Oxygen transmission rate in American oak barrel staves used for wine aging: Influence of wood macrostructure

Abstract

Background

Oak barrels are widely used in wine and spirit aging, where oxygen transmission plays a key role in the development of chemical and sensory properties. In cooperage, stave selection is traditionally based on grain width; however, the extent to which grain and wood macrostructure determine oxygen transmission under barrel conditions remains poorly understood, particularly for American oak.

Results

The macrostructure and oxygen transmission rate (OTR) of 214 American oak (Quercus alba) staves were characterized under conditions simulating barrel use, with wood exposed to air on one side and a model wine solution on the other. Macrostructural parameters – density, porosity, growth‐ring features, earlywood and latewood widths and ratios, medullary ray characteristics, impregnation, and grain – were quantified by image‐based analysis and related to OTR. Compared to previously reported French oak (Quercus petraea), American oak showed higher density, lower porosity, wider growth rings, and greater interannual growth variability. OTR values varied substantially among staves and were on average approximately 68% lower than those reported for French oak. Grain showed only a weak relationship with OTR, and staves with similar grain displayed markedly different oxygen transmission. Multivariate analysis indicated that OTR is not controlled by any single macrostructural parameter but results from the combined effects of wood anatomy, growth patterns, structural orientation, and liquid impregnation.

Conclusion

These findings highlight the limitations of traditional cooperage descriptors for predicting oxygen transmission and provide a basis for improved wood selection targeting specific oxygenation profiles. with direct implications for barrel choice and wine quality. © 2026 The Author(s). Journal of the Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

INTRODUCTION

Oak heartwood is the most widely used wood in cooperage, as its physical, chemical, and mechanical properties make it particularly suitable for barrel production. In practice, winemakers often select barrels based on cooper recommendations or their own experience with wines aged in previous vintages. Traditionally, key criteria in cooperage include oak species, geographical origin, and growth rate, the latter expressed as grain width (in mm yr⁻¹). In cooperage, wood is commonly classified as very fine‐grained (< 1 mm), fine‐grained (1–2 mm), medium‐grained (2–4 mm), or coarse‐grained (> 4 mm), ¹ although simplified classifications distinguishing open grain (> 3 mm) and closed grain (< 3 mm) have also been proposed. ² Grain width is directly related to the annual growth rate of the tree and can be identified by the width of the growth rings.

Numerous studies have focused on the characterization of oak wood used in cooperage, particularly French oak (Quercus petraea), with emphasis on differences among forests and provenances. In contrast, studies on American oak (Quercus alba) have largely focused on comparative analyses of compounds released by American and French oak woods during aging.^{3, 4, 5, 6, 7} The interaction between beverages and wood during barrel aging has been extensively investigated, as this process governs the transfer of wood‐derived compounds to the beverage and, ultimately, its sensory and chemical evolution.^{8, 9} These interactions have been studied for French oak,^{10, 11, 12} American oak,^{13, 14} and other oak species.^{15, 16, 17} The chemical composition of barrel staves is closely linked to wood macrostructure and grain. ¹⁸ Moreover, cooperage processes such as bending and toasting modify wood structure and composition, directly influencing the compounds that may diffuse into the beverage through liquid–wood contact during aging.^{4, 19, 20}

Each annual growth ring comprises earlywood, characterized by large vessels and latewood, which exhibits a flame‐shaped radial pattern and increases in width with growth rate. ¹⁸ Oak wood further contains distinct uniseriate and multiseriate rays, as well as numerous small latewood vessels with thin walls and angular contours, often grouped and partially occluded by tyloses in heartwood. ²¹ Fine‐grained wood has traditionally been associated with higher concentrations of aromatic compounds, whereas coarse‐grained wood is considered to contribute higher tannin levels. ²² However, some studies have reported lower concentrations of wood‐derived aromatic compounds in wines aged in very fine‐grained barrels compared with medium‐grained barrels.^{11, 23} Towey and Waterhouse ²⁴ carried out a very interesting study showing that variability in the barrel toasting process leads to variability in the resulting wines, and that a sufficient number of barrels is therefore required to obtain meaningful results. They demonstrated that the minimum recommended threshold for representative results within ±20% of the population mean is ten barrels per batch, while reaching ±10% requires nearly 40 barrels. Along the same lines, Pfahl et al. ²⁵ focused on evaluating the compounds of aged wines, thereby establishing the barrel‐to‐barrel variability within each cooperage batch. Their results indicated that for some parameters, such as anthocyanins, variability was very high, requiring dozens of barrels to obtain representative results within ±15%, whereas for more stable parameters (color, total phenolics) 5–10 barrels were sufficient. These apparently contradictory findings highlight the intrinsic heterogeneity of wood as a natural material and underline the need for large and representative sample sets to accurately characterize its properties before the barrel is made.

In cooperage, macrostructural features are also critical because they determine the technological suitability of wood. For instance, the lower abundance of tyloses in French oak requires staves to be obtained by splitting the trunk, whereas the higher tylosis content in American oak allows sawing to be used as the primary cutting method.

