A dual geochemical‐phytolith methodology for studying activity areas in ephemeral sites: Insights from an ethnographic case study from Jordan

This study aims to contribute to the interpretation of ephemeral sites by exploring the efficacy of geochemistry and phytolith analysis to identify activity areas in seasonally occupied ethnographic sites. The application of a portable X‐ray fluorescence (XRF) instrument and phytolith analysis to soil samples from six Bedouin campsites at Wadi Faynan, Jordan, provided insights about anthropogenic enrichment patterns and the effects of short periods of abandonment on these. The compatibility of the two analysis techniques and means to combine the results of both are addressed. The results of this study suggest that soil signatures can be found in ephemeral sites following abandonment, even in dynamic and harsh environments. The efficacy of the geochemical analysis to indicate variance within the data was found to be greater than that of the phytolith analysis in these case studies, while certain trends within the phytolith results were more useful in identifying specific activities. Due to the compatibility of the geochemical and phytolith data, it is proposed that a serial or parallel approach should be taken for their statistical analysis.


INTRODUCTION
Ephemeral occupation is characteristic of many pastoral and huntergatherer societies, whose settlement reflects the demands of their highly mobile lifestyles. These short-lived sites do not often preserve well, yet their remains represent many archaeological periods, especially prehistoric ones. Understanding the way in which ephemeral sites were used is important for our knowledge of key developments in human existence, such as the Upper Palaeolithic expansion of humans into challenging new environments, or the transition from mobile hunter-gatherer lifestyles to sedentary farming ones during the Neolithic. These, and other periods characterized by ephemeral occupation, can be difficult to interpret due to the low intensity of occupation, lack of permanent structures, and related poor preservation of many of these sites and the organic remains of which they comprise (Banning & Köhler-Rollefson, 1983;Cribb, 1991;Gifford, 1978). Nevertheless, understanding the use of space in ephemeral structures is vital to their interpretation. This can shed light on past ways of life that are currently underrepresented within archaeological narratives.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. The division of space within human built environments can inform us about subsistence and daily activities, and can also reveal a great deal about notions of cleanliness, sacrality or gender, and relationships with animals or the natural environment (Bourdieu, 1990;Douglas, 1966;Parker Pearson & Richards, 1994).
Current publications summarizing developments in the application of geoarchaeological techniques emphasize the importance of ethnoarchaeological studies (Friesem, 2016) and a multiproxy approach (Canti & Huisman, 2015). In accordance with these observations, this study explores the potential of phytolith and geochemical signatures for understanding the use of space at ephemeral sites through the ethnoarchaeological analysis of soil samples collected from six Bedouin campsites in Wadi Faynan, Jordan. In addition, by considering the variation in length of time after abandonment of the campsites, this research aims to increase our understanding of taphonomic processes influencing these soil signatures within anthropogenic sites.

Description of sites and field methods
The name "Bedouin" refers to populations across the steppes and deserts of the Arab world whose lifestyles were traditionally tied to a nomadic, pastoral existence (Na'amneh, Shunnaq, & Tasbasi, 2008;Saidel, 2009). The mobility of Bedouin populations has decreased over the past century due to external influences, which led to an increased rate of sedentism and modernization, and today herding practices range between transhumance and home range (Na'amneh et al., 2008;Palmer, 2002;Saidel, 2009). Ethnoarchaeological studies of Bedouin campsites became popular in the 1980s as part of efforts to establish a better understanding of different aspects of pastoral and nomadic occupation and to improve our understanding of archaeological pastoral communities. These studies provided insights into the material remains associated with the abandonment of pastoral campsites, the correlation between the location activity areas and refuse, and later studies expanded the methodologies used to identify the layout of Bedouin sites and formation processes related to their abandonment (Banning & Köhler-Rollefson, 1983;Goldberg & Whitbread 1993;Saidel, 2001;Simms, 1988). While micromorphology had been successfully applied for spatial analysis of a Bedouin floor (Goldberg & Whitbread, 1993), other geoarchaeology techniques could also carry potential for the study of activity areas in anthropogenic sites. The ethnographic soil samples analyzed in this study were collected from one occupied and five abandoned sites at Wadi Faynan  (Palmer, Smith, & Daly, 2007).
These data were accompanied by narrative accounts given by the occupants of the area, who provided information about the abandoned campsites and the activities that took place at these. The team conversed with the tent inhabitants in order to gain a better understanding of the use of space at these campsites and where possible, about F I G U R E 2 Photo of occupied Bedouin tent JTW, with a goat and sheep pen on the left and cleared circular areas marking the locations of old animal pens in the foreground Notes. Images of a hospitality hearth and a kitchen hearth can be seen at the bottom left and bottom right (respectively).
the individuals that were living there and the animals owned by them.
During 2000 the same campsites were revisited and studied in greater detail. An artifact distribution study was undertaken, and the soil samples analyzed in this study were collected from chosen sites (Palmer & Daly, 2006). The sampling was aimed at capturing the soil signals of the activity areas at each campsite. Sample locations included the kitchen and hospitality hearths, animal pens, animal pen floors, floor spaces in the hospitality area, and the private activity area and kitchen ( Figure 3). In addition to these, three background  were used was straightforward as features were still recognizable in the field, and samples taken from the animal pens could be described in detail.

