Introduction

Intercropping (IC)—the simultaneous growing of two or more species in the same field for a significant period of their growth but without necessarily being sown and harvested together (Willey 1979)—is known to increase yield (Y) and grain protein concentration (GPC) in spring-sown crops compared to sole crops (SC), particularly in low N input systems (Willey 1979; Ofori and Stern 1987; Vandermeer 1989; Hauggaard-Nielsen et al. 2003). For these reasons, there has been renewed interest in IC in the last decade (Anil et al. 1998; Malézieux et al. 2008).

In our previous companion paper (Bedoussac and Justes 2009) we demonstrated that: (1) IC of winter crops such as durum wheat and winter pea can increase dry weight, yield and N acquisition in low N input systems (<120 kg N ha−1 available); and (2) the efficiency of IC to improve yield and wheat grain protein concentration depends on N availability during early growth. More precisely, we showed that IC permits the use of available resources more efficiently than the corresponding sole crops because the two intercropped species do not compete for exactly the same resource niche and thereby tend to use resources—notably N—in a complementary way. This has also been shown for grain legume–cereal spring-sown intercrops (e.g. Hauggaard-Nielsen et al. 2001a, b). Our previous paper dealing with durum wheat–winter pea intercrops (Bedoussac and Justes 2009) indicated that durum wheat has a higher interspecific competitive ability at high soil N levels than winter pea. This result was in good agreement with data observed for various types of intercrops, such as spring barley–pea (Hauggaard-Nielsen and Jensen 2001) or spring wheat–pea (Ghaley et al. 2005). Moreover, as already observed by Ofori and Stern (1987) and Jensen (1996) for spring-sown intercrops, we demonstrated that winter intercrops such as durum wheat–winter pea intercrop also allowed for more N to be taken up by the durum wheat plant, more N to be remobilized into its grain, and hence wheat grain protein concentration to increase in the intercrop in comparison to the sole crop. Indeed, the cereal produced less dry weight per square metre when intercropped but has access to an almost similar amount of soil inorganic N in both the intercrop and the sole crop. This is due to the increase in the rate of the symbiotic N2 fixation rate in the grain intercrop species (Bedoussac and Justes 2009) as already described by Crozat et al. (1994) on spring-sown intercrops. Such results are of particular interest in low N input cropping systems and in organic farming, where soil mineral N is often limiting for cereals with a high N demand, such as durum wheat. It is also interesting for conventional cropping systems, in order to reduce the high levels of N fertilizer generally applied to fulfil durum wheat N requirements and to obtain maximum yield and grain protein content (Garrido-Lestache et al. 2004).

Species temporal interactions are complex, varying with, for example, nutrient availability or other environmental factors (Connolly et al. 1990). Thus, intercrop competition studies cannot base their conclusions on data from just one single harvest of crops at maturity and grown at one level of N, as, for example, in the study of Dhima et al. (2007). In particular, intercrops are known to be more efficient in intercepting light (Berntsen et al. 2004; Jahansooz et al. 2007) compared to sole crops because of complementary use of space when crops differ in their aerial architecture (e.g. tall vs short crops) and of time when crops have life cycles that differ in timing (e.g. early vs late maturing; Trenbath 1986; Tsubo et al. 2001). Competitiveness of a given species for solar radiation absorption and consequently its dry weight (DW) and yield depends on its green leaf area index (GLAI), foliar architecture, height and timing relative to those of its companion crop (Fukai 1993; Midmore 1993).

While many references are available on cereal–legume spring-sown intercrops dynamics (e.g. Andersen et al. 2004; Ghaley et al. 2005; Corre-Hellou et al. 2006; Jensen 1996) there is no specific reference on wheat–grain legume winter sown intercrops. Now winter sowing is more adapted to the conditions of southern Europe, particularly to avoid water stress during wheat ripening. Moreover, the winter crops growth period is longer than that of spring crops and hence species dynamics and interactions between crops are certainly different throughout crop growth. The aim of this paper was therefore to analyse the dynamics of growth and the resource acquisition throughout the crop cycle. This will allow us to explain the final outcome of a durum wheat–winter pea intercrop in terms of proportion of species, total grain yield and complementary use of N by components. This analysis should provide the knowledge required to propose further optimal strategies in order to improve the management of durum wheat–winter pea cropping sequences. This paper focusses on species dynamics under different N availabilities modified by N fertilization (quantity and splitting of doses) in a durum wheat–winter pea intercrop compared to their respective sole crops. This was achieved through field experiments conducted in 2005–2006 and 2006–2007 with different soil N availabilities in a substitutive design. Throughout the intercrop growth, we analysed: (1) species aerial biomass dynamics; (2) canopy architecture dynamics (GLAI and plant height); (3) N and light use; and (4) the potential advantages for biomass, N acquisition and wheat nitrogen nutrition status.

Materials and methods

Site and soil

The experiment was carried out in the experimental fields of the Institut National de la Recherche Agronomique station in Auzeville (SW France, 43°31′N, 1°30′E) in 2005–2006 (Exp.I) and 2006–2007 (Exp.II). The 25-year mean annual rainfall in Auzeville is 650 mm and the mean annual air temperature is 13.7°C with a maximum monthly mean daily temperature of 21.9°C in August and a minimum monthly mean daily temperature of 6.0°C in January. The rainfall during the growing seasons was 361 mm and 468 mm for Exp.I and Exp.II, respectively, while the 25-year mean was 489 mm for the same period (November–July). Exp.I was characterised by a cold winter and a dry warm spring, whereas Exp.II was characterised by a warm and dry winter and a rather wet spring. Thus, the two years were quite different in terms of climatic conditions, indicating a good level of robustness in the dynamic analysis of intercrops.

Exp.I was carried out on a loamy soil (24% clay, 29% silt and 47% sand) with an available water capacity of 223 mm (0–150 cm). Exp.II was conducted on a clay loam soil (30% clay, 38% silt and 32% sand) with an available water capacity of 207 mm (0–150 cm). For both experiments, phosphorus, potassium and cation exchange capacity values were not limiting (for more details, see Bedoussac and Justes 2009). The four previous crops on the experimental sites were durum wheat (Triticum turgidum), sunflower (Helianthus annuus), durum wheat and sorghum (Sorghum bicolor) for Exp.I, and sunflower, durum wheat, sorghum and sunflower for Exp.II.

