The present study concentrates on locating and compiling data from the two types of inventory survey which provide useful data for management purposes: probabilistic sample surveys and "complete" intensive/systematic surveys. These survey types provide a reliable indicator of where sites are not located - a data type critical to both predictive modelling and informed land management. This data was almost always missing from the large, regional inventory surveys of the 1970's, when over half the current inventory was recorded (Eldridge and Moon 1992:2-3). Of the two groups, the probabilistic surveys are the most useful from a modelling perspective, as these normally included low potential areas in their survey. The systematic surveys are generally confined to narrow shoreline margins along coastlines or rivers, and their results cannot be extrapolated inland; however, they can be extrapolated to similar shorelines elsewhere. Eldridge and Moon (1992:10) found that of most existing surveys:
Few gave detailed accounts [of survey methodologies] and, in many cases, it was very difficult to determine exactly how survey was accomplished, let alone any overall research design. This tends to be particularly true of older reports. Another endemic problem is the lack of information on areas which were surveyed but lacked sites altogether.
In order to find the examples of systematic or probabilistic surveys the Annotated Bibliography (Archaeology Branch, 1992) published on diskette was searched for key words with a word processor. These words included "potentia"[1], "systema"[tic], "predicti"[ctive], "random", ''probabi''llistic], "probaba"pistic (sic)], "quadrat", "transect", "intensi"[sive], and "shoreline". Any reports which included these words in their annotations were found by this method. After the search was completed, searches were made for the names of specific authors whose work had looked particularly useful during the key word search. A few author's names were searched whose work was known to ;be potentially useful but which had not shown up in key word searches.
When a suitable report was found it was appended to a new file containing all the desired reports. Numerous duplications resulting from the Annotated Bibliography's repetitions in different map zones were removed. This resulted in a short list of more than 290 titles. These reports were examined and unsuitable candidates eliminated. A subset of explicitly random or probabilistic reports was then generated and eventually was combined with selected intensive/systematic surveys. A number was assigned to a project or major part of a project and this REF number was applied to all relevant reports on the list and all tables and figures as well as in the text below. The order of the numbers reflects only the order in which projects were processed. The resulting list is attached as Appendix 1 which also includes a few additional sources found during research. This research method does not pretend-to produce an exhaustive coverage, but it was felt to the majority of available B.C. literature, including the "grey literature".
The entire process was made unnecessarily difficult due to the manner of data entry in the Annotated Bibliography. Particularly disabling are irregular formatting, spelling errors, and the frequent substitution of lower case alphabetic "l"'s for numeric "l"'s (which precludes searches for permit and map unit numbers or sorting by date!. Ordinarily, such information would be better entered in a database with error checking capabilities and then reported to a text file for publication or left as a database file for those with the capability to use it. An effective database design could reduce the size of the 1 MByte file by 90% or more.
The reports on the final list were examined in detail. Information about survey methodology, sampling methods, and results were obtained. Differences between the studies in methods of defining study areas, defining sites, and in providing data on methods and results led to a considerable amount of interpretation and reanalysis. Some site types were excluded from calculations since there were such wide variations on how (or even if) these sites were included in the original studies. The principal two exclusions were for historic sites and culturally modified tree sites. Many shoreline lengths had to be determined with a map wheel, often from the less than ideal report maps. Some statistics were recalculated or checked by other means. A few report figures concerning site numbers and locations were checked against 1:50,000 maps in the B.C. Archaeological Inventory and even against original site maps. The data were entered into a spreadsheet, which calculated various statistics concerning sampling fractions, site densities, and so on. The spreadsheets are presented as Tables 1 and 2, located at the end of the main report.
Twenty-one probabilistic surveys have occurred in British Columbia, not including specialized surveys for culturally modified trees. The twenty-one surveys have all occurred in the British Columbia interior, but are spread over much of this part of the province. The greatest number have been conducted in the Chilcotin region, west of the Fraser River (Figure 1). In the following discussion, standard references are supplemented by the reference code number used in the tables and map. Appendix 1 contains more details on these REF's and provides the identifying numbers not found in the references cited section.
