Anthraco-typology as a key approach to past firewood exploitation and woodland management reconstructions. Dendrological reference dataset modelling with dendro-anthracological tools

Abstract Charcoal analysis aims to study different aspects of forest management, techno-economical choices and their specific impact on past landscapes, as well as the impact of climatic events. However, at the present time, charcoal analysis is generally limited to the study of a list of taxa and their relative frequency, as the methods usually employed in dendrochronology to characterize past woodland, based on long tree-ring series, are not suitable for anthracological material. Today, the new challenge for charcoal analysis is thus to develop adapted dendrological tools. In this context, the aim of the ANR DENDRAC project “Development of dendrometrical tools applied to anthracology: study of the interactions between Man, resources and environments” was to characterize modern-day wood stands in accordance with historical woodland practices and convert dendroecological data into parameters adapted to charcoal analysis. The purpose of this study is to define the dendrological features with the help of the anthracological tools without explaining the observed differences between the sampled stands (given the stational variability, age and regeneration modes). The first step consisted in creating dendro-anthracological tools based on morpho-anatomical criteria that help to characterize growth, distinguish heartwood from sapwood and evaluate charcoal-pith distance. The second step involves characterizing three modern-day wood stands (coppice under standard, high forest and young stand formed by a mixture of seeded and coppice trees), defined by their structure, stand density and regeneration modes, using dendrological data measured on fresh wood material and modelled into anthracological data with the dendro-anthracological tools. In this way, anthracological types were defined for each wood stand, forming anthraco-typological models, which area useful for the interpretation of archaeological charcoal assemblages. Finally, an anthracological key is proposed to sort archaeological charcoal fragments in anthraco-groups before data processing.

dendrochronology to characterize past woodland, based on long tree-ring series, are not 23 suitable for anthracological material. Today, the new challenge for charcoal analysis is thus 24 to develop adapted dendrological tools. In this context, the aim of the ANR DENDRAC 25 project "Development of dendrometrical tools applied to anthracology: study of the 26 interactions between Man, resources and environments" was to characterize modern-day 27 wood stands in accordance with historical woodland practices and convert dendroecological 28 R e v i s e d m a n u s c r i p t 1.1.
Anthracology and past woodland reconstruction 48 49 The questions raised by relationships between people and the environment in time 50 and space can be explored by archaeological, ethnographic or environmental approaches. 51 The management of the environment for (plant or animal) food strategies reflects, to some 52 extent, human societies and their organization, their lifestyles, their perception of the 53 environment and the landscape in which they operate. 54 Forest exploitation in order to produce wood material for multiple needs, is perceptible 55 at different scales: the tree, the woodland and the landscape (Michon, 2005(Michon, , 2015; Humans 56 R e v i s e d m a n u s c r i p t and transported by humans, are valuable ecofacts reflecting use, techniques and woodland 66 management, and are themselves conditioned by environmental resources (available wood 67 resources, i.e. biodiversity and biomass). 68 69 In forest science, the criteria characterizing wood stands are the composition 70 (dominant and secondary species), stand density (number of stems per hectare), structure 71 (distribution of age and diameter classes of trees) and modes of regeneration (seeded or 72 vegetative renewal) (Rondeux, 1999). The methods usually employed in dendrochronology 73 to extract this information are not suited to anthracological material. In dendrochronology, 74 samples usually come from timber wood and generally from trunks and/or branches or roots, 75 the wood is not charred, the methods are based on statistical tools that require at least 50 76 consecutive rings and it is possible to individualize the signals (study of distinct elements). In 77 anthracology, fragments derive from trunks and/or branches or roots, the wood is charred, 78 fragmented and incomplete as it is partially reduced to ashes, the fragments present on 79 average less than five rings and result from the exploitation of many indistinguishable 80 individuals (Dufraisse, 2006;Marguerie, 2011). Consequently, in the absence of adequate 81 tools, charcoal analysis is most often limited to the study of a list of taxa and their relative 82 frequency without exploiting the information contained in the wood anatomy. The 83 identification of the morphological characteristics of harvested firewood (part of the tree, age, 84 R e v i s e d m a n u s c r i p t shape, etc.) still raises methodological problems even though it is a fundamental element for 85 characterizing firewood exploitation techniques and reconstructing the populational and 86 environmental parameters of wood stands. 