Higher leaf nitrogen content is linked to tighter stomatal regulation of transpiration and more efficient water use across dryland trees

Summary

• The least‐cost economic theory of photosynthesis shows that water and nitrogen are mutually substitutable resources to achieve a given carbon gain. However, vegetation in the Sahel has to cope with the dual challenge imposed by drought and nutrient‐poor soils.

• We addressed how variation in leaf nitrogen per area (N_area) modulates leaf oxygen and carbon isotopic composition (δ¹⁸O, δ¹³C), as proxies of stomatal conductance and water‐use efficiency, across 34 Sahelian woody species.

• Dryland species exhibited diverging leaf δ¹⁸O and δ¹³C values, indicating large interspecific variation in time‐integrated stomatal conductance and water‐use efficiency. Structural equation modeling revealed that leaf N_area is a pivotal trait linked to multiple water‐use traits. Leaf N_area was positively linked to both δ¹⁸O and δ¹³C, suggesting higher carboxylation capacity and tighter stomatal regulation of transpiration in N‐rich species, which allows them to achieve higher water‐use efficiency and more conservative water use.

• These adaptations represent a key physiological advantage of N‐rich species, such as legumes, that could contribute to their dominance across many dryland regions. This is the first report of a robust mechanistic link between leaf N_area and δ¹⁸O in dryland vegetation that is consistent with core principles of plant physiology.

Introduction

Drylands occupy c. 45% of the Earth’s surface (Prăvălie, ²⁰¹⁶; Maestre et al., ²⁰²¹) and are defined as regions where evapotranspiration is greater than precipitation, leading to water deficit (Huang et al., ²⁰¹⁶). Low soil fertility is yet another limitation for plant productivity in many drylands, especially regarding soil nitrogen and phosphorus content (Noy‐Meir, ¹⁹⁷³; Breman & De Wit, ¹⁹⁸³). However, dryland plants often show higher leaf nitrogen contents than do species from wetter ecosystems (Wright et al., ^{2001, 2003}). Several hypotheses have been proposed to explain the high leaf N contents typically found in dryland vegetation (Prentice et al., ²⁰¹⁴; Adams et al., ²⁰¹⁶). More than 70% of the total N contained in plant leaves is allocated to RuBisCO and other enzymes and proteins involved in photosynthesis (Evans, ¹⁹⁸⁹; Evans & Seemann, ¹⁹⁸⁹; Onoda et al., ²⁰¹⁷; Evans & Clarke, ²⁰¹⁸). Hence, a higher investment in N uptake and allocation to leaves enhances CO₂ fixation and reduces CO₂ concentration in leaf intercellular spaces (c _i) relative to the atmosphere (c _a), leading to low internal c _i/c _a ratios at the sites of carboxylation. Wright et al. (2001, 2003) suggested that plants in low‐rainfall environments increase their N content per unit leaf area (N_area) so that they can save water while maintaining photosynthetic rates similar to plants from wetter environments. This is achieved by optimizing carboxylation and carbon assimilation capacity thanks to high leaf N_area, while at the same time reducing stomatal conductance (g _s) and transpiration (E) rates to minimize water loss, given that a steeper CO₂ diffusion gradient helps to maintain supply of CO₂ into the leaves. This general theory has been named the least‐cost economic theory of photosynthesis (Wright et al., ²⁰⁰³; Prentice et al., ²⁰¹⁴; Wang et al., ²⁰¹⁷).

Remote sensing studies have recently raised awareness about the high woody plant cover and large number of tree individuals present across the Sahelian drylands in northwest Africa (Brandt et al., ²⁰²⁰), despite the dual challenge imposed by low soil fertility and drought stress on plant photosynthesis. Legumes (Fabaceae) are often the dominant plant family in terms of cover and species diversity in Sahelian drylands (Felker, ¹⁹⁸¹; Sprent & Gehlot, ²⁰¹⁰) and are also widely used for livestock feeding. In fact, Adams et al. (2016) showed that high leaf N_mass and N_area, acquired through symbiosis with N₂‐fixing bacteria present in their roots (Powers & Tiffin, ²⁰¹⁰; Vitousek et al., ²⁰¹³; Song et al., ²⁰¹⁵), allows legumes to use water more efficiently than nonlegumes at global scale. Furthermore, production and accumulation of N‐rich osmolytes can also help dryland legumes cope with drought through enhanced internal osmotic adjustment that lowers plant water potentials, thereby increasing plant water uptake from drying soil (Wink, ²⁰¹³). However, the extent to which optimization of water‐use efficiency in dryland woody legumes is achieved primarily through tighter stomatal regulation or via enhanced carbon assimilation capacity (or both) remains unclear (but see Adams et al., ^2018a,b for herbaceous and crop legumes).

Carbon, oxygen and nitrogen stable isotopic composition of leaf material (leaf δ¹³C, δ¹⁸O and δ¹⁵N, respectively), along with xylem water isotopic composition (δ²H and δ¹⁸O), have become key traits that integrate information about plant‐resource acquisition and how plants interact with and respond to their abiotic and biotic environments (Dawson et al., ²⁰⁰²; Barbour, ²⁰⁰⁷; Prieto et al., ²⁰¹⁸; Querejeta et al., ²⁰¹⁸). Leaf δ¹³C and δ¹⁸O in dryland plants can provide reliable proxies of time‐integrated intrinsic water‐use efficiency (WUE_i) and g _s, respectively (Ehleringer, ¹⁹⁹³; Williams & Ehleringer, ¹⁹⁹⁶; Barbour, ²⁰⁰⁷). Leaf δ¹³C in C₃ species is negatively and linearly correlated with the time‐integrated c _i/c _a ratios during photosynthesis (Farquhar et al., ¹⁹⁸⁹) and reflects the relationship between net photosynthetic rate (A) and g _s, thereby providing a robust indicator of time‐integrated intrinsic water‐use efficiency (WUE_i = A/g _s) during the growing season (Dawson et al., ²⁰⁰²). Leaf δ¹⁸O is influenced by the isotopic composition of the water source used by the plant (Sarris et al., ²⁰¹³; Ding et al., ²⁰²¹). The isotopic signal of the water source is thereafter modified by leaf‐level evaporative effects, including stomatal responses to changes in atmospheric relative humidity and soil moisture, making leaf δ¹⁸O a good proxy of time‐integrated stomatal conductance (g _s) and cumulative transpiration in dryland species (Querejeta et al., ²⁰⁰⁶; Barbour, ²⁰⁰⁷; Ramírez et al., ²⁰⁰⁹; Prieto et al., ²⁰¹⁸). The enrichment in ¹⁸O of leaf dry matter above the δ¹⁸O value of the source water used by the plant (leaf Δ¹⁸O) helps to remove the signal of interplant variation in water sources and is thus useful to estimate differences in stomatal regulation among coexisting species exposed to similar environmental conditions (Barbour, ²⁰⁰⁷). Overall, the combined measurement of leaf δ¹⁸O/Δ¹⁸O and δ¹³C (hereafter leaf δ¹³C‐WUE_i) can help in assessing variations in photosynthetic capacity, stomatal conductance, WUEi and overall water‐use strategy (from conservative to profligate) among coexisting dryland species exposed to similar environmental conditions (Moreno‐Gutiérrez et al., ²⁰¹²; Prieto et al., ²⁰¹⁸).

