Introduction

Worldwide, forests absorb approximately two billion tonnes of carbon dioxide (CO2) each year (Köhl et al. 2015) and are vital for offsetting atmospheric concentrations of this gas (Dugan et al. 2018). Many governments and non-governmental organizations (NGOs) worldwide pledged to plant more trees, for example, Government of Malaysia’s 100 million tree planting campaign to run between 2021 and 2025 (Ketsa 2021), and the Trillion Tree Campaign by NGOs proposed at the global level with the support from the United Nations (PFP 2021). Such a concerted effort proposes to use trees as a natural climate solution to restore ecosystems by absorbing carbon in the era of rapid global change (Bastin et al. 2019). However, tree planting projects involve substantial costs, both financially and in terms of time. A primary consideration in large-scale planting is seed source. Specifically, in term of genetic composition which has implications for the success and long-term viability of planting (Jalonen et al. 2018). Tree planting projects typically require 40 to 50 years to achieve the desired level of carbon sequestered or until harvesting for timber is viable. Careful planning at the project level is crucial to avoid serious losses in the long-term. Some difficult experiences show that planting with exotic species can disrupt and threaten local, often unique biological diverse ecosystems (for example, South Africa; Bennett and Kruger 2015). Similarly, following the Indian Ocean tsunami in 2004, several projects were initiated to re-plant mangroves in Sri Lanka. However, monitoring 23 restoration sites, more than 75% had < 10% seedling survival because of poor planning and/or poor site maintenance (Kodikara et al. 2017). In the Malaysian state of Sabah, Acacia mangium Willd. used as a planting source was found to have a narrow genetic base, leading to the establishment of seedlings with restricted genetic diversity and limited adaptive potential (Sim 1984). Therefore, the decision-making process when deciding on species selection requires careful consideration as they carry long-term implications for ecosystem function and to ensure that returns to stakeholders can be realized.

Shorea leprosula Miq., locally known as meranti tembaga, is widely distributed throughout Peninsular Malaysia, Sumatra and Borneo (Symington 2004) and commonly found on well-drained soils in dipterocarp forests up to about 600 m a.s.l. It is a large buttressed species with a tall, well-shaped bole that is regularly fissured. The main features to identify this species are its light copper-coloured, wide, umbrella-shaped crown. S. leprosula is one of the main sources of timber classified as ‘light red meranti’ (Desch 1941). Several studies have been carried out to understand the reproductive biology, genetics, silviculture and pests of these species (Chan and Appanah 1980; Appanah and Weinland 1993; Lee et al. 2000; Ng et al. 2006; Ong et al. 2020). Among species considered locally as ‘red meranti’, its growth rate is the highest for the first 20 years with an average annual diameter increment of approximately 1.2 cm. A 70-cm diameter tree can be achieved in 60 years under favourable conditions (Appanah and Weinland 1993). Widiyatno et al. (2014) showed that a 6-year-old S. leprosula can reach a 14-cm diameter and 8.3-m height with more than 70% survival rate in a logged-over tropical Indonesian rainforest (Widiyatno et al. 2014). The fast growth rate of the species could make it a potential choice for use in reforestation and afforestation, which in turn could reduce the pressure of extracting timber from natural forests. A possible problem of monoculture plantations is that they are more prone to insect damage than mixed stands (Jactel and Brockerhoff 2007). In the present study of S. leprosula, there were pest problems on planted trees, the most serious was scale insect infestation (Pedroniopsis spp.), which had a negative impact on plant health and growth. Pest and disease outbreaks can cause economic losses as shown by the Eucalyptus longhorned beetle (Duarte et al. 2020) and ash dieback in Sitka spruce plantations (Tuffen and Grogan 2019). However, naturally occurring genetic resistance may be present in the species and could offer an alternative to the use of chemicals or other costly pest management solutions (Sniezko and Koch 2017).

