Functional diversification enabled grassy biomes to fill global climate space

Global change impacts on the Earth System are typically evaluated using biome classifications based on trees and forests. However, during the Cenozoic, many terrestrial biomes were transformed through the displacement of trees and shrubs by grasses. While grasses comprise 3% of vascular plant species, they are responsible for more than 25% of terrestrial photosynthesis. Critically, grass dominance alters ecosystem dynamics and function by introducing new ecological processes, especially surface fires and grazing. However, the large grassy component of many global biomes is often neglected in their descriptions, thereby ignoring these important ecosystem processes. Furthermore, the functional diversity of grasses in vegetation models is usually reduced to C3 and C4 photosynthetic plant functional types, omitting other relevant traits. Here, we compile available data to determine the global distribution of grassy vegetation and key traits related to grass dominance. Grassy biomes (where > 50% of the ground layer is covered by grasses) occupy almost every part of Earth’s vegetated climate space, characterising over 40% of the land surface. Major evolutionary lineages of grasses have specialised in different environments, but species from only three grass lineages occupy 88% of the land area of grassy vegetation, segregating along gradients of temperature, rainfall and fire. The environment occupied by each lineage is associated with unique plant trait combinations, including C3 and C4 photosynthesis, maximum plant height, and adaptations to fire and aridity. There is no single global climatic limit where C4 grasses replace C3 grasses. Instead this ecological transition varies biogeographically, with continental disjunctions arising through contrasting evolutionary histories. Significance statement Worldviews of vegetation generally focus on trees and forests but grasses characterize the ground layer over 40% of the Earth’s vegetated land surface. This omission is important because grasses transform surface-atmosphere exchanges, biodiversity and disturbance regimes. We looked beneath the trees to produce the first global map of grass-dominated biomes. Grassy biomes occur in virtually every climate on Earth. However, three lineages of grasses are much more successful than others, characterizing 88% of the land area of grassy biomes. Each of these grass lineages evolved ecological specializations related to aridity, freezing and fire. Recognizing the extent and causes of grass dominance beneath trees is important because grassy vegetation plays vital roles in the dynamics of our biosphere and human wellbeing.

Abstract: Global change impacts on the Earth System are typically evaluated using biome classifications based on trees and forests. However, during the Cenozoic, many terrestrial biomes were transformed through the displacement of trees and shrubs by grasses. While grasses comprise 3% of vascular plant species, they are responsible for more than 25% of terrestrial photosynthesis. Critically, grass dominance alters ecosystem dynamics and function by introducing new ecological processes, especially surface fires and grazing.
However, the large grassy component of many global biomes is often neglected in their descriptions, thereby ignoring these important ecosystem processes. Furthermore, the functional diversity of grasses in vegetation models is usually reduced to C3 and C4 photosynthetic plant functional types, omitting other relevant traits. Here, we compile available data to determine the global distribution of grassy vegetation and key traits related to grass dominance. Grassy biomes (where > 50% of the ground layer is covered by grasses) occupy almost every part of Earth's vegetated climate space, characterising over 40% of the land surface. Major evolutionary lineages of grasses have specialised in different environments, but species from only three grass lineages occupy 88% of the land area of grassy vegetation, segregating along gradients of temperature, rainfall and fire. The environment occupied by each lineage is associated with unique plant trait combinations, including C3 and C4 photosynthesis, maximum plant height, and adaptations to fire and aridity.
There is no single global climatic limit where C4 grasses replace C3 grasses. Instead this ecological transition varies biogeographically, with continental disjunctions arising through contrasting evolutionary histories.