During barrel aging, atmospheric oxygen diffuses through the stave wood in the tangential direction, from the exterior of the barrel toward the beverage. Oxygen transport occurs mainly by diffusion driven by the oxygen partial pressure gradient, in accordance with Fick's law, ²⁶ and involves both gas‐phase diffusion and diffusion through bound water. ²⁷ While wood permeability to liquids has been widely studied, gas transport – particularly in the tangential direction, which is crucial for beverage aging – has received much less attention. Some existing studies assess permeability under pressure‐driven flow conditions, which do not accurately reproduce the situation in barrel staves. ²⁸

To address this limitation, our research group developed and patented a system for measuring tangential oxygen permeability under conditions that reproduce the barrel environment, with wood exposed to air on one side and liquid on the other. ²⁹ This approach is essential because wood moisture content strongly affects gas permeability through capillary condensation, which reduces effective pore size, ³⁰ and because oxygen diffusivity in water is several orders of magnitude lower than in air. Consequently, accurate assessment of oxygen transmission through oak barrel staves requires measurements under realistic barrel conditions that account for the dynamic evolution of permeability over time.^{31, 32, 33, 34} This methodology has proven suitable for both French oak ³⁵ and American oak. ³⁶

The present study provides a comprehensive characterization of Q. alba staves used in barrel manufacture for beverage aging. Wood macrostructural parameters were quantified using image segmentation analysis, and oxygen transmission rate (OTR) was measured under barrel‐like conditions. The parameters evaluated include density and orientation of multiseriate rays, earlywood and latewood proportions and widths, grain and ring‐width variability, porosity, impregnation capacity, and OTR. By combining macrostructural characterization with oxygen permeability measurements, this work aims to improve understanding of the factors governing oxygen transmission through American oak staves under conditions relevant to barrel aging. These findings have direct implications for the selection of oak wood used in cooperage, as variability in oxygen transmission may influence the aging potential and final quality of beverages.

MATERIALS AND METHODS

Oak samples

American oak (Q. alba) fresh staves were supplied by INTONA S.L. (Monteagudo, Navarra, Spain) from Minnesota region. The stave blanks were seasoned outdoors and planed before the start of the study, they did not undergo the toasting process. All samples were stored in a temperature‐ and humidity‐controlled barrel cellar (15 ± 1 °C; 75% ± 3% relative humidity) throughout the experiment. Ambient temperature and relative humidity in the cellar were monitored with a calibrated Vaisala thermo‐hygrometer, model Humicap HMT310 (temperature accuracy of up to ±0.2°C and relative humidity accuracy of up to ±1%; Vantaa, Finland).

A total of 214 staves were selected, as a minimum width of 50 mm is required for OTR measurements. Cylindrical samples (40 mm diameter) with the typical stave thickness (27 mm) were extracted and stored under the same controlled conditions until analysis.

Oxygen transmission rate (OTR) measurement

The OTR was determined using the patented device, ²⁹ specifically designed to reproduce barrel‐like conditions. In this system, one face of the wood sample is in contact with a model solution (12% v/v ethanol), simulating the beverage, while the opposite face is exposed to atmospheric air, with no barometric pressure difference applied between the liquid and the air sides. The experimental procedure was conducted as described in detail in previous studies.^{35, 36}

The method is based on continuous monitoring of the oxygen diffusing through the wood, which acts as a solid membrane separating the two environments. The amount of oxygen accumulated in the device chamber due to permeation through the wood is recorded as a function of time.^{29, 35} Under steady‐state conditions, oxygen permeation through a material of thickness d can be described by Fick's first law:

$$J=P\frac{∆p{\mathrm{O}}_2}{d}$$ (1)

where J (in m³ s⁻¹ m⁻²) is the volumetric gas flow per unit surface area, d (in meters) is the thickness of the sample, and P (m³ mm⁻² s⁻¹ Pa⁻¹) is the permeability coefficient. According to Eqn (1), the oxygen flux is proportional to the oxygen partial pressure difference (ΔpO₂, in pascals) across both sides of the wood, which acts as the driving force for permeation. Steady‐state permeation is indicated by a linear increase in oxygen partial pressure over time. The slope of this linear region corresponds to the OTR, expressed as hPa d⁻¹.

Density measurement

Wood density, defined as the ratio of mass to volume, is a key physical property that has been correlated with the permeability of wood to gases ³⁷ and liquids.^{38, 39} Density measurements were performed on the same cylindrical samples used for OTR determination, under the controlled environmental conditions of the barrel room (15 ± 1 °C; 75% ± 3% relative humidity).

Sample dimensions were measured using a digital Vernier caliper (Digimatic Caliper CD‐6″ASX; Mitutoyo, Sakado, Japan), with each dimension recorded three times and the average value used to calculate sample volume. Sample mass was measured three times using a precision balance (2000/0.01 g, Nahita Blue KBC004 model 5162; Bornova/İzmir, Turkey). All samples were equilibrated under the described environmental conditions (15 ± 1 °C; 75 ± 3% relative humidity). Under these stable hygrothermal conditions, oak wood reaches an equilibrium moisture content (EMC) of approximately 14–15%, as established by published sorption isotherms for oak. Wood moisture content was therefore not measured directly but inferred from this well‐characterized EMC relationship. ⁴⁰

Image segmentation analysis

Wood macrostructure was characterized by image segmentation analysis of the transverse surfaces of the wood samples from which the OTR test specimens were extracted. Images were acquired using a Lumix DMC‐FZ200 camera (Panasonic Co., Osaka, Japan) under controlled illumination provided by a Domo 152/114‐White‐24 V‐3 mm light‐emitting diode (LED) lighting system (Infaimon SL, Barcelona, Spain). Photographs were recorded in JPEG format at a resolution of 4000 × 3000 pixels (180 × 180 dpi; 24‐bit color depth; f/4; 1/640 s; ISO‐100). An example of two different samples is shown in Fig. 1.

