SOIL RESISTANCE AND MULTIVARIATE ANALYSIS AS AN AUXILIARY METHOD FOR SELECTING Calophyllum brasiliense PROGENIES ANÁLISES MULTIVARIADAS COMO UM MÉTODO AUXILIAR PARA A SELEÇÃO DE PROGÊNIES DE Calophyllum brasiliense

– Precision silviculture is being developed to manage and improve Calophyllum brasiliense , a tree species associated with wet and swampy soils. This study estimated genetic parameters for growth traits in response to soil resistance as an auxiliary tool for identifying and selecting progenies adapted to water-saturated soils. This was undertaken in a progeny test of C. brasiliense in which 1,200 seedlings grown from seeds collected from a natural population were planted in a randomized complete block design for single-tree plots. Genetic statistical analysis was conducted using the REML/BLUP method. Signiﬁ cant diﬀ erences ( p <0.01) in diameter at breast height were observed among the progenies. The coeﬃ cients of genetic variation and heritability at the individual and progeny levels were low for diameter at breast height and height, indicating low genetic control for these traits, while high positive and signiﬁ cant genetic and phenotypic correlations were detected between diameter at breast height and height. Theoretical genetic gains and multivariate statistical analysis indicated three groups of progenies with diﬀ erent degrees of tolerance and adaptability to ﬂ ooded soils, which could be useful in future breeding programs for this species, although further analysis at advanced ages is still required. The mean heights of tolerant and sensitive progenies were 12 and 58%, respectively, lower than the theoretical values, indicating that constant soil ﬂ ooding is harmful to this species.


INTRODUCTION
Forest breeding programs present specifi c challenges resulting from the perennial, crosspollinated, and long-lived nature of tree species and their genetic variability in response to environmental eff ects. Precision forestry has emerged to assist in forest management with geospatial information and specifi c deployments for individual production sites to develop forest products with greater embedded technology and aggregated value (Kovácsová and Antalová, 2010).
Calophyllum brasiliense (also known as guanandi) is a monoecious/hermaphrodite openpollinated (mainly by bees and small insects) neotropical Calophyllaceae species that naturally occurs from Central to South America, in areas from sea level to 1,200 m with mean annual rainfall of 1,100-3,000 mm and temperatures from 18 to 27 ºC (Carvalho, 2003;Wrege et al., 2017). The species is often distributed in watersheds (Oliveira-Filho and Ratter, 1995;Mendonça et al., 2014), and seeds and plants remain viable and tolerant to fl ooding even if submerged (Marques and Joly, 2000). Although it occurs in riparian zones, C. brasiliense exhibits good development in well-drained soils (Silva et al., 2018b;Devide et al., 2019), and is sensitive to drought (Campelo et al., 2015;Rocha et al., 2016).
Large swaths of the Brazilian coast (along stretches of the BR-277 highway from Morretes to Paranaguá, for example, as well as in the municipality of Antonina, all in the state of Paraná) are subject to periodic fl ooding, with negative eff ects on plant development. Although C. brasiliense is known to be tolerant to up to 120 days of fl ooding (Oliveira and Joly 2010), persistent hydromorphic conditions can cause growth damage even in this species.
Measuring soil resistance (SR) can identify compaction layers and/or microspatial fl ooding in soils, and together with geospatial techniques can be useful to avoid damage and ensure plant survival and development. Moreover, when combined with assessment of genetic parameters, SR is an innovative tool for identifying and selecting progenies of fl oodtolerant species like C. brasiliense that are better adapted to saturated soils.

