Summary — Different genotypes of white lupin were studied for 2 consecutive years and at 6 sowing periods within each year. The mechanisms of frost resistance in lupins was studied.

The seed yields and maturity dates of an indeterminate cultivar (Lunoble) and a determinate line (CH304/70) of Lupinus albus L. were measured at three locations in France (Lusignan, Dijon and Gotheron) and at Rothamsted, UK, in 1989/90 and 1990/91. Different combinations of sowing dates, plant densities and irrigation treatments were tested at some sites.Averaged over all sites, CH304/70 yielded more than Lunoble (3-26 v. 2-98 t/ha) but there were significant genotype x location interactions for yield and date of maturity. Both genotypes gave similar yields at the three locations in France (313 and 306 t/ha, respectively). But under the cool and wet conditions at Rothamsted, CH304/70 yielded 4-26 t/ha compared with only 202 t/ha for Lunoble. Yields of CH304/70 were more stable than those of Lunoble across locations and years; the coefficients of variation about the overall means were 14 and 22%, respectively. CH304/70 ripened up to 16 days sooner than Lunoble in France, and 6 weeks earlier than Lunoble in the cooler conditions of the UK.In wet growing conditions (Lusignan 1989/90, Rothamsted 1990/91) Lunoble had increased vegetative growth which decreased the proportion of yield produced by mainstem pods and correspondingly increased the proportion borne on second-order branches. This continued branching also delayed harvest, especially in the UK where the crop was not harvestable until early November. In these conditions, CH304/70 outyielded Lunoble at Lusignan (3-4 v. 2-7 t/ha) and at Rothamsted (4-3 v. 2-0 t/ha) and was harvested acceptably early in mid-September. However, under warm, dry conditions, Lunoble produced only one order of branches at Gotheron and two orders at Lusignan in 1990/91. At Gotheron, both genotypes matured at the same time but CH304/70 yielded slightly better (3-1 v. 2-8 t/ha) because it produced a higher proportion of its yield on the mainstem, whereas at Lusignan, Lunoble yielded better than CH304/70 (4.1 v. 3.4 t/ha) because of extra yield from pods on second-order branches.CH304/70 produced seeds of larger mean size and had greater stability of seed size across years and locations than Lunoble did. This was apparently the result of CH304/70's determinate structure, in which c. 75-100% of the yield was from pods on the mainstem and first-order branches, whose seeds were larger and less variable in size than those from pods on higher-order branches of indeterminates.The development of autumn-sown genotypes with a determinate structure that combine adequate yield and early harvest under cool conditions offers the prospect of extending the geographical range of lupins, grown for grain, both to the cool regions of northern Europe and to hotter and drier regions in southern Europe where summer drought limits the yield of conventional indeterminate types. 178

Determinate architecture is of great interest for lupin production in Europe but the extent of genetic and environmental variation in the structure and yield of determinate lupins must be evaluated . Forty-three genotypes of determinate autumn-sown white lupin were studied in four environments in France in 1991/92 and 1992/93 . Wide variation was observed for phenological (dates of mainstem and branch flowering, date of end of flowering and date of maturity), morphological (numbers of branch orders, branches and leaves) and seed characters (seed yield, mean seed weight, number of seeds/m 2 ) . Both genetic and environmental effects contributed to the observed variation . Heritabilities were high for all characters, and the interaction between genotype and location effects was low . There were significant positive correlations of flowering dates with vegetative plant development including the number of branches, number of leaves and number of branch orders . Seed yield and phenological and architectural characters were significantly correlated . Distribution of yield on the vegetative orders showed important variations according the genotype, and the variations were related to variation for the architecture . There are possibilities for yield improvement using morphological characters in determinate autumn-sown lupin .

One means of increasing seed yield in white lupin may be the modification of the harvest index in the reproductive compartment by a reduction of the proportion of pod walls. Genetic and environmental effects on the pod wall proportion and yield components were evaluated. Thirty-five genotypes of spring-sown material were sown in 6 different locations across France in 1996 and 1997, accounting for a total of 10 site × year combinations. The existence of a genetic variation for the pod wall proportion among studied genotypes was demonstrated (0.26–0.34). Highly significant genotypic differences for the pod wall proportion, seed number per pod, seed weight per pod, mean seed weight, flowering time, and seed yield were observed among lupin genotypes. The heritability of pod wall proportion was moderate, the phenotypic correlation between this character and seed yield was significant and negative, and the genetic correlation was high and negative. The environmental variance contributed a major part of the total variation. The genotype × environment effect for the pod wall proportion was small, which suggests that the selection of genotypes with low and stable pod wall proportion in different environments will be possible. The strong negative genetic correlation between pod wall proportion and seed yield supports the feasibility of using the character as a selection criterion for a higher seed yield.

The proportion of pod walls is considered as a potential selection criterion for seed yield in white lupin. It has a negative and highly signi®cant phenotypic relationship with the seed yield and the genetic correlation between both traits is negative and high. The present study over 2 years (1996 and 1998) evaluated the genetic variation of proportion of pod walls in a set of 325 spring-sown white lupin accessions originating from 17 dierent countries. Wide ranges of variation for this trait and other pod components were observed. This variation has a geographical distribution. Egyptian ecotypes had the lowest proportions of pod walls while the Greek and Italian ecotypes had the highest. The between-country range of variation for the proportion of pod walls, was lower than the within-country range (0.25±0.31 and 0.21±0.36, respectively) and the within-country variance accounted for 51% of the total genetic variation. The broad-sense heritabilities of pod wall proportion and proportion of pod walls for one full locule were moderate (0.44 and 0.57, respectively), which indicates that the proportion of pod walls can be used in breeding.