Euphytica 28 (1979) 309-314
Contribution No. 557
Agriculture Canada, Ottawa Research Station, Ottawa, Ontario. Canada
Received 18 May 1978


Heritability in the broad sense and the distribution of levels of winterkill were analyzed in the offspring from hardy diploid and from hardy, semi-hardy and tender tetraploid roses. The heritabilities for different parental combinations ranged from 51 to 92%.

The offspring from crosses of hardy parents were also hardy and showed little variation in hardiness levels. The offspring from crosses of hardy roses with the semi-hardy R. kordesii and the tender 'Queen Elizabeth' survived the winters without coverage but showed a wider variation in hardiness levels.

The desirable level of hardiness, an average winterkill of less than 10%, could be achieved through selection in the first or second generation of breeding, depending on the hardiness levels of the parents. The hypothesis is advanced that winterhardiness in roses is controlled by very few or closely linked genetic factors.


The aim of rose breeding for Canada is the combination of a high level of winterhardiness and disease resistance with a long flowering season and appealing flowers and shrubs. Lack of winterhardiness restricts the cultivation of contemporary garden roses. Without protection, they can be grown along the coast of British Columbia and in the most southern region of the Niagara Peninsula. Only the hardier garden cultivars survive the winters at Ottawa under earth mounds. Generally, hardy species and cultivars have a short flowering season and lack flower quality.

The combination of high levels of winterhardiness and disease resistance with a long flowering season and appealing flowers and shrubs can be achieved only through hybridization of distantly related species and cultivars. Breeding programs were carried out at the diploid and tetraploid levels since winterhardy and disease resistant roses and also roses with a long flowering season and appealing flowers are available at both ploidy levels (SVEJDA, 1975).

Until the present study, no information on the inheritance of winterkill in roses was available. This study was undertaken in support of the breeding program.


The origin of the parents is shown in Table 1. Seven of the nine parents were derived from the present breeding program. Besides hardiness, selection criteria for the parents were: crossability, duration of the flowering period, flower productivity, resistance to mildew, Sphacroiheca pannos WALL. ex. FR./LÉV., and blackspot, Diplocarpon rosae WOLF., and appeal of flowers and shrubs. Cross combinations were carried out at the diploid and tetraploid levels.

Table 1 Origin of parents.

Name Plant
'Martin Frobisher' F06 open pollination of R. rugosa hybrid 'Schneezwerg'
seedling F42 seedling F07, from open pollination of R. rugosa hybrid 'Schneezwerg'
seedling F44 seedling F10, from open pollination of R. rugosa hybrid 'Schneezwerg'
seedling G15 open pollination of R. rugosa hybrid 'Will Alderman'
'Queen Elizabeth' T18 'Charlotte Armstrong' x 'Floradora'
R. kordesii WULFF. K01 R. rugosa x R. wichuraiana hybrid 'Max Graf’
seedling G49 seedling G12, from open pollination of R. rugosa x R. wichuraiana hybrid 'Max Graf'
seedling D08 open pollination of G45, seedling from 'Red Dawn' x 'Suzanne'
seedling L10 R. kordesii WULFF. x G12

The diploid parents were derived from the hardy R. rugosa hybrids 'Schneezwerg' and 'Will Alderman'. They were very hardy (Table 2). 'Martin Frobisher', code F06, (SVEJDA, 1969), and the unnamed seedlings F44 and F42 were derived from 'Schneezwerg'. The seedling G15 was derived from 'Will Alderman' (Table 1). Four cross combinations were carried out: F44 x G 15, F42 x G 15, F06 x F44 and F06 x F42 (Table 4).

Table 2. Hardiness levels of roses used as parents.

Plant code Number of
years tested
Mean winterkill
F44 9 3 a
F42 7 8 ab
F06, 'Martin Frobisher' 9 8 ab
G15 6 16 b
G49 5 4 a
D08 6 16b
L10 8 16 b
K01, R. kordesii WULFF. 8 43 c
T18 'Queen Elizabeth’ 2 90
1 Mean separation, within columns, by Duncan's multiple range test, 5% level.
2 Survived under earth mounds only.

