Estimates of biological nitrogen fixation by Pterocarpus lucens in a semi arid natural forest park in Senegal using 15 N natural abundance method

Nitrogen (N 2 ) fixation by Pterocarpus lucens in a natural semi arid ecosystem, in Ferlo, Senegal was estimated using 15 N natural abundance ( ∂ 15 N) procedure. Other non-fixing trees accompanying P. lucens in the same area were also investigated as control. Results showed an important variation of ∂ 15 N in leaves between the nitrogen-fixing tree ( P. lucens ) and reference plants, whereas no significant differences were recorded in amount of nitrogen (%). The relative ∂ 15 N values (‰) were higher in non-fixing plants than in fixing plants considered. Calculations of %Ndfa gave rates ranging between 26% and 49%. The values of %Ndfa depend largely on soil and the reference plants. The contribution of nitrogen derived from fixation in leaves reached 28.9 kg.N/ha and 10.8 kg.N/ha in ferruginous and in sandy soil respectively.


INTRODUCTION
The Leguminosae is one of the largest and most widely distributed families of flowering plants. Legumes habitats range from rain forests to arid zones throughout the world. These trees are well represented in Savannah and arid regions, where they grow in soils of low fertility unsuitable for most tree species. This dominance of leguminous trees reflects an ecological significance due to their capacity for biological nitrogen fixation. Through symbiosis with root nodule bacteria, legumes play a major role in the functioning of many ecosystems, from natural woodlands to agroforestry systems. Several reports have demonstrated the attractive role of nitrogen-fixing trees as sources of nitrogen and organic matter, and their potential in soil fertility improvement and sustainability (Danso et al., 1992;Sanginga, 1990).
Quantification of nitrogen from biological nitrogen *Corresponding author; e-mail: Samba.sylla@ird.sn, tel: +221-8 49 33 22, fax: +221-8 32 16 75 fixation is a question of considerable relevance. This led to a demand for techniques to assess the contribution of biological nitrogen fixation to trees and shrubs ecosystems. Until now, most of the work undertaken on the evaluation of nitrogen (N 2 ) fixation by tree species has been performed in the framework of short-term experiments where seedlings were grown in containers (Ndoye et al., 1995;Sutherland et al., 2000;Sylla et al., 1998) or in field (Gueye and Ndoye, 2000). The data obtained from such experiments provide an estimation of the N 2 -fixing potential in juvenile stage. Moreover, many environmental factors do interact with N 2 fixation of the trees from the juvenile to the mature stage in natural ecosystems. The N 2 fixation capacity of legume trees is often use to explain their ability to grow better than nonlegume trees on nitrogen-depleted soils and to restore the fertility of such soils. Very few studies attempting to evaluate that potential in natural conditions are actually available, regardless of the estimation method used. The most promising technique, to quantify trees N 2 fixation contribution in natural ecosystems, is the " 15 N natural abundance". It reflects the biological nitrogen fixation for long period for woody perennial plants in natural ecosystem (Boddey et al., 2000).  (Mariotti , 1983;Yoneyama et al., 1987).
Plants that depend on soil nitrogen have positive ∂ 15 N values, which is rather close to those of soil nitrogen. In temperate and cooler areas, however, some plants show very negative ∂ 15 N values (Domenach et al., 1989;Virginia et al., 1989). In legumes, large fractions of plant nitrogen are derived from N 2 , arising from biological nitrogen fixation. Therefore, the ∂ 15 N values of plants grown in soils where nitrogen of positive ∂ 15 N values is available, is diluted when these plants fix nitrogen using biological nitrogen fixation method. Significant success has been obtained when using the 15 N natural abundance in several studies with N 2 -fixing trees including legumes (Hamilton et al., 1993;Ladha et al., 1993;Muofhe and Dakora, 1999;Peoples et al., 1997;Shearer and Kohl, 1986;Yoneyama et al., 1990Yoneyama et al., , 1993 and actinorhizal trees (Domenach et al., 1989;Kurdali et al., 1990;Mariotti et al.1992).
Pterocarpus lucens is a small tree species of the family of Leguminosae, subfamily Fabacea. In Senegal, populations of P. lucens occupy a dominant part of ecosystems in the natural semi arid lowland of Ferlo. In this area, population of Pterocarpus has been studied in terms of structural position and density (Fredericksen and Lawesson, 1992). We have earlier reported that P. lucens is often spontaneously nodulated by rhizobia in its native environment in Senegal (Sylla, 1996). However the contribution of nitrogen fixation towards the nitrogen requirements of this species in situ has never been determined. In the present study, we investigated the variation in 15 N natural abundance between perennial woody plants and assessed the contribution of symbiotic nitrogen fixation by P. lucens in both sandy and ferruginous soil sites in Ferlo, Senegal. Sylla et al. 51