The quality of probabilistic surveys is, with a few exceptions, good. Almost all have some weaknesses in terms of their data presentation or utility for predictive modelling. Many are confined to certain geographic zones. As an example, many of the probabilistic surveys were conducted for B.C. Hydro in areas of proposed hydroelectric developments. These projects almost always were confined to pondage areas within valley bottoms. In two cases in northern British Columbia, the valley bottoms were rated as having much lower potential than the nearby subalpine and alpine biogeoclimatic zones-(e.g., Apland 1980b, 1993 personal communication, REF 3; Mitchell and Eldridge 1981, REF 16). Some studies dealt only with specific site types, such as cultural depressions (Blacklaws 1978).
Many projects did not use subsurface testing. These projects, while providing relatively reliable data for open grassland communities, all show a much lower site density for the higher elevation forested zones (e.g., Eldridge 1975, REF 12; Matson and Ham 1974, Ham 1975, REF 14; Matson, Matson and Ludowitz 1979, Matson et al 1980, Magne and Matson 1984, REF 13; Beirne and Pokotylo 1979, Pokotylo 1976, 1979, REF 7; Pokotylo 1975, REF 17). Each of the projects recognised that the forested zone data were not comparable to the grassland data, but none rectified the situation by subsurface testing. Projects where subsurface testing played a large role in the survey strategy (e.g., Eldridge and Eldridge 1979 REF 10; Spurling 1979, 1980, Alexander 1982, 1983, REF 2; Mitchell and Eldridge 1981, REF 16) did not occur in the bunchgrass/Cariboo aspen biogeoclimatic zones. The Peace River project provides figures for the number of sites found purely by shovel testing. In total "54 undisturbed sites with no surface indications were located using this technique" (Spurling 1980:140). If one adjusts the overall Peace River figures to remove these 54 sites then the total sites found drops from 112 to 58, the sites per square kilometre falls from 5.05 to 3.54, and the predicted number of sites drops from 763 to 533. The effect would be even greater if it were possible to apply these figures just to forested strata. A detailed analysis of the biases introduced without subsurface testing is given in Alexander (1983).
A similar bias affects the best reported example of a probabilistic survey of the alpine/subalpine zones. Alexander and Matson (Alexander and Matson 1986, Alexander et al 1985, REF 1) note that site density is very low in the Alpine zone, but that: (1) sites resulting from brief encampments would not be expected in the areas of bare gravels and rocks where visibility is good; and, (2) sites occurring in the lush meadows and small groves of trees would not be found due to vegetation cover, in the absence of subsurface testing. Most of Alexander's sites in the very high density subalpine zone were cultural depression features, which can be easily identified without subsurface testing. This suggests that the extreme differences in site density between the alpine and subalpine zones may be overstated, and that many more sites could occur in the more sheltered parts of the alpine.
Survey in the Lytton to Botanie Lake area found a high site density in the sub-alpine, but little of the alpine was surveyed due to logistical considerations (Baker 1974, Table 1 this report, REF 4). Judgemental surveys in alpine and subalpine zones have confirmed Alexander's assessment of the relative site density differences (Alexander 1992, no REF). Other alpine studies have found extraordinary densities of archaeological deposits in areas of alpine lithic quarries (e.g. Fladmark 1985; Apland 1977; Choquette and Pickard 1989:133-137). For instance, Fladmark notes that in the Central Plateau region of Mount Edziza every non-vertical area not covered by snow seems to possess some scatter of obsidian. For instance, obsidian flakes were observed essentially continuously in the course of a 4-S km walk across the Central Plateau (1985:48). [recorded sites] are part of a continuous scatter of obsidian items that rarely diminishes to a density of less than 1 piece per 4 m2 over the entire observed area of the Central Plateau. In the actual EP 8 and 21 site areas, however, the surface is basically pure obsidian, with densities of 1000-4000 obsidian items per m2. (Ibid:61).
