How climate, topography, soils, herbivores, and fire control forest–grassland coexistence in the Eurasian forest-steppe

(1) Climate

Most authors attribute the existence of the forest-steppe to intermediate climate, given that it occurs between the temperate forest and the continental steppe, two biomes over which climate exerts considerable control (e.g. Chibilyov, 2002; Pfadenhauer & Klötzli, 2014; Wesche et al., 2016; Erdős et al., 2018a; Wagner et al., 2020). Indeed, around the globe there are many examples of how climate constrains tree growth: arctic and alpine timberlines develop due to low temperature and arid timberlines are the result of low moisture availability (Stevens & Fox, 1991; Breshears, 2006; Bond, 2019). Consistent with these patterns, at the southern edge of the temperate forests of Eurasia, increasing climatic harshness deriving from decreasing precipitation and increasing annual temperature range (increasingly hot summers but still cold winters) plays a major role in constraining forest growth (Walter & Breckle, 1989; Schultz, 2005). This climatic harshness – defined as the combination of hot summers, cold winters, and aridity – is thus hypothesised to control forest distribution by limiting tree germination and survival. In Eurasian forest-steppes, climatic control has been confirmed for some species. For example, Dulamsuren et al. (2008b) found that the seedlings of Larix sibirica, one of the most important tree species in Mongolian forest-steppes, die in the steppe patches due to physiological damage caused by drought and high temperature, even if competition from grassland vegetation is eliminated. Similarly, Pinus sylvestris is limited primarily by low soil moisture (Dulamsuren et al., 2013). Quercus robur acorns in the sandy forest-steppes of the Carpathian Basin are often unable to germinate in grassland patches, and those that do germinate eventually suffer drought-induced mortality (Erdős et al., 2021). In addition to low moisture availability, extreme cold winters, which are typical of the interior of Eurasia due to the large distance from oceans and the dry, seldom overcast sky, can also decrease tree recruitment and growth (D'Odorico et al., 2013). Likewise, heat waves of the continental summers are also detrimental to trees, especially for isolated individuals that lack the protection of cooler microclimates of large forest patches (Shi et al., 2021).

Similar to forests, grasslands have their physiological optima under less harsh conditions, i.e. good water supply and lower temperature extremes. As evidence of this, where temperate or boreal forests are cleared to create hay meadows or pastures, highly productive grasses flourish (e.g. Rychnovská, 1993; Hejcman et al., 2013; Erdős et al., 2019b). With increasing climatic harshness towards the south, the height, density and productivity of grasses decrease; this trend continues throughout the steppe biome until grasslands are no longer viable, and deserts occur (Walter & Breckle, 1989; Schultz, 2005; Smelansky & Tishkov, 2012; Pfadenhauer & Klötzli, 2014; Li et al., 2020; Tishkov et al., 2020). In sum, both forest and grassland vitality decrease along the climatic harshness gradient, but forest vitality declines more sharply (Fig. 2A). At the intersection of the forest and grassland vitality curves, forest gives way to grassland. This Mean Climate Model suggests a sharp transition between forest and steppe, but not mosaics of forest and grasslands across broad geographic and climatic ranges (Fig. 2A).

The idea of mean climate parameters is, of course, a gross simplification of the many components of climate. The climate of forest-steppes is characterised by large interannual variation in precipitation and temperature (e.g. Walter & Breckle, 1989; Chibilyov, 2002), which results in variable levels of climatic harshness for trees. For example, the forest-steppes of the Carpathian Basin (mean annual precipitation = 500–600 mm) regularly experience years with less than 350 mm and years with more than 800 mm precipitation (Tölgyesi et al., 2016), while the long-term limit of tolerance of forests in the region is assumed to be around 500–550 mm. Wet periods may open windows for tree recruitment, whereas drier periods may prevent canopy closure and favour grassland species (Dulamsuren, Hauck & Mühlenberg, 2005b). This means that both forest and grassland vitality can have a certain range of variability along the mean climate gradient, expanding the climatically determined intersection point into a zone where neither forest nor grassland is more vital than the other on a permanent basis (Fig. 2B). As vegetation response to climate variability is often delayed (Yin et al., 2013; Hao et al., 2014), neither the forest nor the grassland can be expected to gain dominance over sufficiently long periods and over large areas, leading to forest–grassland coexistence in a mosaic pattern (House et al., 2003). This climatically determined conceptual model of forest-steppe is often referred to as the zonal forest-steppe in the literature (e.g. Molnár et al., 2012; Pfadenhauer & Klötzli, 2014; Bátori et al., 2018). This Zonal Model can explain forest–grassland coexistence only in a relatively narrow range. Thus, other factors in addition to climate have to be taken into consideration if we are to understand forest–grassland coexistence across the entire distribution of forest-steppe mosaics in Eurasia.

