Study on the variation in and selection of Fraxinus mandshurica provenances and families in northeast China

Fraxinus mandshurica Rupr. is one of the main afforestation species in northeast China, and there is great demand for improved F. mandshurica varieties. The results of an investigation into and analysis of the growth traits of F. mandshurica provenances and families showed that there were significant differences in different periods. However, variations in growth traits decreased year by year. There was a significant negative correlation between tree heights of the provenances and sunshine hours in their areas of origin. The provenances of Xinglong, Hailin and Wuchang were selected based on the volume of 18-year-old trees. The average genetic gain from the selection of fast-growing provenances was 19.4%. Ten superior fast-growing families were selected. The average volume of the selected families was 22.6%, higher than that of all families. The correlation coefficient between heights at 6-year-old and at 18a was 0.838 for provenances, and between heights at 4-year-old and at 18-year-old was 0.303 for families. These results indicate that early selection for height in provenances or families could be performed at 6 years or 4 years, respectively. Early selection for DBH and volume in families could start at 8 years.


Introduction
Fraxinus mandshurica, a deciduous species of Fraxinus in the Oleaceae family, is one of the main afforestation species in northeast Asia (Yuan et al. 2021). It is one of the three hardwood species in northeast China together with Phellodendron amurense Rupr. and Juglans mandshurica Maxim. It is an excellent and valuable fast-growing, adaptable species with high economic value. F. mandshurica is widely found in northeast Asia but with a discontinuous distribution; its central distribution is in northeast China. Natural F. mandshurica often grows in sparse hillside forests or on gently sloping mountain valleys at altitudes of 700-2100 m. It is one of the component species in Korean pine/ broadleaved forests (Xie et al. 2020).
The environmental pressure of natural selection can change the gene frequency of a population, resulting in different geographical variations among the same species (Petrova et al. 2017;Zhang et al. 2020). Due to the limitations of the propagation range of trees, the further apart the provenances are, the less genetic communication, so geographic variation is widespread (Steiner et al. 2019). Through provenance tests, the geographical variations in forest trees may be explored (Lindgren et al. 2008;Pakharkova et al. 2014), and the dominant factors influencing their genetic differentiation (Klisz et al.2019). At the same time, provenance testing can provide the basis for the protection and utilization of natural resources (Calvo et al. 2015; Abstract Fraxinus mandshurica Rupr. is one of the main afforestation species in northeast China, and there is great demand for improved F. mandshurica varieties. The results of an investigation into and analysis of the growth traits of F. mandshurica provenances and families showed that there were significant differences in different periods. However, variations in growth traits decreased year by year. There was a significant negative correlation between tree heights of the provenances and sunshine hours in their areas of origin. The provenances of Xinglong, Hailin and Wuchang were selected based on the volume of 18-year-old trees. The average genetic gain from the selection of fast-growing provenances was 19.4%. Ten superior fast-growing families were selected. The average volume of the selected families was 22.6%, higher than that of all families. The correlation coefficient between heights at 6-year-old and at 18a was 0.838 for provenances, and between heights at 4-year-old and at 18-year-old was 0.303 for families. These results indicate that early selection for height in provenances or families could be performed at 6 years or 4 years, respectively. Early 520 J. Hao et al. Gilbero et al. 2019;Bogunović et al. 2020), seed division (Buras et al. 2020), and the selection of high-yield varieties in various regions (Tsuyama et al. 2020).

3
Because of its range of distribution, F. mandshurica is mainly studied in China, Japan, and South Korea. These studies have focused on physiology (Koike et al. 2001;Ichie et al. 2002;Yang et al. 2021;Yu et al. 2020), ecology, and management (Matsuda et al. 2002). Research on F. mandshurica breeding has been performed mainly in China (He et al. 2021;Zhao et al. 2021b). After studies of cultivation and management strategies have reached a certain point, the issue preventing F. mandshurica from playing a larger role in afforestation is the quality of the material. As one of the main tree species in northeast China, there has been some progress in provenance testing (Xie 2005). However, there is a lag between the selection of improved varieties and their promotion. Early selection can help put improved varieties into production as soon as possible.