Laboratory methods
The geochemical analysis in this study focused on the following with an Ag anode 50 kV, 200 A tube. A helium purge was used in order to lower the detection limits for light elements. The samples were placed in 9 mm plastic cups, covered with a thin polypropylene film, and analyzed using a mobile test stand in order to provide reproducible measurement conditions. The pXRF machine was set to the "mining Cu/Zn mode" and the exposure time for each of the ranges was adjusted to achieve the following settings: the main range was run for 40 s, the high and low ranges for 30 s each, and the light element range for 80 s to allow for reliable readings for elements on the edge of the detection limits of pXRF such as Mg and P. In total, each reading took 180 s. The helium was allowed to flow into the machine approximately 10 min before the first samples were run. when present at more than 5%, except for: Mg, P, and Mn (for a list of precision calculation for NIST standards see Supporting Information).
For Mg precision was only measured for concentrations below 5% and was above 10%, for P precision exceeded 5% when present above 8%, and precision for Mn was better than 10% for concentrations above 5%.
Phytolith extraction was performed using the dry ashing method, where the soil sample is burnt in a muffle furnace in order to remove organic matter and isolate phytoliths (Rosen, 1992

Geochemical analysis
The geochemical analysis of the Wadi Faynan campsites revealed patterns of enrichment and depletion in the context categories most affected by human activity; the hearths, dung sediments, and, to a lesser degree, the animal pen floors (Figures 4 and 5). The variation in abandonment episodes among the Bedouin campsites allowed this study to explore patterns of short-term dissolution.
The only geochemical elements that were found to suffer from a reduc- Generally, most geochemical taphonomic processes are slow, although anthropogenic impact can speed them up in some cases (Mulder & Cresser, 1994).

Phytolith analysis
Trends seen within the phytolith analysis results are more variable and site specific than the geochemical patterns, as was the case in previous phytolith studies of spatial patterning (Portillo et al., 2014 (Table 3)