Experimental design

Durum wheat (W; Triticum turgidum L, cv. Nefer) and winter pea (P; Pisum sativum L., cv. Lucy) were grown as: (1) sole crop (SC) sown at the recommended density (336 and 72 seeds m−2 for durum wheat and winter pea, respectively); and (2) intercrop (IC) where each species was sown at half of the sole crop densities in alternate rows.

In both experiments, different fertilizer-N sub-treatments were evaluated on wheat sole crop and intercrop while pea sole crop was grown without any N application assuming that N2 fixation allows pea to fulfil its N requirements for optimal growth. In Exp.I, three N treatments were compared: (1) no fertilizer-N (N0); (2) low N fertilization (N100) split into two applications of 50 kg N ha−1 at ‘1 cm ear’ (E1cm, Zadoks 30; Zadoks et al. 1974) and ‘flag leaf visible’ (FLV, Zadoks 37); and (3) moderate N fertilization (N180) split into three applications of 30 kg N ha−1 at wheat tillering (Zadoks 23), 100 kg N ha−1 at E1cm and 50 kg N ha−1 at FLV. In Exp.II we carried out four N treatments: (1) no fertilizer-N (N0); (2) one application of 60 kg N ha−1 at FLV (N60) to increase grain protein; (3) one application of 80 kg N ha−1 at E1cm (N80) to increase yield; and (4) a moderate N fertilization (N140) split into two applications of 80 kg N ha−1 at E1cm and 60 kg N ha−1 at FLV. In Exp.II, the previous crop was a rain-fed sunflower grown with four levels of fertilizer N: 50, 150, 0 and 100 kg N ha−1 for N0, N60, N80 and N140, respectively, which led to contrasting dynamics of N availability. As a consequence, the N60 treatment was more than the simple effect of a late N supply due to the previous treatment on sunflower, so we chose to name it N60+. Initial mineral N, early mineral N available and mineral N available throughout crop growth for the two experiments are described in Table 1 (for details, see Bedoussac and Justes 2009).

Table 1 Mineral N (kg N ha−1) and plant densities (plants m−2) for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1). SC Sole crop, IC intercrop (for details, see Bedoussac and Justes 2009)
Full size table

The experimental layout for both experiments was a randomised split-plot design with N application as main plots and crops as subplots, with five replicates in Exp.I (four replicates for wheat sole crop in N0 and intercrop in N180), and three replicates in Exp.II (five replicates for pea sole crop in order to control soil variability). Each subplot consisted of 11 rows spaced 14.5 cm apart. Fungicide-treated seeds were sown on 8 November 2005 (Exp.I) and on 9 November 2006 (Exp.II) using a 6-row pneumatic precision experimental prototype drill with 29-cm row separation. Sowing was done in two passes by moving to the right (14.5 cm) for the second pass and by blocking one row of the drill. The intercrop treatment consisted of six rows of wheat and five rows of pea spaced of 14.5 cm apart, with alternate wheat and pea rows. In Exp.II, 20 mm of irrigation water was applied after sowing because of the low water content in the topsoil. Weeds, diseases and green aphids were controlled as much as possible by application of appropriate pesticides.

Measurements and analysis

The number of seedlings in four rows of 1 m length was counted 1 month after emergence.

Four (Exp.I) and five (Exp.II) samplings were carried out during the growing season at key developmental stages (Fig. 1). Outside rows (the two rows on each side of the plot) were not harvested in order to avoid border effects between plots. At each measurement, seven rows of 0.5 m long were harvested from each main plot by cutting plants just above the soil surface. Sampling points were 0.5 m apart. Final harvesting took place at pea physiological maturity for pea sole crop, and at wheat physiological maturity (Zadoks 92) for wheat sole crop and intercrop. At maturity, the remaining area (4.5 m × 1.015 m) was mechanically harvested to determine total yield. Except for the final date of sampling, the harvested plant biomass was separated into pea and wheat and into green leaf, stem and senescent material. Senescent parts of green leaves were removed and added to senescent material. Green leaf area was determined using a LI-3100 planimeter (LI-COR, Lincoln, NE). Samples were dried at 80°C for 48 h. Dry weight of aerial biomass and N concentration of plants were determined for every crop sample. At maturity, total dry weight and yield were measured on 150 wheat ears and 20 pea plants, together with N concentration of grain and straw. Total N was analysed on sub-samples of finely ground plant material using the Dumas combustion method, with a Leco-2000 analyzer (LECO, St. Joseph, MI). Canopy height was measured manually for each treatment considering its maximum height at: (1) ‘1 cm ear’ stage of wheat (E1cm), (2) beginning of pea flowering (BPF), and (3) wheat flowering (WF).

Fig. 1
figure 1

Major stages and sampling dates during the experimental periods represented on a calendar scale coupled with a degree day scale (base 0°C) scale. Developmental stages: S Sowing, WE wheat emergence, PE pea emergence, WT wheat tillering stage, PD pea leaf development, E1cm ‘1 cm ear’ stage of wheat, SE beginning of pea stem elongation, FL ‘visible flag leaf’ stage of wheat, BPF beginning of pea flowering, WF wheat flowering, EPF end of pea flowering, WR wheat ripening, PSC H final harvest (maturity) for pea sole crop, WSC H final harvest (maturity) for wheat sole crop and intercrop, IC H final harvest (maturity) for intercrop. Wheat stages are indicated with a Z (for Zadoks scale) followed by the number of the stage, e.g. Z 30. N Dates of fertilizer-N application

Full size image

Photosynthetically active radiation (PAR, 400–700 nm wavelengths) was measured using a 30 cm long line sensor PAR-80 (Decagon Devices, Pullman, WA) by orientating it perpendicularly to the rows. Incident PAR on the top of the canopy (PAR0) and PAR transmitted to the soil (PART) were measured orientating the sensor towards the sky. On the other hand, PAR reflected by the ‘soil-crop’ system above the canopy (PARR) and PAR reflected from the bare soil (PARRBS) were measured orientating the sensor towards the soil at a distance of 30 cm from the surface (canopy or soil). Data were collected regularly throughout the growing period between 1300 and 1400 hours (GMT+1) and only when the sky was totally clear.