MATERIALS AND METHODS
Seeds were collected from 30 open-pollinated trees from a natural but highly disturbed remnant population of C. brasiliense in Pontal do Paraná, Brazil ( Figure 1) in areas naturally subjected to seasonal elevation of the water table, and the seedlings were produced at the Embrapa Forestry greenhouse in Colombo, Paraná. The progeny test was conducted at the Embrapa Forestry Experiment Station in Morretes, on the coast of Paraná state (25°26'59'' S, 48°52'07'' W, 21 m). The experimental plot had previously been used for decades in irrigated rice trials, making it the perfect setting to test fl ood tolerance.
Uneven plant development was observed throughout the blocks; this had previously been attributed to the presence of some physically restrictive layer in the soil resulting from sedimentary origins, construction rubble deposited from neighboring roads, or topsoil removal when used for irrigated rice trials. Soil samples were collected from diff erent depths to evaluate compaction layers, but varying drainage conditions during rainy periods revealed signifi cant excess humidity in the soil (evidenced by a gleying process) that could be correlated with low development in some plots. Soil resistance was subsequently measured with a penetrograph device (PenetroLog, Falker) from 0 to 60 cm to investigate the eff ects of soil resistance and soil water saturation on plant development.
Five years after the trial began (2019), the surviving trees were georeferenced (Garmin GPSmap 76CSx) and their circumference at breast height (CBH) and height (H) were recorded. The CBH value (cm) and height (m) were obtained using a graduated tape measure; CBH was transformed into diameter at breast height (DBH).
Values for variance components and genetic parameters were estimated according to the REML/ BLUP method, using unbalanced H and DBH data and Selegen-REML/BLUP genetic statistical software (Resende, 2006;2016). The linear mixed model that considers half-sibling progenies in one site (y = Xr + Za + e) was used, where y: data vector; r: repetition eff ect vector (assumed as a fi xed value) added to the overall average; a: vector of individual additive genetic eff ects (assumed as a random value); and e: residual vector (also assumed as random). Capital letters represent the incidence matrix for these eff ects (Resende, 2016).
Analysis of deviance was implemented to test diff erences among progenies using the likelihood ratio test (LRT), obtained by determining the diff erence between deviance for models with and without the eff ect to be tested (progenies) using chi-square values to test the signifi cance of the models.
The phenotypic and genetic correlations between DBH and H were also estimated using Selegen-REML/BLUP software (Resende, 2006;2016), while standard errors and signifi cance for genetic and phenotypic correlations were estimated with GENES software (Cruz, 2006). Data expression by plotting associations between DBH and H and mean values for these traits was obtained with R scripts (R Core Team, 2019).
The eff ects of soil resistance on plant growth in proposed soil layers were determined by Pearson correlation for each of all 30 progenies, and multiple linear regression curves were fi tted, when possible. Because soil drainage is inversely related to soil Table 1 -Mean values, standard deviations (sd) and estimated genetic gains (u + a and Gain) for height (H) and diameter at breast height (DBH) in progenies of Calyphyllum brasiliense at fi ve years of age. Tabela 1 -Média, desvio padrão (sd) e estimativas de ganhos genéticos (u+a e Gain) para altura (H) e diâmetro à altura do peito (DBH) de progênies de Calophyllum brasiliense aos cinco anos de idade.
Because of the large number of surviving plants (1,127), accuracy varied from 61% (DBH) to 73% (H). A higher accuracy value indicates more trustworthy estimates for genetic parameters and BLUP. However, individual narrow sense heritability (h 2 a ) and heritability within progeny (h 2 ad ) were low (0.04-0.11), while the mean value for progeny heritability (h 2 mp ) was substantial (0.37-0.53), indicating that genetic control of growth traits was very low at the individual level and moderate and high at the progeny level, and coeffi cient of relative variation (CV r ) values were also moderate and high at the progeny level (Table 2). Because h 2 a and h 2 ad were low, expected genetic gains may be capitalized by mass selection in the trial and within the progeny.
Higher additive genetic values were estimated for 15 progenies (1, 4, 7, 9, 11, 16, 20, 21, 22, 23, 27, 29, 34, 35, and 38) (Table 1). However, the low values for genetic parameters indicate that the environment signifi cantly infl uenced these traits. The phenotypic correlation between H and DBH was higher (0.91) than the genetic correlation (0.85), indicating that selecting one of the traits (H or DBH), based on visual observation of the trees (phenotypes) in the trial, can result in the indirect selection of trees that are genetically superior for the other trait (H or DBH).