The widespread distribution of S. leprosula and its high level of genetic diversity promises sufficient genetic variability for seedling selection and tree improvement (Lee et al. 2000). The distribution of genetic variability of this species for adaptive traits can be assessed in common garden experiment. A common garden experiment is a trial design of growing individuals from different localities in a common environment. It has been used as provenance tests to monitor genetic variations for traits of economic and ecological relevance within a set of natural populations (Nabais et al. 2018). The benefit of a common garden experiment is that individuals from different geographic origins will share the same environmental conditions such as sunlight, rainfall, temperature, and soil. Therefore, any average differences among populations will be caused by genetic factors. This enables the study of the genetic basis of complex traits such as life history, morphological, and physiological traits using quantitative genetic tools (de Villemereuil et al. 2016). Common garden experiment can be used to test for signals of local adaptation in traits of interest such as wood properties (Mori et al. 2019), phenology (McKown et al. 2014), and leaf economic spectrum (Ramírez-Valiente et al. 2017). Local adaptation is defined as “native genotypes in each population which have, on average, a higher relative fitness in their local environment than foreign genotypes originating from other environments” (Kawecki and Ebert 2004). Results from studies of this type could be used for decision-making for species or provenance selection for maximising productivity and quality.

To provide insights into traits related to growth performance, S. leprosula seedlings from nine different geographical origins were established at our common garden experimental site at the Bukit Hari, Forest Research Institute Malaysia campus. Specifically, (1) differences in populations growth rates based on height and root collar diameter were described under common environmental conditions; and (2) the effect of scale insect infestation was assessed among different populations. The results could be integrated with genomic and ecological data for future adaptive genetic variation studies. This will help to understand a species evolutionary potential and to use prediction models to predict how species will respond to changing climate conditions across their distribution range (Waldvogel et al. 2020a).

Materials and methods

Establishment of the common garden

In 2013, S. leprosula seedlings were collected from nine naturally growing populations of different forest reserves (Gunung Jerai, Bukit Enggang, Bukit Kinta, Semangkok, Panti, Lentang, Beserah, Hulu Terengganu and Lebir; Table 1). Climate data at each population site was recorded: annual mean temperature and precipitation (Table 1). Seedlings were kept in the nursery with shade netting before exposure to full sunlight. In September 2014, 40 seedlings per population were transferred to a common garden plot in Bukit Hari within the Forest Research Institute Malaysia campus. This area has a typical tropical climate with mostly constant temperature throughout the year; annual mean temperatures and precipitation are 27.4 °C and 2202.5 mm, respectively.

Table 1 Location and climate for the populations of origin
Full size table

The experiment was a randomized complete block design arranged in 40 blocks and of a total of 15 rows (X) × 24 lines (Y); nine seedlings from different populations were randomly assigned within each block (Appendix: Fig. S1). Seedlings were planted at 5 m × 5 m spacing in a 85 m × 130 m area (total area = 11,050 m2). Some individuals were replaced during the period 2014 to 2019 due to mortality and were excluded from the analysis. A buffer zone of S. leprosula (seedlings collected from five mother trees in Semangkok population) was established surrounding the common garden to reduce any edge effects (Murcia 1995). During the first two years, A. mangium was planted between the S. leprosula seedlings as nurse plants, as their canopy decreases light and temperature and increases humidity (Padilla and Pugnaire 2006). In the third year, the nurse plants were removed and the S. leprosula seedlings were exposed to full sunlight.

Measurement of traits

Three years after planting, February 2017, measurements were taken for all seedlings (N = 360 plants). Root collar diameter (D) at 5 cm above the ground was measured using a digital calliper (Digimatic Calliper CD-8″ ASX, Mitutoyo, Kanagawa, Japan) and height (H) measured using a measuring pole (A-8, Senshin Industry, Osaka, Japan). The measurements were repeated in 2018 and 2019. Data for the nine populations were analysed in three steps: (1) comparison of a sample median using box plots; (2) incorporation of spatial distribution information into the growth data of each individual seedling; and, (3) comparison of relative growth rate data between 2017‒2018 and 2018‒2019 using box plots.

