The consequences of land-use change for savanna biodiversity remain undocumented in most regions of tropical Asia. One such region is western Maharashtra, India, where old-growth savannas occupy a broad rainfall gradient and are increasingly rare due to agricultural conversion and afforestation.

To understand the consequences of land-use change, we sampled herbaceous plant communities of old-growth savannas and three alternative land-use types: tree plantations, tillage agriculture, and agricultural fallows (n=15 sites per type). Study sites spanned 457 to 1954 mm of mean annual precipitation—corresponding to the typical rainfall range of mesic savannas globally.

Across the rainfall gradient, we found consistent declines in old-growth savanna plant communities due to land-use change. Local-scale native species richness dropped from a mean of 12 species/m² in old-growth savannas to 8, 6, and 3 species/m² in tree plantations, fallows, and tillage agriculture, respectively. Cover of native plants declined from a mean of 49% in old-growth savannas to 27% in both tree plantations and fallows, and 4% in tillage agriculture. Reductions in native cover coincided with increased cover of invasive species in tree plantations (18%), fallows (18%), and tillage agriculture (3%).

In analyses of community composition, tillage agriculture was most dissimilar to old-growth savannas, while tree plantations and fallows showed intermediate dissimilarity. These compositional changes were driven partly by the loss of characteristic savanna species: 65 species recorded in old-growth savannas were absent in other land uses. Indicator analysis revealed 21 old-growth species, comprised mostly of native savanna specialists. Indicators of tree plantations (9 species) and fallows (13 species) were both invasive and native species, while the 2 indicators of tillage agriculture were invasive. As reflective of declines in savanna communities, mean native perennial graminoid cover of 27% in old-growth savannas dropped to 9%, 7%, and 0.1% in tree plantations, fallows, and tillage agriculture, respectively.

Synthesis: Agricultural conversion and afforestation of old-growth savannas in India destroys and degrades herbaceous plant communities that do not spontaneously recover on fallowed land. Efforts to conserve India’s native biodiversity should encompass the country’s widespread savanna biome and seek to limit conversion of irreplaceable old-growth savannas.

During September and October 2021, the peak growing season for herbaceous plants, we sampled a total of 60 sites, n = 15 in each of the four land use types (Fig. 1). Prior to the field sampling, we randomly selected sites with the goal of stratifying across the rainfall gradient. For the stratification, we created a regional map based on the long-term MAP from a thirty-year period (1970 to 2000; 1 km² resolution; Fick & Hijmans, 2017). While our goal was to sample sites stratified by MAP, we ultimately sampled fewer sites than expected in the 1500 to 2000 mm interval. This was because the wettest portion of the MAP gradient occupied a small area and contained only a few sites that met our selection criteria.  

At each of the 60 sites, we establish three 200-m² sampling plots (20 m × 10 m) oriented in a random compass direction. In western Maharashtra, old-growth savannas and tree plantations typically occur as large patches: our old-growth sites ranged from 6 to 6160 ha with a median of 40 ha; tree plantations ranged from 8 to 403 ha with a median of 66 ha. Given their large area, we were able to randomly locate plots within old-growth and plantation sites without concern they would overlap or extend beyond the site. By contrast, the size of tillage agriculture and fallow sites was small (0.1 to 0.4 hectares). To capture variation in the tillage agriculture and fallow land-use types, we sampled three separate fields, each with one plot that we oriented to fit within the field. For all four land use types, we ensured that the three plots at each site were located within a circular area of diameter < 3 km. At some sites, availability of suitable tillage agriculture and fallowed fields, as well as challenges to acquire permission from landowners, resulted in our sampling of fields that were adjacent to one another. In these cases, we ensured that there were at least 20 m between plots.

We measured herbaceous plant communities in seven 1-m² sub-plots, positioned along the center line of the 200-m² plot, and calculated the mean of the seven sub-plots for subsequent data analyses. We visually estimated percent cover by species (excluding crops in tillage agriculture) and used this data to determine local-scale species richness (i.e., species per 1 m²) and community composition. To measure and identify woody species within each 200-m² plot, we used a variety of sampling techniques suitable for individuals of different size classes (Foster et al., 1998). For trees with a diameter at breast height (DBH, 1.3 m) ≥10 cm (i.e., large trees), we measured DBH of all individuals in the 200-m² plot. For trees of DBH ≥ 1 cm and < 10 cm (i.e., small trees), we sampled a 2 × 20-m subplot (40 m²) positioned along the center line of the plot. To estimate percent cover of shrubs (multi-stemmed woody plants), we used a 20-m line-intercept (Canfield, 1941). Lastly, we quantified woody regeneration by counting the number of seedlings, small tree saplings, and woody resprouts < 1.3 m tall or < 1 cm DBH in the seven 1 m² plots. For identification, nomenclature, and classification of plant functional groups and native/invasive status, we referred to several floras for the region (Supplementary methods).

To characterize soils, we collected two 10-cm diameter samples of the top 10-cm of mineral soil (excluding leaf litter and duff) from the ends of the 20-m center line of the plot, which we pooled for each site. The samples were analyzed by the Soil Science Laboratory of the College of Agriculture, Pune, India for pH, electrical conductivity, organic carbon, available nitrogen, phosphorus, potassium, cation exchange capacity, and soil texture.