Figure 1.

Detail of image analysis with the software Computer Vision Annotation Tool (CVAT). (A) Original photograph of the head of the stave; (B) outline of the entire surface of the stave; (C) detail of the outline of one of the spring wood regions; (D) detail of the outline of a medullary ray; (E) stave with labeled regions; (F) detail of the labeled regions. 1: Left, fine‐grained wood; 2: Right, coarse‐grained wood.

Image annotation was performed using the Computer Vision Annotation Tool (CVAT; https://www.cvat.ai). The entire wood surface was labeled in red (Fig. 1(B)), earlywood regions in green (Fig. 1(C)), and medullary rays in yellow (Fig. 1(D)). Areas labeled as red but not identified as earlywood or medullary rays were defined as latewood. Figure 1(E) shows the original image with all labeled regions superimposed, while Fig. 1(F) presents the final segmented image used for quantitative analysis.

Macrostructural parameters

Based on the segmented images (Fig. 1(F)), the following macrostructural parameters were calculated:

• Average earlywood width (AEW): mean earlywood width per annual growth ring.

• Variation of earlywood width (VEW): standard deviation of earlywood width across rings, reflecting the homogeneity of earlywood growth.

• Earlywood ratio (EWR): ratio of earlywood width to total ring width.

• Average latewood width (ALW): mean latewood width per ring.

• Variation of latewood width (VLW): standard deviation of latewood width across rings, reflecting latewood growth variability.

• Latewood ratio (LWR): ratio of latewood width to ring width, also referred to as texture in cooperage. ¹⁸

• Variation of ring width (VR): standard deviation of total ring width, indicating variability in annual growth.

• Density of multiseriate rays (DMR): number of visible multiseriate rays per centimeter of sample height in the axial plane.

• Multiseriate ray angle (MRA): mean angle of medullary rays measured in the axial plane, weighted by ray length; absolute values were used to eliminate orientation effects.

• Impregnation: wood impregnation capacity, calculated as the percentage increase in mass after barrel wet conditioning, defined here as the initial 7‐day wine‐filling period after which wood degassing has ceased, enabling accurate OTR measurements without interference from air initially trapped in the staves, ³⁵ full hydration of the wood continues beyond this point.

• Grain: mean annual ring width, following the definition of Vivas. ¹

• Porosity: calculated according to Plötze and Niemz ⁴¹ using Eqn (2), assuming a cell wall density of 1.538 g cm⁻³ based on the densities of the main wood chemical components. ⁴²

$$\text{Porosity}=1-\frac{\text{density}}{\text{density of woody material}}$$ (2)

Statistical analysis

Normality tests indicated that the distribution of the analyzed variables did not follow a normal distribution. Consequently, correlations between variables were evaluated using Spearman's rank correlation coefficient. Statistical significance was established at P ≤ 0.05. All statistical analyses were performed using Statistica software (version 10; StatSoft GmbH, Hamburg, Germany).

RESULTS AND DISCUSSION

Description of oak samples

General characteristics of oak samples

Table 1 summarizes the main descriptive statistics of the analyzed parameters for the 214 Q. alba staves, including mean, standard deviation, minimum, maximum and coefficient of variation. Figure 2 shows the frequency distributions of the 15 parameters studied. Unless otherwise stated, the discussion focuses on the range between the 10th and 90th percentiles and includes comparisons with previously reported data, measured under the same conditions, for French oak (Q. petraea) staves used in cooperage.^{35, 43}

Table 1.

Main descriptive statistics of the parameters analyzed: mean value, standard deviation, minimum, maximum and coefficient of variation.

Wood property (n = 214)		Unit	Mean value	Standard deviation	Minimum value	Maximum value	Coefficient of variation (%)
Density	D	kg m−3	770.53	50.92	637.86	894.24	6.6%
Porosity	P	[−]	0.497	0.037	0.416	0.583	6.7%
Average earlywood width	AEW	mm	0.966	0.227	0.472	2.064	23.5%
Average latewood width	ALW	mm	1.296	0.429	0.520	3.280	33.1%
Variation of earlywood width	VEW	mm	0.389	0.266	0.104	1.727	68.6%
Variation of latewood width	VLW	mm	0.727	0.493	0.217	4.777	67.8%
Grain	G	mm per ring	2.260	0.580	1.33	4.67	25.7%
Variation of ring width	VR	mm	0.862	0.504	0.377	4.844	58.5%
Latewood ratio – texture	LWR	[−]	0.566	0.071	0.360	0.761	12.6%
Earlywood ratio	EWR	[−]	0.434	0.071	0.239	0.640	16.5%
Density of multiseriate rays	DMR	rays cm−1	3.169	0.613	1.625	5.266	19.3%
Multiseriate ray angle	MRA	Deg	16.534	10.544	1.696	46.587	63.8%
Impregnation	I	%	3.218	0.700	0.001	6.250	21.8%
Oxygen transmission rate	OTR	hPa d−1	1.87	2.49	0.002	16.95	133.5%

Figure 2.

