B E Frankow-Lindberg, L Salomonsson and A Torstensson
Swedish University of Agricultural Sciences, Dept. Crop Production Science,
P.O. Box 7043, S-750 07 UPPSALA, Sweden
Introduction
Materials and methods
Results
Discussion and conclusions
References
Forage production in Sweden is limited by the number of forage species and the number of suitable cultivars within species that can withstand the harsh winter conditions. An increasing use of white clover would soon exhaust the limited supply of Swedish-bred white clover cultivars and hence increase the need for import of foreign seed material. It is then of vital importance to identify what clover characteristics confer good winter hardiness to a cultivar.
Spring yield (Harris, Rhodes & Mee, 1983) and annual clover yield (Collins, Glendining and Rhodes, 1991) have been found to correlate well with the amount of stolon present in spring. The present work analyses the effect of various stolon characteristics on spring yield of a selected number of white clover cultivars of different geographical origins by the use of principal component analysis (PCA) and partial least square regression (PLSR).
A field experiment was established in 1992 in Uppsala (59°49'N, 17°39'E) and data collected during autumn and spring in 1993/94. Sowing rates were 6 kg ha-1 of white clover + 14 kg ha-1 meadow fescue (cv. Svalöfs Sena). White clover cultivars were Sandra and Undrom (Sweden), AberCrest (Ac 50), AberHerald (Ac 51) and S 184 (UK), Milo (DK) and Grasslands Huia (NZ). The design of the experiment was a complete randomized block. The cultivars are classified as having leaves of an intermediate size, with the exception of AberCrest and S 184 which are small leaved.
Stolon mass (SM) and growing point numbers (terminal buds (TGP) and terminal + axillary buds (TAGP)) were determined from a number of circular turves (346 cm2 plot-1) taken from each replicate on 15 November (SMN, TGPN and TAGPN) and 11 April (SMA, TGPA and TAGPA). Stolon diameter in autumn (SDA) was measured on three randomly collected main stolons and total non-structural carbohydrates (TNC) were determined on randomly collected stolons from each replicate on 15 November (TNCN), 28 March (TNCM) and 28 April (TNCA). Node production in autumn (NPA) was determined from two marked main stolons in each plot. Relative change in some of the characteristics during winter and spring were calculated: LOSS1=SMA/SMN, LOSS2=TGPA/TGPN, LOSS3=TAGPA/TAGPN, LOSS4=TNCM/TNCN and CHANGE=TNCA/TNCM). Clover yield in spring (CLW) was determined by ground-level harvests (386 cm2 plot-1) on 30 May. The weather during the whole period was unusually cold, and a snow cover was present from the middle of November to the end of March.
PCA and PLSR (Martens & Næs, 1989) were performed in the software SIRIUS (Pattern Recognition System A/S; Bergen, Norway). All the variables were centered and scaled to equal variance before further analyses. The PLSR model was validated by cross-validation and the root mean square error of prediction (RMSEP) was calculated. Target rotated loadings (Kvalheim & Karstang, 1989) were calculated with one object (2Milo) excluded as it proved to be an out layer.
Milo had the highest spring yield, while Huia had the poorest yield (Table 1).
Table 1. Clover DM yield in spring
|
Cultivar |
Yield (g/m-2) |
|
Milo |
179.60 |
|
Undrom |
165.86 |
|
S 184 |
155.91 |
|
Sandra |
145.43 |
|
Ac 50 |
124.34 |
|
Ac 51 |
105.23 |
|
Huia |
78.52 |
|
S.E.(+/-) |
17.15 |
|
Significance |
** |
The replicates of the cultivars were more or less well grouped in the two component PCA (Figure 1).
At cross validation the first principal component explained 27% of the total variance. The second component explained 24% of all variance. Component 1 (Figure 2) was dominated by the TNC levels, LOSS1, LOSS3 and CLW. Component 2 was dominated by growing point numbers. This is in agreement with Undrom having had the highest, and Huia the lowest levels of TNC (data not shown) on all sampling occasions. The great distance between S 184 and Milo along component 2 points to their contrasting stolon morphological traits (data not shown).
The target rotated loadings from the PLSR model (Figure 3) reveals that CLW was most positively correlated to the TNC levels in the stolons, and especially so the levels in spring. NPA, and also more surprisingly, SMN were negatively correlated with CLW, while TGPN and TAGPN, SDA, LOSS2 and LOSS4 had little influence on CLW. The optimal PLSR model was obtained with but one component. The RMSEP value of the model was 1.1 compared to 1.5 for standard deviation of CLW, which means that the model was not very strong.
Figure 3.Target rotated loadings for the PLSR-model with one component.
High levels of TNC during the period from autumn to spring, proved the most important characteristic to ensure a good spring yield of clover, rather than stolon weight in spring, while morphological characteristics were of no importance. The Swedish cultivars together with S 184 showed a good ability to accumulate TNC during autumn. Huia and Ac 51 in particular continued to grow late into autumn, a characteristic detrimental to cold hardiness and overwintering (Eagles and Othman, 1981). High TNC levels in spring might imply that the plant retains a good resistance to late frosts even if it seems there is not a strong positive correlation between TNC and frost resistance (Sandli et al., 1993) However, the most important aspect in the Scandinavian environment is possibly the fact that the plant has ample reserves to sustain metabolic processes before a sufficient leaf area has been re-established in spring, since no green leaves survive the winter. Node production rates in spring were indeed slightly higher for the Swedish cultivars, during which period they lost stolon weight. Small relative losses of stolon weight and growing point numbers during winter were also associated with a good spring yield, but neither of these traits alone were good predictors of clover spring yield.
It is concluded that a suitable cultivar for Scandinavia must be able to accumulate sufficient amounts of TNC before the winter, and in order to do so should stop node production early in the autumn. The cultivar ought not to use its TNC reserves (and implicitly, lose neither stolon mass nor stolon length) during winter as they are needed in early spring to re-establish leaf area. The model obtained was not very strong, and this implies that some important characteristic was not measured. We suggest that more information about the role of frost resistance in spring is needed.
COLLINS, R. P., GLENDINING, M. J. and RHODES, I. (1991) The relationships between stolon characteristics, winter survival and annual yields in white clover (Trifolium repens L.). Grass and Forage Science, 46. 51-61.
EAGLES, C. F. and OTHMAN, 0. B. (1981) Growth at low temperature and cold hardiness in white clover. In: Wright, C. E. (ed.) Plant Physiology and Herbage Production. Occasional Symposium of the British Grassland Society, No. 13, pp. 109-113.
HARRIS, W., RHODES, I. and MEE, S. S. (1983) Observations on environmental and genotypic influences on the overwintering of white clover. Journal of applied Ecology, 20, 609-624.
KVALHEIM, O. M. and KARSTANG, T. V. (1989) Interpretation of latent-variable regression models. Chemometrics and Intelligent Laboratory Systems. 7, 39-51.
MARTENS, H. and NÆS, T. (1989) Multivariate Calibration. Chichester: John Wiley & Sons.
SANDLI, N., SVENNING, M. M., RØSNES, K. and JUNTTILA, O. (1993) Effect of nitrogen supply on frost resistance, nitrogen metabolism and carbohydrate content in white clover (Trifolium repens). Physiologia Plantarum, 88, 661-667.
