The experiment was conducted at the Experimental Farm of Tafilalet (SEMVAT) in southern Morocco. In 1997–1998, the annual rainfall averaged 123 mm at the experimental site. The mean annual temperature was 18 °C, with a minimum of 8 °C (January) and maximum of 30 °C (July). The soil of the experimental fields had a clay loamy texture with 29.4% clay, 34.1% loam, 36.5% sand, a bulk density of $1.3479\mathrm{kg}{\mathrm{l}}^{-1}$, a porosity of 49.2%, a field capacity of 27% and a permanent wilting point of 12%. The soil contained 0.06% total N, 0.88% organic matter and a pH of 8.5 (H₂O) in the 0–20-cm horizon.

Plots of 3×3 m were delimited. Before sowing, triple super phosphate was applied at a rate of $124\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ and potassium sulphate at a rate of $84\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$. The plots were seeded with wheat (Triticum durum var. Karim) at the rate of $160\mathrm{kg}{\mathrm{ha}}^{-1}$ on 22 December 1997. Seeding depth was 3 cm. The plants were thinned to have 20 cm between them.

The experimental design was a split-plot, replicated three times. The main plot treatment consisted of four frequency irrigations: I1 plots were irrigated every 10 days; I2 plots were irrigated every two weeks; I3 plots were irrigated every three weeks; I4 plots were irrigated every four weeks. Each dose of irrigation was 70 mm. Each plot contained three sub-plots which received three fertilization treatments: T1 (42 N+42 ¹⁵N+84 N): received $42\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ of ammonium sulphate before seedling, $42\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ of ammonium sulphate enriched with 9.764% ¹⁵N excess at tillering and $84\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ ammonium sulphate at flowering; T2 (42 ¹⁵N+42 N+42 N): received $42\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ of ammonium sulphate labelled with 9.764 at% ¹⁵N excess at seedling, $42\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ at tillering and $42\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ at flowering; T3 (¹⁵R+42 N+42 N): received $4800\mathrm{kg}{\mathrm{ha}}^{-1}$ of wheat residue labelled with 1.504 at% ¹⁵N excess (¹⁵R) and $42\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ of ammonium sulphate before seedling and $42\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ ammonium sulphate at flowering.

The plots were harvested on 3 June 1998. The grain was separated from the rest of the plant. The samples were oven dried at 75 °C for 48 h and then weighed. Plant material (straw, grain and roots) was analysed for total nitrogen and ¹⁵N excess using an automatic N analyser coupled to a SIRA 9 (ANA-SIRA) mass spectrometer [21]. The percentage of ¹⁵N recovered from the fertilizer in the plant samples was calculated according to Zapata [22]:

The percentage of N in the plant derived from the soil and more generally from the unlabelled source of nutrient (%N_dfs) is the difference:

Data were statistically analysed with SPSS (Statistical Package for Social Sciences). Differences between treatments were analysed using ANOVA, followed by least significant difference (LSD) at the 0.05 probability level and T test.

For all irrigation treatments, the N fertilization with $168\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ was not significantly different from the fertilization with $126\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$. For the three irrigation treatments, I1, I2, I3, the results showed that wheat residues plus $84\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ had similar effects on wheat yield than application of $168\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ and $126\mathrm{kg}\mathrm{N}{\mathrm{ha}}^{-1}$ (Table 1). Similar results were found by Vanlauwe et al. [11] with maize, who showed a 50% economy of fertilizer by combining residues with mineral fertilizer in southern Benin. The total N in the grains followed a similar trend as the dry matter content (Table 2).

For treatment I4, a significant difference (P<0.05) was found between the grain yield of the treatment with residues (T3) and without residues (T1). Similarly, Purvis [23] indicated that straw incorporation inhibit wheat growth. This inhibition depended on the quantity and the distribution of rain during the year. According to the author, the depressive effect of residues of crops disappears with increase of humidity. Similarly, Badaruddin et al. [20] showed that straw decomposition was more important under increased soil moisture. The production of dry matter decreased with a lower irrigation frequency. This decrease was significant (P<0.05) between I1 and I4 (Table 1). The interactive effect between irrigation treatments and fertilization was not significant for the straw and grain yield.