(2) Topography

Variations in topography can considerably modify the effect of climate by either decreasing or increasing local temperature and moisture availability in ways that affect the vitality of forests and grasslands (Walter & Breckle, 1989; Chibilyov, 2002; Schultz, 2005; Pfadenhauer & Klötzli, 2014). Topography plays a role in forest–grassland distributions within and beyond the climatically determined forest-steppe zone (Fig. 2B, C). Within the climatically determined (zonal) forest-steppes, topography influences where forest or grassland ecosystem states form and persist. Beyond this climatically determined zone, special topographical circumstances may also result in forest–grassland coexistence (Fig. 2C). This latter situation is frequently called extrazonal (e.g. Zolotareva, 2020), although we know of no substantial difference between the physiognomy of zonal and extrazonal forest-steppes, and their species compositions are similar (e.g. Borhidi, 2004).

The importance of topography is especially evident in the Inner Asian forest-steppe region (Mongolia, north and northeast China, and south Russia), where steep north-facing mountain slopes are usually covered by forests, steep south-facing slopes are occupied by steppes, and less extreme exposures can support either ecosystem state (e.g. Liu et al., 2000; Dulamsuren et al., 2005b; Anenkhonov et al., 2015; Hais, Chytrý & Horsák, 2016; Makunina, 2017). Liu et al. (2012) showed that topography controls forest and steppe distribution mainly through soil moisture. North-facing slopes receive a reduced amount of direct solar radiation, resulting in lower evaporation and, consequently, better soil moisture supply. This local decrease in aridity increases the vitality of forests relative to the steppe (Fig. 2C). By contrast, higher direct solar radiation on south-facing slopes increases temperature and reduces soil moisture. The associated local increase in aridity and heat stress decreases forest vitality relative to steppe vitality.

Ravines, erosion gullies, and depressions have cool and moist microclimates and increased soil water supply. Consequently, they support forests embedded among steppes in West Siberia (Lashchinsky, Korolyuk & Wesche, 2020) and eastern Europe (Walter & Breckle, 1989; Goncharenko & Kovalenko, 2019). Even very small topographical features may permit the formation of forest–grassland mosaics. For example, in the forest-steppes of western Siberia and northern Kazakhstan, shallow saucer-like depressions harbour circular forest patches in a steppe matrix, due to increased moisture input (Lavrenko & Karamysheva, 1993; Rachkovskaya & Bragina, 2012; Lashchinsky et al., 2020). Similarly, small and shallow depressions support forest patches in the Carpathian Basin (Borhidi, Kevey & Lendvai, 2012) (Fig. 2C).

