The distribution of forest and savanna biomes and the interplay of resources (climate and soil) and disturbances (fire and herbivory) in shaping tree-grass dynamics remain complex and geographically variable. This is particularly true in Indian savannas, often misidentified as degraded forests and targeted for tree planting. This study investigates the distribution of these biomes and the factors influencing tree cover across India. Previous studies have highlighted the role of climate and soil in regulating tree-grass dynamics, but the relative importance of these factors versus disturbances like fire and herbivory remains debated, especially in South and Southeast Asia. In India, recent evidence—paleoecological data and discoveries of endemic species—suggests the existence of ancient savanna ecosystems. However, the distribution and drivers of tree-grass ratios within these ecosystems are poorly understood. This knowledge is critical given ongoing tree-planting initiatives intended for climate change mitigation, which may be detrimental if not carefully planned in the context of existing savanna ecosystems and woody encroachment.

Existing literature indicates that resource-based drivers (climate and soil) and disturbance-based drivers (fire and herbivory) regulate tree-grass dynamics in savannas. Climate, especially rainfall and its seasonality, is a key determinant of tree cover globally and in African savannas, constraining the maximum potential tree cover through its effects on tree growth and mortality. Savannas often occur in areas with low rainfall and/or high rainfall seasonality. Soil texture, which affects soil moisture, also influences tree-grass dynamics; sandy soils favor deep-rooted trees, while clayey soils favor shallow-rooted grasses. However, global understanding of the relationship between water stress and tree cover in savannas is limited, especially concerning India. Fire and herbivory are also crucial regulators, with frequent fires preventing canopy closure, while herbivory can either increase or decrease tree cover depending on intensity and type. Previous studies have often neglected the role of domestic livestock, the dominant herbivore in many tropical open ecosystems. While extensive research exists on these drivers in Africa and South America, there is limited research on their effects on Indian savannas.

This study investigates the distribution and drivers of savannas and forests across India using tree cover as a metric. The methodology involved several key steps: 1) Mapping of current tree cover (2000-2020) using MOD44B VCF product at 250m resolution and climatic data including maximum climate water deficit (MCWD) using TerraClimate. 2) Hypothesis testing: The initial hypothesis proposed a multimodal distribution of tree cover across a water stress gradient. Quantile-based generalized additive models tested the relationship between MCWD and tree cover, estimating the climatic maximum potential tree cover. 3) Quantification of tree cover shortfall: this was defined as the difference between the climatic maximum potential tree cover and the current tree cover, providing a measure of divergence. 4) Driver analysis: Generalized additive models (GAMs) were used to assess the relationship between tree cover shortfall and resource and disturbance drivers (sand fraction, fire frequency, fire intensity, herbivory pressure from livestock, Human Population Density), and topographical variables (Heat Load Index and Topographic Position Index). A stratified sampling strategy of 9600 points across India (excluding urban areas, croplands, and high-elevation areas) was implemented to provide a representative dataset. Data on savanna vegetation locations helped refine biome delineation, and Google Earth imagery was used for qualitative assessment of tree cover and the identification of additional drivers like topography and anthropogenic pressures.

The study identified four distinct zones of tree cover across India: a low tree cover zone (MCWD < 1209 mm), a medium tree cover zone (MCWD 845–1209 mm), a mixed tree cover zone (MCWD 486–845 mm), and a high tree cover zone (MCWD > 486 mm). Savanna vegetation was primarily found in the medium and mixed zones. A sigmoidal relationship was observed between tree cover and MCWD, indicating that water stress is a major driver of biome distribution. Climatic maximum potential tree cover was highest in northeastern India and parts of the Western Ghats, and lowest in western India's arid and semi-arid regions. Tree cover shortfall (divergence from climatic potential) was high in central regions (Deccan Plateau, Central Highlands) and low in northeastern India and parts of the Western Ghats. In the medium and mixed zones, analysis of shortfall revealed that it decreased with increased soil sand fraction, increased with human population density, and showed a threshold response to sheep herbivory (decreasing at low levels, increasing at high levels). Topographic variables (HLI and TPI) also showed threshold effects on shortfall. The relationship between fire and shortfall was inconclusive due to data limitations.

This study provides the first India-wide analysis of savanna and forest biome distribution using tree cover. The findings highlight the importance of water stress in determining biome distribution, but also the significant influence of topography, soil texture, anthropogenic pressure, and livestock herbivory in shaping tree cover patterns, especially in the savanna zones. The absence of bimodality in tree cover suggests that savannas and forests coexist in the intermediate zones at finer spatial scales. The impact of livestock grazing on tree cover is complex; moderate grazing may enhance tree growth, but intense grazing reduces tree recruitment. The inconclusive relationship between fire and tree cover likely results from factors such as limited data, low fire intensity and frequency compared to other regions, and historical fire suppression. The study's limitations related to data availability on fire and exhaustive savanna species distribution and the coarse resolution of remotely sensed products need to be considered while interpreting the results.

This study offers valuable insights into the distribution of Indian savannas and forests and the factors governing their tree cover. Understanding the complex interplay between climatic conditions, soil properties, land use practices, and grazing intensities is vital for developing effective conservation and restoration strategies. The identification of intermediate zones with coexisting forests and savannas necessitates moving beyond simplistic tree planting schemes and towards a more nuanced approach. Future research should focus on exploring finer spatial scale patterns, examining wild herbivore impacts, and improving fire data acquisition methods to gain a more comprehensive understanding of the dynamic interplay between these biomes.

The study's limitations include data limitations regarding fire occurrences in India, and incomplete data on the distribution of savanna species. The coarse spatial resolution of remote sensing products may have masked finer-scale patterns in tree cover distribution and the coexistence of forests and savannas. Additionally, the use of domestic livestock density as a proxy for herbivory pressure may not fully capture the complexities of grazing impacts. Further research could focus on obtaining more detailed data and using high-resolution imagery and advanced techniques to address these limitations.