Introduction 1 in the ground layer also affects surface energy, carbon, nutrient and water cycling by, for 25 example, altering rates of decomposition, water infiltration and absorption of sunlight. Grass 26 dominance therefore leads to novel ecological processes and properties in the Earth System, 27 including frequent fire and grazing by mammals (14). 28 During the Cenozoic grasses displaced forests and shrublands by altering disturbance 29 regimes at large scales across tropical and temperate regions (14, 15). The global expansion 30 of grassy vegetation enabled major faunal and floral radiations (14, 16), and is linked to events 31 in human behavioral evolution (17, 18). Today, natural grassy biomes provide grazing lands, 32 water resources and numerous ecosystem services that directly support over a billion people 33 (19). Yet, despite this social and economic significance, and the profound disturbance 34 feedbacks engendered by grassy vegetation (20), understanding of grassy biomes is 35 geographically biased towards few regions (e.g., South and East African savannas, North 36 American grasslands), with the global limits of grassy biomes poorly defined. 37 When considering the limits to grassy biomes, the grass diversity present in a system 38 is generally reduced to a distinction between species using the C3 or C4 photosynthetic 39 pathways. If all else is equal, C4 grasses should outcompete C3 grasses under conditions of high 40 light and temperature as well as low CO2 (21)(22)(23) that focusses on the ground layer contrasts with efforts to map biomes using remotely sensed 63 tree cover or biomass (4, 29). Such studies generally misclassify extensive areas of tropical 64 savanna as forest or degraded forest (30, 31). Global synthesis of grassy biomes has been 65 prohibited as satellite remote sensing does not see through a tree canopy. Therefore, we 66 mapped grassy formations by integrating and re-classifying 20 existing national and regional 67 vegetation maps produced using botanical data and detailed vegetation descriptions (see 68

Methods and SI). 69
What is a grassy biome? We defined vegetation units as grassy where the ground layer 70 is characterized by Poaceae and where grasses comprised > 50% of ground layer cover based 71 on descriptions within vegetation maps and associated literature (see Methods and SI). A 72 The C3 Pooideae occupy regions with lower winter temperatures and shorter droughts 120 than the C4 lineages (Fig. S7). C3 Pooideae dominate grassy biomes to much higher 121 temperatures in the Palearctic than the Nearctic realm, although distributions of C4 122 Andropogoneae and Chloridoideae in these realms are similar (Fig. 3). Conversely, C3 Pooideae 123 are confined to the geographically restricted colder parts of the Afrotropics and Indo-Malay 124 realms, and C4 Andropogoneae dominate at much lower temperatures in these regions (Fig.  125 3). The sorting of C3 and C4 grass species along local and regional temperature gradients is well 126 established (40, 41), and the crossover temperature can be modified by ecosystem factors 127 (e.g., tree cover) (42). However, our observations are broadly consistent with model 128 predictions of carbon assimilation (22, 23, 43), as modeled crossover temperatures under low 129 light conditions and modern CO2 levels occurs at ~20-22 °C. 130 In our data, some species of both Andropogoneae and Chloridoideae lineages have 131 adapted to low mean annual temperatures and may persist in grassy vegetation within cool 132 parts of each realm (e.g. Fig. 3). Given equal investment in the carbon-fixing enzyme Rubisco, 133 a relatively low canopy leaf area and sunny conditions, a C4 canopy can theoretically achieve 134 higher total daily rates of photosynthesis than a C3 at any temperature (37). In this case, the 135 primary limitation on canopy carbon uptake becomes light-mediated damage during low 136 temperature extremes (44), although C4 photosynthesis is energetically expensive. Low 137 temperature tolerance may be absent from most C4 species as C4 photosynthesis evolved in 138 the tropics (38). Andropogoneae (46) and we see this mirrored at a global scale. Grass persistence in these 159 competitive environments relies on the annual production of a new canopy and, in the 160 absence of woody investment, dead biomass must either rapidly decompose, burn or be 161 consumed by herbivores to avoid self-shading (11, 49). Andropogoneae are known to have 162 morphological adaptations enabling tolerances and persistence to fire that are not commonly 163 present in other grass lineages (49). Fire and other forms of repeated disturbance, such as 164 grazing, are therefore crucial for grass-dominated systems to persist in high rainfall 165 environments. While Andropogoneae appears to be the C4 lineage most closely associated 166 with disturbance by fire, multiple lineages in the semi-arid African tropics