The chemical composition of higher plant leaves and other foliar traits, such as leaf mass to area ratio (MA) and leaf longevity, generally co-vary with leaf photosynthetic rate, A (Reich, Walters & Ellsworth 1997; Wright et al. 2004). As photosynthesis is based on biochemical processes, scaling relationships are expected between photosynthetic capacity and tissue chemical composition (e.g. Field & Mooney 1986; Evans 1989; Domingues et al. 2005). As understanding of these important relationships improves, relevant advances in leaf- and canopy-trait representations are gradually being incorporated into fine- and global-scale photosynthesis models (Sellers et al. 1997; Sitch et al. 2003; Ollinger & Smith 2005; Kattge et al. 2009; Xu, Gertner & Scheller 2009).
Despite the importance of globally representative data sets to global models, the bulk of previously published photosynthetic studies have been performed on model species or in temperate ecosystems, with some globally important areas remaining under-represented (Wright, Reich & Westoby 2001; Reich, Wright & Lusk 2007; Kattge et al. 2009). This bias in field data potentially undermines the accuracy of modelling efforts that use leaf traits as a basis for prediction of photosynthesis. A ‘global spectrum’ of positively and anticorrelated leaf traits has been discussed (Wright et al. 2004); however, substantial trait variability is evident within particular regions (e.g. Güsewell 2004; Reich & Oleksyn 2004; Wright et al. 2005a), making it difficult at present to justify a universal set of scaling relationships that functions equally well for all terrestrial ecosystems (Reich et al. 1999; Wright et al. 2006; Kattge et al. 2009). Probably the most important of these relationships is a recognized strong and positive correlation between photosynthetic capacity and leaf N (Field & Mooney 1986). Photosynthetic capacity is often described in plants by the maximum rate of carboxylation, Vcmax, which is in turn determined by the amount and activity of the enzyme ribulose 1·5-bisphosphate carboxylase/oxygenase (RuBisCO) (Farquhar, von Caemmerer & Berry 1980). Examination of the above-mentioned relationship has shown that photosynthetic capacity varies considerably with leaf nutrient content, especially N and P, with some of the biggest differences evident when temperate and tropical systems are compared (Reich & Schoettle 1988; Niinemets 1999; Kull 2002; Meir et al. 2002, 2007; Kattge et al. 2009). Temperate ecosystems are often considered to be limited by N (Schulze et al. 1994). In contrast, several lines of evidence have suggested that many tropical forest soils have excess N (Högberg 1986; Martinelli et al. 1999; Okin et al. 2008), whereas the soils underlying many tropical forests and savannas are often highly weathered and are poor in total and plant available P (Vitousek & Sanford 1986; Lloyd et al. 2001; Hedin 2004; Reich & Oleksyn 2004; Lambers et al. 2008; Quesada et al. 2009a). Notwithstanding this general conclusion, however, examples to the contrary, of strong or partial N limitation in tropical forests exist for some savanna ecosystems (Lloyd et al. 2009), for forests growing on white sand soils such as Arenosols or Podzols (Martinelli et al. 1999; Mardegan et al. 2008), and for forests growing on the oldest and most infertile Ferralsol and Acrisol soil types in north-east and central Amazonia (Quesada et al. 2009b).
Temperate and boreal plants are thus expected to exhibit tight and steep Vcmax–N relationships, because N is generally limiting and its allocation to non-photosynthetic functions is expected to be minimized (Terashima et al. 2005). In contrast, in regions where N is relatively abundant but P availability is low, a weaker Vcmax–N relationship and a stronger Vcmax–P relationship might be anticipated. This has been observed, though not always as clearly as expected (Meir, Grace & Miranda 2001; Niinemets et al. 2001; Wright et al. 2004; Meir et al. 2007; Reich, Oleksyn & Wright 2009). Recently, Reich et al. (2009) presented evidence that high foliar P values are indeed related to a steeper slope of the relationship between saturating assimilation rate and leaf nitrogen content.