For I1, I2, and I3, the plant N is distributed for about 2/3 in the grain and 1/3 in the straw; only about 4% is situated in the harvested roots. The same results were found by Van Cleemput et al. [24] on winter wheat, and on sunflower plants by Atta et Van Cleemput [25] and Corbeels et al. [3]. But, under water-stress conditions, N has a distribution of about 1/2 in the grain and 1/2 in the straw for all treatments. Nitrogen in the roots was higher (13% for treatment with residues). Under these conditions, no percolating moisture may prevent that the applied N is translocated to the root. Similar results were found by Corbeels [26].

The N recovered by the plant is derived from both the soil and the fertilizer. With the use of ¹⁵N labelled fertilizer, the amount of N derived from the fertilizer (N_dff) and the amount of N derived from the soil (N_dfs) were determined (Table 3).

For grain and straw of wheat, N_dff were low while most of the N was derived from the soil for all irrigation and fertilization treatments. Corbeels et al. [27] also reported that soil N was more important than fertiliser N in wheat N uptake.

The proportion of N derived from ¹⁵N-labelled fertilizer (N_dff) in grain and straw were significantly affected (P<0.05) by N application time (Table 3). More N in the plant was derived from fertilizer when applied late in the growing season than applied early in the season. With optimal moisture conditions, the percentage of N in the plant derived from the labelled fertilizer applied at tillering was about 41%, at seeding it was 33% and, for the labelled residues, it was 24%. Probably the early-season application of N fertilizer could also have allowed more time for substitution and resulted in greater isotopic dilution with soil N, which would decrease the N_dff values [27]. The microbial needs could already be satisfied with native mineral soil N and fertilizer N applied at seeding, so that fertilizer N applied at tillering could remain more available to plant [28,29].

The %N_dff values showed no significant (P>0.05) difference between the different plant parts. Similar results were observed by Corbeels et al. [27]. In contrast with these results, other investigators observed significant differences in %N_dff values for the different plant parts, especially for late N dressings. Grain may contain a greater proportion of the fertilizer N than the non-grain plant components, when applied to heading and flowering [30,31].

There was a significant difference (P<0.05) in recovery of fertilizer N at different times of application. For I1, the %N recovery was 63% when the fertilizer N was applied at tillering and 28% for the treatment receiving fertilizers at seedling (Table 4). In I4, the % recovery was 49 and 28% for T1 and T2, respectively. The proportion of fertilizer N applied to seeding was probably more subjected to N loss prior to plant absorption. Root development at tillering stage could enable more efficient N absorption from fertilizer applied at this time. Similarly, Tran and Tremblay [32] showed that the maximum proportion of N fertilizer recovered by wheat was higher for the application at booting (62.1 to 68.4%) than for the application at seeding (37.8 to 45.7%). Hamid and Ahmad [33] found that the amount of N fertilizer used by wheat was lower for application at seedling (33.6%) than at tillering stage (51.5%).

For the treatment with labelled residues and mineral fertilizer, the recovery in the plant was 8% for I1 and 5% for I4.

For I1, the labelled fertilizer was mainly recovered by the grain with all fertilization treatments, followed by the straw, while only a minor part of fertilizer N was recovered in the harvested roots (between 1 and 5%). These low ¹⁵N recoveries in roots were also reported by others [26,33,34]. For an irrigation scheme of 30-d periodicity, the grain and straw recovered nearly equal amounts of fertilizer, although the amount in the roots was generally low.

For I1, I2, and I3, the amount of labelled N in the grain and straw was significantly higher when labelled ammonium sulphate was added at tillering than when ¹⁵N was applied at seedling. This can be explained by some N loss. Thus, on irrigated surfaces, the efficiency of the fertilizer N by the plant depends on the duration of the N application. This result is similar to the findings of Van Cleemput and Hera [35]. The data on uptake of the applied labelled N show that the total amount of ¹⁵N taken by the plant significantly differed depending on whether the treatment was made with or without residues, for all irrigation frequencies.

The results suggested that the %N_dff increased when the percent of recovery decreased. It means that in situation of water stress, roots take the N they need more in the fertilizer than in the soil reserve in comparison with the case of no water stress. In fact, the concentration of N in the soil solution is higher near the fertilizer than in the mean soil.

The non-recovered fertilizer N could be found either as residual N in the soil plus non-harvested roots, or leached out of the profile or denitrified [35,36]. In this context, the fact that the 0–15 cm soil layer in our experiment was dried and re-wetted might have enhanced the denitrification potential [37].