Study area
The experimental forest is in the National Improved Seed Base of Larix spp. in Linkou County, Heilongjiang province (130°57′ E 45°4′ N), on the eastern slope of the Zhangguangcai ridge. The terrain is mainly hilly, with some low mountains. The slope of the test field is between 11° and 25°, and the field is higher in the south. The annual average temperature is 2.5 °C, the accumulated temperature greater than 5 °C is 2732.5 °C, and the average temperature in July is 21.2 °C. The average altitude is 400 m, the annual precipitation 520.1 mm, the annual sunshine hours 2615.9 h, and the annual evaporation 1266.1 mm. Soils are mainly dark brown, with a small amount of meadow soil and swamp soil.

Data collection and analysis
Data from the seedling stage (1-year-old) were collected in 1998. After planting, growth was measured annually in 2000 (3-years) to 2005 (7-years), and in 2007 (10-years), 2008 (11-years), 2011 (14-years), and in 2015 (18-years). Metersticks, tower rulers and ultrasonic height meters were used to measure heights and diameter at breast height (DBH). Volume was calculated by the experimental form factor: where the average experimental shape number for F. mandshurica is f ϶ = 0.40 (Zhao et al. 2021a), g 1.3 is the crosssectional area at breast height, and H is the height.
Family heritability (Xu 2006) was determined by: where h 2 f is the additive genetic variance component between families, 2 fb the block variance, 2 e the error variance component, b the block number, and n is the number of plants per block.
The genetic gain formula was: where S is the selection differential, X the overall average value, and h 2 is the heritability of the trait. Duncan's multiple-range test was used for multiple comparisons. Pearson and Spearman correlation analyses were used for phenotypic correlation and rank correlation, respectively. IBM SPSS statistics 23 and the DPS data processing system were used for data statistics and analysis (Tang 2009).

Analysis of provenance differences
The parameters of the variation in growth traits at different ages were statistically analyzed ( Table 1). The coefficient of variation (CV) of height decreased year by year. At 18-years of age, the CV of volume was the largest, at 80.8%, the CV (3) ΔG = h 2 S X of DBH was 39.7%, and that of height was the smallest (31.1%). The results of the variance analysis of the growth traits of the 15 provenances showed that, at different growth stages, the differences in different growth traits among provenances reached a significant (P < 0.05) or very significant (P < 0.01) level. These results indicated that the growth of F. mandshurica provenances varied greatly at different ages and that provenance selection could achieve good results.

Geographical variation among the different provenances
Correlation between height and provenance longitude/latitude and meteorological data showed that at one year of age, height had a significant negative correlation with the longitude of the provenance but no significant correlation with the other data ( Fig. 1A). At 3 years of age, there was a significant negative correlation between height and average temperature in January but no significant correlation with the other data. From 6 years of age onward, there was a significant negative correlation between height and sunshine hours, i.e., the shorter the sunshine hours in the original area, the better the performance of the population at the test site.