Plant growth in response to soil resistance
Most of the values for soil resistance (SR) in the 1,198 collection points were between 100 and 2,000 KPa (539.9 KPa at 0-10 cm, 1554.7 at 10-20 cm, 1453.6 at 20-40 cm, and 1124.5 at the 40-60 cm soil layer). Lower SR values (<100 KPa) were found in the 0-10 cm soil layer (mainly in blocks 30-33, which were the most aff ected by fl ooding), while the higher values (>2,000 KPa) were found in the 10-20 cm soil layer, randomly distributed throughout the experimental plot.
Pearson correlations between H and DBH were signifi cantly positive for all progenies (0.87 to 0.95, p<0.001), with phenotypic correlation (0.91) higher than genetic correlation (0.85). A signifi cant phenotypic variation in plant development was also resistance (Silva et al., 2016), and most soil resistance values were found up to 2,000 kPa, which is the upper limit for desirable root development (Taylor et al., 1966;Tormena et al., 1998), lower soil resistance values indicate elevated water table and/or fl ooded soils for practically the entire year in the study area, while higher values indicate soils with a deeper water table and/or better-drained soils during the 5-year period.
Soil resistance values throughout the soil layers showed a similar but non-signifi cant pattern for all progenies, but plant growth was mainly aff ected in layers up to 40 cm deep. Multiple linear regression models were fi tted for H growth in response to SR values at diff erent depths (Table 3), according to the correlation matrix values shown in Table 5, and as single models of SR values in the 1-10, 11-20, or 21-40 cm soil layers, double (11-20 and 21-40 cm), triple (1-10, 11-20 and 21-40 cm), and quadruple models (1-10, 11-20, 21-40, and 41-60 cm).
All single and/or multiple linear regression coeffi cients were then used for an initial assessment of potential growth of progenies in extreme drainage conditions, for fl ooding risk (TH min at SR=0 kPa) as well as in well-drained soils (TH max at SR=2,000 kPa). (Table 3) This yielded three sets of fl ood-  Table 3. Figura 2 -Potencial de crescimento em altura da progênies de Calophyllum brasiliense, aos cinco anos de idade, em solos alagados e bem drenados, de acordo com os valores teóricos ajustados nos modelos lineares expostos na Tabela 3.
Constant fl ooding of the soil resulted in 12.2% and 58.6% losses in theoretical maximum growth capacity for tolerant and sensitive progenies, respectively (Table 3). In fact, only a few progenies (1,16,21,31,37) showed real maximum height values that exceeded the theoretical ones (1.7%, 9.7%, 1.6%, 29.8% and 4.1%, respectively), indicating that even among genetic material that is tolerant, sensitive, or capable of proportional growth, the response to environmental conditions varies signifi cantly.

Initial plant growth and genetic parameters of progenies
Genetic diff erences for DBH were observed among the progenies, indicating that superior plants can be selected for breeding purposes. Individual narrow sense heritability (h 2 a ) was low for growth traits (0.06-0.11), which means that the additive genetic eff ects explain only a small proportion of the total variance for these traits (Kvestad et al., 2010). In contrast, mean heritability among progeny was moderate for the traits (0.53 for H, and 0.37 for DBH), indicating that genetic gains can be obtained by selecting the most productive progenies.
Breeding programs for C. brasiliense will require the introduction of new germplasm sources to boost genetic variability in future generations, as well as improved environmental control in experimental designs to minimize environmental infl uences. Among the genetic parameters (and despite our preliminary results at 5 years of age), genotypic variance, heritability, and genotypic and phenotypic correlations demonstrated precise and accurate values. Meanwhile, heritability at the progeny level was moderate to high for growth traits, indicating that genetic gains may be obtained via progeny selection.
Estimates of genetic parameters can shed light on the contribution of genetic and environmental variation to the heritability of traits. Heritability values exceeding 30% indicate that genetic control of a given phenotype can be transferred to the next generation.
Variables of a complex genetic nature are infl uenced by the environment, and their selection is usually complicated due to the signifi cant stochastic eff ects involved (Cruz, 2010); however, signifi cance may be found among these variables, permitting indirect selection. Heritability values for H and DBH in both the restricted and broad sense were low because of environmental infl uences, due to continuous distribution (Baldoni et al., 2020;Braga et al., 2020;Souza et al., 2020).
Genetic and phenotypic correlations between H and DBH in forest species tend to be highly positive, but can change at more advanced ages (Venlovsky and Barriga, 1992;Kien et al., 2009;Sumardi et al., 2016).
Low genetic control of native tropical species is very common in progeny tests, particularly for species that are highly dependent on open pollination. In studies involving REML/BLUP, UPGMA, and PCA, Maia et al. (2016a,b;2018; found positive correlations between fruit yield/quality and growth traits in early selection of native Brazilian fruit species such as Anacardium spp. (cajuí), Hancornia speciosa (mangaba), and Platonia insignis (bacuri), all collected from native populations and/or in the fi rst cycle of recurrent selection. But in all cases these authors emphasized the need for new assessments, since the genotypes came from base populations.
However, it should be noted that unlike the species in the aforementioned studies, the most desirable commercial product from C. brasiliense is not fruit or seeds. For this reason, genetic interactions across recurrent selections with genotypes also coming from a disturbed native population appear less important than the genotype/environment interaction, which results in better tree development and wood production.
The highly positive genetic correlations we have found suggest a common gene is responsible for phenotypic expression (pleiotropy) of H and DBH (Baltunis et al., 2007;Gapare et al., 2015), and that the indirect selection for DBH could yield indirect gains in H, since genetic correlations remove some components of environmental error.
In a previous study focusing on selecting C. brasiliense from commercial stands, also at 5 years of age, Kalil Filho et al. (2012) obtained plants with an average height of 3.69 m and DBH of 18.41 cm, results that exceeded our values (mean H=2.73m, mean DBH=2.75 cm). Although the genotypes came from natural populations, with no breeding stage, they were all planted on well-drained and managed soils.
For this reason, of the environmental error components that can aff ect plant growth, better management of soil drainage could positively impact progenies of C. brasiliense collected from segregating populations, at least at early ages.