The tetraploid parents were of different hybrid origin and ranged in hardiness levels from very hardy to tender (Table 1 and 2). R. kordesii WULFF., and the maternal parent of G 12 were derived from unreduced gametes of the R. rugosa x R. wichuraiana hybrid 'Max Graf' (WULFF, 1951; SVEJDA, 1977). The seedling G49 was derived from open pollinations of G12. It is improved in hardiness, blackspot resistance and duration of the flowering period over G12. L10 was derived from the cross R. kordesii WULFF. X G12 (Table 1). Thus, R. kordesii WULFF., G49 and L10 are related to each other through the common ancestor 'Max Graf'. The hardiness levels of these roses differs significantly (Table 2). The seedling D08, presumably, derived its hardiness from the ancestor Suzanne, a hybrid between the very hardy species R. spinosissima x R. laxa. The grandiflora 'Queen Elizabeth' is tender and survived under earth mounds only. Cross combinations were carried out between the hardy parents G49 X L10, between the semi-hardy R. kordesii and the hardy G49 and D08, and between the tender 'Queen Elizabeth' and the hardy D08 (Table 4).

The hardiness levels of the hardy and semi-hardy parents were obtained from 5 to 9 year trials of genetic stock. Distributions of winterkill of the parental generation and hybrid populations were obtained from 2-year trials. The classes for the ratings are indicated in Table 3 and 4. All roses, except the tender 'Queen Elizabeth' were grown without coverage, in fields exposed to the prevailing winds.

Heritabilities of winterkill were calculated from the results of 2-year trials of hybrids and parents during the second and third year of growth. The trials included different numbers of individual hybrid seedlings and two clones of each parent. The analyses of

Table 3. Distribution of winterkill in parental generation.

Number of
1 2 3 4 5 6 7
    Winterkill (%)
    0 1-5 6-12 13-25 26-50 51-75 76-100
'Schneezwerg' 47 7 10(F06)2
13 15(F07) I I 0
F07 40 2(F42) 9 10 12 7 0 0
F10 13 2(F44) 8 3 0 0 0 0
‘Will Alderman' 11 0 2 7 2(G15) 0 0 0
G12 49 36(G49) 11 2 0 0 0 0
G45 10 1 3 2 4(D08) 0 0 0
K01 x G12 46 1 6 15(L10) 13 11 0 0
1 Seedlings derived from open pollination, except where cross combination is shown.
2 The origin of parents is indicated by the corresponding plant codes, in brackets, following numerical distribution in
different classes of winterkill.

Table 4. Distribution of winterkill in hybrid populations of roses.

Number of
1 2 3 4 5 6 7
    Winterkill (%)
    0 1-5 6-12 13-25 26-50 51-75 76-100
F44 x G15 35 17 15 3 0 0 0 0
F42 x G15 112 69 41 2 0 0 0 0
F06 x F44 15 4 10 0 1 0 0 0
F06 x F42 44 15 20 5 3 1 0 0
G49 x L10 35 9 19 4 2 1 0 0
K01 x G49 55 0 18 19 17 1 0 0
K01 x D08 57 3 22   13 12 6 1 0
T18 x D08 71 0 1   1 8 18 39 4

Table 5. Analysis of variance of hybrids arid parents from one cross.

Source of variation Degrees of freedom
Hybrids h-1
Years y-1
Hybrids vs. years (h-1) (y-1)
Parents p-1
Hybrids vs. parents 1
(Hybrids vs. parents in years 1
Residual 5

 variance for each hybrid population, including parents, were carried out as shown in Table 5. Heritabilities in the broad sense, where the phenotypic variation is envisaged as the sum of the genetic and environmental variations, s2G + s2E, were calculated after the formula:

where M1 mean square (MS) for hybrids, M2 = MS for hybrid versus years, M3 = MS for residual variation.

Heritabilities of winterkill were not calculated for the parental generation because all parents except L10, were derived from open pollination.


Hardiness level of parents. The wood of hardy parents showed an average winterkill between 3 and 16%, over a test period from 5 to 9 years. The wood of the semi-hardy R. kordesii showed an average winterkill of 43%, over a test period of 8 years. The wood of the tender 'Queen Elizabeth' showed an average of 90% winterkill over a two-year period under earth mounds (Table 2). this shows clearly that tender roses barely survive in the Ottawa area. For this reason, the tests are carried out mainly with hardy and semi-hardy roses.

In spite of varied constellations of environmental factors, the wood of very hardy roses, such as F44 and G49 showed consistently low rates of winterkill in different years. Less hardy roses, such as G15, D08 and L10 showed greater variations. The amount of winterkill for F44 and G49 varied between 0 and 10% but for G15, DOS and Lb, the amount of winterkill varied between 3 and 50%. The winterkill for the semihardy R. kordesii ranged from 10 to 75%.