Sites of Investigation
The amount of nitrogen fixed by P. lucens in a forest park, Ferlo located in the sahelian region (15°-16°N; 14°-15°W) in Senegal was estimated. The annual rainfall means range is 400 mm and the temperatures vary slightly around 27°C. In this natural ecosystem, Fredericksen and Lawesson (1992) inventoried all woody plants higher than 1 m with a diameter ≥ 10 cm. A description of tree species, height, density and cover were reported, while phytogeographical affinities and morphopedological types were established. This study concerns two contrasted sites characterized by the vegetation type 4 and type 5 (Fredericksen and Lawesson, 1992). Site 1 is characterized by red ferruginous tropical soils, commonly called laterite. The vegetation type in this site includes a very high density of P. lucens with a significant presence of Anogeissus leiocarpus, Balanites aegyptiaca, Combretum glutinosum, and C. aculeatum. In site 2, the soil is sandy and the vegetation is dominated by P. lucens and B. aegyptiaca.

Field sampling
Sampling of leaves was done in November 1999 when Pterocarpus trees and most of the woody species are still bearing a maximum of foliage, and fructification is just beginning. Occurrence of root nodules has been previously checked in the surrounding soil at 15 cm depth for P. lucens trees. Because of the big size of trees and difficulty in whole plant sampling, the idea of sampling representative and more accessible plant organs were adopted. Sampling was limited to foliage and only young fully expanded leaves were collected. Leaves were obtained from the top of the trees in the upper canopy. Reference species were selected among the non-legume woody species occurring in the same area with P. lucens. These non-legume plants were A. leiocarpus, C. acculeatum, and C. glutinosum in site 1 and B. aegyptiaca in site 2. Ten individual samples from each tree species of both sites were randomly collected. The leaves from the same individual tree were pooled to form a single sample.

Sample analysis
The leaf samples were oven-dried at 60° to 70°C for 48h, ground into fine powder and packed. Analysis of powdered samples was performed in stable isotope facility at the Department of Agronomy and Range Science, University of California Davis, USA. 15 N / 14 N isotope ratio and total nitrogen content were determined using an automated nitrogen analyzer coupled to a mass spectrometer (EA-IRMS, PDZ Europe LTD, CW1 6ZA, UK). Results are reported in percent nitrogen in leaf (N%) and ∂ 15 N ‰. The % Ndfa (%N derived from atmosphere) was calculated according to Shearer and Khol (1986) as follows: The fractionation factor ß equivalent to 0.8% represents the ∂ 15 N value of N 2 -fixing plants (P. lucens) when their only source of nitrogen is atmospheric N 2 . The value of ß was determined in a separate experiment according to Ledgard (1989) and Unkovich et al. (1994). Seedlings of P. lucens were grown in hydroponical condition with nitrogen free nutrient solution (Vincent, 1970). Plants were immediately inoculated (10ml/pot) with a liquid inoculum (containing 10 9 cells/ml) of Rhizobium ORS 206, an effective slow growing strain isolated from P. erinaceus in Senegal (Sylla, 1996 where the standard is atmospheric N2 . The %Ndfa of P. lucens was calculated using the ‰ ∂ 15 N mean value of each non-fixing tree species collected at the sites. The assumptions 1, 2, 3 correspond respectively to the calculations of %Ndfa of P. lucens in ferruginous soil with ‰ ∂ 15 N values of C. glutinosum, A. leiocarpus, C. aculeatum, respectively. Assumption 4 corresponds to the calculations of %Ndfa of P. lucens in ferruginous soil using the average ‰ ∂ 15 N of assumption 1 to 3. This process allowed integrating potential heterogeneity on 15 N uptake on the reference trees (IAEA Tec-doc, 1983). In sandy soil, the %Ndfa of P. lucens, was calculated, with the ‰ ∂ 15 N mean value of B. aegyptiaca which was the only non fixing woody plant accompanying P. lucens in the site.

