stable states in these ecosystems.

Hysteresis, where reversing the driver of change does not lead to recovery (see Glossary), can also be evidence for alternative stable states and tipping dynamics in dryland ecosystems. In Spain (NE, Ebro Valley), past overgrazing was found to interact with droughts to explain the lack of secondary succession or even decreasing normalised difference vegetation index (NDVI, a remote sensing index for vegetation cover) trends (Vicente-Serrano, 2012). Some long-term field studies provide evidence for hysteresis in drylands. For example, in the northern Chihuahuan Desert (US), grasslands shifted into shrublands dominated by Creosote Bush (Larrea tridentata) during a prolonged drought combined with overgrazing, but the recovery of grass productivity did not occur in subsequent wet years (Bestelmeyer et al., 2011). Results also suggest the possibility of crossing critical thresholds for irreversible degradation (i.e. 20 per cent plant cover in Gao et al., 2011).

Long legacy effects are consistent with the existence of hysteresis in drylands. For example, palaeoclimatic legacies, e.g., from the Last Glacial Maximum, influence soil biodiversity (Delgado-Baquerizo et al., 2017), function (Ye et al., 2019) and forest distribution (Guirado et al., 2022). For example, drylands with a wetter past now have greater levels of function and forest coverage than what would be expected for current climatic conditions (Ye et al., 2019). 

The reversibility of ecological transitions in drylands is challenging because plant growth rate is strongly limited by water scarcity and local disturbances. However, it is noteworthy that fast vegetation recovery during rainy periods has been observed at local and regional scales (Holmgren et al,.2006a, 2013). Studies have also found recovery of drylands to strong grazing pressure even at low cover levels in case of favourable weather conditions (Bestelmeyer et al., 2013). Coupling passive and active restoration of drylands to favourable climate swings can open windows of opportunity for dryland recovery (Holmgren and Scheffer 2001; Holmgren et al., 2006b; Sitters et al., 2012).

Timescale for transitions are about weeks to months for tree heat and grazing, months to decades for shrub encroachment Bestelmeyer et al., 2011; Tabares et al., 2019 and abrupt vegetation loss due to droughts Berdugo et al., 2022), and a few decades for the desertification of the Sahara (Shanahan et al., 2015; Claussen et al., 2017; Hopcroft and Valdes 2021; Claussen et al., 1999).

Assessment and knowledge gaps

Dynamical evidence of tipping points in drylands is challenging to find due to the slow dynamics of these ecosystems. Altogether, the knowledge of past transitions shows that relatively rapid changes have occurred in drylands, in particular in terms of vegetation cover, species composition and soil communities, leading to important changes for biodiversity and ecosystem functioning. Further, evidence from positive/amplifying feedbacks between different components of ecosystems, thresholds values in stressors (aridity, fire frequency, grazing) and hysteresis (lack of recovery) suggests the likelihood of future recurrence. We assess that dryland ecosystems can feature local to landscape-scale tipping points towards land degradation (medium confidence) with climate and land use change.

Core ecological questions remain, mainly on the mechanisms by which abruptness appears in drylands. We need long-term dynamical records. This is in particular true for soils; which is very relevant given that several thresholds involve soil transformations (particularly soil fertility losses). This lack of evidence in soils is even more difficult to address given that soils are themselves a slow component in an already slow ecosystem type and, unlike vegetation, can not be assessed with remote sensing. Quantification of the thresholds for herbivory pressure, fire frequency, and logging along aridity gradients is also necessary. Crucially, we need to improve our description and incorporation of social-ecological feedbacks in drylands (Reynolds et al., 2007). Indeed, dryland ecosystem transitions are associated with important social pressures and livelihood dependency, especially in developing countries, making social-ecological feedbacks critical to understand (see Section 2; Walker et al., 2004; Reynolds et al., 2007).

Several biotic mechanisms (e.g. local negative plant-patch feedbacks – Mayor et al., 2019) that confer resilience to dryland ecosystems are still not sufficiently explored, such as plant plasticity or adaptability to drought. Some mechanisms might be able to counteract abrupt changes; for example CO2 fertilisation may confer higher water use efficiency to plants, thus opposing stress caused by lack of water (Zhu et al., 2016) and possibly counteracting aridification (Peñuelas et al., 2017; Zhang et al., 2022). Also, we can refine our understanding of windows of opportunity for restoration in drylands (e.g. taking advantage of temporarily favourable climatic conditions; Holmgren and Scheffer 2001; Holmgren et al., 2006b; Sitters et al., 2012; Walker and Salt 2012).

Bezos Earth Fund University of Exeter logo
Earth Commission Systems Change Lab logo Systemiq logo
Global Tipping Points logo
Share this content