Calculations

The fraction of PAR absorbed by the crop (FPARa) was calculated using the following equations (Hodges and Kanemasu 1977):

$$ {F_{PARa}} = \left( {\frac{{PA{R_0} - PA{R_T} - PA{R_R} + PA{R_{RS}}}}{{PA{R_0}}}} \right) $$
$$ PA{R_{RS}} = PA{R_T} \times {T_S} $$
$$ {T_S} = \frac{{PA{R_{RBS}}}}{{PA{R_0}}} $$

where PAR0 is the incident PAR above the canopy, PART the PAR transmitted to the soil, PARR the PAR reflected by the ‘soil-crop’ system above the canopy, PARRS the PAR reflected from the soil, T S the fraction of PART reflected from the soil, and PARRBS the PAR reflected from the bare soil. The fraction of PAR absorbed by the crop was regressed against thermal time from emergence to wheat flowering using a logistic model:

$$ {F_{PARa}} = {F_{PAR\max }} \times \frac{{{ \exp }\left( {K1 \times \left( {t - K2} \right)} \right)}}{{1 + { \exp }\left( {K1 \times \left( {t - K2} \right)} \right)}} $$

where F PARmax is the maximum value of F PARa when thermal time approaches positive infinity, K2 the thermal time at the point of inflexion (where F PARa is half of F PARmax), K1 the slope of the curve at the point of inflexion and t the thermal time. Parameters were calculated in order to minimise the root mean square error (RMSE).

PAR absorbed by the canopy (PARa) was then calculated as the sum of daily PAR energy absorbed (dPARai) from emergence (day 1) to the last measurement (day n) corresponding to 10 and 5 days before wheat flowering in Exp.I and Exp.II, respectively:

$$ PARa = \sum\limits_{i = 1}^n {dPAR{a_i} = \sum\limits_{i = 1}^n {0.48 \times S{R_i} \times {F_{PARai}}} } $$

where SRi is the total incident solar radiation (MJ m−2) obtained from the meteorological station on day i, FPARai the fraction of PAR absorbed by the crop on day i (calculated from the logistic model) and 0.48 the fraction of PAR of the total solar radiation energy (Varlet-Grancher et al. 1982).

In order to compare the species growth dynamics in the intercrop we defined an index called the comparative absolute growth rate (CGR). The CGR compares the dry weight growth rate (GR) within a given time interval (t 1 to t 2), of one species relative to another. Thus, CGR for a wheat–pea intercrop (IC) is the ratio between the growth rate of wheat (GRW-IC) and that of pea (GRP-IC):

$$ G{R_W} = \frac{{D{W_{W - IC}}{t_2} - D{W_{W - IC}}{t_1}}}{{{t_2} - {t_1}}} $$
$$ G{R_P} = \frac{{D{W_{P - IC}}{t_2} - D{W_{P - IC}}{t_1}}}{{{t_2} - {t_1}}} $$
$$ CG{R_{IC}} = \frac{{GR{}_{W - IC}}}{{G{R_{P - IC}}}} $$

The same concept was used to compare independent sole crops growth rate as following:

$$ CG{R_{SC}} = \frac{{GR{}_{W - SC}}}{{G{R_{P - SC}}}} $$

CGR values were calculated for each time interval between two successive sampling dates during crop growth. At sowing, the total seed weight was taken as total biomass assuming a 1,000-grain standard weight of 50 g and 150 g for wheat and pea, respectively. CGRIC was calculated separately for each IC replicate. The same was done for CGRSC by using the replicate values of wheat DW for the numerators and the mean pea sole crops values across all replicates for the denominators in order to eliminate the variability between each block replicate attributed to sole crop pea DW. Moreover, for CGRIC we considered the same N treatment for the wheat and the pea while CGRSC was calculated with the unfertilized pea sole crop as a reference, hypothesizing that N was not a limiting resource for legumes and would not affect pea DW. CGR > 1 indicates a faster growth rate of wheat compared to pea over the time period considered and vice versa when CGR < 1.

The efficiency of IC was evaluated using the land equivalent ratio (LER), which is defined as the relative land area required when growing sole crop to produce the aerial biomass dry weight or yield achieved in intercrop (Willey 1979). Aerial biomass LER for a wheat–pea intercrop (LERDW) is the sum of the partial LER values for wheat (LERDW-W) and pea (LERDW-P) (De Wit and Van Den Bergh 1965). LER was calculated for each sampling in order to analyse the dynamics of resource use for wheat and pea. Calculations, assumptions and significance are given in our previous paper (Bedoussac and Justes 2009). By analogy, we calculated the LER by considering the shoot N accumulated, and named it LERN.

The nitrogen status of the wheat was characterised using the nitrogen nutrition index (NNI)—an indicator of crop N nutrition status that reveals a possible N deficiency in the crop at the time of observation—which is defined as the ratio between the actual crop N concentration in shoots (%Nact) and the critical crop N concentration in shoots (%Nc), i.e. the minimum N concentration in shoots that allows the maximum shoot biomass production (Lemaire and Gastal 1997):

$$ NNI = \frac{{\% Nact}}{{\% Nc}} $$

%Nc for a sole crop winter wheat (%NcW-SC) is defined by the critical N dilution curve proposed by Justes et al. (1994) and validated for durum wheat (Justes et al. 1997).

$$ \% N{c_{W - SC}} = 4.4\% {\hbox{ if D}}{{\hbox{W}}_{{\rm{W - SC}}}} < {1}{\hbox{.55 t h}}{{\hbox{a}}^{{ - 1}}}{\hbox{ else 5}}{.35} \times {\hbox{D}}{{\hbox{W}}_{{\rm{W - SC}}}}^{{ - 0}{.442}} $$

where DWW-SC is the shoot dry weight per unit area achieved in sole crop for wheat. In intercrops, the biomass of the neighbouring species has to be taken into account because it contributes to: (1) the canopy light capture and attenuation; and (2) the N dilution according to growth and stages. Hence, Soussana and Arregui (1995) proposed to calculate %Nc for an intercrop (%NcW-IC) using its total biomass (i.e. wheat plus pea). Thus, for the intercropped wheat we used the following equation:

$$ \% N{c_{W - IC}} = 4.4\% {\hbox{ if D}}{{\hbox{W}}_{\rm{IC}}} < {1}{\hbox{.55 t h}}{{\hbox{a}}^{{ - 1}}}{\hbox{ else 5}}{.35} \times {\hbox{D}}{{\hbox{W}}_{\rm{IC}}}^{{ - 0}{.442}} $$

where DWIC is the shoot dry weight per unit area of the whole intercrop.