(3) Soil

Soil properties also profoundly influence water and nutrient availability for plants and thus are able significantly to influence forest and grassland distribution (Schultz, 2005; Pfadenhauer & Klötzli, 2014; Zech, Schad & Hintermaier-Erhard, 2014). Similar to topography, soils can modify both forest and grassland vitality within the climatically determined forest-steppe zone, and also broaden the forest-steppe zone in both directions along the harshness gradient (Fig. 3). In mosaics of the forest-steppe, soils beneath forests usually differ from those below grasslands, but it is often difficult to determine if these differences are primarily due to substrate or caused secondarily by the vegetation itself (Walter & Breckle, 1989). There are some cases in which primary soil characteristics apparently play a decisive role in forest versus grassland occurrence. For instance, gravelly soils within the Mongolian forest-steppe usually support the forest ecosystem state (Wallis de Vries, Manibazar & Dügerlham, 1996; Dulamsuren et al., 2009), apparently because coarse-texture soils permit rapid infiltration of precipitation to deeper soil layers where it is accessible by deep rooted woody plants, but not grassland species (Fig. 2C). Coarse soil texture can also contribute to the emergence of forest-steppe beyond its climatically determined interval (Fig. 2C). In the Naurzum Nature Reserve of Kazakhstan, a vast sandy forest-steppe occurs surrounded on all sides by pure steppic grassland matrix associated with loamy and clayey soils (Rachkovskaya & Bragina, 2012; Bátori et al., 2018). In a reversal of this pattern, in high-precipitation regions with a preponderance of temperate forest, shallow rocky soils often support patches of steppe-specialist plant species (Erdős et al., 2014b; Boch et al., 2019).

(4) Herbivory

Herbivory by large mammals is regarded as one of the main factors controlling the relative abundances of woody and herbaceous plants in savannas and forest–grassland mosaics. In tropical savannas grazers tend to increase, while browsers tend to decrease, woody cover (Roques, O'Connor & Watkinson, 2001; Augustine & McNaughton, 2004; Sankaran et al., 2005; Bond, 2008; Archer et al., 2017). Such effects may be dependent on herbivore pressure: Sankaran, Ratnam & Hanan (2008) found that grazers of African savannas increase woody abundance only at high grazing pressure, while low and medium grazing pressure have an opposite effect. Similarly, for semi-arid African savannas, Asner et al. (2004) and Archer (2010) concluded that heavy grazing increases woody plant abundance. In contrast to African ecosystems, the distinction between grazers and browsers is less clear in temperate regions (Owen-Smith, 2008). In the Eurasian forest-steppe, there is no evidence of grazer-induced woody encroachment. Here, in addition to browsers such as various species of deer (Cervidae) and goats (Capra spp.), animals that are typically considered grazers such as horses (Equus spp.), cattle (Bos taurus Linnaeus), European bison (Bison bonasus Linnaeus), and sheep (Ovis spp.) also feed on woody plants. Such browsing by ‘grazers’ combined with their trampling, wallowing, and uprooting of trees limits forest expansion into grasslands (Walter & Breckle, 1989; Wallis de Vries et al., 1996; Sankey, 2012). Grazers may also alter soil moisture availability indirectly by preventing the accumulation of dead plant material, which increases evaporation from the topsoil, rendering grasslands less suitable for tree seedlings (Walter & Breckle, 1989).

In addition to wild native herbivores, domestic ungulates are important to the ecology of the forest-steppe. Sheep, cattle, goats and horses are all regarded as limiting factors for tree establishment and survival in livestock-producing areas of Eurasia (e.g. Wallis de Vries et al., 1996; Smelansky & Tishkov, 2012; Hais et al., 2016; Török et al., 2018). In Mongolia, Khishigjargal et al. (2013) found that livestock grazing can effectively limit forest encroachment at grassland edges by reducing sapling number through trampling. In temperate pastures of Mongolia, goats consume tree saplings even when fresh herbs are available (Lkhagvadorj et al., 2013). In both Hungary and Mongolia, livestock prevent shrub establishment in grazed grasslands, whereas in areas with herbivore exclusion, shrubs can establish and survive (Varga et al., 2015; Takatsuki et al., 2018).

The capacity of large native herbivores to push forest–grassland balance towards grasslands is generally accepted in the temperate zone of Eurasia (e.g. Lavrenko & Karamysheva, 1993; Vera, 2000; Wagner et al., 2020) and other temperate regions (Bredenkamp, Spada & Kazmierczak, 2002). Great populations of now-threatened or extinct Holocene herbivores such as tarpan (wild horse, Equus ferus Boddaert), takh (Przewalski's horse, E. przewalskii Poliakov), onager (Asian wild ass, E. hemionus Pallas), wild ox (Bos taurus primigenius), Eurasian elk (Alces alces Linnaeus), and saiga antelope (Saiga tatarica Linnaeus) once inhabited the Eurasian forest-steppe and certainly influenced forest–grassland dynamics (Walter & Breckle, 1989; Chibilyov, 2002; Pfeiffer, Dulamsuren & Wesche, 2020; Török et al., 2020; Wagner et al., 2020). Although the historical population sizes of these large native herbivores are unknown, some authors assume that low densities of domestic livestock may serve a similar ecological function to maintain grasslands (Wallis de Vries et al., 1996; Wesche & Treiber, 2012; Pfeiffer et al., 2020).