87 88 In order to address this need to learn more about forest exploitation and practices, the 89 ANR DENDRAC project "Development of dendrometrical tools used in anthracology: study of 90 the interactions between Man, resources and environments" aims to convert 91 dendroecological data measured on fresh wood material from modern-day oak wood stands -92 corresponding to different types of historical woodland practices -into parameters adapted to 93 charcoal analysis, using a method similar to that developed by A. Billamboz,termed 94 dendrotypology (Billamboz, 2011(Billamboz, , 2014. His method consists in establishing a typological 95 classification of tree-ring series according to their growth patterns. The application of a 96 similar method in anthracology involves associating the identification of the taxa with the 97 examination of dendrological and anatomical parameters; a concept that leads to the notion 98 of dendro-anthracology (Marguerie et al., 2010). Deciduous European oak (Quercus 99 petraea/robur) was chosen for its abundance in temperate forests, its anatomy with clearly 100 identifiable growth rings and its representativity in anthracological spectra. 101 In the present study, we postulate that the characteristics of an assemblage of tree-102 rings can be exploited, without taking into account tree-ring series in terms of time series like 103 in dendrochronology. The first step consisted in developing dendro-anthracological tools 104 based on morpho-anatomical features. The second step was to convert dendroecological 105 data to form an anthraco-typological grid, which could then be used as a key approach for 106 the interpretation of archaeological charcoal assemblages. This approach was applied to 107 three modern-day wood stands: a coppice under standard, a high forest and a young stand 108 formed by a mixture of seeded and coppice trees. Analysis was conducted at different levels: Growth rate is a widely used dendro-anthracological parameter, but the successive 115 tree-rings width series of each charcoal fragment must be localized as precisely as possible 116 on the stem cross-section. In that aim, different dendro-anthracological tools are proposed in 117 order i) to distinguish sapwood from heartwood which provides information about the minimal 118 age of the wood (heartwood formation i.e. duraminisation starts when deciduous oak is 119 around 25 years old) ii) to localize the tree-ring series in respect to the center of the stem, iii) 120 to model dendroecological data from modern wood stands into dendro-anthracological 121 parameters adapted to charcoal analysis. 122 123 1.2.1. The Heartwood-Sapwood discriminating tool 124 125 In some species the coloration of heartwood due to the deposition of lignins and 126 polyphenols makes heartwood recognizable, but the charcoalification process that occurs 127 during carbonization obliterates the colour difference, making this feature unusable in 128 anthracology. Fortunately, in some Angiospermae, such as deciduous European oak 129 (Quercus petraea/robur), the formation of tyloses (cellulose walls expansions) in earlywood 130 vessels is an important feature of the changeover of sapwood to heartwood. However, tylosis 131 formation also occurs in sapwood and increases with the formation of heartwood, from 0% of 132 tyloses in the cambial region and close to 100% in the heartwood. Thus, we quantify the 133 number of vessels sealed by tylosis in order to establish discriminating thresholds between 134 sapwood and heartwood (Fig.1a) (Dufraisse et al., 2016). Trunks and branches of ten 135 deciduous oak trees from 15 to 60 years old were sampled in three stations in order to 136 evaluate the number of earlywood vessels with tylosis in sapwood and heartwood. For an 137 application to archaeological charcoal (tyloses are preserved until 800°C), at least one tree 138 ring and 15 vessels must be counted. The best strategy is to count 50 vessels spread over 3-139 4 tree rings. Thresholds of less than 65% for sapwood and up to 85% for heartwood are 140 R e v i s e d m a n u s c r i p t significant. Besides the discrimination of sapwood and heartwood, the process of heartwood 141 formation starts when deciduous oak is about 25 years old. The absence of heartwood is 142 thus an indication of the exploitation of young wood (trunks or branches). 143 144 2.2. The pith estimation tool 145