In this study, we addressed how interspecific variation in leaf N_area modulates leaf carbon and oxygen isotopic composition across 34 woody species encompassing diverse phylogenies and leaf habits in the Sahel. In particular, we hypothesized that leaf δ¹³C and δ¹⁸O values will be positively correlated across dryland species as a result of the shared dependence of both traits on stomatal conductance, and will converge towards a relatively narrow range of values constrained by the harsh climatic and soil conditions (Paillassa et al., ²⁰²⁰); and that higher leaf N_area will be associated with higher leaf δ¹⁸O and δ¹³C values across dryland species, indicating greater carboxylation capacity, tighter stomatal regulation with lower conductance and higher WUE_i (Wright et al., ^{2001, 2003}). We further aimed to elucidate whether higher leaf δ¹³C‐WUE_i with increasing leaf N_area in dryland species might be achieved through higher carboxylation capacity, allowing lower stomatal conductance for any given photosynthetic rate (resulting in a strong influence of leaf N_area on both δ¹⁸O and δ¹³C), or primarily through enhanced carboxylation capacity but with little or no impact on stomatal conductance (resulting in weak or no influence of leaf N_area on leaf δ¹⁸O). These hypotheses were analyzed and tested with structural equation modeling (SEM; Fig. 1) using a dataset of leaf and stem traits collected on 34 Sahelian woody species. We thereafter compared legumes (Fabaceae) vs nonlegumes (Table 1) and hypothesized that legumes would exhibit higher leaf δ¹⁸O and δ¹³C‐WUE_i values and greater drought tolerance than nonlegumes thanks to higher leaf N_area achieved through symbiotic atmospheric N₂ fixation. We also hypothesized that dryland legumes would fix large amounts of atmospheric N₂, resulting in distinct leaf δ¹⁵N values near 0‰, whereas co‐occurring nonlegumes would show high leaf δ¹⁵N values typical of plants using ¹⁵N‐enriched soil N sources in hot and dry environments (Amundson et al., ²⁰⁰³; Aranibar et al., ²⁰⁰⁴; Craine et al., ^{2009, 2015}).

Fig. 1.

Structural equation model based on the hypothesized effects of leaf N per area on time‐integrated stomatal conductance (g _s) and intrinsic water‐use efficiency (WUE_i) at leaf level in dryland woody species (Wright et al., ²⁰⁰¹; Adams et al., ²⁰¹⁶) in the Sahel. A priori set of hypotheses and relationships among variables (a) are described in the ‘Materials and Methods’ section. In this model, leaf N_area is considered as a surrogate of carboxylation capacity (V _cmax), leaf δ¹⁸O is a surrogate of stomatal conductance (g _s) and δ¹³C is a surrogate of WUE_i. Continuous and dashed black arrows (b) indicate positive and negative relationships between variables, respectively. Numbers adjacent to arrows indicate the effect size (standardized path coefficients, analogous to partial regression weights) and significance (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ^¥, marginally significant, P = 0.074) of the path; arrow thickness is proportional to the effect size (n = 168). Numbers within circles indicate squared multiple correlations for the variables. Overall goodness‐of‐fit tests (χ², comparative fit index (CFI), incremental fit index (IFI) and root mean square error of approximation (RMSEA)) are shown at the bottom of the model. SLA, specific leaf area.

Table 1.

General characteristics of the tree and shrub species sampled in two semiarid agroforestry systems in the Sahel (Mali and Senegal).

Species	Family	Growth type	Leaf habit	DBH (m)	Height (m)	Mali (no. of trees)	Senegal (no. of trees)
Acacia nilotica (L.) Willd. ex Delile	Fabaceae	T	D	0.71 ± 0.11	5.80 ± 0.28	1	6
Acacia senegal (L.) Willd.	Fabaceae	T	D	0.19 ± 0.01	7.25 ± 1.75		2
Acacia seyal Delile	Fabaceae	T	D	0.52 ± 0.08	4.94 ± 0.30	6	2
Acacia tortilis spp. Raddiana (Forssk.) Hayne (Savi) Brenan	Fabaceae	T	D	0.27 ± 0.03	5.92 ± 1.05		6
Adansonia digitata L.	Malvaceae	T	D	2.46 ± 0.64	10.43 ± 0.88	7	6
Annona senegalensis Pers.	Annonaceae	S	D		0.97 ± 0.05		6
Anogeissus leiocarpus (DC.) Guill. & Perr.	Combretaceae	T	E	1.54 ± 0.07	10.18 ± 1.19	6
Aphania senegalensis (Juss. ex Poir.) Radlk	Sapindaceae	T	E	0.29 ± 0.04	4.13 ± 0.88		2
Balanites aegyptiaca (L.) Delile	Zygophyllaceae	T	E	0.69 ± 0.12	5.98 ± 0.70	6	6
Bauhinia rufescens Lam.	Fabaceae	T	D	0.14 ± 0.00	4.25 ± 0.52		3
Boscia senegalensis (Pers.) Lam. ex Poir.	Capparaceae	S‐T	E		1.71 ± 0.23		6
Celtis integrifolia Lam.	Ulmaceae	T	E	0.55 ± 0.08	11.42 ± 1.04		6
Combretum glutinosum Perr. ex DC.	Combretaceae	T (S‐T)	E	0.75 ± 0.17	7.11 ± 0.75	6	6
Combretum micranthum G. Don.	Combretaceae	S	D	0.17 ± 0.06	2.68 ± 0.28	6
Cordia sinensis Lam.	Boraginaceae	T	E	0.11	6.00		1
Crateva religiosa Forst. f.	Capparaceae	T	E	0.33 ± 0.02	8.08 ± 0.96		3
Diospyros mespiliformis Hochst. ex A. DC.	Ebenaceae	T	D	1.34 ± 0.13	6.91 ± 0.81	7
Faidherbia albida (Delile) A. Chev.	Fabaceae	T	D	0.83 ± 0.17	9.04 ± 0.70	6	6
Ficus gnaphalocarpa (Miq.) Steud.	Moraceae	T	D	2.11 ± 0.43	8.53 ± 1.11	6
Guiera senegalensis J. F. Gmel.	Combretaceae	S‐T	D	0.12 ± 0.02	3.02 ± 0.10	6
Maytenus senegalensis (Lam.) Exell	Celastraceae	S‐T	D	0.12 ± 0.00	2.39 ± 0.44		5
Neocarya macrophylla (Sabine) Prance	Chrysobalanaceae	T	E	0.38 ± 0.04	6.08 ± 0.98		6
Piliostigma reticulatum (DC.) Hochst.	Fabaceae	T	E	0.79 ± 0.13	4.99 ± 0.26	6	2
Prosopis africana (Guill. & Perr.) Taub.	Fabaceae	T	D	2.00 ± 0.24	9.53 ± 1.36	6
Prosopis juliflora (Sw.) DC.	Fabaceae	T	E	0.26 ± 0.03	8.17 ± 0.76		6
Pterocarpus erinaceus Poir.	Fabaceae	T	D	1.38 ± 0.11	7.78 ± 0.36	6
Saba senegalensis (A. DC.) Pichon	Apocynaceae	T	E		4.20 ± 0.72	6
Sclerocarya birrea (A. Rich.) Hochst.	Anacardiaceae	T	D	0.77 ± 0.13	7.98 ± 0.40	6	6
Tamarindus indica L.	Fabaceae	T	E	1.20 ± 0.23	8.78 ± 0.91	6	7
Tamarix senegalensis DC.	Tamaricaceae	T(S‐T)	E		2.20 ± 0.21		6
Terminalia laxiflora Engl.	Combretaceae	T	D	1.70 ± 0.23	9.23 ± 1.14	6
Vitellaria paradoxa C.F.Gaertn.	Sapotaceae	T	D	1.70 ± 0.14	9.83 ± 1.40	6
Vitex doniana Sweet	Labiatae	T	D	2.02 ± 0.12	8.62 ± 0.62	5
Ziziphus mauritiana Lam.	Rhamnaceae	S	E	0.43 ± 0.18	5.42 ± 0.44	3	6
Total						124	106

