NEIGHBOUR effect of pollen density on yield evaluation of non 100% fertile genotypes OF OILSEED RAPE (Brassica napus).

Jean-Jacques Schott¹

¹ GEVES (Groupe d'Etude et de Contrôle des Variétés et des Semences), La Minière, 78285 GUYANCOURT. E-mail : jean-jacques.schott@geves.fr

Abstract

The objective of this study is to estimate the influence of 100% fertile oilseed rape (Brassica napus) genotypes on yield evaluation of non 100% fertile genotypes (varietal association between 80% of a sterile hybrid and about 20% of one or two fertile lines) when tested together in small plot trials.

In a spring oilseed rape trial of 1998, we show, using erucic acid content as a marker, that for varietal associations, cross-pollination from both adjacent plots is about 70%. For fertile lines, this ratio is only 13%. The rate of cross-pollination not due to the neighbour plots is about 2% only.

For winter oilseed rape, covariance analysis shows that yield of varietal associations is frequently overestimated when bordering plots are fertile genotypes.

In 1995, varietal associations were tested in 6 locations in France with either 60% (high level of pollen density) or only 7% (low level of pollen density) of fertile genotypes inside the trials with 30 different entries. Significant competition effects due to heterosis of the hybrids could be demonstrated in these small plots (width of 1.5 m). The analysis shows that when there is one adjacent 100% fertile plot, yield of the varietal association is significantly enhanced of about 3% in the low pollen level trials. This increase can be completely assigned to pollen availability because comparisons are made with similar competition conditions between plots.

In French registration trials of 1996 and 1997, estimation of the bias is 1.6% increase of yield for a varietal association bordered by one fertile plot in comparison to no fertile adjacent plot. For this estimation, we used only 10 trials with non harvested border rows in order to control competition. Similarly, in the post-registration network of 1997, the value estimated on 19 trials is 1.9% of yield. By using covariance analysis, we show that on average the bias is 4.5% of yield overestimation when a varietal association is bordered by two adjacent plots containing fertile lines.

Keyword : brassica napus; oilseed rape; pollen dispersal; varietal association; neighbour effect

Introduction

Varietal associations (VA) are composed of 70 to 80 % male sterile hybrids which need the pollen of the 20 to 30 % fertile lines (FL) to produce seeds. Since 1992, this kind of varieties are tested in the French official network for national registration by the CTPS (Comité Technique Permanent de la Sélection des plantes cultivées). Due to heterosis, VA induce competition effects in small plot trials. In 1995, non harvested border rows or sufficiently wide plots were imposed in the trials to reduce significantly this bias (Schott et al, 1994). Yield assessment of VA was primarely realised in field trials containing principally 100% FL of different flowering precocity. These experiments optimise the cross-pollination conditions and introduce some bias in productivity evaluation. Since 1996, in the official trials, VA are all tested together in isolated trials containing only a few (2 or 3) standard FL. The objective of this study is to estimate the bias on yield evaluation of pollen dispersal in oilseed rape trials.

Material and method

Trials

The field experiments results used in this study are issued from four different private or official networks including VA and FL in the same trials :

i) One spring oilseed rape trial realised in 1998 by CETIOM (Centre Technique Interprofessionnel des Oléagineux Métropolitains) in Nancy which contains 12 genotypes (2 VA and 10 FL) and 4 replicates of plots with 5 rows harvested out of 7. Randomised complete block (RCB) design is used. One VA and two FL are high erucic acid cultivars (É 50% erucic acid), the others are low ones (< 2%). Erucic acid content is measured on each plot by chromatography. Frequency of cross-pollination (outcrossing) is estimated using erucic acid content of the seeds as a marker in accordance with the assumption that heterozygous seeds contain about 25% erucic acid. Values observed on self pollinated plants are theoretical references. For VA these references are obtained in 5 isolated fields.

ii) Private network from Cargill Génétique Europe company realised in 1995 and 1996 in 6 different locations in the West, Centre and East regions of France (Table 1). Different VA are tested in the same location in either a high level pollen context (HP) or a low level pollen context (LP) depending on the rate of FL inside the trials. Influence of neighbour plot pollen density is estimated by comparing the yields of the same VA in the different pollen environments.