146
The pith estimation tool is used to measure the distance between the charcoal 147 fragment and the center of the stem (or the missing pith), named the "charcoal-pith distance". 148 This measurement is taken with the trigonometric pith estimation tool based on 149 measurements of the angle and the distance between two ligneous rays ( Fig.1b) (Dufraisse 150 and Garcia Martinez, 2011;Paradis-Grenouillet et al., 2013). This tool was evaluated on 151 fresh and carbonized oak wood discs with different angle values and distances between 152 ligneous rays. This work enables us i) to propose exclusive criteria (angle < 2° and distance 153 < 2 mm) for reducing the margin of error and improving results in archaeological applications, 154 ii) to establish correction factors linked to the trigonometric tool itself (underestimation of 155 distance values between 5 and 10 cm) and the shrinkage which occurs during 156 charcoalification, iii) to highlight that there are no reliable measurements for charcoal-pith 157 distances beyond 12.5 cm, i.e. diameter of 25 cm (Dufraisse and Garcia Martinez, 2011;158 Garcia Martinez and Dufraisse, 2012). 159 The values were ordered into diameter classes chosen to be compatible with 160 standards used in dendrometrical plans by foresters (Gaudin, 1996 Therefore, an Analysis Diameter model (ADmodel) was developed, based on the fact that a 172 trunk is biologically considered to be a stack of cones ( Fig. 1c)  The general analytical protocol consists in sampling modern-day oak woodlands 187 corresponding to specific archaeological questions, removing logs from felled trees, cutting 188 wood discs from logs and producing experimental charcoal assemblages (Fig. 2). Various 189 kinds of datasets were produced: i) dendrometrical plans to characterize tree morphology 190 and wood stands (composition, structure stand density, regeneration modes), ii) 191 dendrochronological data from wood discs, iii) anthracological data modelled with the With respect to historical woodland practices and to answer to specific archaeological 197 issues such as the distinction branch/ trunk or coppice /high forest three "contrasted" 198 deciduous oak stands managed by the National Forestry Office (ONF) in France were 199 chosen (Fig. 3). The first one is " Les Cagouillères ", located in the Vienne department, on a 200 limestone plateau (altitude: 115m). It is in an old abandoned coppice woodland, about 62 201 years old, currently undergoing conversion to high forest. The second stand is "Bogny-sur-202 Meuse" located in the Ardennes department. This is a coppice-under-standard growing on an 203 acidic brown soil on schists, about 68 years old. The third stand is "Le Bois de l'Or", also 204 located in the Ardennes department, near Bogny-sur-Meuse. This is a young stand, about 15 205 years old, formed by a mixture of even aged seeded and coppice trees (altitude: 350m). The dendrological information for each tree, such as diameter, age, growth rate and 217 radial growth trend, was defined at breast height on the field and from disc located at 1.30 m 218 above ground, as is usual in dendroecology. However, the nature and representativeness of 219 archaeological samples are different in dendroecology and anthracology. Consequently, for 220 R e v i s e d m a n u s c r i p t the conversion in anthracological data according to anthracological constraints, the 221 dendrological data were measured in the whole tree. 222 For the study of tree ring-climate relations in sessile oak, six is the number of optimal 223 trees to sample. For our purpose, and taking into account our archaeological questions, one 224 to five trees were felled and registered meter by meter, one dominant tree in the coppice-225 under-standard at "Bogny-sur-Meuse", four dominant stems from distinct multi-stem trees at 226 "Les Cagouillères" and five coppice shoots and five seeded trees at "Le Bois de l'Or". In order to estimate the relative proportion of trunk and branches for each tree and 234 each stand, each tree was cut into logs of 1-metre-long including branches with a diameter of 235 more than 4 cm. A code was attributed to each log according to its position in the tree 236 (height, number of branches, location in the branch). Length and circumference (at three 237 points) of each log were measured to calculate the mean diameter and the volume. Branches 238 with a diameter of less than 4 cm were packed into bundles according to two diameter 239 classes; 0-2 cm and 2-4 cm. Each bundle was weighed. Sub-samples of wood were 240 collected from each bundle to estimate the density of the wood and then to calculate the 241 volume of each bundle. 