Species (N = 34), families, growth type (T, tree; S, shrub; S‐T, spp. that can grow as a shrub or tree depending on the environmental conditions), leaf habit (E, evergreen (N = 15 spp.); D, deciduous (N = 19 spp.)), mean (± SE) diameter at breast height (DBH) and plant vegetative height (height) and number of sampled individuals.

Materials and Methods

Study sites

The study was conducted in silvopastoral and agroforestry ecosystems of the western Sahel region. Sampling sites were selected near Louga (15°37′N, 16°13′W) in northwest Senegal (seven sites), and near Ségou (13°27′N, 6°16′W) in south‐central Mali (eight sites). In this region, rain falls mainly during the monsoon season (June–October), followed by a dry season between November and June (Supporting Information Fig. S1). Louga has a semiarid sub‐Canarian climate (Wade, ¹⁹⁹⁷) with a mean temperature of 24.9°C and average annual rainfall of 330 mm (1950–2000). Both rain and the warmest temperatures occur mostly during the monsoon season (Edmunds & Gaye, ¹⁹⁹⁴; NOAA, ²⁰¹⁵). Ségou has a continental semiarid climate with lower relative humidity than Louga, but has higher mean annual temperature (27.8°C) and precipitation (566 mm; NOAA, ²⁰¹⁵; Fig. S1). Daily maximum temperatures in both areas frequently reach above 50°C. Soils in Louga and Ségou are highly weathered acidic sands on ancient dunes, inter‐dune depressions or plains that typically show low water‐holding capacity and organic matter content, and are particularly poor in phosphorus and other nutrients (Bitchibaly et al., ²⁰¹²). Relatively shallow groundwater can be found in both the Louga and Ségou sites studied. Vegetation of these agroforestry ecosystems is an open savannah with sparse trees and shrubs scattered across a grassland matrix and interspersed with croplands where typical management practices include harvesting of trees and shrubs, grazing and farming (IER, ²⁰¹⁰; Konaté, ²⁰¹⁰).

Sampling and trait data collection

We sampled leaves and stems from 230 individual trees and shrubs of 34 species, including 11 species from the Fabaceae family (legumes, Table 1). Both legumes and nonlegumes include evergreen species that retain a full canopy throughout the year and drought‐deciduous species that remain leafless or partially leafless for several months during the dry season. The only exception is Faidherbia albida, a deciduous legume that sheds leaves during the rainy season (Diémé et al., ²⁰¹⁸). Species were assigned to leaf habit categories based on local expert knowledge (Diémé et al., ^{2018, 2019}) and descriptions in Arbonnier (2004). Plant samples were collected in the early dry season of 2011 before leaf senescence of deciduous species. Sampled trees and shrubs were healthy‐looking and were at least 15 m far apart from each other (ranging between 0.015 and 60 km apart within each country; http://wms.nina.no/FunciTree/) and with their crowns fully exposed to sunlight.

From each plant, we sampled two sun‐exposed branches from the eastern side of the canopy before dawn. One branch was placed in a sealed plastic bag within a dark hermetic bucket and was used for measuring stem predawn water potential (Ψ_pd) with a Scholander‐type pressure bomb. Leaf thickness (mm), specific leaf area (SLA; m² kg⁻¹), and leaf relative water content (RWC; g g⁻¹) were measured in fully expanded, mature, damage‐free fresh leaves. Thickness was measured in three points in each leaf with a digital caliper and the mean value was recorded. The leaf collected to measure RWC was first weighed (FW), then fully rehydrated overnight in the dark, weighed again (hydrated weight, HW) and then scanned, and leaf area was then measured with ImageJ. Leaves were then oven‐dried at 60°C for 72 h and weighed again (DW). SLA is the one‐sided area (leaf area, LA) of the fully rehydrated leaf divided by its dry mass (SLA = LA/DM), while leaf RWC is the difference between leaf FW and DW divided by the difference between fully hydrated weight and leaf DW (i.e. RWC = (FW − DW)/(HW − DW)). Dry leaves were then ground using a ball‐mill to determine C and N concentrations (mass based) and δ¹³C, δ¹⁵N and δ¹⁸O composition. Leaf N_area (mg cm⁻²) was calculated as the ratio between leaf N_mass and SLA. From the second branch, we cut a terminal 8 cm leafless woody stem which was immediately placed in a screw‐cap polypropylene vial and sealed with Parafilm. Vials were transported in a cooler to the laboratory within 4 h and stored frozen. Xylem water was extracted using cryogenic vacuum distillation (Ehleringer & Osmond, ¹⁹⁸⁹). The oxygen isotopic composition of xylem water (xylem water δ¹⁸O) helps in assessing the approximate depth of soil water uptake by roots in dryland ecosystems where steep vertical gradients in soil water δ¹⁸O develop during rainless periods (Moreno‐Gutiérrez et al., ²⁰¹²). Evaporation from upper soil during hot, dry periods leads to heavy isotopic enrichment of the remaining soil water near the surface, which decreases steeply with depth (Allison & Hughes, ¹⁹⁸³). Higher xylem water δ¹⁸O values indicate uptake of isotopically enriched water from shallower soil layers exposed to intense evaporation, whereas lower xylem water δ¹⁸O values indicate utilization of non‐enriched water from deeper, less evaporated water sources (Querejeta et al., ²⁰⁰⁷; Ding et al., ²⁰²¹).

Foliar δ¹³C can be used to estimate long‐term ratios of the intercellular to ambient CO₂ values (c _i/c _a) if the carbon isotope ratio of atmospheric CO₂ (δ¹³C_air) is known (Farquhar et al., ¹⁹⁸⁹). To calculate long‐term c _i/c _a ratios, we first calculated Δ¹³C as:

$${\mathrm{\Delta }}^{13}\mathrm{C}=\frac{1000\left({\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{air}}-{\mathrm{\delta }}^{13}{\mathrm{C}}_{\text{leaf}}\right)}{1000-{\mathrm{\delta }}^{13}{\mathrm{C}}_{\text{leaf}}}$$ (Eqn 1)

where δ¹³C_air is the C isotopic composition of atmospheric CO₂ (−8.45‰, Mauna Loa records; http://www.esrl.noaa.gov/gmd/dv/ftpdata.html) and δ¹³C_leaf is the C isotopic composition of leaf material. Then, from Δ¹³C values, we calculated c _i/c _a as:

$$\frac{c_{\mathrm{i}}}{c_{\mathrm{a}}}=\frac{{\mathrm{\Delta }}^{13}\mathrm{C}-a_{\mathrm{s}}}{\overline{b}-a_{\mathrm{s}}}$$ (Eqn 2)

where a _s is the fractionation factor of gaseous diffusion (4.4‰) and $\overline{b}$ represents effective fractionation as a result of carboxylation (27‰), estimated empirically (Farquhar et al., ¹⁹⁸²).