A value of NNI of at least 1 indicates that crop growth was not limited by N supply at this time whereas when NNI is significantly below 1, N was limiting for crop growth at that specific date.

Statistics

Analysis of variance was carried out using the AOV procedure of the 2.7.1 version of R software (R development Core Team 2007) for each year and date of sampling separately, considering N treatments as the main factor, crops as a sub-factor and interaction between N treatments and crops. All data were tested for normal distribution using the Shapiro–Wilk test, and pairwise comparisons were performed using a two-tailed t-test (P = 0.05 or P = 0.10) to compare N treatments within crops and crops within N treatments. According to Sheskin (2004), the significance of differences between treatments can be estimated using simple planned comparisons when these have been planned beforehand, regardless of whether or not the omnibus F value is significant. Finally, confidence intervals for the means of CGR, LER and partial LER values were calculated from replicates assuming normal distribution according to Sheskin (2004) in order to compare the means of CGR and LER with 1 and partial LER values with 0.5.

Results

Growth dynamics

Emergence and plant density

For both experiments, wheat emerged earlier than pea and the percentage of the cereal emergence was lower than that of the legume (63% and 76%, respectively). On average, for both species, emergence was similar for the two experiments (70% and 66% for Exp.I and Exp.II, respectively) but it was slower in Exp.I. Moreover, for both species, no significant difference (P > 0.10) was found between sole crops and intercrops. Plant densities are given in Table 1.

Green leaf area index

Without N fertilization (Fig. 2a,d), GLAI of the pea sole crop was higher than that of the wheat sole crop at the intermediate growth stages (BPF and WF) but lower for earlier stages (WT and E1cm). Moreover, pea sole crop GLAI reached zero (1,700 and 2,050°C day−1 after wheat emergence for Exp.I and Exp.II, respectively) while wheat sole crop GLAI was still significantly positive.

Fig. 2a–g
figure 2

Dynamics of green leaf area index (GLAI) of sole crops (SC) and intercrops (IC) of pea and wheat for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1) and dates of sampling. Values are means (n = 3–5). For each sampling date, vertical bars on the top of the figure correspond to standard deviation of the total IC, wheat IC, pea IC, wheat SC and pea SC (N0 only), respectively. Asterisks on the x-axis correspond to fertilizer-N application. a–c Experiment I (Exp.I), d–g Experiment II (Exp.II)

Full size image

Overall, GLAI of the whole intercrop was greater than that of the wheat sole crop (P < 0.10) and less than that of the pea sole crop (P < 0.10), except in the initial growth stages (WT and E1cm) where it was comparable to that of the wheat sole crop and slightly greater than that of the pea sole crop (P < 0.10). N fertilization increased wheat GLAI significantly both in sole crop and intercrop (Fig. 2b,c,e,f,g), in particular from E1cm to BPF (P < 0.05). Conversely, in Exp.I, no difference between N treatments was observed for the intercropped pea GLAI while in Exp.II the maximum was obtained for N0 (P < 0.05). Thus the GLAI of the whole intercrop was slightly but not significantly affected by N fertilization (P > 0.10).

Growth in height

Wheat and pea differed in height, but this difference changed over time (Table 2). Overall, wheat was taller than pea in both the initial (E1cm) and the later (WF) growth stages, but no difference was found at the intermediate growth stage (BPF). No significant difference was found between sole crop and intercrop canopy height for both wheat and pea at any of the growth stages (P > 0.10), except for pea at the BPF stage in Exp.II (P < 0.05). N fertilization had no effect on pea height except at the BPF (Exp.II; P < 0.05) and WF (Exp.I; P < 0.05) growth stages, and only a slight but significant effect on wheat height at the BPF and WF growth stages (P < 0.05). Moreover, no pea lodging was observed in intercrop, although this occurred in sole crop for both experiments, with a high proportion in Exp.II.

Table 2 Canopy height (cm) of sole crops (SC) and intercrops (IC) measured at ‘1 cm ear’ stage of wheat (E1cm), beginning of pea flowering (BPF) and wheat flowering (WF) for the unfertilized plots (N0) or considering the mean of all N treatments. Values are means (n = 3–5) ± standard error
Full size table

Dry weight of aerial biomass and interspecies dynamics

Without N fertilization (Fig. 3a,d), the aerial biomass of sole cropped pea was lower than that of the sole cropped wheat until BPF, and then higher (P < 0.05). Overall, the biomass of the whole intercrop was similar to the sole cropped pea (Fig. 3a,d) and higher than that of the sole cropped wheat for low N availabilities from the WF stage onwards (Fig. 3a,b,d,e). Upon increasing the amount of fertilizer N (Fig. 3c,f,g), the biomass of the whole intercrop was similar or lower than that of the sole cropped wheat throughout the whole crop growth (P < 0.10) except at the WF stage for N140 (P > 0.10).