Plant growth development in response to soil resistance
Soil resistance was highest in the 20-40 cm layer, which can result from physical changes in the soil structure (Tormena et al., 2002) due to compaction from agricultural machines (Silva et al., 2000;Gülser and Candemir, 2012), root density, water/soil or O 2 / CO 2 imbalance, and soil hydric potential across fi eld capacity (100 kPa) and permanent wilting point (1,500 kPa), mediating non-limiting and least limiting water range, and optimal hydric interval of soils (Letey, 1985;Silva et al., 1994;Orellana et al., 1997), which limit plant development at lower or upper values (Silva et al., 1994). A slight linear response to soil resistance was observed in tree height, which may be explained by the large number of observations (N=1,127) as well as the large genetic variability among progenies, environmental characteristics, and the trees' initial response to water/soil stress. However, of all the signifi cant correlation indexes, at least 30% of the variability in height and DBH values result from environmental conditions, which could be potentially related to soil resistance.

Integrated plant growth development in response to genetic potential and soil resistance
Genetic variation is essential for survival and adaptation to environmental changes (Gribel, 2000), and although soil resistance is not an intrinsic factor for cell metabolism or plant physiology, it does aff ect biochemical relations and aspects of root morphology and has been investigated for its signifi cant correlation with plant development (Gubiani et al., 2013). Water saturation in the soil has various adverse eff ects on plant development, all related to hypoxia, such as stomatal closure, photosystem-II damage, reduced photosynthesis capacity, and reduction of carbon assimilation mechanisms, including drastic decreases in Rubisco enzyme and ethylene growth regulator concentrations (Koslowski, 1997;Pezeshki, 2001;Herrera et al., 2008).
The most common fl ooding tolerance mechanisms are linked to foliar senescence and abscission, reduction of new leaves, and changes in growth and stem/root anatomy (increased size, presence of lenticels, development of adventitious and diageotropic roots, formation of aerenchyma) (Davanso-Fabro et al., 1998;Medri et al., 2012;Oliveira and Gualtieri, 2017). Despite its strong adaptative capacity to occupy riparian soils (Kawaguici and Kageyama, 2001;Percuoco et al., 2014), C. brasiliense can suff er severe damage if the time of exposure or recurrent stress surpasses 120 days, with height losses reaching 14% (Oliveira and Joly 2010).
For this reason, integrative analysis of estimated genetic gains (Table 1) and development growth potential of progenies in response to soil drainage ( Figure 2) could off er useful solutions. All the genetic evaluations performed in this study at 5 years of age indicated signifi cant gains in height for 14 progenies, which should be selected to establish clonal seed or seedling orchards to supply commercial plantations or establish the next generation for breeding, particularly to maintain genetic variability, which is already low in commercial stands. By simulating the eff ects of soil resistance on tree development, this could lead to advances in establishing diff erent sets of C. brasiliense progenies, both confi rming genetic estimated gains (progenies 1, 4, 21, 22, 23, 29, and 38), as well as contradicting them (7, 11, 17, 20, 27, 35, 36, 37, and 39).

Final Considerations
Because of the original nature of this study investigating how low and/or high SR values aff ect the development of tree species, some questions remain unanswered, and the specifi c eff ects of certain SR values on given soil layers and specifi c progeny still remain unknown. Still, the results were considered in a highly conservative manner to avoid making unreliable extrapolations.
Additional approaches at more advanced ages are needed, including new mathematical and statistical tools such as PCA, geostatistical mapping, and the constant integration of new data. As part of precision silviculture, the results of this study also indicate at least two diff erent approaches for the future: maintain environmental stress in order to assess breeding for fl ood tolerance over time, or improve soil drainage to increase growth of C. brasiliense and boost wood production.

CONCLUSIONS
The development of Calophyllum brasiliense varied signifi cantly among progenies. Despite high positive genetic and phenotypic correlations between growth traits, which indicate the possibility of an indirect selection strategy for the species, genetic control was low, particularly at the individual level.
The innovative combination of mathematical and analytical tools presented in this study demonstrates precision silviculture's potential to determine which genetic materials are best suited for diff erent environmental conditions and also to assist in decision making for forest management to avoid harmful environmental eff ects.