Cryogenic vacuum distillation and stable isotope analyses of leaf and water samples were conducted at the Stable Isotope Ratio Facility for Environmental Research, University of Utah (USA). Leaf N and C concentrations and δ¹³C, δ¹⁵N were measured with an isotope ratio mass spectrometer (Finnigan Mat Delta+ IRMS, Waltham, MA, USA) coupled to an elemental analyzer (EA; Carlo Erba CHN EA1110, Waltham, MA, USA). Leaf δ¹⁸O was measured with a Finnigan TC/EA IRMS. The δ¹⁸O isotopic composition of xylem water was measured using a laser water isotope analyzer (Picarro L2130i, Santa Clara, CA, USA).

Data analysis

To investigate the influence of leaf N_area on δ¹⁸O and δ¹³C‐WUE_i in dryland vegetation, we built an a priori structural equation model (SEM) of hypothesized relationships within a path diagram (see the ‘Introduction’ section and Fig. 1a), allowing a causal interpretation of the model outputs (Grace, ²⁰⁰⁶). This a priori set of hypothesized relationships were as follows: high leaf N concentration (N_mass) and low SLA (i.e. higher LMA) both increase leaf N_area, which in turn enhances carboxylation capacity, allowing tighter stomatal regulation of transpiration and lower time‐integrated stomatal conductance (Wright et al., ^{2001, 2003}). A key underlying assumption in this SEM model is that a higher leaf N_area enhances the leaf carboxylation capacity (V _cmax) and thus enables the leaf to achieve a given carbon assimilation rate with lower stomatal aperture and conductance under dry environmental conditions. In our a priori SEM model, leaf N_area is considered a proxy of V _cmax (Wright et al., ^{2001, 2003}), leaf δ¹⁸O is considered a proxy of time‐integrated stomatal conductance (inversely related, 1/g _s; Barbour, ²⁰⁰⁷), and leaf δ¹³C is a proxy of WUE_i (Farquhar et al., ¹⁹⁸⁹). The SEM was first tested with the experimental data collected in 34 dryland tree and shrub species from the Sahel (n = 168 individuals for which both SLA and N_area data were available; Fig. 1). Then, to rule out any potential confounding effects as a result of pooling together species of different leaf habits (deciduous vs evergreens) and with different proportions of legumes (eight legume species were deciduous and two were evergreen), we reanalyzed the dataset excluding evergreen species (Powers & Tiffin, ²⁰¹⁰). Goodness of fit of the SEM model was assessed using the traditional χ² goodness‐of‐fit test, but because of its sensitivity to sample size, the Bentler comparative fit index (CFI) and the incremental fit index (IFI) and root mean square error of approximation (RMSEA) were also considered (Grace, ²⁰⁰⁶). For the SEM analysis, contrary to other statistical analyses, model P‐values > 0.05 in the χ² and RMSEA indices (Schermelleh‐Engel et al., ²⁰⁰³), and values close to 1 (> 0.90) for CFI and IFI indices are required to guarantee an acceptable fit (Hu & Bentler, ¹⁹⁹⁹).

We did not account for spatial variability among sampling sites in the SEM, but performed pairwise regression analyses between traits using linear mixed regression models with sampling site included as a random factor. Differences in plant traits between legumes and nonlegumes were analyzed with linear mixed models were the main fixed factor was Fabaceae/nonFabaceae. Geographic area (i.e. sampling site) was included as a random factor. Normality of residuals and homogeneity of variances assumptions were assessed by graphical inspection of residuals, and when these assumptions were not met (i.e. for Ψ_pd), we used a model correction for heterogeneity of variance (varExp), which represents an exponential structure of the variance–covariance matrix (Gałecki & Burzykowski, ²⁰¹³). Post hoc differences were tested with Fisher least significant difference tests.

All statistical analyses were performed with R software (R Core Team, ²⁰¹⁹) interfaced by InfoStat statistical software v.2020 (Di Rienzo et al., ²⁰²⁰) using the packages lme4 (Bates et al., ²⁰¹⁵) and nlme (Pinheiro et al., ²⁰¹⁴). Structural equation modeling analyses were carried out with the AMOS extension in Spss (Arbuckle, ²⁰¹⁴).

Results

Trait coordination across dryland woody species

Our pool of 34 woody species (mean height = 7.3 m; Table 1) encompassed a remarkably wide range of leaf N_area, N_mass, δ¹⁸O and δ¹³C values, thereby revealing large interspecific differences in carboxylation capacity, stomatal regulation of transpiration, c _i/c _a ratios and WUE_i (Table 2). Sahelian trees and shrubs also showed large interspecific variability in SLA, leaf thickness, leaf RWC, δ¹⁵N and Δ¹⁸O, xylem water δ¹⁸O, and predawn water potentials (Fig. 2; Table S1). Whereas interspecific differences accounted for most of the variability in trait values (Fig. 2), intraspecific variability across sites was also rather high for some traits (Table S1 and indirectly shown in Fig. 2), suggesting large phenotypic trait variability in response to environmental heterogeneity. We did not detect any significant phylogenetic signal (λ) in any of the measured traits (Table S2).

Table 2.

Description, number of trees and range of values (min to max) for the 10 functional traits measured in 34 woody species in the Sahel (Mali and Senegal) and their key role in plant functioning.

Trait	Abbreviation	Units	N	Range	Key role in plant functioning/interpretation
Leaf carbon isotopic composition	Leaf δ13C	‰	217	−31.5 to −24.6	Time‐integrated water‐use efficiency
Ratio of intercellular and ambient CO2 concentrations	c i/c a	Unitless	217	0.57 to 0.91	CO2 carboxylation capacity, stomatal conductance
Leaf nitrogen concentration	Leaf Nmass	%	216	1.32 to 4.80	Photosynthetic and carboxylation capacity, osmoregulation, nutritional quality
Leaf nitrogen area	Leaf Narea	mg cm−2	169	0.08 to 0.98	Carboxylation capacity
Leaf oxygen isotopic composition	Leaf δ18O	‰	211	19.02 to 34.53	Time‐integrated stomatal conductance, cumulative transpiration
Leaf oxygen isotopic enrichment above source water	Leaf Δ18O	‰	199	23.58 to 42.08	Time‐integrated stomatal conductance, cumulative transpiration
Predawn stem water potential	Ψpd	MPa	224	−5.59 to −0.28	Soil water availability, rooting depth, osmolyte accumulation in plant tissues
Relative water content	RWC	%	221	61.1 to 96.4	Leaf water status and hydration
Leaf thickness	Leaf Thickness	mm	207	0.12 to 0.62	Leaf gas exchange, water retention
Specific leaf area	SLA	m2 kg−1	175	4.25 to 18.63	Photosynthetic rate, light capture, relative growth rate
Leaf nitrogen isotopic composition	Leaf δ15N	‰	217	3.42 to 12.9	Atmospheric N2 fixation, soil N sources
Oxygen isotopic composition of xylem water	Xylem water δ18O	‰	211	−8.13 to 1.20	Water sources, depth of soil water uptake

Description of traits and key role in plant function are based on Wright et al. (2001, 2005); Barbour (2007); Bernard‐Verdier et al. (2012); Perez‐Harguindeguy et al. (2013), and Craine et al. (2015).