Histogram of the frequency of the value of each parameter analyzed in the 214 samples.

The density of the American oak staves showed a wide variability, spanning approximately 260 kg m⁻³ (Fig. 2(A)). Samples between the 10th and 90th percentiles exhibited a mean density of 770 kg m⁻³, with values ranging from 707 to 831 kg m⁻³. These values are substantially higher than those reported for French oak (Q. petraea), which showed an average density of approximately 621 kg m⁻³ and a range 543–735 kg m⁻³. ³⁵ By contrast, Quercus robur has been reported to exhibit a much wider density range (384–864 kg m⁻³). ⁴⁴

Porosity, which is closely related to density, varied between 0.41 and 0.59 across the full dataset (Fig. 2(B)). Within the 10th and 90th percentile range, American oak exhibited a significantly lower mean porosity (0.497; range 0.458–0.538) than French oak, which showed an average porosity of 0.59 and values between 0.52 and 0.65. These differences are consistent with the higher density observed in American oak staves.

The AEW exhibited substantial variability among samples, with values differing by up to 2 mm (Fig. 2(C),(D)). Within the 10th and 90th percentile range, AEW varied from 0.697 to 1.214 mm, with a mean value (0.948 mm) significantly higher than that reported for French oak (mean 0.687 mm; range 0.483–1.009 mm). Similar trends were observed for ALW (Fig. 2(E),(F)). In American oak, ALW ranged from 0.802 to 1.809 mm, with a mean value of 1.250 mm, compared with a mean of 0.890 mm in French oak (range 0.479–1.347 mm).

Overall, the American oak staves analyzed showed, within the 10th and 90th percentile range, a 38% greater earlywood thickness and a 40% greater latewood thickness than French oak. This resulted in a larger grain and a higher contribution of latewood‐derived material. Variability parameters further highlighted these differences: VEW ranged from 0.193 to 0.650 mm (mean 0.337 mm) in American oak, compared with a mean of 0.139 mm in French oak, while VLW ranged from 0.394 to 1.025 mm (mean 0.637 mm), more than four times the variability observed in French oak staves.

These findings indicate that American oak exhibits both greater growth in spring and summer and greater interannual variability in growth than French oak. The variability of earlywood and latewood growth was higher than in French oak staves. This behavior contrasts with classical descriptions for wood in general, in which annual ring width is largely determined by latewood growth and earlywood width is assumed to be relatively constant.^{18, 45} In oak wood, however, and consistently with previous reports for Q. petraea,^{43, 46, 47} both earlywood and latewood growth were found to vary substantially, with earlywood width varying by more than 1.5 mm in some samples (Fig. 2(D)).

American oak staves exhibited grain widths ranging from 1.4 to 4.7 mm (Fig. 2(G)). According to cooperage classification, 11 staves were classified as extra‐fine grain (< 1.5 mm), 65 as fine grain (1.5–2 mm), 119 as medium grain (2–3 mm), and 19 as coarse grain (> 3 mm). Thus, 56% of the staves exhibited medium grain, while approximately 30% showed fine or very fine grain and 19% coarse grain. Within the 10th and 90th percentile range, American oak staves exhibited a mean grain of 2.19 mm (range 1.61–2.88 mm per ring), higher than that reported for Q. petraea, which ranged from 1.031 to 2.301 mm with a mean of 1.6 mm. This confirms that American oak staves used in cooperage exhibit, on average, a grain approximately 37% larger than French oak staves.

VR further reflected this behavior, with American oak exhibiting values between 0.4 and 4.9 mm (Fig. 2(H)). Within the 10th and 90th percentile range, American oak staves presented a mean 0.766 range between 0.496 and 1.272 mm, compared with a significantly lower mean of 0.374 mm in French oak (range 0.234–0.534 mm). This indicates that interannual growth variability is approximately 100% higher in American oak staves.

Despite these differences in absolute growth, the proportion of LWR and EWR in American oak was comparable to that observed in French oak. Within the 10th and 90th percentile range, American oak staves presented a LWR mean value of 0.568 (range 0.470–0.647), while EWR averaged 0.430 (range 0.352–0.518) (Fig. 2(J)). French oak showed a slightly higher EWR (mean 0.434), indicating a marginally greater proportion of earlywood and, conversely, a slightly higher proportion of latewood in American oak staves.