The protective snow cover varies inversely with the height of the shrub and might influence the assessment of hardiness levels. G49, D08 and 'Martin Frobisher' have shrubs from 1.5 to 2.0 min height. F44, F42, and G15 5 have dwarf shrubs from 0.5 to 0.75 min height. R. kordesii and L 10 have trailing canes, generally well protected by the snow cover. Over the duration of the test, the differences in survival rates could not be explained by differences in height. G49 showed less winterkill than 'Martin Frobisher' and D08. Similarly, F44 showed less winterkill than F42 and G15 (Table 2). L10 was hardier than R. kordesii but comparable in growth habit and height.

The results from the trials with 2- and 3-year old plants and the results from the trials of genetic stock with more mature plants were comparable but discrepancies occurred (Tables 2 and 6). In the 2-year trials, no differences in hardiness levels between G49 and L 10 and between R. kordesii and D08 were indicated (Table 6) while the trials of genetic stock showed significant differences (Table 2). The lack of sensitivity of the 2-year trials might be explained by differences in growth habit and height. The trailing growth habit of R. kordesii and L10 might have ensured a complete snow cover while the higher shrubs of D08 and G49 were largely exposed to frost and prevailing winds. The 2-year test showed a significant difference in hardiness levels between R. kordesii and G49 (Table 6).

Table 6. Heritability of winterkill in roses.

Significance level
of difference between
vs. years
parents hybrids
F44 x G15 P<0.05 n.s. 0.64 0.36 0.50 -
F42 x G15 P<0.05 n.s. 0.24 0.02 0.30 -
F06 x F44 P<0.05 n.s. 0.56 0.26 0.30 -
F06 x F42 n.s. P<0.01 2.06 0.85 0.10 92
G49 x L10 n.s. P<0.05 0.98 0.64 0.12 74
K01 x G49 P<0.05 n.s. 0.63 0.53 070 -
K01 x D08 n.s. P<0.05 1.71 1.37 0.33 51
T18 x D08 P<0.01 P<0.05 1.58 0.85 0.28 72

The method of rating emphasized the differences between hardy plants, decisive advantage for the separation of very hardy from hardy individuals.

Hardiness levels of hybrids. The offspring from crosses of hardy roses were also hardy and showed little variation in winterkill between sister seedlings while the offspring from crosses of the semi-hardy KO l with hardy roses, and from the cross of the tender T18 with a hardy rose, were less hardy and showed a wider range of variation in winterkill between sister seedlings (Table 4). All seedlings from the cross T18 x D08 survived the winters without protection and about 14% appeared to be as hardy as the hardy parent (Tables 2 and 4). As might be expected, the offspring from the crosses of the hardy D08 with the semi-hardy KO 1 and the tender T 18 showed significant differences in hardiness levels between sister seedlings (Tables 4 and 6). A comparison of the offspring from K01 x D08 with T18 x D08 and from G49 x L10 with K01 x G49 showed greatly improved hardiness levels in the offsprings from hardier parents, namely K01 x D08 and G49 x L10 (Tables 4 and 6).

In certain combinations of hardy parents, the hardiness levels of sister seedlings was significantly different, such as for the crosses F06 x F42 and G49 x L10 (Table 6).

Heritability of winterkill. Calculations of heritabilities in the broad sense (Table 6), indicated that between 51 and 95% of the variation in different hybrid populations was due to genotypic variations (Table 4). These findings were substantiated by the consistency of results from 2-year test results (Table 3) with prolonged test results (Table 2). Three parents were selected from the hardiest classes of sister seedlings: F42, F44 and G49 (Table 3). Each proved to be very hardy in further tests (Table 2). The slightly less hardy parents G15, D08 and L10 retained their position in relation to the very hardy parents in later tests. Thus, 2-year tests are adequate for the selection of hardy seedlings.

The hardiness level can be improved in successive generations of breeding. The mean winterkill of seedlings from the crosses K01 x G12, K01 x G49 and G49 x L10 were 25, 10 and 4% respectively (Table 3 and 4). The parental means were 25% for K01 and G12, 23% for K01 and G49 and 10% for G49 and L10. Further studies are required to ascertain dominance or epistatic effects of genes.


Depending on the hardiness levels of the parents, very hardy offspring, with less than 10% mean winterkill, can be obtained in one to three generations of breeding. This suggests that winter-hardiness in roses is controlled by very few major genes or closely linked genetic factors. The lack of variation in hardiness levels among the offspring from hardy parents of different genetic origin at the diploid and the tetraploid level supports this hypothesis.