Fig. 2.

Estimated variance decomposition for 12 functional traits measured in woody species in the Sahel: (a) in the 34 woody species measured; and (b) in nine species common to the two sites sampled (Ségou, Mali, and Louga, Senegal; see Table 1 and Supporting Information Table S1). Variance partitioning between sites (15 sample sites, brown) and species (interspecific, green) is shown; residual variance is also shown (light gray). Leaf trait abbreviations and units are as follows: leaf N_mass, leaf nitrogen concentration on a mass basis (%); leaf N_area (mg cm⁻²), leaf N content per leaf area; SLA (m² kg⁻¹), specific leaf area; leaf thick, leaf thickness (mm); leaf δ¹³C, leaf carbon isotopic composition (‰); c _i/c _a, ratio of intercellular and ambient CO₂ concentrations (unitless); leaf δ¹⁸O, leaf oxygen isotopic composition (‰); leaf Δ¹⁸O, leaf ¹⁸O enrichment above source (xylem) water (‰); xylem δ¹⁸O, xylem water oxygen isotopic composition (‰); RWC, leaf relative water content (g g⁻¹); Ψ_pd, predawn stem water potential (MPa); leaf δ¹⁵N, leaf nitrogen isotopic composition (‰).

Leaf δ¹⁸O was unrelated to xylem water δ¹⁸O across species and sites, indicating that variation in leaf δ¹⁸O was not primarily driven by inter‐plant differences in source water isotopic composition or root water uptake depth. Instead, both foliar δ¹⁸O and Δ¹⁸O were positively related to leaf δ¹³C (Fig. 3; Table 3), indicating that foliar oxygen isotopic composition primarily reflected leaf‐level evaporative effects related to stomatal regulation and time‐integrated stomatal conductance and cumulative transpiration (Fig. 3). Moreover, both foliar δ¹⁸O/Δ¹⁸O and δ¹³C correlated positively with leaf N_area, revealing tighter stomatal regulation and higher WUE_i with increasing carboxylation capacity (i.e. leaf N_area; Fig. 3; marginally significant for leaf N_area‐δ¹³C relationship). Furthermore, leaf δ¹⁸O and Δ¹⁸O correlated closely with RWC (Fig. 3; Table 3), suggesting that tighter stomatal regulation and lower time‐integrated stomatal conductance are linked to improved leaf hydration during the early dry season in dryland species.

Fig. 3.

Relationships between different pairs of traits measured in 34 woody species in the Sahel (n = 168–230 individuals): relationship between leaf N_area and leaf carbon isotopic composition (δ¹³C) (a), leaf oxygen isotopic composition (δ¹⁸O) (b) and predawn stem water potential (Ψ_pd) (c), between leaf δ¹³C and leaf δ¹⁸O isotopic composition (d), and between relative water content (RWC) and leaf δ¹⁸O (e) and leaf oxygen isotopic enrichment above source (xylem) water (Δ¹⁸O) (f). Solid lines represent the model‐predicted relationships, and dotted lines the 95% confidence intervals. Colored dots depict different species. Log‐transformed N_area values (negative scale) are shown in panels (a–c).

Table 3.

Results from linear mixed regression models for pairwise relationships between key traits analyzed in this study (many of them included in the a priori structural equation modeling; see Fig. 1) for the whole dataset (N = 34 species).

Predictor variable	Response variable	Slope	SE	F‐value	P‐value
Leaf Narea	Leaf δ13C	1.162	0.684	2.887	0.091
Leaf Narea	Leaf δ18O	5.247	1.869	7.885	0.005
Leaf Narea	Leaf Δ18O	5.614	2.362	5.648	0.019
Leaf Narea	Ψpd	−2.869	0.513	31.258	< 0.001
Leaf Narea	RWC	0.020	0.033	0.368	0.545
Leaf Narea	Leaf δ15N	1.762	0.520	11.46	< 0.001
Leaf Nmass	Leaf δ13C	0.312	0.116	7.269	0.007
Leaf Nmass	Leaf δ18O	1.674	0.533	9.883	0.002
Leaf Nmass	Leaf Δ18O	1.377	0.654	4.437	0.036
Leaf Nmass	Ψpd	−0.514	0.125	17.048	0.000
Leaf Nmass	RWC	−0.008	0.008	−0.923	0.357
Leaf Nmass	Leaf δ15N	0.447	0.205	4.756	0.030
Leaf δ18O	Ψpd	−0.007	0.015	0.177	0.674
Leaf δ18O	RWC	0.004	0.001	17.57	< 0.001
Xylem water δ18O	Leaf δ18O	−0.057	0.122	0.218	0.641
Leaf δ13C	Ψpd	−0.053	0.052	1.048	0.307
Leaf δ13C	RWC	−0.002	0.003	0.342	0.559
Leaf δ13C	Leaf δ18O	0.758	0.213	12.641	< 0.001

Shown are the slope and standard error of the slope (SE) and F‐ and P‐values (significant relationships, i.e. P < 0.05, are in bold, and marginal significant relationships are in italics). Trait abbreviations are as in Table 2.

Modulation of water‐use efficiency by leaf N_area is achieved through both enhanced carboxylation capacity and tighter stomatal regulation in dryland trees

The SEM analysis (Fig. 1) indicated that both higher leaf N concentration on a mass basis (N_mass) and lower SLA (i.e. higher LMA) contributed to enhance leaf N_area in dryland trees and shrubs. As predicted by theory and global datasets, leaf N_mass and SLA (1/LMA) were positively related to each other across species (e.g. Wright et al., ²⁰⁰⁴). Furthermore, the SEM analysis revealed two simultaneous pathways that explained the positive relationship between leaf N_area and δ¹³C‐WUE_i in dryland species (Fig. 1b). A main SEM pathway linking leaf N_area, δ¹⁸O and δ¹³C highlighted a strong influence of leaf N_area on foliar δ¹⁸O (β = 0.20, P < 0.01), indicating tighter stomatal regulation and lower time‐integrated stomatal conductance with increasing leaf N_area, which in turn contribute to enhanced WUE_i through leaf δ¹⁸O/stomatal effects on δ¹³C‐WUE_i (β = 0.22, P < 0.01). Leaf N_area also had a smaller direct positive effect on leaf δ¹³C (β = 0.14, P = 0.07; marginally significant; Fig. 1b) that was unrelated to stomatal effects, which we interpret as an indication of enhanced carboxylation capacity with increasing N_area across species (as predicted by theory) that further contributes to increase WUE_i.