Scale insect infestation survey

While surveying the seedlings, branch dieback was observed on some individuals (Appendix: Fig. S2). This was caused by an infestation of Pedroniopsis sp. scale insects (Hemiptera: Sternorrhyncha) (Ong et al. 2020). To evaluate the impact of the infestation on growth, a field survey was carried out in 2018 to count the number of scale insects found on each tree. A pesticide of Alika ZC (12.62% w/w Thiamethoxam + 9.49% w/w Lambda-cyhalothrin; Syngenta) was applied to control the infestation after the survey, and a second survey was conducted the following year. The infestation was not uniform among the branches of an individual tree, and therefore, three branches per seedling with the highest density of infestation were selected for examination. For each branch, a 30-cm area which had the highest density of scale insects was selected and the number of scale insects recorded (Appendix: Fig. S3). The total number of scale insects from the branches was divided by three and this average used for box plot comparison. Subsequently, the spatial distribution information was incorporated into the number of scales insects of each individual seedling.

Statistical analysis

To evaluate the growth performance of individual seedlings, the relative growth rate (RGR) in biometrics X, such as collar diameter (D) and height (H), was calculated as follows:

$$RGR=\frac{ln{X}_{2}-ln{X}_{1}}{{t}_{2}-{t}_{1}}$$
(1)

where, X1 and X2 represent X at time t1 and t2, respectively.

To test the differences in RGR of D or H among nine populations, a generalized additive mixed model (GAMM) was employed, considering the effects of scale insect infestation, measurements repeated in each individual, and spatial auto-correlation, where the effects of population, initial size, scale insect density, measurement year and interactions among populations were treated as fixed effects and individual tree (as Tree individual ID) and spatial coordination (as planting row X and line Y) were random effects. Spatial coordination was incorporated into the model as a smooth term based on a polynomial function of X and Y. To conduct the analysis using GAMM, the gamm function in the package of Mixed GAM Computation Vehicle (mgcv) ver. 1.8‒35 (Wood 2019) was employed with an open source statistical software R ver. 3.6.1 (R 2019). Model selection was conducted according to Akaike Information Criterion (AIC) using dredge function in the package of Multi-model Inference (MuMIn) ver. 1.43.17 (Barton 2019). Multiple comparison was done using the results of the gamm function and General Linear Hypotheses (glht) function in the package of multcomp ver. 1.4‒10 (Hothorn et al. 2016). ANOVA was conducted with the results of GAMM.

Results

Root collar diameter and height

Average root collar diameters at 5-cm above the ground and seedling heights were variable among the populations but were not obvious because the inter-quartile range of diameter or height specific to the population overlapped with each other in each period (Fig. 1). The differences between the median maximum and minimum values for diameter among the nine populations were 10.7 mm (2017), 12.5 mm (2018) and 8.2 mm (2019). Similar minor differences were also observed between the median maximum and minimum values for height, recorded as 62.0 cm (2017), 78.0 cm (2018) and 62 cm (2019). When the spatial distribution data were incorporated into the growth measurements, diameter and height tended to be larger as row X became closer to zero in each period (Fig. 2).

Fig. 1
figure 1

Box plots for root collar diameter (D) and height (H) for seedlings of each of the nine populations (P1–P9) in 2017, 2018 and 2019. Minimum, maximum, sample median, and first and third quartiles are shown with outliers

Full size image
Fig. 2
figure 2

Spatial distribution of root collar diameter (D, mm) and height (H, cm) in 2017, 2018 and 2019; the contour lines denote smoothed values estimated using polynomial functions. X and Y are row and line numbers of seedlings, respectively

Full size image

For relative growth rate of root collar diameter (RGRD), the highest median value was recorded in the Hulu Terengganu population (P8) (median = 0.49) for 2017‒2018 and for the Beserah population (P7) (median = 0.33) for 2018‒2019 (Fig. 3). However, differences among populations were minor in each measurement period and the differences between the median maximum and minimum values were 0.10 (2017–2018) and 0.17 (2018–2019). In addition, the inter-quartile range of RGRD overlapped for each population in each measurement period. For relative height growth rate (RGRH), the differences between the median maximum and minimum values were minor and noted as 0.10 (2017–2018) and 0.10 (2018–2019). The inter-quartile range of RGRH specific to population also overlapped among the populations in each period. The highest median value of RGRH was in the Beserah population (P7) for 2017‒2018 (median = 0.28) and 2018‒2019 (median = 0.34), (Fig. 3).

