Euphytica 26(3): 703-708 (1977)
Breeding for improvement of flowering attributes of winterhardy Rosa kordesii Wulff hybrids
Felicitas Svejda


The results from tetraploid hybrids of R. kordesii showed that flowering attributes can be improved by a suitable combination of modifying factors as was shown previously by investigations with diploid hybrids from R. rugosa. A longer flowering period than 7 weeks indicated always the ability for recurrent bloom in these investigations but a shorter flowering period did not always indicate non-recurrent bloom.

The offspring from the cross of the recurrent R. kordesii with the non-recurrent seedling G12 segregated into non-recurrent and recurrent types in a ratio of 3:1 for a tetrasomic inheritance, assuming a duplex segregation and complete dominance.

The flowering attributes could be improved without loss of a high level of winterhardiness.


Recurrent flowering in roses is controlled by the recessive allele for non-recurrent flowering and it segregates in simple Mendelian ratios (SEMENIUK, 1971). A study of flowering attributes in the offspring from winterhardy and recurrent R. rugosa hybrids showed that the duration of the flowering period and flower productivity could be improved through breeding (SVEJDA, 1977). The present study analyses the flowering attributes in the offspring from winterhardy tetraploid R. kordesii hybrids and compares the results with previous results from diploid R. rugosa hybrids.


R. kordesii Wulff was used as a pistillate parent in combination with the two unnamed seedlings G12 and DO8. R. kordesii is a tetraploid that arose by spontaneous chromosome duplication of the R. rugosa x R. wichuraiana hybrid 'Max Graf' (Wulff, 1951). G12 was obtained from open pollination of 'Max Graf' and a cytological examination by Dr D. R. Sampson, of this Station, found it to be tetraploid (2n = 4x = 28). G12 differs from R. kordesii in that it is very hardy at Ottawa where it shows little or no winterkill. It flowers non-recurrently and produces fewer flowers. It has single, pink flowers like 'Max Graf'. R. kordesii is regularly killed to the snow-line at Ottawa, it flowers recurrently and is more floriferous than G12. Unlike 'Max Graf, R. kordesii has double flowers.

The duration of the flowering period was determined by examining each seedling weekly in the 16 week period from June till September. The amount of bloom was estimated as of the surface area covered. The ratings of amount of bloom and % winterkill were as described previously (SVEJDA, 1977).

As in the previous study, the observations on the total seedling populations were carried out for a 2-year period. The data in Table 1 and 2 were obtained from the second year, the data in Table 3 from the average of 2 years. The observations were continued for an additional 2-year period on 6 selected seedlings from each population. The selection criteria were flowering characters, hardiness, disease resistance and appearance of flowers and shrubs.

Table 1. Flowering period, from June till September, of F1 seedlings from the crosses of (A) R. kordesii x G12 and (B) R. kordesii x DO8.

Cross Flowering in weeks
  l-6   7-16   total  
  number % number % number %
A 73 82.0 16 18.0 89 100
B 34 64.2 19 35.8 53 100

Table 2. Amount of flowers in F1 seedlings from crosses of (A) R. kordesii x G12 and (B) R. kordesii x DO8.

  Average amount of flowers, in % coverage, during 16 weeks from June till September
Cross   1-5 6-12 13-25 26-50 51-75 76-100 total
A Number of
B Number of



Table 3. Hardiness levels of F1 seedlings from crosses of (A) R. kordesii x G12 and (B) R. kordesii x DO8.

    Average amount of winterkill in %
Cross   0 1-5 6-12 13-5 26-0 51-5 76-100 total
Number of
Number of

Table 4. Flower characters and winterhardiness levels of F1 seedlings from crosses of (A) R. kordesii x G12 and (B) R. kordesii x DO8.

Seedling Flowering period in weeks
from June till Sept.
Amount of flower
in % coverage
Winterkill in %
  mean range test1 mean range test1 mean range test1
A4 12.0 a 12 bc 11 b
A5 10.8 a 13 abc 11 b
A1 9.8 ab 17 abc 19 a
A3 6.0 c 25 ab 11 b
A6 6.0 c 9 bc 4 b
A2 5.5 c 17 abc 23 a
R. kordesii 7.3 bc 30 a 45 a
G12 5.5 c 10 bc 10 b
B4 12.0 l 11 m 8 m
B1 8.5 lmn 15 lm 7 m
B5 6.8 mn 13 m 11 m
B2 5.3 n 25 lm 14 m
B3 5.3 n 16 lm 5 m
B6 5.0 n 38 l 10 m
R. kordesii 7.3 lmn 30 l 45 l
DO8 10.3 lm 19 lm 14 m
1 Means followed by the same letter(s) are not significantly different at P ≤ 0.05 as determined by Duncan’s Multiple
Range Test.


Comparison of progenies. The χ2 for heterogeneity between the progenies from R. kordesii x G12 and R. kordesii x DO8 was 5.62 for the duration of the flowering period, 21.60 for flower productivity and 25.53 for winterkill. Such values would be obtained in less than 2% and in less than 0.1% and therefore, the distribution of flowering attributes and hardiness levels for the two progenies were considered significantly different (Table 1, 2 and 3).