242 243 In order to characterize each tree and then each wood stand at different levels (whole 244 tree, and trunks and branches separately), one disc was removed from the extremity of each 245 log. In the present study, a subsample of the set of discs was taken by selecting discs at 246 different heights in the trunk and in the crown (23 discs for the four trees at "Les 247 R e v i s e d m a n u s c r i p t Cagouillères", 14 discs for the tree at "Bogny-sur-Meuse" and 77 discs for the 10 trees at "Le 248 Bois de l'Or" (Table 2a, 2b). 249 The tree-ring widths (discriminating earlywood and latewood) of each disc were 250 measured to the nearest 0.01 mm using a LINTAB measurement device and associated 251 TSAP software (Frank Rinn, Heidelberg, Germany) along 5 radii and averaged in order to 252 reduce intra-tree variability. 253 Each tree-ring was then associated with a diameter class (calculated by the 254 cumulated ring widths) and sapwood/heartwood. Thus, the proportion of sapwood and 255 heartwood was characterized by averaging tree-ring numbers, tree-ring width and wood 256 volume. 257 The usual dendro-anthracological parameters were first independently considered to 258 obtain a "whole tree" estimation, and then the trunks and branches were separated. i) the 259 distribution of growth-ring width, ii) the proportion of sapwood/heartwood, iii) the distribution 260 of the decomposed unburnt wood diameters (UWD) were recorded. 261 diameter of branches less than 7 cm in diameter. Thus, the tree at "Bogny-sur-Meuse" is less 275 R e v i s e d m a n u s c r i p t slender than trees at "Les Cagouillères" and "Le Bois de l'Or" (see the height/diameter ratio, 276 table 1). 277 In the three sampled stands, trunk volume is always predominant and branches are 278 poorly represented. The diameter 20-40 cm class is the best represented at "Bogny-sur-279 Meuse" whereas the 10-20 cm diameter class characterizes "Les Cagouillères". The main 280 volume at "Le Bois de l'Or" is distributed between 7-10 diameter but a few trees reach 11 cm 281 and thus belong to the 10-20 cm diameter class. 282 Radial growth rate and growth trends are different in each stand. Tree-ring widths 283 average 1.23 mm/year at "Les Cagouillères" coppice, and the growth trend has been 284 decreasing over the past 20 years due to strong competition between shoots, intra-tree and 285 between stools. At "Bogny-sur-Meuse", growth-ring widths average 1.35 mm/year and the 286 growth trend has been decreasing slightly over the past 20 years. At "Le Bois de l'Or", 287 growth-ring widths average 2.99 mm/year and are marked during the 1 to 10 first years by a 288 steady increase in the coppice trees while seeded trees are characterized by narrowest rings 289 than coppice from around the pith to 6-7 years, followed by an intensive growth period before 290 a relatively sudden decrease (for more details, see Girardclos  The distribution of the growth rates indicates differences at stand and tree levels ( Fig.  296 5a). First, the difference in growth rate observed in § 3.1 and only based on one disc 297 localized at 1.30 m in the trunk, is conserved when the whole tree is taken into account, what 298 is more realistic for anthracology. The growth rate at "Les Bois de l'Or" is the highest, 299 followed by "Bogny-sur-Meuse" and "Les Cagouillères". For a same stand, we also note a 300 significant difference between trunks and branches, the latter being characterized by a lower 301 rate. Moreover, considering the different parts of the trunk (base, top, upper part in the 302 crown), we note that the annual ring-width in the top of the bole is wider than in the lower 303 R e v i s e d m a n u s c r i p t part, and that the growth rate of the trunk localized in the crown is comparable with branches 304 (Fig. 5b). However, this latter observation is less clear at "Le Bois de l'Or". 305 306 3.2.2. Sapwood/heartwood 307 308 The trees at "Le bois de l'Or", less than 15 years old, are characterized by the 309 absence of heartwood, contrasting with "Bogny-sur-Meuse" and "Les Cagouillères" (Fig. 6). 310 However, at "Les Cagouillères", heartwood formation is not yet initiated in branches. 311 Conversely, the trunk and branches of the dominant tree at « Bogny-sur-Meuse » are 312 characterized by heartwood and sapwood. The relative proportion of sapwood in trunk is less 313 important at "Bogny-sur-Meuse" than at "Les Cagouillères". Likewise, the average number of 314 sapwood tree-rings is less important at "Bogny-sur-Meuse". Nevertheless the average 315 sapwood ring width is higher at "Bogny-sur-Meuse" reflecting more vigorous growth. 