Fig. 3
figure 3

Box plots showing relative growth rate (RGR) of root collar diameter (D) and height (H) for 2017–2018 and 2018–2019; the minimum, maximum, sample median, and first and third quartiles are shown with outliers

Full size image

Scale insect infestation

For scale insect infestation, the Beserah (P7) (median = 0.67) and Semangkok (P4) (median = 0.67) populations had the lowest median value for number of infestations in 2018 and 2019 (Fig. 4). The scale insect infestation was more serious from line number Y1 to Y10 (number of scale insects > 8) in 2018 (Fig. 5). However, the infestation was reduced slightly in 2019 and plants with > 8 scale insects were found only at the lower left corner of the plot (row numbers X1 to X3 and line numbers Y1 to Y5).

Fig. 4
figure 4

Box plots for the number of scale insects (30 cm–1 per branch) among the nine populations (P1–P9) in 2018 and 2019, where the minimum, maximum, sample median, and first and third quartiles are shown with outliers

Full size image
Fig. 5
figure 5

Spatial distribution of the number of scale insects (30 cm–1 per branch) in 2018 and 2019. The contour lines denote smoothed values estimated using polynomial functions. X and Y represent row and line numbers of planted trees, respectively

Full size image

Model selection

Root collar diameter

Model selection based on AIC values (‒335.2) suggested that the best model for evaluating RGRD involved the year of measurement (Year) and the initial root collar diameter (Initial D) as fixed effects without the interaction between Year and Initial D (Table 2). The random effects as ‘spatial effects’ (as a smooth term based on polynomial function of row X and line Y) and ‘individual trees’ were measured repeatedly. The first year was set as reference point and the standard deviation of random effect of individual trees was 2.90 × 10–6 among individuals and 0.177 within an individual. The smooth term was significant (p < 0.001) with the estimated degree of freedom (edf) as 6.21. Overall adjusted coefficient of determination R2* was 0.252 (n = 574). AIC of the best model was ‒335.2 and lower than ‒304.9 in the full model and ‒463.2 in the null model. According to ANOVA, the effect of Year was not significant (p = 0.138) whereas Initial D showed a significant effect (p < 0.05). As summarized in Table 2, the coefficient for Initial D was negative, considering its 95% confidence interval (95%), e.g., 95% confidence interval did not involve zero.

Table 2 The best model based on generalized additive mixed model (GAMM) for relative growth rate as measured by root collar diameter, RGRD; values in parentheses are 95% confidence intervals
Full size table

Height

The best model (AIC value = ‒902.2) for evaluating RGRH involved the population (Population), measurement year (Year) and initial height (Initial H) as fixed effects without the interaction between Population and Year (Table 3). The random ‘spatial effects’, (as a smooth term based on polynomial functions of row X and line Y), and ‘individual trees’ were measured repeatedly. The first year was set as reference point and the standard deviation of random effect of individual trees was 0.0369 among individuals and 0.0984 within an individual. The smooth term was significant (p < 0.001) with an estimated degree of freedom (edf) of 14.0. Overall adjusted coefficient of determination R2*was 0.262 (n = 574). AIC of the best model was ‒902.2 and lower than ‒871.8 in the full model, and ‒817.9 in the null model. The effects of Population, Initial H and Year were significant (p < 0.05).