Medullary rays, impregnation, and cooperage

The lower abundance of tyloses in French oak compared with American oak has important implications for cooperage practices. French oak staves are typically obtained by splitting along medullary rays, which constrains ray orientation, whereas American oak staves are sawn, allowing greater variability in ray orientation. In the American oak staves studied, the DMR reached up to 5 rays cm⁻¹ (Fig. 2(L)). Within the 10th and 90th percentile range, DMR varied between 2.366 and 3.896 rays cm⁻¹, with a mean value of 3.159 rays cm⁻¹, approximately 20% lower than that reported for French oak staves. ⁴³

The MRA exhibited particularly high variability, ranging from 1.696° to 46.578° (Fig. 2(M)). The 10th and 90th percentile range varied between 3.94° and 32.82° with a mean of 15.6°, significantly higher than the mean value of 9.5° reported for split French oak staves. This increase (approximately 64%) reflects the sawing process used in American oak cooperage and has direct implications for liquid penetration and oxygen transmission.

Consistent with these structural features, impregnation values for American oak staves between the 10th and 90th percentiles ranged from 2.5% to 4.2%, corresponding to beverage absorption values of approximately 0.8–1.3 g during conditioning (Fig. 2(N)). These results reflect the capacity of the beverage to flood the wood and access its internal structure.

Oxygen transmission rate (OTR) and comparison with French oak

The OTR of American oak staves (between the 10th and 90th percentiles) varied widely, from 0.266 to 4.951 hPa d⁻¹, with a mean value of 1.306 hPa d⁻¹. This corresponds to an oxygen transmission of 0.013–0.234 mg L⁻¹ d⁻¹, with an average of 0.062 mg L⁻¹ d⁻¹, calculated assuming an oxygen solubility in the liquid equivalent to that of water. These values are substantially lower than those reported for un‐toasted French oak staves, which exhibited an average oxygen transmission of 0.191 mg L⁻¹ d⁻¹ within the 10th and 90th percentile range. ³⁵

Overall, American oak staves exhibited an average OTR approximately 68% lower than that of French oak. Nevertheless, within the population of 214 Q. alba staves, some samples exhibited OTR values exceeding 10 hPa d⁻¹ (Fig. 2(Ñ),(O)), highlighting the substantial heterogeneity of oxygen permeability within American oak.

Grain correlation analysis

The correlation coefficients and their significance (P < 0.05) among the analyzed parameters are summarized in Table 2. Figure 3 presents selected scatter plots illustrating the relationships between representative macrostructural parameters and grain, which remains the primary criterion used in cooperage for stave classification.

Table 2.

Correlation coefficient and its significance between each of the parameters analyzed

Variable	D	P	AEW	ALW	VEW	VLW	EWR	LWR	DMR	MRA	G	VR	I	OTR
D		−1.000	−0.297	0.271	0.043	0.191	−0.605	0.605	0.137	−0.023	0.084	0.201	−0.292	0.002
		***	***	***		**	***	***	*			**	***
P	−1.000		0.297	−0.271	−0.043	−0.191	0.605	−0.605	−0.137	0.023	−0.084	−0.201	0.292	−0.002
	***		***	***		**	***	***	*			**	***
AEW	−0.297	0.297		0.512	0.188	−0.003	0.249	−0.249	0.092	0.392	0.773	0.051	0.155	−0.178
	***	***		***	**		***	***		***	***		*	**
ALW	0.271	−0.271	0.512		−0.132	0.477	−0.671	0.671	0.225	0.130	0.940	0.393	−0.016	−0.169
	***	***	***		*	***	***	***	**		***	***		*
VEW	0.043	−0.043	0.188	−0.132		−0.003	0.338	−0.338	0.137	−0.150	−0.024	0.355	−0.120	0.051
			**	*			***	***	*	*	*	***
VLW	0.191	−0.191	−0.003	0.477	−0.003		−0.491	0.491	0.194	−0.244	0.351	0.920	−0.078	−0.049
	**	**		***			***	***	**	***	***	***
EWR	−0.605	0.605	0.249	−0.671	0.338	−0.491		−1.000	−0.140	0.153	−0.398	−0.345	0.151	0.038
	***	***	***	***	***	***		***	*	*	***	***	*
LWR	0.605	−0.605	−0.249	0.671	−0.338	0.491	−1.000		0.140	−0.153	0.398	0.345	−0.151	−0.038
	***	***	***	***	***	***	***		*	*	***	***	*
DMR	0.137	−0.137	0.092	0.225	0.137	0.194	−0.140	0.140		−0.124	0.201	0.213	0.017	−0.040
	*	*		**	*	**	*	*			**	**
MRA	0.042	0.023	−0.399	−0.250	0.319	0.240	−0.038	0.038	0.110		0.255	0.325	0.209	−0.046
			***	***	***	***					***	***	**
G	0.084	−0.084	0.773	0.940	−0.024	0.351	−0.398	0.398	0.201	0.255		0.311	0.048	−0.198
			***	***		***	***	***	**	***		***		**
VR	0.201	−0.201	0.051	0.393	0.355	0.920	−0.345	0.345	0.213	−0.263	0.311		−0.118	−0.032
	**	**		***	***	***	***	***	**	***	***
I	−0.292	0.292	0.155	−0.016	−0.120	−0.078	0.151	−0.151	0.017	0.209	0.048	−0.118		−0.003
	***	***	*				*	*		**
OTR	0.002	−0.002	−0.178	−0.169	0.051	−0.049	0.038	−0.038	−0.040	−0.046	−0.198	−0.032	−0.003
			**	*							**

Note: n = 214 (***P < 0.001, **P < 0.01, *P < 0.05). Abbreviations according to Table 1.

