Plant essential macronutrients like nitrogen (N) and phosphorus (P) can limit savanna tree growth and are important determinants of savanna vegetation dynamics, along with rainfall, fire, and herbivory. How nitrogen and phosphorus shape tree-grass competition and their coexistence remain unclear, hindering our ability to predict how savannas may respond to altered nutrient cycling.

Here, we evaluate (1) if trees and grasses respond differently to N vs. P availability, or (2) if grasses are more competitive in low nutrient environments while trees are more competitive in high nutrient environments. To do this, we grew saplings of 6 tree and 1 grass species from Kruger National Park, South Africa, for 16 weeks under fully factorial nutrient and competition treatments (with/without competitors, low/high rate of N supply, and low/high rate of P supply) under a watering regime designed to mimic wet season rainfall in a mesic savanna.

Trees and grasses foraged most aggressively for nitrogen and allocated biomass differently depending on nitrogen availability. Overall, tree growth decreased in competition with grass, even in high nutrient environments where they grew faster. Grasses were always better belowground competitors, utilizing aggressive nutrient foraging strategies, including high root phosphatase activity in response to nitrogen and large root biomass allocation.

Synthesis: In low nutrient environments (e.g., on nutrient-poor sandy soils), nutrients may limit tree growth. Nutrient rich environments enable tree growth, but grasses continue to compete effectively with trees. Understanding what this means for ecosystem responses to nutrient availability is not trivial, especially in the context of fire and herbivory. However, it is clear that soil nutrients likely affect tree and grass growth and competition in savannas, which suggests that future changes in nutrient cycling, such as N deposition, may have important effects on savanna vegetation.

The primary goal of the experiment was to understand how trees and grasses respond to different nutrient and competition regimes. Six tree species and one grass species were selected based on phylogenetic diversity and nitrogen fixation ability for this experiment (Table 1). The grass species chosen for this experiment was Melinis repens, the only annual grass species common in Kruger National Park (Wigley-Coetsee and Staver 2020), one of the largest natural reserves that is situated in the low-lying savannas of South Africa.

Plants were grown in a full-factorial nutrient-addition experiment with water availability held constant across treatments. Each pot was watered three times a day for a total of 400 mL per day, designed to mimic a wet growing season in the Pretoriuskop region of Kruger National Park, which receives about 750 mm of rainfall per year. The four nutrient treatments were (1) low N and P (), (2) high N and low P (), (3) low N and high P (), and (4) high N and P (). These nutrient treatments were applied factorially to trees grown with and without grasses and to grasses grown without trees for a total of 260 pots. Each nutrient and competition treatment was replicated five times in a randomized block design (except Schotia brachypetala, of which we had enough seedlings for only three replicates each).

Grass and tree seeds were collected in Kruger National Park or purchased from Silverhill Seed Company in South Africa in 2019 and stored until germination. In July-August 2020, tree seeds were clipped and soaked in water for 24-72 hours before planted in potting soil. After sowing, seeds were watered and placed in growth chambers set at 28°C and 500 mmol/m²/s of light for the duration of germination. Grass seeds were germinated on watered potting soil in the same growth chambers. In September 2020, germinated tree seedlings and grasses were transplanted into 30-liter tree pots (Stuewe & Sons, Inc., 0.25 m diameter x 0.60 m height, sold as 8-gallon pots) filled with turface (All-Season Turface, SiteOne Landscaping, New Haven CT). One tree sapling was planted per pot without grass competition, and two grass tufts were planted per pot without tree competition. For competition treatments, one tree sapling and two grass tufts were planted per pot (tree sapling in the middle, one tuft on either side).