Fig. 3a–g
figure 3

Evolution of total shoot dry weight (DW) production (t ha−1) of SC and IC of pea and wheat for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1) and dates of sampling. Values are means (n = 3–5). For each date of sampling, vertical bars on the top of the figure correspond to standard deviation of the total IC, wheat IC, pea IC, wheat SC and pea SC (N0 only) respectively. Asterisks on the x-axis indicate fertilizer-N applications. a–c Exp.I, d–g Exp.II

Full size image

During crop growth, sole cropped and intercropped wheat biomass increased significantly with fertilizer N amount in Exp.I (Fig. 3b,c) from BPF stage (P < 0.05). In Exp.II, sole cropped wheat biomass at WF stage and maturity increased significantly (P < 0.10) from N0 to N80 while wheat biomass of intercrop was highest (P < 0.05) in N60+ (Fig. 3e) and clearly lowest (P < 0.05) in N0 (Fig. 3d). For both experiments, the biomass of intercropped pea was significantly reduced at WF stage and maturity with fertilizer-N compared to N0 (P < 0.05) except for N100 (Fig. 3b) at WF stage (P > 0.10). As a consequence, total intercrop biomass at WF stage and maturity was increased with N fertilization in Exp.I, while in Exp.II the highest total intercrop biomass at maturity was found in N60+ (Fig. 3e) and the lowest in N140 (Fig. 3g). Crop growth was very slow from emergence until E1cm, but increased rapidly from E1cm to WF stage. Moreover, in Exp.II sole crop wheat biomass remained stable between WF and maturity. Finally, and more surprisingly, the biomass of the sole cropped and intercropped pea decreased—in particular for N0 (Fig. 3d) and N140 (Fig. 3g)—probably due to the high leaf losses.

The CGR values were significantly affected by N fertilization (P < 0.05) and changed over time (P < 0.05) for both sole crops and intercrops.

From sowing (S) to ‘ear 1 cm’ stage of wheat (E1cm), the biomass growth rate of wheat was 2.5 to 4.8 times and 4.5 to 11.7 greater than that of pea in intercrop for Exp.I and Exp.II, respectively (Table 3). Similar dynamics were observed for SC, with average CGR values of 1.8 and 4.0 for Exp.I and Exp.II, respectively, indicating that competition between wheat and pea intercropped in rows were weak until E1cm stage.

Table 3 Comparative absolute growth rate (CGR) calculated for the time interval corresponding to the dates of sampling for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1) of experiments I and II. Values are means (n = 3–5) ± standard error
Full size table

Afterwards, for both experiments and all treatments, CGR values in intercrop were close to or higher than 1 during the two following growth periods (E1cm to BPF and BPF to WF) except for N0 between BPF and WF (Table 3), indicating that the growth rate of wheat was similar to, or slightly higher than, that of pea in fertilized intercrops. However, values significantly lower than 1 for N0 treatments indicated higher growth rate of pea in intercrop from BPF to WF. Concerning the sole crops, CGR values were close to 1 or more from E1cm to BPF (1.1 and 1.2 on average in Exp.I and Exp.II respectively) and close to or below 1 (0.7 for both experiments) from BPF to WF, indicating similar or higher pea growth rate than wheat during this later period.

Next, from WF to harvest maturity, wheat growth rate was higher than that of pea in Exp.I for both sole crop and intercrop, even though intercrop CGR values were significantly different from 1 only for N0. On the contrary, in Exp.II, CGR values were around 0 for both sole crop and intercrop (except for N60+ in intercrop) due to uncontrolled pests, diseases and lodging in late growth stages.

Considering the whole growth period, CGR values were lowest and close to 1 in N0 for both experiments in sole crop and intercrop. CGR values were always significantly higher than 1 in N fertilized plots, with no significant difference between N treatments, indicating that wheat growth rate was higher than that of pea with N fertilization. Globally, CGR were higher in Exp.II than in Exp.I (3.4 and 2.4, respectively), indicating pedoclimatic conditions more favourable to wheat during the second experiment.

Dry weight biomass land equivalent ratio

At the four key stages sampled (E1cm, BPF, WF and maturity), LER values calculated from aerial biomass dry weight (LERDW) were roughly 1 or more in all treatments (P < 0.05; Fig. 4) except for N180 (Fig. 4c) at E1cm and BPF and for N140 (Fig. 4g) at maturity. These results indicate that, throughout the whole crop growth, resources were used for aerial biomass production up to 36% more efficiently in intercrops than in sole crops in low N conditions. On the whole, LERDW values were reduced with N fertilization, particularly for the N180 (Fig. 4c) and N140 (Fig. 4g) treatments, mostly because of the reduction in pea partial LERDW (LERDW-P). LERDW-P values were always equal to or significantly below 0.5 (P < 0.10), except for N0 in both experiments (P > 0.10). For all N treatments and stages, wheat partial LERDW values (LERDW-W) were always at least 0.5, and often much higher (P < 0.10). Moreover, LERDW-P values remained more or less stable throughout the whole crop growth, and similar results were found in Exp.I for LERDW-W, while in Exp II, LERDW-W increased until pea flowering and then decreased, in particular for N0 (Fig. 4d) and N60+ (Fig. 4e).

Fig. 4
figure 4

Partial land equivalent ratio (LER) of wheat and pea calculated from dry weight (LERDW) and N accumulated (LERN) for the two experiments (a–c Exp.I, d–g Exp.II) and N treatments (Nx where ‘x’ represents N applied in kg N ha−1) for the different sampling dates: WT wheat tillering, E1cm ‘ear 1 cm’ stage of wheat, BPF beginning of pea flowering, WF wheat flowering, H harvest maturity. Values are means (n = 3–5) ±standard error. +, * LER significantly different from 1 at P = 0.10 and P = 0.05, respectively

Full size image

Resource use dynamics

Light absorption

The fraction of PAR absorbed (FPARa) by sole crops and the whole intercrops increased with crop growth from emergence to WF following a logistic model (Fig. 5). The RMSE values (Table 4) were very low (3.6% at maximum) and the bias was nil (0.6% at maximum), indicating a good fit of the data by the model. The parameter values of the logistic model (Table 4) indicate that the thermal time at the point of inflexion (K2) of the pea sole crop was greater than that of the intercrop, itself higher than that of wheat sole crop. The slope of the curve at the point of inflexion (K1) was highest for the pea sole crop and that of the intercrop was higher than that of the wheat sole crop for N60+ and N140 (Exp.II) and lower for N0 in Exp.I.