316 The unburnt wood diameters (UWD) were decomposed with the ADmodel, according 320 to the relative volume of each hollow cone composing the logs (Fig. 1c, 7). 321 The raw dendrological data indicate that there is little overlap between the diameters 322 of branches and trunks. In fact, the low proportion of the trunk represented in the smallest 323 diameter classes corresponds to the upper part of the trunk localized in the crown. Therefore, 324 for each wood stand, the distribution of the decomposed UWD of branches is clearly distinct 325 from the trunk. Besides, as the volume of branches is weak, the wood diameter pattern for 326 whole trees does not show clear differences with that of the trunk. 327 The first combination consisted in assessing growth trends by characterizing each 332 wood stand. In dendroecology, growth trends are obtained by combining tree-ring width and 333 cambial age. Given that i) the analysis of tree-ring patterns in segment of cambial age is 334 considered relevant for studying forest dynamics and development (Haneca et al., 2005) ii) 335 the distance of the charcoal from the pith can be estimated by the charcoal-pith tool, we 336 combined tree-ring width with diameter classes. For the dendrological data, average tree-ring 337 width was calculated for each cambial age. For the modelled anthracological data, average 338 tree-ring width was calculated for each diameter class (Fig. 8). 339

340
The radial growth trends of the three wood stands are different and the modelled 341 anthracological data correspond well to their dendrological characteristics. Even though the 342 anthracological data are smoother because of the calculation of average ring width per 343 diameter class, the radial growth trend is consisting of i) a strong increase in the radial 344 growth of trees at "Le Bois de l'Or", reflecting a free juvenile growth, ii) the increase followed 345 by a decrease at "Les Cagouillères" due to the high density of trees over a long period of 346 time, iii) a slight decrease in the life of the tree at "Bogny-sur-Meuse" due to a managed 347 coppice-under-standards. However, the differences observed between seeded and coppice 348 trees at "Le Bois de l'Or" are no longer evident. 349 The branches at "Les Cagouillères" and "Bogny-sur-Meuse" are characterized by a 350 lower growth rate than in the corresponding trunks (cf. § 3.2.1.) and by a downward growth. 351 In contrast, the young seeded and coppice trunks at "Le Bois de l'Or", with diameters 352 comparable to the branches, are characterized by a clearly higher growth rate and a more 353 upward growth. 354 The radial growth rate of whole trees is lower in the first diameter classes than in the 355 trunk considered separately, as it includes the low growth rates of branches. Then, radial 356 growth increases from the boundary of the step between diameters of trunks and branches. The second combination aims to improve the interpretation of the distribution of the 361 decomposed UWD by associating them with the presence or absence of heartwood, and the 362 sapwood/heartwood ratio in each diameter class, as decomposed by the ADmodel (Fig. 9). 363

364
The distribution of heartwood/sapwood according to the diameter classes shows 365 specific patterns for the different wood stands and possible exploitation modes (whole trees, 366 trunks/branches separately). 367 At "Les Cagouillères", where branches are characterized by the absence of 368 heartwood, the volume of the trunk is mainly distributed in the penultimate diameter class. 369 The pattern of the whole trees is similar to that of the trunk, as branches only represent 370 9.58% of the volume. At "Bogny-sur-Meuse", the same pattern is observable but the main 371 volume is represented in the last two diameter classes. However, regarding the whole tree, 372 sapwood is better represented in the small diameter classes than at "Les Cagouillères", as 373 branches account for 37.4% of the tree volume. The third combination consists in combining tree-ring width with the decomposed 386 UWD and their respective affiliation to sapwood or heartwood (Fig. 10). 387 Globally, the pattern between whole trees and trunks from a same stand is similar. 388 This is less obvious at "Bogny-sur-Meuse" where no disc from the upper part of the trunk 389 without heartwood has yet been studied. However, we can expect the same pattern, 390 characterized by sapwood and heartwood in all the diameter classes, and a lower average 391 tree-ring width in sapwood corresponding to the external rings, which is coherent with the 392 growth dynamic of trees (Fritts, 1976). 393 The exploitation of branches only is clearly distinct, with a low growth rate and the 394 absence of heartwood in the case of young branches, as at "Les Cagouillères". If branches 395 are a little older as at "Bogny-sur-Meuse", heartwood is absent in the largest classes of 396 diameter. Lastly, the young vigorous seeded and coppice trees are characterized by a high 397 growth rate in sapwood, while heartwood is absent. 