Figure 3.

Scatterplot among some of the representative parameters of the wood, where the grain of each sample was represented with a different shape and the oxygen transmission rate (OTR) group of each sample with a different color according to the legend.

Wood density showed significant correlations with several structural parameters. Density was positively correlated with features associated with a higher proportion of dense tissues, including LWR (r = +0.605), ALW (r = +0.271), VLW (r = +0.191), and VR (r = +0.201). A positive, although weaker, correlation was also observed with the DMR (r = +0.137). Conversely, density was negatively correlated with earlywood‐related parameters, including AEW (r = −0.297) and EWR (r = −0.605), reflecting the lower density associated with earlywood‐dominated structures.

Despite these relationships, grain itself was not significantly correlated with density (r = 0.084; Fig. 3(A)). Woods with identical grain values (e.g., 2.5 mm) exhibited a wide density range (approximately 675–900 kg m⁻³), indicating that grain alone is not a reliable predictor of density. High density heterogeneity was observed across all grain classes, particularly within fine‐ and medium‐grained groups, likely reflecting the complex internal structure of latewood and the contribution of medullary rays. Similar behavior has been reported for French oak (Q. petraea) and Q. robur. ⁴⁴

Earlywood‐related parameters showed strong internal coherence. Samples with higher AEW exhibited a higher proportion of EWR (r = +0.249) and greater VEW (r = +0.188). Grain was strongly correlated with AEW (r = +0.773; Fig. 3(B)), consistent with observations previously reported for French oak.^{18, 43} This relationship (EWR and grain, r = −0.398) in American oak wood samples varied especially in staves classified as extra fine and fine grain (less than 2 mm), resulting in an earlywood proportion of 30% in some woods and 65% in others, while in coarse‐grained samples (more than 3 mm), some woods were found with EWRs of 25% or 45% (Fig. 3(C)).

Latewood characteristics exhibited even stronger correlations with grain. ALW increased markedly with grain (r = +0.940) (Fig. 3(D)), exceeding the strength of the grain–AEW relationship. VLW was of similar magnitude across grain classes, indicating that while latewood thickness increases with growth rate, its interannual variability is relatively independent of grain.

The combined analysis of AEW and ALW revealed that grain variability is governed by both earlywood and latewood growth (AEW–ALW; r = +0.512; Fig. 3(E)). Notably, in fine‐ and extra‐fine‐grained woods (< 2 mm), earlywood width occasionally exceeded latewood width, leading to pronounced variability in both LWR (0.35–0.67) and EWR (0.30–0.65). In contrast, in coarse‐grained woods (> 3 mm), increased grain was more clearly associated with latewood development. These findings confirm that, in oak wood used for cooperage, annual growth affects both earlywood and latewood formation, particularly in slow‐growing trees. This behavior has been reported previously for French oak and is here confirmed for American oak.

Grain was positively correlated with LWR (r = +0.398; Fig. 3(F)) and negatively with EWR (r = −0.398; Fig. 3(G)). Coarser‐grained woods also exhibited higher densities of multiseriate rays, suggesting that faster growth promotes both latewood formation and ray development. Additionally, ALW showed significant correlations with medullary ray properties, being positively related to DMR (r = +0.225) and negatively related to MRA (r = −0.250), reflecting the influence of sawing practices in American oak cooperage.

During barrel aging, beverage–wood interactions are governed by the ability of the liquid to access the wood structure and by the diffusion of atmospheric oxygen through the stave. Impregnation showed a weak but significant negative correlation with density (r = −0.144) and a positive correlation with porosity (r = +0.144), confirming that lower‐density woods tend to absorb more liquid. Impregnation was also positively correlated with MRA (r = +0.136) (Fig. 3(H)), indicating that a greater deviation of medullary rays from the tangential plane enhances liquid penetration. In other words, the greater the angle of the medullary rays, the greater the impregnation of the wood, since the rays are more impermeable to liquids than early and latewood. In addition, the interaction of the beverage with atmospheric oxygen that passes through the wood into the interior of the barrel is also relevant. ⁴⁸ Therefore, the ability of wood to transmit oxygen is an important property, as it defines the properties that the beverage acquires during this process.

In contrast, the relationship between grain and OTR was negative (r = −0.198; Fig. 3(I)). Woods with fine or medium grain exhibited OTR values spanning nearly the entire observed range (from near zero to 10 hPa d⁻¹), demonstrating that grain alone does not determine oxygen permeability. Similar behavior was observed across all grain classes, including very fine‐ and coarse‐grained woods.

