To the best of our knowledge, this is the first India wide study examining the distribution and drivers of savanna and forest biomes using a remotely sensed tree cover metric. First, we demonstrate that there are four distinct zones of tree cover- a low, high and medium tree cover zones where maximum climate water deficit (MCWD), a measure of severity of water stress regulates tree cover and a mixed tree cover zone where MCWD does not have any influence on tree cover. The medium and mixed tree cover zones have savanna vegetation (Fig 1). Second, using the tested relationship between tree cover and MCWD, we predicted the climatic maximum potential tree cover (Fig 2 a). Consequently, we quantified the shortfall of current tree cover from the climatic maximum potential. This shortfall is high in the central regions of the Deccan Plateau and is low in north-east India and specific sections of the Western Ghats (Fig 2 b). Third, in the medium and mixed tree cover zones, we show that shortfall decreases with increase in soil sand fraction (Fig 3 c). High levels of grazing and anthropogenic pressures increase the shortfall (Fig 3 d & e). At the coarse scale of our analyses, we demonstrate the distinct role of topography in regulating shortfall (Fig 3 f & g). Furthermore, we could not ascertain a relationship between fire and shortfall due to limited data on fires (Fig 3 a & b).
Distribution of tree cover driven by severity of water stress (MCWD)
The low and high tree cover zones clearly delineate the desert and tropical forest biomes respectively49, while the biomes in the zones characterized by medium and mixed tree cover are less clear. The unclear biome status in these two zones is supported by the lack of bimodality in tree cover, typical of co-existing forest and savanna biomes50. However, as shown in this study, these zones contain evidence of herbaceous life forms of graminoids and forbs that are characteristic of savannas (from Ref9), indicating the co-existence of both savannas and forests in these zones. South Asian savannas include dipterocarp savannas, mixed savannas, fine leaved and spiny savannas and pine savannas with dominant tree clades spanning the entire range of tree canopy cover and physiognomies similar to forest trees8. Similarly, different forest types, especially that of secondary regrowth due to intensive and extensive land use and land cover changes, have a wide range of tree canopy cover. This wide range of tree cover across the variety of forest and savanna formations is difficult to assess using coarse scale remote sensed products contributing to the lack of bimodality and clear biome delineation. We suspect that forest and savannas coexist in the medium and mixed tree cover zones at fine spatial scales, i.e. at the landscape scale, such as in southeast Asia17 and the forest-grassland mosaics in north-eastern India51.
Overall, the sigmoidal response of current tree cover to MCWD confirms that severity of water stress is an important driver of savanna biome distribution. This is evident by the clear delineation of the desert and forest biomes at high and low water stress respectively and in the remaining two zones containing savanna herbaceous vegetation. The medium tree cover zone (~40% tree cover) is in relatively drought prone areas and the mixed tree cover zone (~0-80% tree cover) in relatively less drought prone areas, akin to findings globally about the distribution of the savanna biome and possibility of alternative biome states with forests36,52,53. This is because water stress drives differences in survival and growth rates of forest and savanna trees and the savanna herbaceous layer2. Mechanistically, trees and grasses coexist in savannas by partitioning limited resources such as water, i.e. there is hydrologically driven resource competition2. Additionally, climate change variability and climate change driven disturbances (including water resource availability) cause demographic bottlenecks at various life stages of trees, thereby maintaining tree-grass coexistence2. Hence, predicted changes in water stress due to climate change54 are likely to drive future changes in the distributions of savanna and forest in south Asia55.
Topography and anthropogenic pressures drive tree cover shortfall
Topography regulates tree-grass dynamics in the medium and mixed tree cover zones containing savanna vegetation. In areas with water availability, forest trees persist irrespective of terrain, i.e. tree cover converges with its climatic maximum in hilly terrain (including upper slopes and ridges) and valleys as shown. Indeed, our findings here agree with those of similar studies across tropical South America, which show that areas with high rainfall have more forests than savannas irrespective of the depth of the water table56. Low hill slopes might have alternating periods of waterlogging and drought stress amplified by rainfall seasonality; however, savannas can withstand these stresses as shown in tropical South America56, corroborating the increasing trend in shortfall in low slopes and flat areas shown in this study. Furthermore, temperature gradients and the resulting heterogeneity in energy distribution created by topography explains the pattern of shortfall57, with shortfall being less in north-easterly and south-westerly facing areas. We posit the interactive effects of seasonality and aspect will explain the threshold response of tree cover shortfall and Heat Load Index (HLI), with increased seasonality limiting tree cover in a wide range of terrain aspects. Clearly, interactive effects of topography and rainfall seasonality regulate tree and grass co-existence56,58,59. These effects can be further explored by assessing tree-grass dynamics across topographic and elevational gradients58, in turn accounting for water depth56.