398 The dendrological characteristics of each wood stand, discriminating branches, trunks 402 and whole trees, were defined with the help of the dendro-anthracological tools. The dendro-403 anthracological parameters (growth rate, heartwood/sapwood, diameters) were recorded 404 independently of each other and then combined, forming anthraco-types (Fig. 11). 405 406 First of all, annual ring width was considered individually. Considering the whole tree 407 and the trunk, ring width distribution is significantly different among stands. However, the 408 distribution between seeded and coppice trees at "Le Bois de l'Or" is not significantly 409 different. Likewise, the distribution between branches at "Les Cagouillères" and "Bogny-sur- The association of growth rates with the sapwood/heartwood ratio can provide 464 information about the vigour of wood stands and tree morphology. For example, the 465 proportion of sapwood is higher in trunks from "Les Cagouillères" (high forest) than in the 466 trunk of the dominant tree at "Bogny-sur-Meuse" (coppice-under-standard). However, 467 average sapwood ring-width and sapwood width are higher at "Bogny-sur-Meuse" than in 468 "Les Cagouillères" (Fig. 7). This observation shows that i) for a same age (Bogny: 68 years 469 R e v i s e d m a n u s c r i p t old, Cagouillères: 62 years old), the most vigorous trees have a more extensive sapwood 470 surface (Lebourgeois, 1999) ii) sapwood width is higher in coppice-under-standard than in 471 high forest (Dhôte et al., 1997). Thus, the under-representation of sapwood in the trunk of the 472 tree in "Bogny-sur-Meuse" is probably due to a larger tree diameter, 33 cm as opposed to 473 20.75 cm. 474

475
The third combination consists in associating tree-ring width and diameters 476 (distribution of the decomposed unburnt wood diameters). For an application to charcoal 477 assemblage, each tree ring is associated with a charcoal-pith distance, then to a diameter 478 class and finally an average tree-ring width is calculated for each diameter class. Radial 479 growth trends appear to be preserved keeping with dendrological radial growth. An original 480 pattern marked by a low growth rate along the smallest diameter classes followed by a 481 higher rate in the largest diameter classes may be a characteristic of the exploitation of 482 whole trees. However, as it is often the case in dendroecology, one pattern may correspond 483 to several scenarios. Here for example, a partial clearing of the wood stand could lead to a 484 comparable growth trend. Thus interpretations have to be associated with the results 485 established by other disciplines. In addition, an initial distinction between young trunks 486 (coppice) and young branches becomes possible as their growth rate and growth trend differ 487 (high rate and upward trend for coppice, low rate and downward trend for branches). 488 However, no further distinction is visible between coppice and seeded trees at "Le Bois de 489 l'Or". In fact, only the proportion of earlywood is only significant when radius is up to 1,6 cm 490 (Girardclos et al., 2016). forming anthraco-groups is proposed (Fig. 12). 514 515 Each oak fragment is characterized by a charcoal-pith distance, sapwood/heartwood 516 affiliation and annual tree-ring width. The first division at the threshold of a diameter of 7 cm 517 is often used by foresters and corresponds to the diameter limit between branches and 518 trunks in deciduous oak forest. Concerning tree-ring width, charcoal fragments with regular 519 and irregular tree-ring width series are taken into account separately. For example in 520 northern France, according to V. Bernard (1998, p. 96), narrow rings are less than 0.7 521 mm/year and large rings are between 0.7 and 3 mm/year for deciduous oak. Very large rings, 522 up to 3 mm, can be also considered (i.e. 12 groups). 523 The use of this anthracological key enables us to sort charcoal fragments according 524 to their position in the tree. Then, measurements of each batch can be processed separately. 525 R e v i s e d m a n u s c r i p t To close, it is important to make several remarks concerning the dendro-527 anthracological tools and their applications. 528 i) The application of dendro-anthracological tools requires a minimum 529 transversal plane size of about 4 mm x 4 mm and at least one whole growth-530 ring. The optimal number of fragments to analyze is around 100 per sampling 531 unit (structure, layer, etc. according to the problematic). 