Although there were 23 species in the world about 16 million years ago, there are only two species now in Liriodendron(Parks et al., 1990; Wolf, 1985).One is Liriodendron chinense (Hemsl.)Sarg.and the other is Liriodendron tulipifera L, generally called yellow poplar in the eastern part of USA, and both are Vicariad species pairs (Wolfe, 1985).L. chinense is one of endangered species in China, which is sparsely distributed within 11 provinces.Recently L. chinense has attracted great attention due to its rapid growth and potential wood value.L. chinense has been listed as an extended species in NAP (National Afforestation Project) and its provenance trials have also been established since 1990.However, very few studies have been conducted on wood variation of L. chinense, while many studies have been reported (Harold et al., 1980; Taylor, 1979)in L. tulipifera.

The objectives of this study were to determine the radial variation pattern of basic density (BD) and fiber length (FL) of the plantation in L. chinense, the transition from juvenile to mature periods, the age-age correlation and the feasibility of earlier selection.These results were suggested to be useful for tree breeders and forest managers.

1 Materials and Methods 1.1 Materials

33 sample trees were randomly selected from a 17 years old L. chinense plantation, only including one provenance from Zhejiang Province, near the formerly studied 7 years old plantation with 15 provenances (Li Bin, 2001) in Shanxia Forestry Farm, Fenyi county, Jiangxi Province (27°34′41″N, 114°33′47″E, 250 m elevation).The 5 mm increment core samples were taken from north to south at the breast height.The height (H) and Diameter at breast height (DBH) of each tree were measured (Tab.1).

1.2 Method of Measurement and analysis 1.2.1 Measurement of basic density and fiber length

The ring basic density (RBD), ring fiber length (RFL) and growth ring width (GRW)were measured by conventional methods.The individual' s BD and FL at breast height were calculated by the weighted areas of the corresponding growth ring (Gu et al., 1994)(Tab.1).

1.2.2 Variation trend analysis of mean ageage correlation coefficients with age interval increase

Age-age correlation coefficients are always applied in the research on wood earlier selection and prediction (Wang et al., 2000), but the variation patterns of age-age correlation coefficients with the increase of age interval (the age interval distance from young to old) have not been reported (Zobel et al., 1989). Because the sample number of age-age correlation coefficients was different in the different age intervals, the variation of age-age correlation coefficients should be expressed by the average of those, namely the mean age-age correlation coefficients(MACC)for k age interval value(r_k).

Definition :i and j is the cambial age (CA), i, j =1, 2, 3, 4, 5 …n.Age interval (AI), k = i -j, i ≠j. The matrix of age-age correlation coefficient =r_{i, j}, 1 ≤i < j, i < j ≤n.

Then, MACC for k AI(r_k)can be given by:

The radial variation trend of RBD, RFL, GRW with CA was performed to describe the radial profiles, with diverse non-linear imitation methods provided by Excel 2000.The analysis of variance and regression was performed by SAS general model (SAS Institute, 1999).The demarcation analysis between juvenile and mature wood was conducted by Fisher cluster method (Lang et al., 1986).

2 Results and discussion

Variance analysis showed that both BD and FL varied significantly among trees and among years (F_BD & F_FL among trees were 11.83^** and 9.83^**.FBD & FFL among years were 73.91^** and 29.51^**).BD varied from 0.380 to 0.438 g·cm^-3 and FL varied from 1.433 to 1.787 mm among the 33 trees (Tab.1).Among years the varied range of BD and FL were 0.358 5 to 0.452 6 g·cm^-3, 1.279 6 to 1.820 1 mm, respectively.The CV of growth-ring basic density (RBD) and growth-ring fiber length (RFL)varied from 3.6 %~ 13.1 %, 4.7 %~ 15.8 %, respectively.There was a weakly negative relationship between BD and growth factors, and a weakly positive relationship between BD and FL. The relationship was weakly negative between FL and H, and positively significant between FL, DBH and V.

2.1 Radial profile

RBD for CA 1 was the lowest, followed by a rapid increase CA 2, and then gradually decreased till CA 5, followed by a slightly wavelike increase till the bark.Compared with its sister species, L. tulipifera, the RBD of the first year in L. chinense was much lower(Harold, 1980).This was attributed to the fact that the materials of L. chinense in this study came from plantation but L. tulipifera materials in Harold' s study came from natural stands since the pith in natural stands was much smaller than that from plantations.Hence the data for the first ring containing pith should be erased.The reestablished pattern of RBD (Fig. 1)appeared a "V" type curve(y =-1E -06x⁵ +7E -05x⁴ -0.001 4x³ +0.013 1x² -0.054 7x +0.478 7 R² =0.972 7).The result was basically in agreement with that of L. tulipifera (Harold, 1980).It was easily observed from Fig. 1 that RBD had begun steadily increased since CA 9 to CA 10, but the increase rate seemed not slow down.