Table 1 – Locations and composition of Cargill’s trials

Department N°	10	17	18	21	28	89
Width plot (m)	1.5	1.5	1.5	1.5	1.5	1.5
Length plot (m)	6.7	8.0	6.0	6.0	7.5	6.0
1995	HP + LP^3	HP^3 + LP^3	HP^3 + LP^3	HP + LP	HP^3 + LP^3	HP + LP
1996	2LP + 1HP	4LP (1^3)	4LP (2^3)	3LP	5LP (2^3)	2LP

³ : trial with 3 replicates (others have 2 replicates only)

HP (high pollen level density type of trial)         : 12 VA + 18 FL

LP (low pollen level density type of trial)          : 28 VA +   2 FL

4LP (2³) = 4 trials of LP type with 2 having 3 replicates

iii) CTPS network from 1996 and 1997 composed of 37 trials in RCB design with 4 replicates. Trials contained more or less VA (Table 2). Some mixed hybrids (MH) 50% fertile are present in these trials. In the analysis they are considered as a third category. In these trials, competition is controlled by non harvested border rows or sufficiently wide plots as proposed by Schott et al (1994).

Table 2 – Description of the CTPS network.

Year	1996	1997	Total
Nb trials	21	16	37
Nb varieties	20	22	42
Type of variety
MH	1	6	7
VA	15	14	29
FL	4	2	6

MH : mixed hybrid 50% fertile            

VA : varietal association 20-30% fertile          

FL : 100% fertile line

iv) Post registration network of 1997 realised by CETIOM in the main rapeseed growing areas in France. It includes 19 trials in a RCB design with 4 replicates and 17 different varieties (3 FL + 13 VA + 1 MH).

Analysis

A classical analysis of variance model with replicate and genotype effects is fitted on each trial from the three networks. Residuals representing the genotype * replicate interaction are used to compare the different varietal behaviours in relation to the type (genetic structure) of neighbour plots. Student's t test is used to compare the means of the different situations.

In a second step, covariance analysis is used to quantify roughly the influence of neighbour pollen density level on yield of VA experimented in low pollen level conditions. The model used is :

U_{w  =} q_t(w) + l.(dP)_w + b_b(w) + e_{w (dP1)}

where (dP)_w = P_w - (P_w-1 + P_w+1) / 2 is the difference of the pollen rate between each plot and the average pollen rate of its two neighbours. The estimated values of l quantify the bias on yield due to pollen availability,

U_w is the yield of the plot _w , q_t(w)  is the mean effect of genotype t on the plot _w , b_b(w) is the effect of block b , and  e_w is an error term.

Results

The main results of outcrossing are presented below for the different approaches.

i) Table 3 presents the cross pollination rate for the different kinds of spring rapeseed varieties. We observe that for VA, outcrossing is important, about 70% when plots are bordered by two FL. Extrapolation from the case of one FL leads to approximately the same rate. For FL, outcrossing rate is of about 12 to 18 %. A background contamination level of 2.1% is recorded on FL. Some publications, Scheffler et al. (1993) and Mesquida and Renard (1982) indicate that the major part of pollen dispersal is limited to a few meters from the source. This corroborates our hypothesis of a preferential nearest neighbour pollination trend. Hühn and Rakow (1979) used the same marker and estimated outcrossing to be from 5% to 15% only in winter oilseed rape depending on the cultivar and the distance between plots.

Table 3 – Cross pollination rate for the different kinds (erucic/non erucic) of spring rapeseed varieties.