Pairwise regression analyses between leaf traits further supported the key role of a high leaf N_area in enhancing carboxylation capacity, thereby allowing a tighter stomatal regulation of transpiration (i.e. lower time‐integrated stomatal conductance) and higher WUE_i. It is worth highlighting that, in addition to the positive pairwise relationship between leaf N_area and δ¹³C (Figs. 1, 3; marginally significant), there was also a stronger positive pairwise relationship between leaf N_mass and δ¹³C (Table 3). Leaf N_area was also positively correlated with leaf δ¹⁵N, and negatively correlated with predawn water potential (Table 3), the latter suggesting greater tree ability to sustain more negative water potentials probably through internal plant osmoregulation linked to accumulation of N‐based osmolytes (Fig. 3c). Pairwise regression analyses also revealed a strong positive relationship between leaf δ¹³C and δ¹⁸O across dryland species (Fig. 3d), further evidencing a shared stomatal control on both traits. Moreover, tighter stomatal regulation and lower time‐integrated stomatal conductance (inferred from higher leaf δ¹⁸O and Δ¹⁸O values) were linked to improved leaf hydration and higher leaf relative water content in dryland woody species (Fig. 3e,f).

The majority of N‐fixing species in our dataset where drought deciduous (Table 1), so we thereafter tested the robustness of the described links among traits within the subset of drought deciduous species only, in order to ascertain that the potentially confounding effects of contrasting leaf phenology (evergreen/deciduous) were not biasing the results of the SEM analyses (while at the same encompassing the full range of N_area and N_mass values present in the dataset). Within the subset of deciduous species (i.e. excluding evergreens), SEM analyses revealed very similar links among traits; however, the overall fit of the SEM analysis was poorer as a result of low statistical power (n = 87; data not shown). Anyhow, pairwise relationships between variables for deciduous species were also similar to those found in the whole dataset (Table S3), which further confirms the robustness of the described patterns.

Differences in plant traits between dryland legumes and nonlegumes

Dryland legumes (Fabaceae) had, on average, higher leaf N_mass, SLA and δ¹⁸O/Δ¹⁸O values, and thinner, less sclerophyllous leaves compared with nonlegumes (Fig. 4). By contrast, legumes and nonlegumes did not differ significantly in leaf N_area, δ¹³C (Fig. 4) or leaf RWC (F _1,212 = 0.67; P = 0.414), predawn water potential of stems (F _1,215 = 1.35; P = 0.247), xylem water δ¹⁸O values (F _1,202 = 1.64; P = 0.202) or average tree height (F _1,219 = 1.19; P = 0.276).

Fig. 4.

Differences in average trait values between legume and nonlegume species from the Sahel. From left to right panels and from the top to the bottom panels: leaf N content on a mass basis (leaf N_mass), leaf N content per leaf area (leaf N_area), specific leaf area (SLA), leaf thickness, leaf oxygen isotopic composition (leaf δ¹⁸O), leaf oxygen enrichment above source (xylem) water (leaf Δ¹⁸O), leaf carbon isotopic composition (leaf δ¹³C), and leaf N isotopic composition (leaf δ¹⁵N) measured in nonlegumes (nonFabaceae, nonFab) and legumes (Fabaceae species, Fab). Values included in each panel are F‐Fisher results (F _{n1, n2}, where n ₁ and n ₂ are degrees of freedom) of the linear mixed models testing differences between nonFab and Fab species and model significance (**, P < 0.01; ***, P < 0.001; ns, nonsignificant differences). Results are shown as boxplots that include different components: midline, median; upper and lower box edges, first and third quartiles; whiskers, 5% and 95% confidence intervals; points, outliers.

When considering only deciduous species (n = 19 spp.), legumes had marginally higher leaf δ¹³C‐WUE_i (Fig. S2) than nonlegumes, along with higher leaf N_mass and δ¹⁸O/Δ¹⁸O values and thinner leaves. Differences in leaf N_mass between legumes and nonlegumes were larger in deciduous species than in the whole dataset, which includes evergreens (0.75% vs 0.39% higher N_mass in Fabaceae, deciduous vs all spp., respectively). Compared with deciduous nonlegumes (nonFab), deciduous legumes (Fab) also had significantly lower predawn water potentials (−1.41 ± 0.36 vs −2.02 ± 0.36 MPa, nonFab vs Fab; F _1,114 = 22.82, P < 0.0001) and higher xylem water δ¹⁸O values (−4.75 ± 0.47 vs −3.78 ± 0.51‰, nonFab vs Fab; F _1,106 = 4.04, P = 0.047), suggesting colonization of drier microsites and use of shallower, more evaporatively enriched soil water sources by deciduous legumes.

Discussion

Variation in SLA, leaf N_mass, N_area and water‐related isotope traits (δ¹³C, δ¹⁸O, Δ¹⁸O) among Sahelian woody plants was strikingly large, revealing large interspecific differences in the relative costs of nitrogen and water acquisition. This finding can be interpreted in light of the least‐cost economic theory of photosynthesis, showing that nitrogen and water are mutually substitutable resources to achieve a given photosynthetic carbon gain, so that the cost associated with acquiring one of them can be alleviated by increased supply of the other (Wright et al., ^{2001, 2003}; Prentice et al., ²⁰¹⁴; Wang et al., ²⁰¹⁷). Therefore, a high photosynthesis can be achieved by either high leaf N_area or high stomatal conductance, so that the total cost per unit photosynthetic carbon gain associated with carboxylation and transpiration is minimized under the environmental conditions prevailing in the plant's habitat (Paillassa et al., ²⁰²⁰). Within this optimality framework, the positive foliar N_area–δ¹⁸O and N_area–δ¹³C correlations found across Sahelian plants are in agreement with the theory's prediction that a high leaf N_area should allow higher carboxylation capacity, tighter stomatal regulation of transpiration for any given carbon assimilation rate and more efficient use of water in dryland vegetation (Wright et al., ^{2001, 2003}). The structural equation model analysis revealed that the link between leaf N_area and δ¹³C‐WUEi was driven by an enhanced carboxylation capacity but even more strongly by a tighter stomatal regulation of transpiration (inferred from leaf δ¹⁸O) with increasing leaf N_area. To our knowledge, this is the first study demonstrating a robust mechanistic link between leaf N content per area and foliar oxygen isotope composition in dryland vegetation, which is fully consistent with first principles of plant physiology (Wright et al., ^{2001, 2003}; Dong et al., ²⁰¹⁷).

Our pool of 34 phylogenetically diverse tree and shrub species spanned c. 70% of the range of δ¹³C values commonly encountered in C3 vegetation at global scale (−22 to −32‰; Kohn, ²⁰¹⁰; Maire et al., ²⁰¹⁵; Cornwell et al., ²⁰¹⁸; Paillassa et al., ²⁰²⁰). The hot and dry climate with a short rainy season and the nutrient‐poor sandy soils and flat terrain of low elevation (< 300 m asl) that prevail across our sampling sites appear to offer limited potential for environmental heterogeneity and niche segregation that could favor such wide c _i/c _a divergence among coexisting Sahelian woody species. However, more subtle environmental heterogeneity regarding microtopography, groundwater accessibility to roots, soil fertility or microclimatic conditions across and within sampling sites may create opportunities for niche segregation and physiological divergence among species (Álvarez‐Yépiz et al., ²⁰¹⁷). Environmental heterogeneity could also explain the large within‐species variability encountered for some key traits, including leaf N_area, δ¹⁸O and δ¹³C (Donovan et al., ²⁰⁰⁷). Nonetheless, the strikingly large interspecific differences in trait values encountered in this study primarily reflect idiosyncratic differences in leaf physiology and resource use strategies among coexisting species, and must therefore have an evolutionary basis (Fig. 2).