RFL increased from pith to bark.It was abided by the index curve (Y =1.754 7 -0.52 ×0.75 ^x R² =0.926)(Fig. 2) and the result agreed with that of L. tulipifera (Taylor, 1979).The radial variation trend was very steady and the variation pattern kept unchanged whether we gave up the first year' s RFL or not.That was comprehensible because there were hardly any fibers measured in pith.It was visually obvious from Fig. 2 that RFL had kept basically steady and tended to be mature since CA 9 to CA 10.

The Fisher cluster analysis was conducted with three characteristics, including RBD, RFL and GRW.The result stated that the first 9 years were clustered in the first group and the late 6 years were in the second group.It indicated that the demarcation age was between the ninth year and the tenth year as a whole.However, the demarcation for different characteristics might be different.The juvenile age of fiber length would be younger than basic density' s in this study.To basic density, it was probably still in its juvenile period before CA 15, but it would be sure that CA 15, 17 years old if considering that the seedling of L. chinense generally took two years to reach at the breast height (1.3 m), was more than a half of cut age for the plantation of L. chinense.

The analysis result above stoned the base for age-age relationship analysis and for feasibility analysis of its earlier selection.

2.2 Feasibility analysis of early selection 2.2.1 Age-age correlation analysis and the feasibility of earlier selection

The correlation coefficients between different ages for BD and FL were listed in Tab.2. Result showed the age-age correlation coefficients of FL above CA 4 were all significant (α=0.05) and those of BD CA 5 were all significant as well, but the ages that correlation coefficients reached to significance at 0.01 level were more than CA 6 and CA 7 respectively for FL and BD (Tab.2).

Age-age relation analysis results above indicated that the earlier selection was feasible when the individual age was not younger than 5-year-old at breast height for basic density and fiber length in L. chinense(α=0.05).Sevenyear-old (α=0.01)would be more dependable for both of two characteristics (Tab.2).Considering that it generally took two years for the trees to grow more than 1.3 m in height, the tree age of earlier selection was at least seven, and more dependably nine.

2.2.2 Earlier selection feasibility of 7-year-old plantation with 15 provenances

To apply the result of earlier selection feasibility into 7-year-old provenance trial plantation, including 15 provenances(Li et al., 2001), the two groups' data of 7-year-old and 17-year-old during the same age range of CA 1 ~ 5 were compared and T-test.The result showed that therewere significant differences between two groups for both of BD and FL. For calling off the differences and further comparing the variation pattern, the two groups' data were respectively centralized.Then the regression analysis was conducted with the data at the same age, and the results showed that there was significant linear relation between the 7-year-old group and the 17-year-old group, and the regression coefficients were very near to 1 (Tab.3), which indicated that the two group data were basically on parallel and both of them were of the same variation trends.Hence, the result of feasibility analysis for earlier selection above was of important application value in L. chinense plantation.

2.2.3 The variation patterns of MACC with AI and earlier selection age

The variation trend of MACC with AI was very obvious, MACC decreased when AI increased (Fig. 3), which implicated that some genes, which controlled wood characteristics, might express under time sequences and the interaction of G ×E.The negative relationship between MACC and age interval indicated that the smaller age interval was, the larger MACC was.It would be easy to find the least age interval (LAI)which could make the MACC significantly (11 a for FL and 8 a for BD in this paper).The earlier selection age (ESA)wasn' t a constant value according to the LAI.We could calculate the ESA according to the difference between the objective age (OA) and LAI (ESA =OA -LAI).Supposing the OA was 17 a in this paper, the ESA would be 6 a (=17 a -11 a)for FL and 9 a (=17 a -8 a)for BD.

To valuate variation' s significance and its source, grouped the MACC with different starting age and different AI(Tab.4, Fig. 4).Variance analysis showed that there were also significant differences among groups and among AIs (Tab.5).The result indicated that MACC varied regularly with AI, and meanwhile MACC varied with different starting ages in an identical AI.The MACC was larger when the starting age was younger according to Fig. 4.This perhaps meant the selection result would be more dependable when selection age became older.

3 Conclusions

The radial variation pattern of BD showed a curve like "V" type when the 1^st year with pith was cut off, and the pattern was basically agreed with L. tulipifera(Harold, 1980).The radial variation pattern of fiber length in L. chinense displayed a general index curve, being similar with that of L. tulipifera (Taylor, 1979), and the pith did not produce obvious effect on FL.

The earlier selection was feasible as a whole when its cambial age was not younger than 5 years at breast height.

This paper illuminated the variation trend of MACC with age interval. There was a negative relationship between MACC and age interval, which was a important basis for deciding the earlier selection age.