Table 3 The best model based on the generalized additive mixed model (GAMM) for relative growth rate as measured by height RGRH; values in parentheses are 95% confidence intervals
Full size table

In Table 3, Initial H showed a significantly negative effect and Year a significantly positive one compared to the first measurement year, i.e., the second measurement year showed significantly higher RGRH than the first measurement year. Bonferroni correction was used to control for multiple comparisons of coefficients in the effect of Population among nine populations (Fig. 6a) revealed that the Beserah population (P7) had the highest coefficient for RGRH, and was significantly higher than that of the Bukit Enggang (P2), Semangkok (P4), and Lebir (P9) populations.

Fig. 6
figure 6

Comparison of (a) the means of relative height growth RGRH (yr‒1) and (b) means of the number of scale insects (30 cm‒1 branch) among nine populations (P1–P9) using the coefficients of the best fit model. Error bar denotes 95% confidence interval. Different symbols (α, β, γ) between two populations indicates significant difference between them (Bonferroni post-hoc test, p < 0.05)

Full size image

Scale insect infestation

According to the model selection, the best model (AIC value = 3684) for evaluating scale insect infestation involved Population and Year as fixed effects without the interaction between Population and Year (Table 4). The random ‘spatial effects’ (as a smooth term based on polynomial functions of row X and line Y), and ‘individual trees’ were measured repeatedly. The first year was set as a reference point and the standard deviation of the random effect of ‘individual’ tree was 0.920 among individuals and 5.768 within an individual. The smooth term was significant (p < 0.001) with an estimated degree of freedom (edf) of 2.0. Overall adjusted coefficient of determination R2* was 0.128 (n = 574). AIC of the best model was 3684 and lower than 3686 in the full model and 3722 in the null model. The effects of Population and Year were significant (p < 0.05).

Table 4 Results of generalized additive mixed model (GAMM) for scale insect infestation; values in parentheses are 95% confidence intervals
Full size table

As summarized in Table 4, Year showed a significantly negative effect with the first measurement year as a reference point, i.e., the second measurement year showed significantly less scale insect infestation than the first year. Bonferroni correction controlled for multiple comparisons of coefficients in the effect of P among nine populations (Fig. 6b), revealing that the Beserah population (P7) had the lowest coefficient for scale insect infestation and showed significantly less infestation than the Bukit Kinta (P3) and Lebir (P9) populations.

Discussion

Plant growth is influenced by genetics, environment, and by interconnected processes such as resource acquisition (light) and resource utilization efficiencies (leaf architecture) during development (Nunes-Nesi et al. 2016). A large array of variation in plants is found because every individual is genetically unique, except for clonal replicates, and there are infinite environmental variations. Natural environments in the tropical forest are highly heterogeneous and therefore some of the variations observed among populations in this study might be due to environment rather than inherited genetics. There was an attempt to control this heterogeneous factor by taking first year measurements as a baseline in the model analysis. By growing S. leprosula in a common garden experiment plot, an assumption was made that environmentally induced variations in phenotype may be diluted by the sharing of a common growing environment. The design of the current common garden study was to determine how much of the variations in phenotypes, namely height, root collar diameter and scale insect infestation was due to genetic differences.

Growth data was not recorded in the first two years after planting. This was to provide ample time for the seedlings to acclimatize for stable and reliable measurement in the measurement years 2017‒2019. Spatial distribution of root collar diameter and height showed that S. leprosula in rows X1 to X4 consistently outperformed those in rows X5 to X15 (Fig. 2). This might be due to heterogeneity of factors such as water, nutrients, and light availability. The control of this heterogeneity was attempted by utilising 40 replicates in the randomized complete block design of the common garden. The spatial distribution of scale insect density indicated that the infestation was widespread in seedlings in lines Y1 to Y10 in 2018. This could be due to an initial outbreak of infestation starting from individual trees from these lines. After the recording was completed, a pesticide was applied to control the infestation. The efficacy of the pesticide was realized the following year (2019) and the infestation was reduced.