Ratio of seedlings with non-recurrent and recurrent bloom. Assuming that G12 is heterozygous for 2 allels, the expected ratio between non-recurrent and recurrent types should be 3:1 for a tetrasomic inheritance, assuming a duplex segregation and complete dominance. The χ2 for goodness of fit between observed and expected ratios was 2.35. Since such a value would be obtained in 13% of repeated trials, the hypothesis was not rejected.

The observed ratio for the cross with DO8 cannot be explained on the same basis. Since both parents flowered recurrently all seedlings from this cross should have flowered recurrently. An examination of seedlings B2, B3 135 and B6 with a shorter flowering period than 7 weeks, showed that each seedling flowered recurrently in one or two of the four years of observation. This agrees with previous findings (SVEJDA, 1977) whereby a relatively short flowering period was not always indicative for non-recurrent bloom. On the other hand, an examination of the seedlings A2, A3 and A6 from the cross with G12 showed that only A6 flowered recurrently.

Correlation of flowering attributes. The correlation coefficients for the length of the flowering period and amount of flowers were 0.0306 for the cross with G12 and 0.2017 for the cross with DO8. These values are not significant.

Flowering period. The comparison of parental means alone showed significant differences between R. kordesii and DO8 at P = 0.05 but not between R. kordesii and G12. Simultaneous comparison of means from sister seedlings showed 2 seedlings from the cross with G12 with a significantly longer flowering period than R. kordesii (Table 4). One seedling from the cross with DO8. had a longer flowering period than DO8, the longer flowering parent but the difference was not significant (Table 4).

Flower productivity. The parental means were significantly different between R. kordesii and G12 at P = 0.05 but were not significantly different between R. kordesii and DO8. No seedling from either cross was significantly increased in flower production over R. kordesii (Table 4). Generally, the offspring from the cross with DO8 produced a greater proportion of floriferous individuals than the offspring from the cross with G12 (Table 2).

Hardiness level. The proportion of very hardy seedlings was higher in the seedling population from the cross R. kordesii x DO8 (Table 3). R. kordesii showed an average winterkill of 45% over the 4 year period from 1971 to 1974. Over the same time period G12 and DO8 showed an average of only 10 and 14% (Table 4). All but 2 of the selected seedlings were comparable in hardiness level to the hardy parent. The 2 seedlings were intermediate in hardiness level between R. kordesii and G12, the hardy parent (Table 4).


The present results are comparable to previous findings from crosses with diploid R. rugosa hybrids (SVEJDA, 1977). A flowering period of 7 weeks or longer was always indicative for recurrent bloom. A flowering period of less than 7 weeks was not always indicative for non‑recurrent bloom. In different seedling populations, the frequency of seedlings with a flowering period between 7 and 16 weeks varied. From the cross of the rugosa hybrids 'Schneezwerg' x 'Frau Dagmar Hartopp', 84% of the seedlings flowered between 7 and 16 weeks but from the cross of R. kordesii x DO8 only 36% and from the cross of R. kordesii x G12 only 18% were in this class. Since 35% of the seedling from rugosa flowered between 12 and 16 weeks, the selection level could be increased to 12 weeks. In the kordesii hybrids the selection level for a prolonged flowering period could not be raised as yet but this could be achieved through further breeding efforts. One seedling from each of the 2 seedling populations flowered for an average period of 12 weeks (Table 4).

The simultaneous improvement of both flowering attributes was more successful n the first generation from the cross of rugosa cultivars since the longest flowering seedlings produced also more flowers than the more floriferous parent. The greater range in flower production in the kordesii, 1 to 100% compared with 1 to 25% in the rugosa, indicated that further improvement of flower production should be more successful in the kordesii.

The improvement of flowering attributes would not be very useful without improvement or retention of a high hardiness level. The present garden roses combine a long flowering period with a high level of flower productivity but they lack winterhardiness in most parts of Canada. Both studies with tetraploid kordesii and with diploid rugosa hybrids showed that a long flowering period can be combined with a high level of hardiness.

In the beginning of this breeding endeavour it was not certain whether winterhardiness could be combined with a prolonged flowering season in one organism since the one attribute depends, among other things, on the cessation of growth while the other depends on continued growth. In hardy roses, dormancy and cessation of growth are controlled by endogenous factors but in tender roses they are induced by ambient factors. In an environment favourable for growth, tender and everblooming roses are evergreen and continue to flower throughout the year without visible loss of vigour. In spite of favourable conditions, hardy roses revert to the dormant state but they resume vigorous growth and flowering after a period of exposure to low temperatures.

Generally, the offspring from crosses of very hardy parents is also very hardy. The hardiness level of the offspring from crosses between a very hardy and a less hardy parent is variable. All seedlings from the cross of the rugosa hybrids were very hardy and showed little variation in hardiness level. The frequency of very hardy offspring, showing between 0 and 12%) winterkill was different for the 2 crosses with kordesii even though the hardiness levels of the hardy parents G12 and DO8 were comparable. G12 is, presumably, an autotetraploid of the R. rugosa x R. wichuraiana hybrid Max Graf'. It can be assumed that it derived its hardiness from the rugosa ancestor. DO8 is a complex hybrid. Presumable, it derived its hardiness from the ancestor 'Suzanne', a hybrid between the very hardy species R. spinosissima x R. laxa.