In addition to large ungulates, other important groups of animals in the forest-steppe are rodents and insects. Hamster (Cricetus cricetus Linnaeus), marmots (Marmota spp.), and voles (e.g. Microtus spp. and Myodes spp.) (Walter & Breckle, 1989; Lavrenko & Karamysheva, 1993; Chibilyov, 2002) consume seeds and seedlings of trees, and thus may limit tree establishment in the grassland ecosystem state and at the forest edge (Dulamsuren et al., 2008b; Hauck et al., 2008). Insects such as orthopterans and gypsy moth (Lymantria dispar Linnaeus) contribute to tree mortality by defoliating seedlings in the grassland ecosystem state (Dulamsuren et al., 2008b) and damaging both seedlings and mature trees at the forest edges (Hauck et al., 2008).

In sum, where herbivory disproportionately damages woody plants relative to grasses and forbs, forest vitality is reduced and grasslands may occupy areas where the climate is humid enough and soil moist enough theoretically to support forests. In light of the extensive evidence that the forest-steppe developed under the influence of a rich assemblage of Holocene large herbivores, and is now maintained by both native animals and domestic livestock, we suggest that our understanding of the coexistence of forests and grasslands should incorporate herbivory (Fig. 3), not just climate, soils, and topography (Fig. 2C).

(5) Fire

Most grasses and forbs are able to resprout after a fire event relatively quickly from underground organs and regenerate from the seedbank, whereas woody species, except some fire-tolerant or resprouting ones, need decades if not centuries to reestablish (Bond, 2008). Although few Eurasian studies examine the effects of fire on vegetation in general, and on the forest–grassland balance in particular (Valkó et al., 2014), fire is regarded as being capable of limiting woody vegetation, even in moist sites that would otherwise permit development of forests (e.g. Walter & Breckle, 1989; Korotchenko & Peregrym, 2012). According to Kertész et al. (2017) and Ónodi et al. (2021), severe wildfires are able to eliminate the forest ecosystem state from the forest-steppes, shifting the forest–grassland balance in favour of grasslands. Forest patches containing Juniperus communis L. are particularly vulnerable to fires, as juniper is highly flammable and cannot resprout (Kertész et al., 2017; Ónodi et al., 2021). Erdős (2014) found that wildfires in forest-steppes can open up the canopy layer, and the regeneration of the forest may take several decades. Pinus sylvestris of large diameter are able to withstand surface fires of low to medium intensity (Wirth, 2005), but not high-intensity crown fires; Pinus sylvestris stands killed by fire can be very slow to recover, requiring decades to regrow (Ivanova et al., 2010; Barrett et al., 2020).

Because humans are responsible for many fires today, the current frequency of fires in the forest-steppe is often regarded as ‘unnatural’. While it is true that fire has long been used by humans to prevent woody encroachment into grasslands and to maintain pastures for livestock (Smelansky & Tishkov, 2012; Valkó et al., 2014; Novenko et al., 2016; Unkelbach et al., 2018), burning by humans may be viewed as perpetuating fire as an ancient ecological process in the region. Indeed, palaeoecological evidence suggests that natural (lightning-ignited) wildfires regularly occurred in many regions of the forest-steppe, including the Carpathian Basin (Magyari et al., 2010); the Mongolian Altai (Unkelbach et al., 2018), and European Russia (Novenko et al., 2018). This may not be recognised, because fires today are usually suppressed near human settlements. But in remote forest-steppe regions fire continues to play an important ecological role to maintain grasslands in places that could otherwise develop into forests (e.g. Kertész et al., 2017; Erdős et al., 2018a; Kolář et al., 2020; Wagner et al., 2020). In contrast to tropical savannas of C₄ grasses, which can burn annually, wildfires are much less frequent in forest-steppes: recent research indicates that fire-free intervals in Eurasian forest-steppes have ranged from several years to a couple of decades or even centuries during the Holocene, with considerable temporal variations due to climatic modifications and human activity (Ivanova et al., 2010; Hessl et al., 2012, 2016; Feurdean et al., 2013; Novenko et al., 2018; Rudenko et al., 2019; Kolář et al., 2020). Generally, fires in forest-steppes are more frequent than in boreal forests but less frequent than in open grasslands of the steppe biome (Barrett et al., 2020).