Kinf of neighbour variety (erucic/non erucic)		Different				Same
Type of neighbour variety	0 FL		1 FL	2 FL	2 FL
VA	-		34.6 (11.9) [4]	70.9 (1.9) [3]	11.2 [1]
FL	2.0 [1]		6.1 (4.3) [16]	18.5 (4.6) [7]	2.1 (2.2) [16]

( ) standard deviation of the mean value                       [ ] number of observations

ii) Table 4 presents yield results obtained under different pollen level contexts in Cargill's network of 1995 and 1996. The general average yield is 43.6 dt / hectare. Student's t test is used to compare the means. In 1995, for FL in HP context, only the values for 0FL and 2FL are significantly different (prob : 0.035). This demonstrates that some competition effects occur in these trials so that it's wise to do comparisons only between different pollen contexts and within the same type of neighbour. Competition estimated on FL is approximately of 1dt/ha (0.74 + 0.24). For VA, there is no yield difference between the two contexts when two VA are bordering. But, when there is one adjacent FL, the difference becomes highly significant (prob 0.002), it's estimated to 1.3 dt/ha (1.28 + 0.04). The results of 1996 confirm existence of competition.

Table 4 – Average yield performance of different types of varieties grown in a low or a high level pollen context in Cargill's network. Yield is expressed in dt/ha.

Year		1995
Type of neighbour variety		0 FL	1 FL	2 FL
Pollen context	Type of variety
LP	VA	-0.28 (2.47) [366]	1.28 (2.99) [81]	0.58 [1]
	FL	0.11 (3.44) [30]	-2.56 [1]	-0.72 [1]
HP	VA	-0.56 (3.16) [36]	-0.04 (2.65) [95]	0.32 (2.75) [73]
	FL	-0.74 (2.38) [42]	0.02 (2.54) [150]	0.24 (2.64) [144]
Year		1996
LP	VA	-0.09 (1.81) [1295]	0.68 (1.77) [187]	-0.09 (1.79) [2]
	FL	-0.22 (1.63) [96]	2.16 (1.67) [10]	-
HP	FL	-	-0.43 (2.01) [4]	0.03 (2.48) [56]

( ) standard deviation of the mean value                       [ ] number of observations

LP and HP : low and high pollen level context

iii) Table 5 presents yield results obtained in the CTPS network of 1996 and 1997 in a low pollen context (less than 20% of the varieties are 100% fertile). The general average yield is 42.3 dt/ha. If all the trials are considered, we observe a significant (prob = 0.001) difference of 0.37dt/ha (0.87% of yield increase) using Student's t test for VA between 0 and 1 FL as border. The difference between 1 and 2 FL is not significant whereas the difference between 0 and 2 FL is significant (prob = 0.041) and represents 1.63% of yield increase. For MH and FL there is non significant difference. This indicates also that competition effects are generally well controlled on average in this kind of lay-out. Nevertheless, by using a covariance analysis model, see Schott et al. (1994), we show that some trials present a significant effect of the "plant height" covariate indicating potential presence of competition between plots. On the remaining 10 trials with a high control of competition, we observe 1.60% (prob = 0.001) of yield increase between 0 and 1 FL for VA and 5.52% (prob = 0.015) between 0 and 2 FL as border.

iv) In the CETIOM network, in Table 5, the general average yield is 39.4 dt/ha obtained in a low pollen context (18% of the varieties are 100% fertile). Plots are sufficiently wide to avoid inter-plot competition. We observe a significant (prob = 0.001) difference of 0.73 dt/ha (1.86% of yield increase) for VA between 0 and 1 FL as neighbour. Other effects are estimated with too little observations to allow more conclusions.

Covariance analysis shows that only about 30% of the trials give a significant (prob a <10%) estimation of the bias due to neighbour pollen density. It can be estimated to 0.2 to 0.3 dt/ha (0.57% of the VA average mean) more yield for 10% neighbour pollen increase. Thus, a VA's yield can be overestimated by 4,5% on average when it's bordered by 2 FL. In some trials it can become more important, the maximum observed in one trial is 10%.