Fig. 5a–g
figure 5

Evolution of the photosynthetically active radiation (PAR) fraction absorbed of SC of pea or wheat, and of the whole IC measured throughout crop growth for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1). Values are means (n = 3–5) and curves are regressed non-linearly against thermal time using the following logistic model: \( {F_{PARa}} = {F_{PAR\max }} \times \frac{{{ \exp }\left( {K1 \times \left( {t - K2} \right)} \right)}}{{1 + { \exp }\left( {K1 \times \left( {t - K2} \right)} \right)}} \). For each date of sampling, vertical bars on the top of the figure correspond to standard deviation of the total IC, wheat SC and pea SC (N0 only), respectively. Asterisks on the x-axis indicate fertilizer-N application. a–c Exp.I, d–g Exp.II

Full size image
Table 4 Values of fitted parameters, cumulative photosynthetically active radiation (PAR) absorbed, root mean square error (RMSE), bias and PAR land equivalent ratio (LER) calculated from the logistic model: \( {F_{PARa}} = {F_{PAR\max }} \times \frac{{{ \exp }\left( {K1 \times \left( {t - K2} \right)} \right)}}{{1 + { \exp }\left( {K1 \times \left( {t - K2} \right)} \right)}} \) for the whole intercrop (IC), pea sole crop (P SC) and wheat sole crop (W SC) for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1)
Full size table

Without N fertilization (Fig. 5a,d), FPARa of the pea sole crop was lower than that of wheat sole crop in early stages (until E1cm), itself slightly greater than that of the whole intercrop. Then, from E1cm to WF, the FPARa of intercrop became lower than that of pea sole crop and greater than that of wheat sole crop. When N fertilizer was applied, no significant difference was observed between intercrop and wheat sole crop FPARa (Fig. 5c,e,f,g) and the maximum values of FPAR, indicating that the potential of PAR absorption was reached (Table 4). Finally, the cumulative PAR absorbed (Table 4) from sowing to WF was lowest for the pea sole crop and highest for the wheat sole crop except for N0 where the maximum was observed for the whole intercrop.

Nitrogen acquisition and wheat nitrogen nutrition status

Without N fertilizer, and until BPF stage (Fig. 6a,d), the pea sole crop accumulated significantly (P < 0.10) more N in shoots than the wheat sole crop and than the whole intercrop. In N-fertilized treatments, the whole intercrop accumulated more N than the pea sole crop in Exp.I (Fig. 6b,c) and a similar or smaller amount in Exp.II (Fig. 6e,f,g). The whole intercrop accumulated more N than the wheat sole crop except in early stages (WT and E1cm) and at maturity for N180 (Fig. 6c) and N140 (Fig. 6g) (P > 0.10).

Fig. 6a–g
figure 6

Total shoot N accumulation (kg N ha−1) of SC and IC of pea and wheat for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1) and sampling dates. Values are means (n = 3–5). For each date of sampling, vertical bars on the top of the figure correspond to standard deviation of the total IC, wheat IC, pea IC, wheat SC and pea SC (N0 only) respectively. Asterisks on the x-axis indicate fertilizer-N application. a–c Exp.I, d–g Exp.II

Full size image

The wheat N accumulation in sole crop and intercrop were positively correlated with N fertilization in Exp.I (Fig. 6a,b,c). In Exp.II, early N fertilization applied at E1cm (Fig. 6f,g) increased the wheat N accumulated in sole crop and intercrop between E1cm and BPF stages. Conversely, N fertilizer applied at flag leaf visible (Fig. 6e,g) increased the wheat N accumulated in sole crop and intercrop between BPF and WF stages only for the N60+ treatment (Fig. 6e). The N accumulated by crops increased throughout the growth period in Exp.I while in Exp.II the wheat N uptake of sole crop and intercrop remained more or less stable between WF and maturity, whereas that of the pea intercrop decreased.

Last but not least was the finding that N accumulated by intercropped wheat was 70–85% that of the sole crop despite the fact that intercropped wheat was sown at half the plant density of the wheat sole crop (Fig. 6). The proportion of N uptake in intercrop compared to sole crop was higher for N0 treatment, indicating a higher N availability per wheat plant in intercrop.

In Exp.I, for all N treatments, wheat nitrogen status (NNI) was always significantly below 1 (P < 0.05) for both sole crops and intercrops (Fig. 7) indicating a significant N deficiency throughout the whole growth period. In Exp.II, NNI values of both wheat sole crops and intercrops were significantly above 1 (P < 0.05) at wheat tillering, roughly 1 or slightly less at E1cm stage of wheat and significantly lower thereafter (P < 0.05) except at the beginning of pea flowering (BPF) in N80 (Fig. 7f) for the intercrop and in N140 (Fig. 7g) for both sole crop and intercrop. The main result is that the intercropped wheat NNI was always higher than, or at least statically equal to, that of the sole crop for both experiments and all N treatments, indicating that intercropping improved the wheat N status during the growing season. Without N fertilization, the wheat NNI of both sole crop and intercrop decreased with development stages and aerial biomass increase, but in Exp.II the wheat NNI remained quite stable from BPF to wheat flowering (WF). Each fertilizer-N application increased the wheat NNI or reduced its decline.