Plant biomass estimates
Plant leaf biomass (P) was assessed from diameter at breast height (dbh) according to the allometric equation where P = k.dbh a (Brown, 1997;Lescure et al., 1983;Puig et al., 1990).
With k = 0.00873 and a = 2.1360 for leaf biomass and k = 0.05635 and a = 2.7248 for total biomass.

Statistical analysis
Data were analyzed statistically. Separate one-way analysis of variance was performed. Means and standard deviation were calculated at the P ≤ 0.05 levels.

RESULTS
The %N and ‰ ∂ 15 N values of P. lucens leaves and that of different reference woody plants used are shown in Table 1. The %N in leaves does not seem to vary between the different woody species occurring in the two soil types. The ‰ ∂ 15 N in the leaves were significantly different between the woody non-fixing plants. In the ferruginous soil and sandy soil sites, the ‰ ∂ 15 N values of P. lucens were 4.06±0.57 and 5.54±0.26 respectively. These values are lower than those of neighboring nonfixing tree species (average value of 6.65±1.33 and 7.24±0.37, respectively). This indicates occurrence of N 2 fixation by P. lucens in both sites. P. lucens showed significant differences in ∂ 15 N between ferruginous soil and sandy soil. The %Ndfa values ranged from 35 to 49% for ferruginous soil under the assumptions 1, 2, 3 and 4 (Table 1). These values are higher compared to those for the sandy soil, 26% using assumption 4. In the ferruginous soil %Ndfa values were obtained by taking into account each reference plant ‰ ∂ 15 N (assumptions 1, 2 and 3) and the average of these values (assumption 4). In the sandy soil, only one species B. aegytiaca was used as reference plant. The standard deviation of the %Ndfa ranged from 6 to 11% for P. lucens (Table 1). These values are comparable to the 5-10% described in previous studies (Shearer and Kohl, 1986).
Leaf biomass per plant and leaf biomass per hectare (Table 2) were estimated according to the allometric equation of Lescure et al. (1983). The leaf biomass per hectare found for P. lucens was 3.000 t/ha and 2.286 t/ha in ferruginous and sandy soils, respectively. P. lucens roots were checked for nodulation, and root nodules were observed only on those growing in the ferruginous soil; no nodule was recorded in P. lucens growing in sandy soil at 15 cm depth.  (Lawesson, 1990;Fredericksen and Lawesson, 1992). **Leaf biomass values were estimated according to the allometric equation of Lescure et al. (1983) and Puig et al. (1990). Sylla et al. 53 The average amount of nitrogen accumulated in leaves was 0.054 kg N/plant and 0.043 kg N/plant for P. lucens in the ferruginous and sandy soils, respectively (Table 3). Taking together all the P. lucens plants in this study, the amount of nitrogen contained in leaves was 66.65 kg N/ha and 49.66 kg N/ha in the ferruginous and sandy soils, respectively (Table 3). Significant differences in Ndfa were recorded when different species were used as reference trees (Table 3). In the ferruginous soil, the Ndfa value varies from 23 to 32 kg N/ha, depending on the reference plant. In the sandy soil, where only one reference plant was used, the Ndfa was 13 kg.