DISCUSSION
The dual geochemical-phytolith method used in this study was able to identify several correlations between known activities and soil signa- derived from the preparation of bread. The dominance of cereals associated with food preparation could provide an indication for cooking hearths and areas archaeologically, though such trends depend on local preferences for food consumption, the availability of identifiable cereal material, and abandonment processes. The hearth of JTW did not contain a high enrichment of wheat even though bread was prepared on it, as the last activity that took place prior to sampling was the addition of dung cakes, which dominated the phytolith signature for this kitchen hearth.
This example illustrates the influence of abandonment processes and the importance of sampling, which was found to work best when specific activity locations were sampled at the Wadi Faynan sites. The majority of chemical elements and phytoliths measured in this research do not appear to be affected by taphonomic processes within the short span of time differentiating the periods of abandonment of the Bedouin campsites. However, Cl and K concentrations, presum-ably derived from salts, decrease over time more rapidly within dung deposits than with other context categories (Figure 8). This could be in part due to the organic nature of the dung and dung-related sediments they are found in, but probably also as a result of exposure to moisture and sunlight.
Within the phytolith data, the lack of clear influence by the shortterm abandonment might relate to the depositional environment at Wadi Faynan, an arid region characterized by yellow steppic soils (Palmer et al., 2007). Arid conditions have long been considered favorable for phytolith preservation due to the high rates of evaporation under these conditions, which contribute to silica consolidation in the plant cell and a lesser degree of loss of phytolith material in the soil through water seepage (Hillman, 1984). In addition, the initial amount of available silica and the depth of burial will also have an impact on the chemical dissolution (diagenesis) of phytoliths in archaeological sites (Cabanes et al., 2012).
Beyond the depositional environment, the characteristics of the phytoliths themselves contribute to their preservation. The degree of silicification, shape, and surface area all influence durability, and there is evidence to suggest that phytolith dissolution rates vary among different plant taxa and even within a single plant (Bartoli & Wilding, 1980;Piperno, 2006;Wu, Yang, Wang, & Wang, 2013). A recent study by Cabanes and Shahack-Gross (2015) indicated that the dissolution of various morphotypes differs depending on their surface area to bulk ratio. Within the Wadi Faynan material, relative phytolith morphotype composition is similar across sites, and does not appear to reflect taphonomic trends. WF916 stands out in comparison to the other sites, but this is considered to reflect the differences in use of fuel rather than preservation (elaborated below). Similarly, the relative abundance of elongate dendriform at WF940 would not reflect dissolution, as this morphotype would preserve worse, not better than the bulliform type ones less abundant at this site. Overall, there does not appear to be a clear decrease through time in the presence of morphotypes, which would be expected to suffer from dissolution in a way that would exclude nontaphonomic explanations.
F I G U R E 7 Decision tree created for all of the sites based on geochemistry, 77% of cases correctly classified TA B L E 2 Associations between chemical elements and anthropogenic related activities found in earlier studies and in the analysis of the site of Wadi Faynan

Associated activity in this study
P Food preparation and consumption (Fernandez, Terry, Inomata, & Eberl, 2002;Parnell and Terry 2002;Vyncke et al., 2011), burning and food storage (Middleton, 2004), refuse areas (Fernandez et al., 2002), excrements (Vyncke et al., 2011), byres (Wilson et al., 2008), meat (da Costa & Kern, 1999) Hearths, animal dung Mg Wood ash (Middleton & Price, 1996), cooking hearths, food preparation and consumption (Fernandez et al., 2002), meat (da Costa & Kern, 1999) Hearths, animal dung Ca Cooking hearths (Fernandez et al., 2002), food storage and preparation (Middleton, 2004;Vyncke et al., 2011), lime use (Middleton & Price, 1996) Hearths K Wood ash (Middleton & Price, 1996), cooking hearths, food preparation and consumption (Fernandez et al., 2002) Hearths, animal dung Mn Burning (Middleton, 2004), vegetable (da Costa & Kern, 1999) Hearths S Not measured in previous studies Hearths, animal dung Sr Hearths (Wilson et al., 2008), excrements and food preparation (Vyncke et al., 2011), lime use (Middleton & Price, 1996) Hearths Zn Hearths and Byres (Wilson et al., 2008), refuse areas (Fernandez et al., 2002), vegetable (da Costa & Kern, 1999), meat (Tripathi et al. 1997) Hearths (higher concentrations in kitchen hearths) and animal dung Cl Not measured in previous studies Animal dung, hearths, animal pens Fe Craft production (high levels in combination with burning, Middleton, 2004), burning (Vyncke et al., 2011) Background Ti Background (Middleton, 2004) Background Al Background (Middleton, 2004) Background The discrepancies between WF916 and the other campsites, described in the geochemical and phytolith analysis results sections, are probably related to a preference for other fuel sources above dung cakes at this site. The occupants of WF916 belonged to a paramount group, the Rashayda Tribe, who use well-made stone lined hospitality hearths. The members of this tribe prefer the use of wood to dung cakes. The latter are sometimes avoided in the hospitality hearths of all campsites as they produce much smoke, yet preferred in the kitchen hearth because the burning temperature is lower and more consistent than with wood. This difference in fuelling between the two hearth types is most prominent within WF916. This observation also indicates that hearths at the other campsites were enriched with K and P through the use of dung cakes, as their concentrations are significantly lower in the WF916 hospitality hearth (for details see Vos, 2017).
The nature of each method of analysis affects the way in which their results can be processed and combined. One aspect related to this is the level of universal applicability of each technique. The geochemical patterns can generally be directly correlated to known activities such as burning and food preparation, which are associated with specific chemical elements (see overview in Table 2). However, the phytolith trends must be explored within the context of the site since phytoliths derived from activities such as burning or animal husbandry may vary across sites depending on the local availability and use of plants and other materials leading to an indirect phytolith signature (such as the use of dung or wood for construction or fuel).
In addition, differences in the form the results take for each type of proxy influence their degree of compatibility. The measurement level of chemical elements was ppm. This allowed for one type of comparison within the geochemical data, one that is based on the concentrations of elements in the soil. The phytolith assemblages, on the other hand, could be compared through counts of phytolith types, taxonomic identifications, related attributes such as silica aggregate material or weight percent, and also through exploring ratios between related categories based on the phytolith counts such as multi-to single-celled phytoliths, or plant parts. This means that there are different levels of comparison within the phytolith data.
The efficacy of the geochemical analysis in identifying activity areas was found to be greater than that of the phytolith analysis. Decision trees created for each of the two techniques, and including the results of both of them, suggest that combining the variables from both analyses does not provide a better classification of cases than the geochemistry alone. The latter was able to classify 77% of all cases correctly, compared to 60% for a decision tree combining the results of both methods, while the phytolith decision trees classified a third of the cases correctly (Figures 7 and 10). In addition, the PCA scatterplots created for the geochemical data generally showed a better degree of clustering than the PCA scatterplots presenting the phytolith analysis results, and explained a higher degree of variance (Figures 6 and 11).
If the geochemical analysis generally provides the best certainty of identification of activity areas, why bother using the phytolith analy-