appear linked to 167 grazing tolerance (Fig S8, (50, 51)), and this may be due to the strength and form of 168 environmental filtering associated with fire versus grazing, as well as the antiquity and 169 biogeography of grazing pressure relative to fire. 170 171 Implications. The Andropogoneae, Chloridoideae and Pooideae grass lineages dominate 172 globally, via mechanisms encompassing plant production and competition, resilience to 173 drought, freezing and disturbance. Why do three of the most diverse grass lineages 174 characterise grassy biomes? Does diversity beget ecological success or does success beget 175 diversity? Early diversification may have enabled ecological success, such that ecological 176 speciation allowed each lineage to radiate across broad environmental envelopes (an 177 ecological mechanism). Alternatively, a neutral mechanism of a long history of diversification 178 may have led to high diversity as Andropogoneae and Chloridoideae are the oldest C4 lineages. 179 Across our dataset, evidence for this is equivocal. We list 8.8% of all grass species and within 180 lineages: Andropogoneae, 14.5%; Chloridoideae, 6.5%; Pooideae, 10.8%. Perhaps ecological 181 success facilitated diversification, such that large geographical ranges enabled by unique 182 adaptations made the isolation of populations and allopatric speciation more likely (a 183 geographic mechanism). The rapid spread of the cosmopolitan Themeda triandra from Asia to 184 Africa in < 500,000 years supports this idea (52). Resolving the relative role of these 185 mechanisms requires comparative phylogenetic analyses of the relationships among ecology, 186 functional traits, range sizes and diversification rates. 187 The biogeographic contingencies described here in crossover temperatures align with 188 emerging evidence that regional evolutionary and environmental histories have been 189 important modifiers of biome-climate relationships (9, 53). However, the rapid rates of considered grassy deserts where the total above-ground biomass was considered <50 g m 2 , 256 or where total ground cover <25%, throughout the year. Finally, we retained all formations 257 where grasses were the dominant component of the ground layer, irrespective of tree cover. 258 Numerous grassy biomes, such as tropical savannas and woodlands, may be characterised by 259 up to 80% tree cover, but behave functionally as savannas due to a contiguous grassy ground 260 layer (13, 35). Where necessary, we sourced additional information from published vegetation 261 descriptions and analyses to attribute key grass species to a grassy vegetation unit. WorldClim dataset (www.worldclim.org) was used to obtain species median values of 307 minimum temperature (BIO6 variable) and seasonal drought length (calculated as the number 308 of successive months where mean annual precipitation was below 30mm). These species level 309 data were used to construct frequency histograms to examine lineage level variation in fire 310 regimes and climate extremes (Fig S7). 311 Environmental data used in global analyses. Our analysis aimed to elucidate lineage, climate 312 and disturbance relationships, and whether biogeography impacts the C3-C4 crossover 313 temperature. We used the WorldClim dataset at a 0.5 degree resolution to match the 314 vegetation map, and extracted mean annual precipitation (MAP), rainfall seasonality, mean 315 annual temperature (MAT) and temperature seasonality (www.worldclim.org). We used a 316 rainfall concentration index to describe rainfall seasonality based on (35). Growing season 317 temperature (GST) was calculated for each grid cell to quantify regional and global C3-C4 318 crossover temperatures. GST was calculated as the mean temperature across months with a 319 greater than or equal to 5 degree mean temperature and at least 25 mm rainfall, and was 320 calculated using WoldClim monthly climate normals (65).  showing 95% confidence intervals. The right-hand axis is the global relationship between fire 405 return interval and MAP for grassy biomes and is inverted to reflect the inverse relationship 406 with MAP. The global peak in fire activity coincides with the global peak in dominance of 407  should be of concern to many. 599 600 We undertook a validation process between plot data describing in situ grass abundance and 601 our global species list. Using publicly available data that intersected with vegetation unit 602 descriptions we found that, at the level of independent evolutionary lineages of grasses (i.e., 603 subfamily), we had strong confidence in the geographic and environmental relationships we 604 elucidate here (Fig. S9). To validate the classification of common grass species across regions, 605 we compared the species list in each vegetation unit to a plot level database developed for 606 validation purposes (Fig. S9). Plot data were sourced from the literature and vegetation