532 ii) The choice of the diameter classes, chosen to be compatible with standards 533 used in dendrometrical plans by foresters, seems to be relevant. However, a 534 charcoal fragment can be classified in a class or the other when the value of 535 the charcoal-pith distance is close to a limit but usually the interpretation is not 536 affected. 537 iii) Given that it exists a boundary between the diameters of trunks and branches 538 within a wood stand and that the part of the trunk located in the crown 539 presents the same dendrological characteristics as a branch, it is more 540 relevant and accurate for charcoal analysis to distinguish bole from crown 541 than trunk from branch when considering oak and probably more generally 542 Angiospermea. However, by Gymnospermea, the trunk can be easily followed 543 until the apex with a clear separation of the branch material. Thus this 544 distinction bole/crown or trunk/branch has to be adapted according to the 545 architecture of the tree. In addition, variations in growth rates are often 546 considered and interpreted in terms of environmental (light, soil or climate) 547 and human factors (clearings or woodland management). However, we have 548 to keep in mind that they can also result from a change in exploitation 549 techniques (whole trees, trunks, branches). The use of the anthracological key 550 may allow for the classification of the growth-ring width signal and thus bring 551 more accurate information. 552 R e v i s e d m a n u s c r i p t iv) Shrinkage during charcoalification leads to lower tree-ring width. This process 553 is not consistent, depending on sapwood/heartwood and charcoal-pith 554 distance. A preliminary study on shrinkage offers promising results in order to 555 propose correction factors (Garcia Martinez and Dufraisse, 2012). 556 v) The relative frequency of the different taxa in charcoal assemblages is 557 representative of the used biomass (wood volume). In the same way, the use 558 of the dendro-anthracological parameters is based on the assumption that 559 charcoal fragments represent the different parts of trees proportionally to their 560 volume, with their dendrological characteristics (growth, ratio 561 sapwood/heartwood, diameter). That is why the ADmodel is based on wood 562 volume (and not on the number of fragments). However, we stress that, this 563 model cannot reconstruct the quantity of initially burnt wood. 564

vi)
As for the interpretation of tree-ring width (Marguerie, 1992 Besides the measurement of tree-ring width, the present study is based on the 581 development of three anthracological tools consisting in i) measuring charcoal-pith distance, 582 ii) discriminating heartwood/sapwood and iii) modelling dendrological data to make them 583 compatible with charcoal analysis. Three dendro-anthracological parameters i.e. growth ring 584 width, charcoal-pith distances and heartwood/sapwood, modelled with ADmodel, were tested 585 on modern-day oak wood stands chosen with respect to historical woodland practices: a 586 coppice-under-standard, an old coppice undergoing conversion to high forest and a young 587 stand formed by a mixture of seeded and coppice trees. For a more realistic representation 588 of dendrological data according to anthracological constraints, different levels of analysis 589 were considered: the whole tree, and trunks and branches separately, allowing us to further 590 consider various modes of wood exploitation. 591 The dendro-anthracological parameters taken into account independently of each 592 other provide interesting results but rather limited interpretation, especially for tree-ring width 593 or sapwood/heartwood. Indeed the dendrological information cannot be interpreted in the 594 same way depending on its position in the tree. For example, growth conditions and thus 595 paleo-environmental information are essentially recorded in the trunk. In contrast, the 596 combination of the dendro-anthracological parameters highlights specific patterns between 597 organs, stands and regeneration modes, and enables us to establish an anthraco-typology 598 forming an interpretative grid. A major result here is the identification of the position of the 599 charcoal fragment belonging to young woods or internal/middle/external parts of mature 600 woods and the distinction between branches and young trunks when associated with the 601 tree-ring width. These results lead to the establishment of an anthracological key aiming to 602 sort charcoal fragments into anthraco-groups according to their position in the tree and their 603 growth rate.  Table 1 Dendrological characteristics of each wood stand and sampled trees. 618 Table 2a Analyzed wood discs and dendrological characteristics: Bogny-sur-Meuse; Les 619 Cagouillères 620 Table 2b Analyzed wood discs and dendrological characteristics: Bois de l'Or.