In sum, fire is able to limit forest vitality, and thus modify forest–grassland proportions anywhere in the forest-steppe, reducing tree cover below the potential allowed by climate, soil, and topography. For our understanding of the wide climatic and geographic distribution of the forest-steppe, the effects of fire are most important at the humid end of the climatic harshness gradient (Fig. 3). Here, fire is not just a modifier but, alongside herbivory, is essential to prevent canopy closure, and enable long-term forest–grassland coexistence.

(6) Vegetation feedbacks and alternative ecosystem states

Emerging theory on grassland–forest coexistence and the distribution of savanna and forest biomes details how vegetation feedbacks that reinforce either grass or tree dominance contribute to the stability of alternative ecosystem states under the same climate (Staver et al., 2011; Hirota et al., 2011; Murphy et al., 2016; Staal et al., 2018a,b). In the tropics, these ideas have focused on the distinct and generally opposite influences of grasses and trees on ecosystem flammability (fire), forage quantity and quality (herbivory), resource availability (e.g. light, water, nutrients), microclimate (temperature and humidity), and tree establishment and survival (Hoffmann et al., 2012; Murphy & Bowman, 2012; Pausas & Dantas, 2017). Based on our review of literature from the forest-steppe, we suggest that vegetation feedbacks are also important for understanding the distributions and stability of grassland–forest mosaics in Eurasia. These feedbacks are critical to the interpretation of our hierarchical models, in which grassland and forest plant communities are not merely passive entities whose distributions are determined by combined effects of climate variability, soils, topography, herbivores, and fire. Instead, we view trees and herbaceous plants of the forest-steppe as active ecosystem engineers, who themselves influence forest and grassland vitality across a wide geographic range in Eurasia.

Trees of the forest-steppe have strong feedbacks on local conditions beneath their canopy. Tree canopies intercept solar radiation, leading to low light availability, cooler diurnal temperature and higher relative air humidity at the forest floor, and the canopy reduces heat loss at night compared to the steppes (Breshears et al., 1997; D'Odorico et al., 2013; Tölgyesi et al., 2018, 2020; Süle et al., 2020). Microclimatic extremes are also tempered within forest patches by the edges acting as wind breaks and thus attenuating evaporation compared to adjacent grasslands (Davies-Colley, Payne & van Elswijk, 2000). The altered conditions impose a strong filter, limiting the growth of light-demanding plant species, while facilitating shade-tolerant and drought-sensitive species, for which the steppe does not offer suitable habitat (Erdős et al., 2014a; Lashchinskiy et al., 2017; Tölgyesi et al., 2018).

As for soil moisture availability, the effects of trees are rather mixed in the forest-steppe, and it is difficult to separate a priori moisture differences caused by topography and soil structure from true forest–moisture feedbacks. The proportion of precipitation intercepted by tree canopies and the leaf litter can be high (up to 70% of each rainfall event; Yang et al., 2019), especially in coniferous forests, where interception captures not just rain, but also causes considerable amounts of snow to sublime before reaching the ground. At the arid southern edge of the forest-steppe in Kazakhstan, mid-summer topsoil can be drier under forest tree canopies than in adjacent open steppes (Tölgyesi et al., 2018). In climatically less harsh sites, such as the sand regions of the Carpathian Basin, forest topsoil tends to be moister than that of the steppe patches (Erdős et al., 2018b, 2021) but deeper soils are desiccated, with the rate of desiccation dependent on whether trees are deciduous or evergreen (Tölgyesi et al., 2020). It is an open question though, whether the moisture surplus in the topsoil is solely a consequence of the reduced evaporation due to the cool shaded microclimate or if trees bring deep water up to the topsoil via hydraulic lift, as occurs in many semi-arid regions (Yu & D'Odorico, 2015).