All pots were fertilized once a week for 16 weeks at their assigned levels from October 2020-January 2021. Low nutrient treatments were designed to replicate nutrient limitation, while high nutrient treatments were designed to alleviate nutrient limitation while avoiding nitrogen or phosphorus toxicity. To determine the appropriate concentrations of N and P for each treatment, we derived monthly estimates of fertilizer from Tomlinson et al. (2019) along with field nutrient estimates from Ludwig et al. (2004). All  treatments received 100 mg N and  treatments 10 mg P per pot per week to mimic field conditions via ammonium nitrate and potassium phosphate, with  treatments receiving 800 mg N and treatments 100 mg P per pot per week. Along with N and P fertilization, each treatment received the same quantity of N- and P-free Hoagland’s micronutrient solution, which included all essential micronutrients: potassium, calcium, sulfur, magnesium, chloride, sodium, manganese, zinc, copper, molybdenum, boron and iron. Pots were located in a single greenhouse bay at Yale Marsh Botanical Gardens in a block-randomized design, kept at 28°C for the duration of the experiment, and supplemented with grow lights for 12 hours a day to replicate a sub-tropical growing season. Soil moisture was measured throughout the experiment to ensure that soil moisture remained constant across pots.

Plant Responses

At the time of transplanting, 10 individuals of each tree and grass species were harvested for initial size and biomass estimates. For tree seedlings, we recorded starting biomass, stem height, basal diameter, diameter at 10 cm height, root diameter, root length, wet and dry leaf stem mass, and fine (sorted into two diameter classes: 0-1 mm and 1-2 mm) and coarse root (> 2 mm) dry mass. For grasses, we recorded basal circumference, maximum culm and sward height, root diameter, and wet and dry weight of aboveground and belowground grass biomass.

Throughout the experiment, measurements were taken every four weeks. Tree measurements consisted of stem height, basal diameter, and diameter at 10 cm. Grass measurements consisted of basal circumference and sward height. Along with tree and grass measurements, light availability was measured using an fPAR meter (fraction of incident photosynthetically active radiation, 400-700 nm) at three different heights (tree base, tree canopy height, and grass height). At the final harvest in January 2021, we repeated all routine measurements and measured wet and dry weights of each plant tissue type. Time to harvest did not differ among species or blocks nor with respect to any experimental treatment (n = 260).

To calculate mass gain, we subtracted the mean total dry weight of the 10 tree seedlings or grasses harvested at the start of the experiment from the final total dry weights at the end of the experiment. We used allometric relationships to evaluate whether competition and nutrient treatments influenced the growth form of trees. To do this, we calculated the allometric exponent b  as follows (Feldpausch et al. 2011): for both shoot and root systems. For shoots, we used stem height and basal diameter (mm). For roots, we used taproot length for ‘height’ and root diameter (mm). We refer to the traits represented by these allometric exponents as ‘shoot elongation’ and ‘root elongation’, respectively.  Competition intensity was calculated by dividing the average mass gain of a tree or grass grown with competition by the average mass gain of a tree or grass grown out of competition and subtracted by one for each nutrient treatment within one species (i.e., only individuals of the same species were compared). Negative values (< 0) indicate competitive effects while positive values (> 0) indicate facilitative effects (Carlyle et al. 2010, Jabot and Pottier 2012).

To quantify root phosphatase activity, we used a method adapted from Turner et al. (2001), using para-nitrophenyl (pNPP) as an analogue for phosphomonoesterase, a root phosphatase that hydrolyzes simple phosphate monoesters to release a phosphate ion for plant uptake (Browman and Tabatabai 1978, Tabatabai 1994). For each plant within a pot, we collected three subsamples of fine roots for measuring phosphomonoesterase activity and one for measuring color production without substrate added. For each plant, 500 mg of fresh root was added to a glass vial with 9 mL of 50 mM sodium acetate-acetic acid buffer (pH 5.0) and shaken in a water bath at 28°C for five minutes to simulate wet season surface soil temperatures at Kruger National Park. To initiate the reaction, we added 1.0 mL of 50 mM pNPP (5.0 mM final concentration). The reaction was terminated by mixing 0.5 mL of buffer/substrate solution with 4.5 mL of 0.11 M NaOH, which was then vortexed and measured for absorbance at 405 nm against paranitrophenol (pNP) standards. Two sodium acetate-acetic acid buffer and substrate blanks were run for every assay. Roots were removed from the vials and dried at 65°C for 3 days to express the activity as μmol pNP g^-1 dry root mass h^-1.