The average c _i/c _a ratio of Sahelian woody species (0.74) was unexpectedly high considering their rather high leaf N concentrations (> 2% on average), N_area values and the high vapor pressure deficit (VPD) prevailing across sampling sites (for comparison, see the average c _i/c _a ratios typically found in drylands as reported by Prentice et al. (2014) and Paillassa et al. (2020)). High leaf N_area and carboxylation capacity combined with high VPD would be expected to favor tight stomatal regulation of leaf gas exchange, leading to very low c _i/c _a ratios (Cernusak et al., ²⁰¹³; Cornwell et al., ²⁰¹⁸; Grossiord et al., ²⁰²⁰; Paillassa et al., ²⁰²⁰). However, extremely hot temperatures combined with the high O₂ partial pressure at low elevations may boost photorespiration in C₃ plants, thereby constraining photosynthetic nitrogen‐use efficiency and preventing further decreases in c _i (Paillassa et al., ²⁰²⁰; but see also Terashima et al., ¹⁹⁹⁵). Moreover, low availability of soil P or other essential nutrients for photosynthesis as a result of poor soil fertility may further constrain carbon assimilation capacity despite high leaf N_area (Augusto et al., ²⁰¹⁷; Du et al., ²⁰²⁰; Paillassa et al., ²⁰²⁰). Alternatively, the high c _i/c _a ratios observed in our study may simply reflect that most of the annual carbon assimilation in Sahelian species takes place during the short rainy season when high soil water availability may help to sustain high stomatal conductance, but with little additional carbon assimilation during the drier part of the year (Choat et al., ²⁰⁰⁶). Finally, taproot access to relatively shallow groundwater may have further contributed to higher‐than‐expected average c _i/c _a ratios, particularly for some phreatophytic species that can behave as profligate water users under hot and high VPD conditions (Gries et al., ²⁰⁰³).

The leaf δ¹⁸O and Δ¹⁸O values of Sahelian woody species were generally lower than those of trees and shrubs from other drylands with less extreme temperature, VPD and more fertile soils (e.g. semiarid Mediterranean ecosystems; Cernusak et al., ²⁰⁰⁵; Moreno‐Gutiérrez et al., ^{2012, 2015}). This suggests rather high stomatal conductance during the short growing season, possibly linked to the need for evaporative leaf cooling to prevent leaf overheating and damage of the photosynthetic machinery under extremely hot conditions in the Sahel (Helliker & Richter, ²⁰⁰⁸; Aparecido et al., ²⁰²⁰). Moreover, high stomatal conductance may be needed to enhance nutrient harvesting through transpiration‐driven mass flow of soil nutrients to roots in these low‐fertility sandy soils (Cramer et al., ²⁰⁰⁹; Matimati et al., ²⁰¹⁴; Salazar‐Tortosa et al., ²⁰¹⁸). Modeling studies at a global scale have shown that stomatal conductance should reach peak levels in dry tropical vegetation (savannah trees; Lin et al., ²⁰¹⁵). The high leaf N_area combined with low leaf δ¹⁸O, Δ¹⁸O and δ¹³C values of Sahelian woody plants suggests that they are capable of achieving high rates of transpiration and carbon assimilation during the short rainy season (Cornwell et al., ²⁰¹⁸; Sibret et al., ²⁰²¹). Sahelian woody vegetation may thus contribute substantially to global primary productivity, despite the severe water and nutrient limitations typical of these drylands (Ahlström et al., ²⁰¹⁵; Smith et al., ²⁰¹⁹; Brandt et al., ²⁰²⁰).

High foliar N content can reflect the production and accumulation of N‐based osmolytes, such as proline, that enable plants to endure and sustain more negative internal water potentials during drought (Wink, ²⁰¹³; Adams et al., ²⁰¹⁶), as suggested by the strong negative correlation found between foliar N_mass and predawn water potential across species (Table 3). Accumulation of N‐based osmolytes enhances internal plant osmotic adjustment and thus the ability of N‐rich species to extract water held at progressively lower soil moisture potentials during dry periods, potentially conferring a competitive advantage in drylands (Giordano et al., ²⁰¹¹). Lower predawn water potentials with increasing foliar N content may thus indicate greater drought tolerance and a superior ability to establish in drier microsites in N‐rich species such as legumes.

Limited N₂ fixation in dryland legumes

The wide range of leaf δ¹⁵N values encountered among Sahelian woody species revealed the existence of a wide diversity of N‐acquiring mechanisms and N sources (Craine et al., ²⁰⁰⁸). Sahelian trees and shrubs exhibited remarkably high foliar δ¹⁵N values, suggesting high rates of gaseous N losses from the soil system (i.e. through ammonia volatilization, nitrification and denitrification) that could explain the heavy ¹⁵N enrichment of leaf nitrogen pools despite vegetation dominance by N₂‐fixing legumes (Craine et al., ^{2008, 2015}; Ruiz‐Navarro et al., ²⁰¹⁶). High rates of gaseous N losses from the soil system may be greatly favored by hot temperatures and transiently moist soil conditions during the short Sahel rainy season, which are known to stimulate the abiotic (e.g. ammonia volatilization) and biotic (e.g. microbial denitrification) processes driving gaseous N losses from soil (Craine et al., ²⁰¹⁵). The large inputs of ¹⁵N‐enriched livestock manure in Sahelian silvopastoral systems may further stimulate heavy gaseous N losses and ¹⁵N enrichment of the remaining soil and vegetation N pools. Interestingly, highly enriched plant δ¹⁵N values were also reported in the Kalahari drylands dominated by woody legumes growing on nutrient‐poor sandy soils (Aranibar et al., ²⁰⁰⁴). Contrary to expectations, the remarkably high leaf δ¹⁵N values found in legumes were indistinguishable from those of nonlegumes, which suggests that atmospheric N₂ fixation may not be their primary source of nitrogen, possibly as a result of the strong water and P limitation imposed on biological N₂ fixation by the weathered sandy soils of the Sahel (Breman & De Wit, ¹⁹⁸³; Henao & Baanante, ¹⁹⁹⁹; Vitousek et al., ²⁰¹⁰). Sahelian legumes may instead deploy other effective N‐acquiring mechanisms that may contribute to their higher average leaf N concentration as a group (McKey, ¹⁹⁹⁴; Dovrat et al., ²⁰²⁰). Moreover, the livestock habit of seeking shelter under the shade of large trees and shrubs in this hot climate may provide an abundant external N subsidy of animal origin from the surrounding grasslands in the silvopastoral systems of the Sahel. This ‘fertility’ island effect could be even greater under legume trees and shrubs producing nutritious pods or fodder that are attractive for livestock (Casals et al., ²⁰¹⁴; Hoosbeek et al., ²⁰¹⁸).