CONCLUSIONS
The use of the dual geochemical-phytolith methodology was found useful for distinguishing between activity areas in ephemeral sites, confirming associations between chemical elements and phytolith attributes identified in earlier ethnoarchaeological studies. The results of the ethnoarchaeological analysis support the hypothesis that geochemical and phytolith signatures can be found in the soil at the locations where activities took place, and suggest that soil signatures at ephemeral sites can be preserved in harsh and dynamic environments, in our case those of the Near East. While the surfaces of the Bedouin campsites studied in this research were left exposed to wind erosion and rain after the tents covering them were moved to a different location, they retained phytolith and geochemical soil signatures for at least 15 years.
Within this study, the efficacy of the geochemical analysis was found to be higher than that of the phytolith analysis when it came to identifying activity areas. Decision trees and PCA scatterplots created for the geochemical results of both ethnographic and archaeological data provided a higher percent of correctly identified instances and explained a higher percent of variance within the data than those incorporating the results of the phytolith analysis. This is probably related to the use of dung cakes for fuel, a dominant activity that makes it difficult to distinguish between areas and may mask other traces. Nevertheless, adding information from both methods was found to be more useful in identifying activity areas than only one.
While geochemistry may explain more variance within the data than the phytolith results, the two methods complement each other and provide information about different aspects of activities.
This study suggests that while the results of multiple geoarchaeological proxies cannot always be readily integrated due to differences in the nature of the used methods, a parallel analysis carries much potential and helps combat issues of equifinality and equivocality when studying traces of human activities in soils. Geoarchaeological analyses of activity areas are a fairly recent development, and further studies looking into the compatibility and integration of such techniques will help determine the best approach for studying sites of different scale, date, nature of habitation, and taphonomic disturbance.

ACKNOWLEDGMENTS
We thank the anonymous reviewers and the editors whose com- Faynan Project. We also thank Helen Smith, who collected the soil samples as part of the WFLS and provided much support and guidance during the analysis. We are indebted to Jouma' Aly Za'noon and his family for their friendship, support, and wisdom over many years.