Height variation among the different provenances
The mean height and CV for the different provenances at different ages are shown in Table 2. In terms of height, the average height of 1-year-old seedlings was 4.2 cm, and the average height of 18-year-old trees increased to 712.6 cm. Before the age of 8 years, the XL provenance was consistently the tallest at 67.8%, 31.8%, 35.2%, 38.1%, and 29.1% taller than the overall averages at 1-, 3-, 6-, 7-and 8-years of age, respectively. At 14-years of age, the average height of the HL provenance was 419.8 cm, only 1.4 cm (3.4%) higher than that of the XL provenance. At 18-years of age, the HL provenance had the highest average height, 840.90 cm, and the height of the XL provenance was the second highest. The average height of the HL provenance was 17.1 cm (2.1%) higher than that of the XL provenance, 18.0% higher than that of the experimental forest, and 46.4% higher than that of the LSH provenance, which was the shortest. The provenances of HAN, DFH and LSH exhibited a declining performance throughout the study period. The coefficients of variation indicate that the overall variation in the height of the experimental forest gradually decreases. At 1-year of age, the CV was 63.5%. HL had the largest CV (63.1%), and HR the smallest CV (25.6%). By the age of 18, the overall CV decreased to 31.1%. The LJ provenance had the largest variation, with a CV of 40.2%. The TWH provenance had the smallest variation (17.6%). The experimental forest exhibited great variation, indicating that appropriate provenance selection has the potential to result in substantial production gains.
There was no significant correlation between the height of the 1-year-old seedlings and heights at each age in the later growth periods (Fig. 1B). After the age of six, correlations between heights at different ages became significant, indicating that early selection could be carried out.

Cluster analysis of height and evaluation of provenance stability
Based on the cluster analysis of the heights of the 15 provenances at different ages, the provenances were divided into three types: fast, medium and slow growth (Fig. 1C). In the different ages, the proportion of fast-growing provenances was small, ranging from 6.7% to 33.3%, with an average of 15.0%. These results indicate that fast-growing provenances selected at different ages would generate higher gains. From 1-year-old to 18-years old, the XL provenance was always categorized as fast-growing, while the HL and WC provenances were also relatively fast growing.
The results of the statistical analysis show that the provenances of HL and XL exhibited excellent performance in Linkou County and the surrounding areas. The HL provenance performed well between the age of 3 to 8; its performance was not significantly different from that of the XL provenance which was ranked first. At 14-years and 18-years, the HL provenance ranked first. The height at 18-years of age was 840.9 cm, 18.0% higher than the overall average. The XL provenance showed excellent performance from 1-year to 18-years of age; it attained a height of 823.8 cm by the age of 18, which was 15.6% higher than the overall average. These two provenances are suitable for this region. The ranking of the WC provenance for height was stable and slightly lower than those of the first two, and is also suitable for this region. The provenance ranking of TWH fluctuated during the study period the variation  within the provenance is low, and the height tends to be stable; these factors should be considered in the selection. The results of the yield and stability analyses show that the HL provenance was superior and that the XL, WC and DL provenances were good (Table 3). The provenances with poor performances included DFH, LSH, HAN, LJ, WQ and others. These provenances consistently exhibited lower than average performance, indicating that they are not suitable for the local environment and for afforestation.

Fast-growing provenance selection and gain evaluation
Provenance selection with timber as the main objective should give priority to volume. The average heights, DBH and volumes of the 18-year-old experimental provenance forest were 7.2 m, 5.0 cm and 1.05 × 10 −2 m 3 , respectively. When the selection rate was 20%, three provenances, XL, HL and WC, were selected. The XL provenance had the best performance, with an average height of 8.5 m, an average DBH of 6.0 cm, and an average volume of 1.47 × 10 −2 m 3 , which were 17.0%, 18.7% and 39.5% higher than those of the experimental forest, respectively. The average height, DBH, and volume of the three provenances were 8.3 m, 5.7 cm, and 1.36 × 10 −2 m 3 , respectively. Compared with the mixed provenances, the genetic gains of the three provenances were 10.6%, 7.9%, and 19.4%, respectively (Table 4).