Oxygen transmission rate (OTR) correlation analysis

The study of oxygen transmission through barrel stave wood must be conducted under conditions of use, that is, with the wood in contact with air on one side and liquid on the other. However, few studies have evaluated wood under these conditions.^{35, 36} When barrel wood is dry, gas flow through its different anatomical structures indicates that a higher proportion of latewood results in lower overall gas permeability, a relationship that may also apply to gas diffusivity. ⁴⁹ This behavior can be explained by the fact that the small vessels present in latewood are relatively impermeable to gas, unlike the large vessels found in earlywood. ⁵⁰ From a macrostructural perspective, earlywood cells contain many pores, whereas latewood contains fewer pores that are also smaller in size. In addition, pit membranes in latewood are often heavily encrusted, further limiting fluid transport. Consequently, wood is naturally structured in such a way that it could be expected that earlywood can be up to 100 times more gas permeable than latewood. ⁵¹

According to Vivas et al., ²⁸ in French oak barrels made from fine‐grained Q. petraea, dry wood is more permeable to oxygen than coarse‐grained wood. However, when the barrel is filled with liquid, these mechanisms change substantially, and oxygen ingress is altered because flooding of the wood structures modifies oxygen diffusion pathways.^{52, 53} On the one hand, flooding of earlywood, which contains larger vessels, can reduce oxygen diffusion, since oxygen diffuses much more slowly in liquid than in gas. On the other hand, earlywood vessels may be partially or fully blocked by tyloses,^{18, 54} a phenomenon that is particularly pronounced in American oak due to its high tylosis content. Together, these factors govern the entry of oxygen into the wine.

Previous studies have demonstrated a strong correlation between the thickness of flooded wood (above the fiber saturation point, with free water present) and oxygen diffusivity in Q. petraea samples in the tangential direction, a situation comparable to that of a barrel stave. ⁵⁵ These studies also showed that, at the beginning of aging, oxygen release is greater in latewood than in earlywood areas.^{48, 53, 55} This seemingly contradictory result, given the higher density of latewood compared with earlywood, may be explained by the presence of tyloses in earlywood vessels, which hinder complete flooding of the large earlywood vessels and prevent displacement of the air they contain.^{53, 56} As aging progresses, earlywood subsequently exhibits slightly higher oxygen content. Moreover, the role of different structures within latewood has been highlighted, showing that darker latewood zones release more oxygen than lighter zones, mainly due to their lower density and consequently higher porosity. Overall, the diffusion of atmospheric oxygen into a barrel is a complex process that depends on wood structure, its arrangement within the stave, and the stave's exposure to the beverage during aging.

Analysis of OTR under conditions of use in a population of American oak staves allowed three OTR levels to be defined: low (OTR ≤ 0.36 hPa d⁻¹), observed in 18% of the samples (n = 39); medium (0.36–1.44 hPa d⁻¹), observed in 49% of the staves (n = 105); and high (OTR ≥1.44 hPa d⁻¹), observed in 33% of the staves (n = 70). Comparison of OTR levels with other properties measured in the same wood pieces highlighted the contribution of each parameter to explaining OTR variability. Figure 4 presents scatterplots relating oxygen permeability rate to six properties analyzed.

Figure 4.

Scatterplot between eight representative parameters of the wood and the oxygen transmission rate (OTR), where the symbols represent the grain of each sample according to the legend.

Samples with similar OTR values exhibited wide ranges of ALW and AEW. For example, some American oak samples with OTR values below 1 hPa d⁻¹ had an ALW of 0.5 mm, whereas others exceeded 2.5 mm. Similarly, samples with higher OTR values (around 5 hPa d⁻¹) showed ALW values ranging from 1 to 2 mm (Fig. 4(A)). Comparable patterns were observed for AEW: samples with an OTR of approximately 1 hPa d⁻¹ showed AEW values between 0.5 and 1.5 mm. However, at higher OTR values, AEW variability decreased, ranging from 0.8 to 1.2 mm (Fig. 4(B)).

For American staves, OTR was negatively correlated with AEW (r = −0.178, P = 0.009), indicating that wood with narrower earlywood within each growth ring (i.e., finer grain) exhibited lower oxygen transmission. Similarly, higher ALW values were associated with lower OTR, with a correlation of r = −0.169 (P = 0.013; Table 2). Differences between earlywood and latewood, particularly regarding the presence or absence of tyloses, are known to influence gas and liquid diffusion. In addition, wood water status²⁸ and the degree of flooding^{53, 55} significantly affect diffusion kinetics. These findings are supported in American oak by the significant negative relationship between grain and OTR (r = −0.198, P = 0.004), a trend also reported for French oak (r = −0.1172, P = 0.1712). ³⁵ In contrast, VR was not related to OTR, as staves with OTR values below 1 hPa d⁻¹ displayed VR values ranging from 0.3 mm to over 3 mm (Fig. 4(C)). Grain variability was markedly higher in American oak staves than in French oak staves, where VR did not exceed 0.9 mm, compared with values above 4 mm in some American oak samples.

Density showed no meaningful relationship with OTR (r = +0.002; Fig. 4(D)), confirming that oxygen transmission through oak wood cannot be predicted solely from density. This observation is consistent with previous findings for French oak, where density also showed no significant correlation with OTR. The high abundance of tyloses in American oak, combined with sawing‐induced variability in structural orientation, likely contributes to the wide dispersion of OTR values.

Similarly, DMR and MRA showed negligible correlations with OTR (r = −0.040 and −0.046, respectively; Fig. 4(E),(F)), despite the generally low permeability of rays in hardwoods. ⁵⁷ Porosity also exhibited no relationship with OTR (r = −0.002), likely because the beverage side of the wood samples were above the fiber saturation point, where liquid‐filled pores strongly limit oxygen diffusivity.