Fig. 7
figure 7

Nitrogen nutrition index (NNI) of wheat in SC and IC for the different N treatments (Nx where ‘x’ represents N applied in kg N ha−1) and sampling dates. Values are means (n = 3–5) ±standard error. Asterisks on the x-axis indicate fertilizer-N application

Full size image

Land equivalent ratio for Nitrogen acquisition

Whatever the date of sampling or N fertilization regime, LER values calculated from N accumulated in shoots (LERN) were always higher than, or at least statistically similar to, those calculated from aerial biomass (Fig. 4). Throughout crop growth, LERN values were roughly 1 or more in all treatments (P < 0.05) except for N140 at maturity (Fig. 4g). This indicates an advantage of intercrop in comparison to sole crop for N acquisition throughout crop growth more than for biomass production. On the whole, LERN values were reduced with N fertilization, in particular for N180 (Fig. 4c), N80 (Fig. 4f) and N140 (Fig. 4g). For all N treatments and stages, wheat partial LERN values (LERN-W) were always above 0.5 (P < 0.10) except at E1cm in Exp.I, wheat tillering in Exp.II and for N140 at all stages. On the other hand, the values of LERN-P were always equal to or below 0.5 and lower when N fertilizer was applied. The total LERN values and partial LERN values changed significantly over time, particularly in Exp.II. LERN values of wheat were increased until pea flowering and then decreased, particularly for the N0 (Fig. 4d), N60+ (Fig. 4e) and N80 (Fig. 4f) treatments. In Exp.I, LERN values of wheat remained more or less stable throughout crop growth, and similar results were observed for LERN values of pea. In Exp.II, LERN values of pea were significantly (P < 0.10) reduced between WF and maturity.

Discussion

Various authors have indicated that sequential measurements of crop growth allow a better understanding of the dynamics and mechanisms of competitive interactions in intercrop and sole crop compared to a single measurement of final yield (e.g. Connoly et al. 1990; Andersen et al. 2004). In our study, we sampled four (Exp.I) and five (Exp.II) times during the growing season at key development stages. The following discussion will be organized largely according to the growing periods between these stages because they determined key periods for interspecific competition and complementarity between durum wheat and winter pea.

Our data has shown that for limited early N availability (no fertilizer or a small amount applied after the wheat booting stage) there were significant complementarities between intercropped durum wheat and winter pea for light and nitrogen use. This was clearly due to species differences in their dynamics of growth and to differences in species phenology and physiology, as discussed below.

Early growth dynamics

We observed that wheat growth was earlier than that of pea for both sole crops and intercrops and the better start of wheat was largely the result of faster seedling emergence. Furthermore, wheat early advantage over pea was probably due to: (1) its faster and deeper root growth; and (2) higher N demand than those of pea (Fujita et al. 1992; Corre-Hellou 2005; Hauggaard-Nielsen et al. 2003; Corre-Hellou and Crozat 2005). It was also certainly due to the establishment of costly pea nodulation (in energy and nutrients) according to e.g. Vocanson et al. (2005) and Voisin et al. (2002), which could have reduced early pea growth. Crop emergence and early growth were slower in Exp.II due to: (1) the coarse soil structure; (2) the low soil water content at sowing, only partially compensated by irrigation; and (3) the low winter temperatures. In particular, pea was affected by the coarse soil structure and the low soil water content at sowing in Exp.II possibly explained by a higher water demand due to greater pea seed size. Also, and particularly in Exp.II, pea emerged later and hence at lower temperatures than wheat, increasing the differences between the early growth of the two species as illustrated by higher CGR values in Exp.II than in Exp.I.

In early stages of growth (until wheat stem elongation corresponding to ‘ear 1 cm’ stage), wheat was much more efficient than pea for light absorption, due to its earlier growth, greater height during winter, tillering and higher GLAI, as illustrated by CGR values much higher than 1 from sowing to E1cm. Similar dynamics were observed for the sole crop and the intercrop, indicating that our results are in good agreement with species’ phenological and physiological differences. However, pea growth was lower in intercrop than in sole crop during early stages as illustrated by CGR values higher for the intercrop than for the sole crops. This is certainly due to less incoming PAR being available for the intercrop pea, being partly shaded by the durum wheat.

Intermediate period of growth dynamics

From wheat ‘1 cm ear’ stage to wheat flowering, the intercrop as a whole was more efficient for light absorption than wheat sole crop thanks to light absorption by the intercropped pea in late stages, in particular without N fertilization. Indeed, legume efficiency for light absorption in late stages was due to: (1) the later growth of the legume; and also (2) the small effect of senescence on pea light interception; this behaviour of pea was first shown for sole crop by Guilioni and Lecoeur (2005), who demonstrated that pea leaves did not fall, maintained their orientation and reduced their area by only 5–10%, allowing higher light absorption compared to wheat close to cereal flowering stage. The later growth of the legume was observed for both sole crop and intercrop, as illustrated by lower CGR values compared to the period from sowing to ‘ear 1 cm’ stage of wheat indicating that phenological difference can play a central role in shaping the observed dynamics (Berntsen et al. 2004).

When N fertilizer was applied early (wheat tillering or stage ‘ear 1 cm’ of wheat), we observed a slightly increase in cereal growth in winter but a strong increase in early spring. Hence, when the growth of the intercropped legume was expected to increase markedly—as in sole crop—there was less light available to it, which reduced its aerial biomass production and consequently that of the whole intercrop. However, N fertilization had only a small effect on the FPARa of intercrop because it was already close to the potential without N fertilization, indicating a good complementary absorption of incoming PAR of the two species under these N conditions. This suggests that the efficiency of intercrop in N fertilized treatments was due mostly to wheat light absorption—certainly related to the effect of N fertilization for wheat GLAI and biomass increase—which markedly reduced pea growth due to light competition. This result was coherent with the positive correlation observed in both experiments between wheat sole crop and intercrop GLAI and N fertilization, as already described by many authors for other graminaceous crops (e.g. Gastal et al. 1992; Wilman and Pearse 1984).

On the contrary, in low N availability conditions, no harmful competition of wheat for light was observed on the legume as indicated by low CGR values, because wheat early growth was limited by a lack of available mineral N in the soil. Then, unlike spring barley–pea intercrop (Corre-Hellou 2005), durum wheat–winter pea allowed a similar or better use of light compared to sole crops throughout the growth period, especially with low N supply. These conclusions are strengthened by the fact that similar results were obtained for both experiments while weather conditions were very different. This also confirms that, in N fertilized treatments, wheat has a higher and pea a lower interspecific competitive ability when intercropped, and this difference was increased as soil N availability was enhanced.