DISCUSSIONS
The study was undertaken in a natural park in Senegal, where Pterocarpus exploitation is prohibited. No data are available in literature on the values of leaf biomass of adult P. lucens trees in their natural habit. To avoid tree destruction, our method of estimation of tree leaves biomass was similar to those of Lescure et al. (1983) and Puig et al. (1990). Other authors have also used this method to estimate total leaf and or total above ground biomass of trees in tropical forest ecosystems (Roggy et al., 1999;Scatena et al., 1993). The amount of leaf biomass in their studies ranges from 3.7 to 6.4 t/ha. In our study, which was conducted in a tropical semi-arid ecosystem, values of leaf biomass were about 3.0 t/ha. Due to the non-variation of the average values of diameter of woody vegetation on both sites of this study (Lawesson, 1990), results obtained by using allometric equation showed uniform values of leaf biomass per plant within each site. Therefore, the leaf biomass per hectare changes with the plant density per hectare. Many authors recommend that if possible entire plants or at least whole shoots should be sampled to assess overall 15 N abundance for estimating biological nitrogen fixation. With trees, sampling entire plant is usually extremely difficult. And from the limited data available, variation between tree plant organs may often be less than that observed in annual legume species. Ladha et al. (1993) found that there were no significant differences in 15 N abundance of trunk, branch and leaves taken from nodulated Gliricidia sepium or from Senna spectabilis. Data indicates that in many nitrogenfixing trees (both greenhouse and in-field studies), %Ndfa did not differ significantly between different plant parts and the whole tree Gueye et al., 1997;Sylla et al., 1998)  measured the %Ndfa in leaves to estimate the contribution of P. lucens adult trees in their natural ecosystem.
The reference plant is unquestionably the 'Achilles' heel' of the methodology within the context of isotopic non-uniformity. According to Danso et al. (1992), the socalled 'selection criteria' do not permit identification of a single valid reference plant. We adopted the strategy which used several reference plants as advocated by a number of authors (Bremer et al., 1983;Boddey et al., 1995Boddey et al., , 2000. Root nodules occur more often in ferruginous soil conditions than in sandy soil conditions in P. lucens. Moreira et al. (1992Moreira et al. ( , 1994 mentioned similar observation in the frequency of leguminous tree root nodules between the oxisols and the spodsols in Brazilian primary rain forest. Our observation is also comparable to the N 2 fixation variability of some nitrogen-fixing tree species reported among habitats in Sonorian desert and Tanzania (Högberg, 1986;Shearer et al., 1983). Johnson and Mayeux (1990) postulated that, in a number of sites, nodules cannot be detected in some nitrogen-fixing tree roots even when compatible rhizobia are present in the soil because they decomposed too rapidly, are too small to be detected, or are located at 5-10 m depths. In the case of Pterocarpus trees in Ferlo (Senegal), these hypotheses require further investigation.
The fractionation factor, ß for P. lucens was 0.8‰. This value appear to range between -2.0 and + 1.0 ‰, for most tree legume species (Boddey et al., 2000). Indeed estimates may vary between and within plant species.  (Ledgard, 1989;Virginia et al., 1989), and the forms of nitrogen available to plants (Yoneyama et al., 1990). We did not measure the exact rooting depth of the different species. The increase in ∂ 15 N in the sandy soil could indicate that there was a considerable loss of nitrogen from the soil.
Among the plants investigated, B. aegyptiaca in sandy soil and C. aculeatum in ferruginous soil have the highest ∂ 15 N values (7.24 ‰ and 7.1 ‰) in their leaves. P. lucens (the fixing plant), possessed the lowest values of ∂ 15 N in leaves (4.06 ‰, in ferruginous soil and 5.54 ‰ in sandy soil). This suggest that P. lucens is less dependant on soil nitrogen nutrient than the nonleguminous trees, due to the potentialities of this tree species to fix nitrogen in symbiosis with rhizobia. In natural condition, as in the semi arid zone of Ferlo, the calculation of nitrogen derived from atmosphere in P. lucens tree leaves indicate that Ndfa may represent 35-49 % of total nitrogen accumulated. Using the soil 15 N labeling method in a greenhouse, we found Ndfa of 35% in the whole plant of P. erinaceus and P. lucens at the juvenile stage, in earlier work (Sylla et al., 1998). As previously reported by Högberg (1990) and Shearer and Kohl (1986), the %Ndfa values calculated depends on the reference plant. The estimated %Ndfa for P. lucens is 44.23% in ferruginous soil and 26.32% in sandy soil. Schulze et al. (1991) estimated %Ndfa ranging from 10 to 30% in woody Mimosaceae species along an aridity gradient in Namibia. The % Ndfa obtained for P. lucens in ferruginous soil exceeded most of what is reported for other N 2 -fixing species occurring in tropical regions. Shearer et al. (1983) reported that N 2 fixation might contribute 43-61% to the nitrogen content for Prosopis under drastic conditions of very low mineral fertility. In Tanzania 44% of the nitrogen content of potentially N 2fixing species originated from N 2 fixation (Högberg, 1986). Coté and Camire (1984) estimated fixed-N 2 to be 53 kg N/ha in both mixed and pure stands of alder grown in Canada. Under field conditions, Sougoufara et al. (1990) estimated fixed nitrogen to be 40-60 kg N/ha for Casuarina equisetifolia. Our results compare very well with these reports. Roggy et al. (1999), however, found very low estimates (7 kg ha-1) for tree species in a rain forest in French Guyana.