Root collar diameters and heights were indicators of growth performance. Differences in these indicators were not generally apparent, based on the differences between the median maximum and minimum values among the nine populations (Fig. 1). This might be attributed to the juvenile age of the plants over five years at the site (2014‒2019). Comparing the relative growth rate between 2017 and 2018, the Hulu Terengganu (P8) and Beserah (P7) populations had the highest median RGRD and RGRH values although these differences were not significant among all the populations. Some significant differences were found in RGRH when multiple comparisons of coefficients were conducted in the modelling analysis (Fig. 6a). The Beserah (P7) population had the highest coefficient for RGRH, significantly higher than that of the Bukit Enggang (P2), Semangkok (P4), and Lebir (P9) populations but not significantly different from that of the other populations. In contrast, there was no significant difference for RGRD in modelling analysis. Nevertheless, for 2018‒2019, the Beserah (P7) population had the highest median for both RGRD and RGRH (Fig. 3). This might indicate an allometric relationship of root collar diameter and height to describe the strategies of the species to gain maximum light resources (Iida et al. 2011; Osunkoya et al. 2007). The performance of RGRH in the Beserah (P7) population indicates that some individuals might have specific genetic predisposition controlling growth. To discover and understand the growth traits in these individuals, genome wide association studies (GWAS) could be conducted by utilizing the present common garden experimental site. Several GWAS studies on growth traits have been successfully conducted with other forest genera, namely Eucalyptus (Resende et al. 2017), Pinus (Lu et al. 2017) and Populus (Du et al. 2016). Once the underlying genetic architecture of the traits has been unravelled, indispensable information may be provided for marker-assisted selection in dipterocarp breeding programmes.

In addition to the highest median RGRH in the Beserah (P7) population, this population also had the lowest infestation of scale insect. Infestation on the trees led to branch dieback in the upper part of the crowns and impaired growth performance. The scale insect was Pedroniopsis sp., with the specific name unclear (Ong et al. 2020). A low-density scale insect infestation in the Beserah (P7) population indicates that individuals might exhibit natural genetic resistance. This could encompass a wide range of responses, from partial to complete resistance. The biology of host-insect interaction at different points throughout the insect life cycle should be fully understood in order to develop screening methods to identify host plant resistance (Sniezko and Koch 2017). Resistance can be manifested in different phenotypic traits such as antixenosis (non-preference) that repel insect herbivory, antibiotic traits to reduce insect survival, and tolerance to withstand or recover from insect damage (Smith and Clement 2012). The insect infestation of S. leprosula could be also pleiotropic such as multiple disease resistance in the light of other biotic stresses that potentially impacted the common garden experiment in future (Capador-Barreto et al. 2021). Similar to growth performance, there is the potential for GWAS to identify associations between genetic markers and resistance traits across populations. Selection of desired resistance phenotypes through markers could be achieved at a relatively young age and therefore reduce both cost and time in conventional breeding programs. Evidence of genetic resistance to insects in forest species have been reported for Sitka spruce (Picea sitchensis (Bong.) Carr.) with resistance against white pine weevil (Pissodes strobi Peck; Alfaro et al. 2013) and American beech (Fagus grandifolia Ehrh.) with resistance to the scale insect (Cryptococcus fagisuga Lind.) (Koch and Carey 2014). The presence of a genetic resistance trait is an advantage for forest manager. It may provide an alternative to the use of chemicals or other costly management methods that might be detrimental to the environment.

The findings from this study could be integrated with genomic and ecological data to study adaptive genetic variation in future. For instance, eco-evolutionary models of adaptation to climate changes that consider life-history traits and genetic characteristics of a species, can predict how a species might respond to changing environments and inform policymakers and conservation strategies (Waldvogel et al. 2020a). By incorporating genetic-estimated adaptive values in modelling, studies have shown reduced range loss projections and the importance of identifying the climate adaptive potential of populations, as well as an increase of landscape connectivity between populations to enable the spread of adaptive genetic variation (Razgour et al. 2019; Sebestyén et al. 2021).