N accumulated by the intercropped wheat was at least 70% and often more of that of the wheat sole crop for half plant density, particularly until the beginning of pea flowering and when little or no N fertilizer was applied. This confirms that the cereal had access to a more than proportionate share of soil mineral N in intercrop per plant as compared to sole crop (Jensen 1996). This result is supported by the following evidence: (1) the increase in pea symbiotic N2 fixation rate demonstrated in our companion paper (Bedoussac and Justes 2009), and in agreement with several other studies for spring intercrops (e.g. Corre-Hellou 2005; Hauggaard-Nielsen et al. 2003; Corre-Hellou and Crozat 2005); and (2) the lower durum wheat biomass per square metre in intercrop due to light competition exerted by pea. Consequently, and as expected, the intercropped wheat N status (NNI) was always higher than that of the sole crop except when a large amount of N was applied. This confirms that intercrop can improve wheat N status compared to sole crop, and consequently the grain protein content in particular, with little or no N fertilizer. Moreover, the advantage of intercrop over sole crop seems to increase throughout the growth period due to higher N available per durum wheat plant. The N status of sole cropped and intercropped wheat were mostly below 1, in particular in late stages and when little or no N fertilizer was applied, underlining the existence of a nitrogen deficiency in the crop throughout the growing period, which confirmed that we were in low N input conditions. However, the intercropped wheat had a lower N deficiency, which allowed better photosynthesis and hence higher radiation use efficiency (e.g. Justes et al. 2000).

Late growth dynamics

From WF to maturity, durum wheat seems to be the most efficient, certainly due to: (1) the earlier initiation of seed filling and maturity of pea (physiological maturity of the legume was observed 3 weeks before that of the durum wheat); and (2) the effect of the diseases and pests of pea, which were not perfectly controlled by pesticide applications. However, it was not possible to determine precisely if the greater efficiency of wheat in late stages occurred between WF and pea maturity, or between pea maturity and wheat maturity because wheat biomass was not measured at the pea maturity stage in our experiments. Nevertheless, in late stages, wheat took advantage of the intercrop by remobilizing more N into its grain due to: (1) lower grain yield; and (2) almost similar N available compared to the sole crop according to the high legume N2 fixation described in our companion paper (Bedoussac and Justes 2009).

Conclusion

The objective of our paper was to answer the following question: how can the final outcome of the durum wheat–winter pea intercrop be explained? Thus, in our experiments, we demonstrated that the use of environmental resources for plant growth changed over time, indicating that the ability of a crop to exploit resource niches and thereby capture resources varies significantly during the growth cycle. Throughout crop growth, resources were used up to 36% more efficiently in intercrops than in sole crops for aerial biomass in low N conditions, and the efficiency of intercrop for N uptake was greater than for biomass production throughout crop growth at all N fertilization levels. Regardless of N availability, wheat took advantage of intercrop for aerial biomass production more than pea (LERDW-W > LERDW-P) by using N, light and certainly CO2 and water more efficiently throughout the growth cycle due to its earlier growth, greater height and higher N demand. CGR and LER dynamic analysis indicated that pea took advantage of intercrop mainly between pea flowering and wheat flowering (near the end of pea grain filling), while wheat took advantage of intercrop between the emergence and the beginning of wheat stem elongation (E1cm stage) and until pea flowering.

Wheat benefited from N fertilization more directly than pea by its faster growth, improving light and water capture, and hence suppressed pea growth regardless of the amount and date of N application (Fujita et al. 1992; Jensen 1996; Bulson et al. 1997). Moreover, as shown by many other studies on spring peas, winter pea has a higher interspecific competitive ability at low N levels because of: (1) the reduction in N2 fixation with N supply (Voisin et al. 2002; Ghaley et al. 2005); and (2) the amplification of differences in species growth dynamics with N supply (Hauggaard-Nielsen and Jensen 2001; Ghaley et al. 2005). Durum wheat and winter pea showed interspecific complementarity in good agreement with other studies on cereal–legume spring-sown intercrops (Corre-Hellou and Crozat 2005; Hauggaard-Nielsen et al. 2006, 2009).

However, intercrop efficiency and crops interspecific competitiveness ability depended on the dynamics of N availability throughout the growth cycle of intercrops. N fertilizer applied at the beginning of wheat stem elongation clearly leads to a significant and early decrease in pea N acquisition due to the quantitative reduction in N2 fixation described in the companion paper (Bedoussac and Justes 2009), with the opposite result seen with no or late N fertilization. As a consequence, the advantage of intercrop for biomass production, yield or accumulated N was maximal under low N fertilization treatments (no or late fertilization) due to good complementarities of durum wheat and winter pea for GLAI and growth dynamics over time.

Another point concerns the reduction of pea lodging, which is of great importance in order to obtain good yield. Indeed, pea is very sensitive to lodging, mainly because of diseases like pea ascochyta (Mycosphaerella pinodes), which usually attacks the stems. Hence pea sole crop must be harvested soon after maturity, which may be difficult in bad weather or with severe lodging. In intercrop, pea maturity occurred around 3 weeks before durum wheat and, due to lodging, pests and diseases, this can be very prejudicial to pea harvest efficiency. However, under our conditions, no lodging was observed for pea in intercrop because the legume clings to the durum wheat with its tendrils, creating a dense and stable network.

As a practical conclusion, our results indicate that durum wheat–winter pea intercrop productivity can be optimised by adapting cropping sequences, particularly N fertilization, but also by choosing crop species of widely different phenologies and/or morphologies in order to maximise capture and minimise competition for solar radiation. It is essential to prevent an adverse effect on legume growth (involving N2 fixation), which very much influences the final outcome and total yield potential of the intercrop.

Further research is needed to investigate the effect of wheat and pea cultivars on interspecies dynamics, hypothesizing that short or late wheat cultivars would be more suited for intercrop than tall or early ones and vice versa for pea cultivars. Furthermore, intercrop competition analysis—considering various legumes species, crop densities, sowing dates (possibly shifting wheat and pea sowing date) and N fertilization—seems critical in order to optimise these innovative agro-ecosystems. Moreover, defining specific objectives (grain or biomass production, grain protein concentration increase, reduction of N leaching and chemical use, etc.) would lead to different choices for the optimal combination of cultivars of the two species, their density and N-fertilization strategies.