Differences in water‐use traits between legumes and other species

The higher average leaf δ¹⁸O and Δ¹⁸O values of legumes compared with nonlegumes revealed a tighter stomatal regulation of transpiration with lower time‐integrated stomatal conductance, which may represent a key physiological advantage of Fabaceae that could contribute decisively to their dominance across the Sahel and other tropical dryland ecosystems (Felker, ¹⁹⁸¹; Sprent & Gehlot, ²⁰¹⁰; Pellegrini et al., ²⁰¹⁶; Gei et al., ²⁰¹⁸). However, dryland legumes as a group exhibited higher leaf N concentrations (N_mass) but did not exhibit significantly higher average leaf N_area or δ¹³C‐WUE_i values than nonlegumes, even though higher leaf N_mass was strongly linked to higher δ¹³C‐WUE_i across the whole dataset (Table 3). This was an unexpected result, as tighter stomatal control of transpiration in N‐rich legumes (inferred from their high leaf δ¹⁸O and Δ¹⁸O values) would be expected to also lead to higher δ¹³C‐WUE_i. Moreover, higher leaf N_mass in legumes would be expected to enhance carboxylation capacity and thus lead to greater drawdown of CO₂ concentration at the sites of carbon fixation, thereby widening the c _a – c _c gap (i.e. lower c _c/c _a ratios) and further enhancing δ¹³C‐WUE_i (Wright et al., ^{2001, 2003}; Prentice et al., ^{2011, 2014}; Onoda et al., ²⁰¹⁷). However, it is important to note that average N_area values did not differ significantly between legumes and nonlegumes, because much higher average LMA in nonlegumes (i.e. thicker leaves with lower SLA) compensated for their lower leaf N_mass, thereby leading to converging and similar N_area and δ¹³C‐WUE_i values between these two plant functional groups. In other words, Sahelian legumes and nonlegumes achieved similar average N_area and δ¹³C‐WUE_i values through contrasting trait combinations: high SLA and leaf N concentrations in legumes vs lower N concentrations with lower SLA values (i.e. higher LMA) in nonlegumes.

Several other nonmutually exclusive hypotheses could further contribute to explain the unexpected lack of differences in average δ¹³C‐WUE_i values between legumes and nonlegumes: a substantial proportion of leaf N in Sahelian legumes may be invested in nonphotosynthetic proteins, such as N‐based osmolytes and/or defensive compounds against herbivores (Adams et al., ²⁰¹⁶); the thin, high‐SLA, nitrogen‐rich leaves of legumes may exhibit high mesophyll conductance to CO₂ (g _m) favoring high CO₂ concentration at the sites of carboxylation (c _c) (Seibt et al., ²⁰⁰⁸; Niinemets et al., ²⁰⁰⁹; Onoda et al., ²⁰¹⁷; Ma et al.,  ²⁰²¹); legumes could be more prone to photorespiration burdens under hot conditions if their tighter stomatal regulation of transpiration and lower time‐integrated stomatal conductance hamper evaporative leaf cooling aimed at alleviating overheating. A higher vulnerability to leaf overheating and photorespiration in legumes would constrain carbon assimilation capacity and impede any further decreases in c _i despite their high leaf N concentrations (Paillassa et al., ²⁰²⁰). All this would help to reconcile the lower stomatal conductance of legumes as a group (inferred from their higher leaf δ¹⁸O and Δ¹⁸O) with the unexpected lack of differences in c _i/c _a ratios between legumes and nonlegumes, along with the similar N_area between both functional groups (Warren & Adams, ²⁰⁰⁶). Nonetheless, it should be noted that the subset of deciduous legumes still had marginally higher mean leaf δ¹³C‐WUE_i value and higher N_area as a group compared with deciduous nonlegumes (Fig. S2), as previously reported at a global scale (Adams et al., ²⁰¹⁶). Achieving any given leaf N_area value through high SLA and N_mass (as in dryland legumes) could be more effective for enhancing carboxylation and photosynthetic nitrogen‐use efficiency than achieving it through lower SLA (i.e. higher LMA) with lower N_mass (as in nonlegumes), given that a larger proportion of leaf N is usually invested in nonphotosynthetic structural tissues and compounds in low SLA leaves (e.g. cell walls; Poorter & Evans, ¹⁹⁹⁸; Onoda et al., ²⁰¹⁷).

In conclusion, the Western Sahelian drylands harbor woody species with strikingly diverse leaf δ¹⁸O and δ¹³C values and water‐use strategies that appear to be similarly adaptive to cope with the dual challenge imposed by drought and nutrient‐poor soils (Wright et al., ²⁰⁰¹). Our findings provide strong experimental support for the least‐cost economic theory of photosynthesis in which water and nitrogen are mutually substitutable resources to achieve a given carbon assimilation gain in dryland ecosystems (Wright et al., ²⁰⁰³). Structural equation modeling revealed that foliar N_area is a pivotal trait linked to multiple other traits related to water use in Sahelian woody vegetation. In particular, higher leaf N_area is linked to both higher carboxylation capacity and tighter stomatal regulation of transpiration, which in turn is associated with higher WUE_i and a more conservative water‐use strategy in dryland trees. Legumes possess several key functional traits that may help them thrive under the harsh environmental conditions of the Sahel. Their efficient N‐acquiring ability enables them to achieve both high carboxylation capacity and tight stomatal control of transpiration and may also enhance their drought tolerance through improved osmotic adjustment and ability to extract water from drying soil. Finally, our study highlights the promise offered by leaf δ¹⁸O and δ¹³C measurements as useful functional traits for screening the water‐use strategies of multiple species in remote dryland locations where extensive gas exchange measurements are often not feasible.

Author contributions

GMR and FIP conceived the FUNCiTREE project and designed the general field campaign. JIQ designed the hypotheses to be tested with the help of IP and CA, and proposed new variables to be measured. IP led the statistical analyses with the help of FC and CA. Field data collection was led by GMR, CA, MD, JSD, HY and BK, and CA organized the laboratory analyses, and assembled and curated the database. JIQ, IP and CA wrote the first draft of the manuscript, and all authors contributed substantially to revisions. JIQ, IP and CA contributed equally to this work.

Supporting information

Fig. S1 Climograms for the studied regions in the Sahel.

Fig. S2 Differences in trait values between deciduous legume and deciduous nonlegume woody species.

Table S1 Summary statistics of 12 functional traits measured in 34 woody species in the Sahel (Mali and Senegal).

Table S2 Phylogenetic tree for the 34 target species and phylogenetic signals in measured traits.

Table S3 Pairwise relationships between all traits analyzed in legumes and deciduous tree species.

Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

The data that support the findings of this study are available from the corresponding author (cris@eeza.csic.es) or Graciela M. Rusch (graciela.rusch@nina.no) upon reasonable request. Trait data per species (mean and SE) are available in the Table S1 and number of replicates per species and site (n) are in Table 1. The data are deposited in the FUNCiTREE database (https://doi.org/10.15468/nye57x) at the Norwegian Institute for Nature Research (NINA).