Conversely, both growth and pest tolerance traits presumably exhibit a polygenetic basis and strong genotype by environment (G × E) interactions (Guevara-Escudero et al. 2021). Therefore, it is important to estimate trait heritability scores and G × E interactions in future studies. It is insufficient to rely solely on provenance differences found at the original source in Malaysia. Heritability scores describe the likelihood that offspring will replicate the parental phenotype and the G × E interactions reflect the adaptability of species to factors such as temperature, rainfall, soil, and pests. These genetic parameters are important for selection strategies and for genetic gain in tree improvement and breeding programs. It is also imperative to contrast the results from the common garden trial with expectations from in situ environment associations using predictive genomics approaches which rely on genomic and phenotypic data (Cortés et al. 2020). Predictive genomics offers a feasible way to predict adaptation from allele frequencies in many genes of low effects underlying polygenic traits such as growth adaptive responses in this study (Cortés et al. 2020).

Future studies will explore more quality traits such as crown width and leaf parameters. Models based on infinitesimal additive algorithms will be utilized with the assumption of a more complex polygenic adaptive architecture in which all loci are liable to improve the modelling of the adaptive response (Sebestyén et al. 2021). However, efforts to identify useful variations in species through common garden trials remains a major challenge for identifying complex traits due to a complex inheritance of abiotic stress tolerant phenotypes involving many genes of low effects and several environmental interactions (Guevara-Escudero et al. 2021). This problem could be overcome by applying modern analytical approaches such as genomic prediction and machine learning to speed up more accurate pre-breeding efforts to increase in adaptability and yields (Guevara-Escudero et al. 2021). Data of genotype-environment associations could be deposited in a standardized, structured format to enable meta-analyses across multiple species to generate insight in species adaptability to environmental variables, especially climate (Waldvogel et al. 2020b). Simultaneously, reciprocal transplant experiments could be carried out to incorporate estimates of G × E interactions in addition to provenance effects. This can improve understanding of local adaptation in the species by comparing traits between environments to determine whether there is a genetic component to the differences (Sedlacek et al. 2015). Presently, it is difficult to concisely interpret pre-adaptation at the phase of antagonistic biotic interactions because the pest infestation was not homogenous in the common garden (Fig. 5) and there was widespread growth heterogeneity within the field (Fig. 2). The subsequent establishment of common garden at different sites will allow comparison of results and a more conclusive interpretation.

It is recommended that forest managers use local provenances or match the climates of seed collection and planting zones when neither long-term tests are available. The common garden of S. leprosula is a long-term experiment and the results will be collected over a period of 10 to 20 years, and alternative options for planting may be identified. There are three options for forest managers to decide on the planting zone. First, they may choose one or several provenances to be used if no provenance differences were found. Second, they may introduce a few provenances on all planting sites if large provenance differences in the absence of G × E interactions were found. Third, forest managers need to ensure that the provenances are matched to edaphoclimatic regions in the zone if large provenance differences and large G × E interactions were found. The current common garden experiment may serve for future genetic assisted gene flow (AGF) or assisted migration strategies for the adaption of populations to new climatic conditions (Aitken and Whitlock 2013; Flanagan et al. 2018). An AGF strategy is defined as “the managed movement of seed sources between populations within the historical range of species to mitigate local maladaptation in the short- and long-term” (Fig. 1 in Aitken and Bemmels 2016). Once the long-term, field-based provenance trial data is available for the species from this study, AGF strategies can integrate information on provenance productivity and other traits with climate change projection (Aitken and Bemmels 2016). Assisted gene flow can compensate for climate change and also increase productivity and carbon sequestration, as well as expanding planting areas for exceptionally productive provenances (Aitken and Bemmels 2016). Conversely, assisted migration is moving seed material to a target climate niche outside the historical range where they are anticipated to be well-adapted with expected future warming.

Conclusions

This study reports on the growth performance of nine populations of S. leprosula from different geographical origins under controlled, common garden conditions. Preliminary results show that the Beserah (P7) population had the highest height growth and also the lowest scale insect infestation. The underlying complex traits of the association between the genotype and phenotype in the Beserah population should be further explored. This common garden experimental site is the first reported for dipterocarp species and it can serve as a platform to unravel the genetic basis of complex quantitative traits in dipterocarps through genome-wide association studies.