Analysis of family variation and the selection of superior families
Fifty families belonging to 3 provenances were considered: 13 families from the FZ provenance, 28 from the WC provenance and 9 from the XL provenance. There were no significant differences in growth traits among the three provenances, but there were among the families in the FZ and WC provenances. Only height was significantly different among the XL families (Table 5). There were significant differences among the 50 families at various ages (Fig. 2). Based on the volume data, at age 18 the families were ranked. When the selection rate was 10%, the genetic gain in volume was 38.8%. The top five families were WC25, FZ17, WC21, XL8, and FZ16. Their average volume was 1.28 × 10 −2 m 3 , 30.7% higher than the overall average. The WC25 family ranked first for volume, and its ranking was consistent for many years. At 18-years, the volume of WC25 was 1.45 × 10 −2 m 3 , 48.3% higher than the overall average; the CV was 57.9%, which was lower than the average. The volume of the FZ17 family was 1.38 × 10 −2 m 3 , second to that of WC25. The ranking of FZ17 for volume was relatively high for several years, and its variation was lower than the average. The WC21 and FZ16 families ranked third and fifth, respectively, and their ranking fluctuated over the years. The coefficients of variation for these families were higher than the overall average, indicating that their stability was slightly low. The XL8 family ranked fourth, and its ranking showed an upward trend for several years, from 28th at 10 years to fourth. These results indicate that this family has good growth potential. When the selection rate was 20%, the top 10 families were selected, i.e., the top five families as well as FZ13, FZ4, WC19, WC30 and XL1. Their average volume was 1.20 × 10 −2 m 3 , 22.6% higher than the overall average, and the genetic gain was 31.3%. The WC7, WC1, WC9, WC3 and FZ14 families showed poor performance and their rankings for several years were consistently poor. They can be eliminated from consideration to reduce operational and selection costs and to improve selection efficiency.

Early-late correlation analysis and family selection
Correlation analysis showed that family rankings remained consistent in the different ages. Starting at 5-years of age, the correlation among the families reached a significant level, and the correlation coefficient was > 0.62 (Fig. 3). This indicates that from the age of 5 years, the family ranking remains relatively stable, and early selection could be carried out. The phenotypic correlations among heights in different ages showed that there were significant positive correlations between heights at 18-years of age and that in other years. At 4-years of age, the inter-year correlations of height in different families reached a significant level in each year, and the correlation coefficient between heights at 4 years and 18 years was 0.303. At 5-years of age, the inter-year correlations of height became very significant, and the correlation coefficient between heights at 5 and 18 years was 0.639. This shows that early selection for height can be started at 4 years but the correlation improves significantly by 5 years and the rank correlation is also very significant at this age. Selection at 5 years will be more accurate than earlier selection. This finding is consistent with the results of the tests of early selection for the provenances.
The trend for diameters was like that for volume. Because most seedlings did not reach 1.3 m height before 8 years of age, the analysis of DBH and volume started at 8 years. The results show that DBH and volume at 8 years had significant positive correlations in the following years, which indicates that early selection could be performed at 8 years. With increasing age, the correlation between the volume in each year and that at 18 years became stronger, which indicates that the later the selection time, the higher the accuracy of selection. Therefore, the intensity of early selection should not be too high. Considering the stability of the family rankings, early selection should be performed mainly to eliminate families showing poor performance.