The overall effect of trees on grassland species seems to be negative, with a sparser herbaceous layer in forests compared to grasslands (Erdős et al., 2014a; Tölgyesi et al., 2018). The herbaceous layer species compositions in grasslands and forests show little overlap, thus it is unclear whether the trees directly exclude steppe species, or do so indirectly by allowing the growth of species that are competitively superior in shaded conditions. Conditions beneath forests, which are unsuitable for grassland species, can facilitate tree recruitment by attenuating heat and water stress during the summer, and reducing cold stress in winter and early spring (Dulamsuren et al., 2008a,b; Erdős et al., 2021). In addition, the sparser herb layer in the forests is less flammable, limiting the spread and intensity of wildfires compared to the grasslands. Saplings are thus more likely to survive fires inside the forest, but this has not been tested. Such fire protection may not apply to forests composed of highly flammable conifers (Pinus spp. or Juniperus spp.), which can burn intensely and regenerate slowly if their crown catches fire (Kolář et al., 2020; Ónodi et al., 2021). Shaded conditions in the forest patches are likely to limit tree saplings too, but less than by the grassland species, since most forest-steppe trees are widespread components of closed-canopy temperate and boreal forests where there has been strong evolutionary selection for shade tolerance (Valladares & Niinemets, 2008).

Parallel to the favourable recruitment conditions of trees inside forests, conditions in the grassland state promote the recruitment and persistence of steppe species for a number of reasons. Fire, which can suppress saplings in the steppe, causes little harm to the belowground organs or the seedbank of grasses and forbs, for which the conditions after the fire provide excellent opportunities for regeneration via resprouts, clonal spread, or seed germination (Ónodi et al., 2021). Contributing to a positive fire feedback, after burning, aboveground plant productivity is enhanced relative to pre-fire levels (Valkó et al., 2016). Herbaceous plants in steppes benefit from a sharper drop in nocturnal temperature relative to temperatures in forests, which often leads to dew formation (Lellei-Kovács et al., 2008; Tölgyesi et al., 2018), which is an important moisture source for herbaceous plants in water-limited ecosystems (Agam & Berliner, 2006). Tree saplings in the steppes are less able to benefit from dew because they have few superficial roots. Furthermore, there is evidence that the belowground competitive effects of grasses can directly constrain tree growth in the Eurasian forest-steppe (Walter & Breckle, 1989; Peltzer & Köchy, 2001). However important direct grass–tree competition may be, competition alone is not necessarily strong enough to exclude trees completely from invading grass-dominated communities (Wilson & Peltzer, 2021). In Eurasian forest-steppes, competitive effects of grasses on trees are probably best viewed a minor vegetation feedback, relative to the strong influence of the steppe microclimate, fires, and herbivores in limiting tree establishment.

The effective recruitment of trees and grasses in association with the forest and the steppe ecosystem states, respectively, stabilises their position and distinctness, contributing to the mosaic vegetation structure. The resulting stability of the forest edges is also reflected by distinct, species-rich edge communities in forest-steppes (Erdős et al., 2014a; Bátori et al., 2018). This overall pattern means for our hierarchical conceptual model that in sites where climate as well as topography, soil, herbivory and fire allow the co-existence of forest and steppe, vegetation feedbacks further stabilise spatial patterns by hindering state transitions (i.e. hysteresis; Ratajczak et al., 2018). This stable patch pattern has been confirmed for Hungarian forest-steppes by historical map interpretation (Erdős et al., 2015). The stabilising feedbacks may lend considerable resilience of both forest and grassland ecosystem states to environmental changes, as highlighted by Xu et al. (2017) for Siberian forest-steppes.