Several reasons have been proposed for the wide observed range in the proportion of woody versus herbaceous vegetation in tropical savannas, including the influence of rainfall, soil fertility, soil texture and moisture, and the local history of fire and grazing (e.g. Furley 1992; Lloyd et al. 2008). Although leaf-level properties do not necessarily explain differences in the distribution of vegetation types (Gifford & Evans 1981), two of these factors have provided the main focus for leaf gas exchange studies: fertility and rainfall, often with measurements taken over a transect spanning either or both environmental axes (e.g. Wright et al. 2001; Prior, Eamus & Bowman 2003; Midgley et al. 2004). There are, however, few leaf gas exchange measurements available for seasonally dry tropical biomes (Prado and de Moraes 1997; Franco 1998; Wright et al. 2001; Prior et al. 2003), especially for Africa (Mooney et al. 1983; Tuohy, Prior & Stewart 1991, Midgley et al. 2004, Meir et al. 2007). The few data that do exist show examples of lower area-based maximum photosynthetic rate at saturating irradiance (Asat-A) per unit N at drier sites (Wright et al. 2001, Midgley et al. 2004) and indicate that P is sometimes, but inconsistently, more strongly associated with Asat-A than N (Tuohy et al. 1991; Wright et al. 2001, 2006; Prior et al. 2003; Meir et al. 2007).
Reliable estimates of the relationship between photosynthetic capacity and leaf nutrient content are particularly important for parameterizing terrestrial biosphere models (e.g. Woodward, Smith & Emanuel 1995; Niinemets 1999; Sitch et al. 2003; Reich et al. 2006; Raddatz et al. 2007). Extending earlier work that examined the global co-ordination of leaf traits (Reich et al. 1997; Wright et al. 2004), Reich et al. (2007) highlighted the potential to estimate Asat globally from phylogeny, plant growth form and leaf phenology, coupled with climatic data and easily measured leaf traits, such as MA and N. Kattge et al. (2009) developed this approach for terrestrial biosphere models, focusing on the need to specify biochemical parameters such as Vcmax (Farquhar et al. 1980). A lower Vcmax per unit nitrogen has been reported for tropical forest leaves (Carswell et al. 2000; Domingues, Martinelli & Ehleringer 2007; Meir et al. 2007), and a potentially large negative impact of phosphorus-deficient tropical soils such as Ferralsols or Acrisols on the Vcmax–N relationship highlighted (Kattge et al. 2009). Although a lack of data hampers our understanding of how N and P constrain photosynthesis in tropical woody ecosystems, the data assimilation and modelling analysis approach of Kattge et al. (2009) yielded a much lower gross primary productivity for tropical forests compared with those obtained from earlier, more indirect, global estimates of photosynthetic parameters (e.g. Beerling & Quick 1995). This, in conjunction with the recent findings of Paoli & Curran (2007) and Quesada et al. (2009c) that soil phosphorus availability may be the key modulator of the above-ground net primary productivity of tropical forests in Borneo and Amazonia, respectively, underscores a strong need for more information on the role of phosphorus in modulating rates of carbon acquisition for tropical tree species, especially for ecosystems that experience seasonal water stress.
Here we present data on MA (or specific leaf area, S = MA−1, m2 g−1), leaf N and P concentrations and leaf biochemical photosynthetic capacity (Vcmax and the electron transport capacity, Jmax) for woody vegetation along a 900 km north–south transect in West Africa, from Mali to Ghana, a globally important vegetation transition, yet one still poorly characterized. Although the biochemical transformations governed by Vcmax and Jmax both require nitrogen and phosphorus, we expected Vcmax to be more strongly controlled by the amount of nitrogen present, because of the high N requirement by RuBisCO (Evans 1989). On the other hand we expected a stronger phosphorus constraint for Jmax because of the many transformations of phosphorus-rich molecules (ATP, NADP and sugar-phosphates from the Calvin cycle) that occur for the regeneration of ribulose-1, 5-bisphosphate, RuBP (Farquhar et al. 1980; Woodrow & Berry 1988). Phosphorus is also an important component of phospholipid membranes, which might act as temporary storage for this element, changing membrane characteristics as phosphorus deficiency takes place (Tjellström et al. 2008). We examined vegetation types ranging from open savanna to different forms of woodland, and covering a large range in annual precipitation, with the (strong) dry season (<50 mm rainfall month−1) ranging from 2 to 10 months (data from Climatic Research Unit, University of East Anglia, UK; New, Hulme & Jones 1999). We examined the following questions: (1) are data from this transect consistent with existing global leaf trait data sets?; (2) how strong are the Vcmax and Jmax versus N relationships on both a dry-weight and an area basis?; (3) does P have a stronger influence over Vcmax and/or Jmax than N?; (4) can we devise a general model to interpret and specify potential constraints on Vcmax and Jmax by either N or P?