Discussion
Variation among provenances is common (Risk et al. 2021). The performance of breeding materials from different sources is different from that at experimental sites because the trees are adapted to the climate and environment of their production areas (Memišević Hodžić et al. 2020). This phenomenon in tree breeding is interpreted as the interaction of genes and the environment (Mwase et al. 2008;Escobar-Sandoval et al. 2021). The results of many provenance trials showed that (the same test materials exhibited significant differences in growth performance among different test sites (Yang et al. 2020;Gao et al. 2021;Smolnikar et al. 2021). This shows that a single provenance test cannot be used to determine that some provenances are good and others are inferior. Test results from local or adjacent areas should be used as the main reference for practical applications.
In this study, the growth traits of F. mandshurica were significantly different among provenances and families. The coefficients of variation decreased year by year, and the CV between families was always lower than between provenances. The 50 families in the experimental forest were from three provenances: XL had the best performance in terms of volume and good stability; WC had high growth and moderate variations; and FZ had poor growth and large variations. There were significant differences among the 50 families but none among families within the same provenance. Among the 10 families selected, there were four from the FZ provenance, four from the WC provenance and two from the XL provenance. Based on the number of families tested, 30.8%, 22.2%, and 14.3% of the FZ, XL, and WC families were selected, respectively (Fig. 4). From the results of the provenance test, the FZ provenance did not perform as well as WC and XL regarding volume and stability. However, because of its large variation and the notable family differentiation within the provenance, superior families may occur within the FZ provenance. If the 50 families were classified into five ranks, the proportions of "good" and "bad" families within the FZ provenance would be larger (30.8%) than those in the other families. In the provenance test, XL provenances with superior performance also had good stability. The number of superior families in the XL provenance was twice the number of poor families, but most exhibited mediocre performance. The performance of WC provenances was more ordinary, with similar proportions in the top 10 and the bottom 10 (approximately 15%), and more families with moderate performance. The results show that the performance of various provenances in trials cannot be used to indicate whether the families in the provenance will perform well or poorly, that is, families in superior provenances are not necessarily better than those in average provenances, so separate provenance tests and family tests need to be carried out (Kang 2019).
Tree breeding is long, and the value of its final products often cannot be determined until the forest has been growing for decades (Moreira et al. 2020). Early selection can shorten the breeding cycle and improve breeding efficiency (Mihai et al. 2019). The results of the correlation analysis for height showed that early selection of provenances could be initiated from the age of six and that the early selection of families could be carried out at 4 years. This is because the variation within families is less than within provenances, and the performance of families is more stable among the different ages. With increasing tree age, the correlation between each year and later years gradually becomes stronger, indicating that the later the selection time, the higher the accuracy of selection (Lu et al. 2016). This study suggests that the volume of F. mandshurica families can be selected at 8 years. Zhao et al. (2015) also suggested that the selection time of F. mandshurica should not be too early. Due to the long-term nature of forestry and the different growth periods of different strains (Waldy et al. 2021), long-term followup observations on experimental forests is necessary and early research results constantly revised. At present, there are relatively few methods of early selection used in forestry, and most use year-on-year correlations for analysis. There are also some studies that use model prediction (Xu et al. 1995) or greenhouse studies to reflect traits at maturity (Luo et al. 2020). However, the former approach needs large amounts of tracking data for the model to develop reliable observations; the latter is too different from actual afforestation environments, which makes it challenging to obtain selection results by "matching the species with the site". The family rank correlation in this study reached a very significant level, indicating that family ranking remained relatively stable throughout the 18 years of growth. This stability was observed for both the top-and bottom-ranked families. Therefore, in early selection, the focus should be to eliminate families with slow growth. Selection should not be carried out only once in a rotation period but should be carried out in steps to narrow the genetic basis and thereby avoid misjudgments or omissions.

Conclusions
There were significant differences in growth traits among 15 provenances and 50 families of F. mandshurica at different ages. The variation in growth traits decreased year by year. There was no significant geographic variation in the height growth of provenances, but there was a significant negative correlation between height growth and the number of sunshine hours at the provenance locations. Three fastgrowing provenances suitable for afforestation in the region were selected based on their volume at 18-years of age, and an average genetic gain of 19.4% was achieved. These provenances were XL, HL and WC. Ten superior families, WC25, FZ17, WC21, XL8, FZ16, FZ13, FZ4, WC19, WC30 and XL1, were selected. Their average volume exceeded the overall average by 22.6%, and WC25 and FZ17 had the best performance. Provenance and family selection should be carried out independent of each other. Early-late correlation analysis showed that selection for height by provenance and family could be started at 6-and 4-years of age, respectively, and that early selection for DBH and volume could be started at 8 years.
Acknowledgements This work was supported by the staff of Qingshan Larch National Improved Seed Base in Linkou County.
Funding Promotion project of State Forestry Administration (2017) 06, China.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.