Overall, these results indicate that OTR is not controlled by any single macrostructural parameter but emerges from the combined effect of wood anatomy, moisture state, structural orientation, and the degree of flooding during barrel use. This complexity explains the difficulty in predicting oxygen transmission behavior based solely on conventional cooperage descriptors.

Multivariate analysis

Principal component analysis (PCA) was carried out to obtain a reduced number of linear combinations of the parameters analyzed which explained the greater variability of the data, so 13 linearly independent parameters were selected. The first four principal components (PCs) explained 73.18% of data variability, PC1 31.94%, PC2 18.93%, PC3 13.83% and PC4 8.48%. The weight of each parameter analyzed for each component is shown in Table 3. The first principal component (PC1) was strongly positively related to LWR, ALW, VLW, grain, density, VR, and negatively to EWR, while the parameters of most weight in the second principal component (PC2) were: MRA and AEW with a positive charge (Fig. 5(A)).

Table 3.

Parameter loading, eigenvalue and variance (%) for the first four principal components (PCs).

Wood property		PC1	PC2	PC3	PC4
Density	D	0.261	−0.275	−0.137	0.463
Average earlywood width	AEW	0.064	0.600	0.092	0.133
Average latewood width	ALW	0.425	0.275	−0.099	0.045
Variation of earlywood width	VEW	−0.052	0.056	0.572	0.394
Variation of latewood width	VLW	0.362	−0.067	0.311	−0.369
Latewood ratio – texture	LWR	0.426	−0.202	−0.246	−0.024
Earlywood ratio	EWR	−0.426	0.202	0.246	0.024
Density of multiseriate rays	DMR	0.145	0.051	0.228	0.087
Multiseriate ray angle	MRA	−0.052	0.352	−0.331	0.162
Impregnation	I	−0.080	0.226	−0.111	−0.624
Grain	G	0.339	0.439	−0.038	0.087
Variation of ring width	VR	0.318	−0.046	0.490	−0.196
Oxygen transmission rate	OTR	−0.068	−0.164	0.043	−0.066
	Eigenvalue	4.152	2.461	1.798	1.102
	Variance (%)	31.94	18.93	13.83	8.48

Figure 5.

Projection of the parameters on the two first principal component planes (a) and projection of the samples on the two first principal component planes, where the grain of each sample was represented by a different shape and the oxygen transmission rate (OTR) group of each sample with a different color according to the legend (b). (abbreviations according to Table 1).

Figure 5(B) shows that the woods do not differ in groups; those with a finer grain (square and asterisk) are on the left‐hand side, while those with faster growth, medium or coarse grain (wide grain, triangle) are on the right‐hand side. Considering the position of the parameters and samples, the finer‐grained samples (extra‐fine grain and fine grain) are positioned on the negative axis of PC1 and are related to EWR, VEW, and OTR. This confirms the importance of very slow growth in obtaining a higher OTR, as described for French oak. ⁴³ Coarse‐grained woods, located on the positive side of PC1, were defined by parameters related to the width of growth rings, especially latewood (ALW, VLW, LWR), by grain, and by density (Table 3). The medium‐grain wood samples (circle) are in the center. These results demonstrate that low OTR is related to faster wood growth, that is, wood with larger growth rings and coarser grain should have a lower OTR. However, this relationship is not so obvious, as Fig. 2(H) shows that staves with the same grain had very different OTRs and that there were samples across the entire OTR range within each grain group.

CONCLUSIONS

For the first time, a study based on a sufficiently large sample set has characterized the macrostructure of Q. alba wood and evaluated its role in the transmission of atmospheric oxygen to beverages under barrel‐like conditions. The results highlight the pronounced heterogeneity of American oak wood used in cooperage and demonstrate the relevance of macrostructural features – particularly earlywood and latewood widths, their interannual variability, and growth‐ring characteristics – in defining wood behavior during barrel aging.

Grain, the traditional criterion used in cooperage to classify Q. alba staves, was shown to be insufficient to explain the variability observed in OTR among staves of the same grain class. Although population‐level trends indicate that slower‐growing woods tend to exhibit higher OTR, individual staves with similar grain displayed markedly different oxygen permeability. This confirms that OTR is an emergent property resulting from the combined influence of wood anatomy, growth patterns, structural orientation, moisture state, and the degree of flooding during use.

The comparison with French oak (Q. petraea) further emphasizes species‐specific differences in growth behavior, macrostructure, and oxygen transmission, underlining the need to consider wood origin and anatomical characteristics beyond conventional descriptors. The results also confirm that OTR cannot be reliably predicted from single macrostructural parameters or simple correlations, highlighting the limitations of traditional selection criteria.

Overall, this work provides a robust experimental basis for advancing toward predictive models of oxygen transmission in barrel staves, integrating multiple macrostructural descriptors. Such models could support the rational selection and design of barrels with tailored oxygenation properties, contributing to improved control of beverage aging processes and final quality characteristics.

FUNDING INFORMATION

This study was supported by the Ministerio de Ciencia e Innovación, la Agencia Española de Investigación (Project‐PID2022‐139694OB‐I00), and the Fondo Europeo de Desarrollo Regional, funded by MCIN/AEI/10.13039/501100011033/FEDER, EU.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.