Anthropogenic pressures such as small-scale clearing and fuel wood collection limit tree cover60 explaining the linear increase in shortfall from the climatic maximum potential shown. Hence, in areas where forest and savanna biomes coexist in India, anthropogenic pressure is a regulator of tree cover, like regional evidence from Africa24 and South America61.
Resource and disturbance factors regulate tree cover shortfall
Within the medium and mixed tree cover zones, sandy soils allow tree cover to reach the climatic maximum potential i.e. decrease in shortfall, similar to findings in Africa27,31,62 and southeast Asia17. By increasing drainage, high sand content can drain moisture in shallow soil layers but can lead to an accumulation of water in deeper layers accessed by deep-rooted forest trees2,33,63. However, the interactive effects of soil moisture and rainfall seasonality can have varying effects on tree cover across forests and savannas when considering rainfall intensity31. Lastly, a better understanding of the functional rooting characteristics of all life forms of forests and savannas can yield interesting insights about the factors responsible for the distribution of forests and savannas. Root traits such as rooting depth of forest and savanna trees and savanna grasses can explain the extent to which varying water depths are being tapped across different soil types and rainfall regimes.
Herbivory directly affects tree cover shortfall and hence regulates forest and savanna biome distribution in the medium and mixed tree cover zones. Generally, at moderate densities, grazers such as sheep, can increase tree growth rates by consuming grasses1,4,28,40, explaining the initial decreasing trend in shortfall. However, at intense levels of grazing, high densities of sheep limit tree recruitment into adult tree size classes thereby promoting savanna grasses39,40. Complimentary to livestock, it would be valuable to understand the impacts of wild herbivores on biome distribution, considering the relatively distinct and last remnant populations of wild herbivores in south and southeast Asia64. This is crucial because functional changes in herbivore communities i.e. wild native herbivores to domestic livestock can have differential impacts on distribution of forests and savannas65 and in turn on ecosystem functions such as accumulation of soil carbon66.
Although there is a significant and decreasing trend in tree cover shortfall with increasing fire frequency in the medium and mixed tree cover zones, these results are difficult to interpret for multiple reasons. For one, the information about fire in the medium and mixed tree cover zones is highly skewed. The fire return interval in south Asia is significantly shorter and the fires are of smaller spatial extent and of reduced intensity, compared to vegetation fires in Africa and possibly South America67,68. For example, in 2001-2017, 16% of the African savanna area burned as opposed to 1.6% in south and southeast Asia69. Additionally, small fires accounted for ~90% of the burned area in south and southeast Asia versus 30% in Africa70. Hence, small and less intense fires might continue to maintain or even increase tree cover. Second, even though we have used complementary satellite derived information about fire intensity and frequency, the small extent and reduced intensity of fires in India might be difficult to detect accurately in remote sensed products71–73. Third, reduced fire activity has been reported in savannas and grasslands, mainly because of fire suppression associated with agricultural expansion and intensification globally74 and in savannas of South America75. India’s extensive and intensive historic and current conversion of land especially for agriculture76,77 might have altered fire regimes contributing to uncertainty in detection and the consequent impacts on vegetation.
India’s complex historic and contemporary practises of land management for forests, fire suppression and ongoing anthropogenic pressures including high densities of domestic livestock and ambitious tree-planting initiatives are important factors to consider for appropriate conservation and restoration of its savannas. Here, considering these factors we show that there are intermediate zones of tree cover with savanna vegetation, indicating possible co-existence of forests and savannas. Furthermore, once climatic drivers of tree cover are accounted for, topography, soil texture (sandy soils), anthropogenic pressures and sheep domestic livestock herbivory regulate tree cover in the intermediate tree cover zones. This information can help us to go beyond the ‘forest centric’ approaches to ecosystem restoration13,78. For example, our results of delineation of intermediate zones of tree cover containing savanna vegetation and the low tree cover zone can inform the design of ‘no-go’ areas for tree planting schemes. When considering woody encroachment, insights from our study in India about the role of intense sheep herbivory in reducing tree cover provides useful evidence for the design of appropriate livestock grazing management regimes to control the biomass of palatable dominant trees, thereby maintaining healthy tree-grass dynamics and the persistence of the savanna biome. Hence, by understanding the distribution of the savanna biome and the drivers of its life forms (tree and grass component), we can design nuanced ecosystem restoration strategies that go beyond simplistic tree-planting initiatives in the UN Decade on Ecosystem Restoration.