Table 5 – Average yield performance of different types of varieties in relation to the type of bordering plot in both CTPS and CETIOM networks. Yield is expressed in dt/ha.

Type of neighbour variety	0 FL	1 FL	2 FL
CTPS
Type of variety	37 trials in 1996 and 1997
VA	-0.11 (2.37) [1516]	0.26 (2.23) [605]	0.58 (1.72) [29]
FL	-0.01 (2.12) [358]	-0.04 (2.14) [104]	0.74 (2.05) [9]
MH	-0.02 (2.03) [379]	0.14 (2.04) [78]	-1.54 (0.08) [2]
	10 trials selected without any significant competition effect
VA	-0.25 (2.37) [388]	0.44 (2.17) [188]	2.14 (1.18) [6]
FL	-0.00 (2.29) [111]	0.04 (2.46) [24]	-0.38 [1]
MH	0.18 (2.23) [79]	-0.70 (2.53) [20]	-
CETIOM
	19 trials in 1997
VA	-0.14 (2.29) [532]	0.59 (2.50) [131]	-1.26 (0.32) [2]
FL	0.05 (1.86) [73]	-2.01 (3.20) [2]	0.47 [1]

( ) standard deviation of the mean value                       [ ] number of observations

Table 6 – Results of the covariance analysis model (dP1) on 12 selected trials where the covariate is significant at a < 10% for VA genotypes.

Year	Network	Nb trials / total	Mean yield dt/ha	l (X10)	Pr > |t|
1996	CTPS	7/21	38.46	-0.19 ± 0.16	2.98
1997	CETIOM	5/19	38.58	-0.25 ± 0.19	3.76

l (X10) : mean regression coefficient multiplied by 10 (average ± standard deviation)

Pr > |t| : mean probabilities associated to the value of Student's t test for the covariate dP.

Conclusion

These biometrical analysis show that cross-pollination is very important in winter and spring oilseed crops. In the case of varietal associations outcrossing is necessary for seed production. In small plot trials, pollen exchange is optimised between plots because of proximity and of different precocities of the genotypes. In these conditions, VA productivity is always overestimated when tested together with fertile lines in the same trials. The relative position of the plots within the replicates has direct influence on the productivity and induces bias. Even if it depends also on the climatology, we demonstrate that this bias is of about 5% of the yield on average when a VA is bordered by two FL. To avoid this, trials for VA evaluation should be isolated as much as possible from all the sources of pollen and should exclude all 100% fertile genotypes in it. A large width of the plots reduces physical inter-plot competition effects and perhaps also pollen exchange between experimental plots. But this has already to be confirmed by some new experiences.

Acknowledgments

This research was supported by the French Ministry of Agriculture. We thank CETIOM, Cargill and the different companies involved in the CTPS network for providing the data used in this study.

References

SCHOTT JJ., BAR-HEN A., MONOD H. and BLOUET F. (1994). Compétition entre génotypes dans des essais variétaux de colza oléagineux d’hiver. Cahiers Agriculture, 3, 377-383.

HÜHN M. and RAKOW G. (1979) Einige experimentelle Ergebnisse zur Fremdfruchtungsrate bei Winterraps (Brassica napus oleifera) in Abhängigkeit von Sorte und Abastand. Z. Planzenzüchtg. 83, 289-307.

SCHEFFLER J.A., PARKINSON R. and DALE P.J. (1993). Frequency and distance of pollen dispersal from transgenic oilseed rape (Brassica napus). Transgenic research 2, 356-364.

MESQUIDA J. and RENARD M. (1982). Etude de la dispersion du pollen par le vent et de l’importance de la pollinisation anémophile chez le colza (Brassica napus L., var. Oleifera Metzger). Apidologie 13 (4), 353-366.