The quality of the reporting (as opposed to the research in general) varies greatly. Some reports are written with vague discussions of research design or methods, and the impression is given of inexperience (e.g., Blacklaws 1978, REF 22; Richards 1981, REF 19; McIntyre and McIntyre 1983, REF 15), while another consists only of a six page interim report (Baker 1974, REF 4). In one report, there were major discrepancies between the text and maps in terms of number and size of sample units surveyed, the number of units in the sampling frame, the map scale, and the number of sites located (Ball 1978, REF 6). However, this work is also the only example of a probabilistic survey in which a simple model was created correlating archaeological site location with environmental data and the data were then extrapolated throughout the study area to obtain zones where similar archaeological site locational patterning would be expected. Some reports failed to present data critical to evaluating site density, such as number of previously recorded sites which lay within sampling units, and these data had to be obtained from the B.C. Archaeological Inventory (Burley 1975, REF 9). Many reports only gave statistics related to sampling units: while this is mathematically sound, a conversion to standard density values and population estimates in terms of number of predicted sites in the study area would have been much more useful (e.g. Matson et al 1980, REF 13).
Of peripheral interest are the ubiquitous minor errors, ranging from use of sample sizes for sampling without replacement when actually sampling with replacement (Matson, Ham and Bunyan 1979, REF 14), to miscounting the number of quadrats surveyed (e.g, Mitchell and Eldridge 1982, REF 16), and many simple minor errors of calculation. Most of these minor errors were revealed in the spreadsheet when calculated values were slightly different to those presented in the Originals.
A discussion of the site density values of the probabilistic surveys is presented following the discussion of the intensive surveys.
No probabilistic surveys for archaeological sites have taken place on the coast of British Columbia, with the exception of specialised CMT surveys which did not explicitly search for buried archaeological sites (e.g., Eldridge and Eldridge 1988; Arcas Associates 1986). A major factor which led to the rejection of probabilistic surveys by coastal archaeologists is the dense coastal vegetation. This can make foot travel extremely difficult and restricts the amount of mineral soil which is visible. The low potential for finding sites in these environments using visual foot sweeping techniques led to the development of subsurface testing. In the interior, the subsurface testing often was combined with probabilistic survey methods. On the coast shoreline inventories were stressed, probably because of the ease of boat access and the heavy pressure of development and modern use of this zone.
Relatively systematic coastal survey began in the southern Strait of Georgia (e.g., Acheson, Cassidy and Claxton 1975 REF 48), where a long term Provincial government inventory program concentrated on shorelines. At much the same time intensive surveys of interior lakes and rivers were starting as part of the same movement to build up the inventory (e.g. Mohs 1974, 1975 REF 39). In much of the Strait of Georgia, erosion from subsidence is combined with a relatively dry environment to make many shore zone sites highly visible. Systematic intensive surveys of the uplifting outer coasts, where many sites are completely vegetated and stranded, created the need for subsurface testing. This need was met with the use of subsurface soil probes which could detect anthropogenic soils (e.g. Hobler 1976:46). Probe testing was eventually combined with explicit standardized survey methods to ensure even coverage of the study area (Haggarty and Inglis 1983, 1984 REF 30). The methodology has become a standard followed for many inventories throughout the B. C. coast (e. g., Archer 1984, 1990a, 1990b, 1991, REF 35; Mackie 1982, REF 28, 1986b, REF 23; Acheson and Zacharias 1985, REF 25; Zacharias and Wanagun 1992, REF 26; Eldridge, Mackie, and Wilson in preparation, REF 27; Marshall 1992, REF 24; McMillan and St. Claire 1991, 1992, REF 34; Eldridge et al 1988, REF 53), and the data resulting from these projects are generally comparable.
An analysis of old and new coastal survey results by Maclcie (1986b:3-5) reveals a wide range of biases generated by older survey methods on the westcoast of Vancouver Island. He notes (Ibid:3) an increase in sites found (excluding CMT's) in resurveyed areas of 140% for the early phase of Archer's work in Prince Rupert, 270% on the Ohiaht Project, 350-600% on the Pacific Rim Project and 560% from the 1982 Meares Island Survey. Recently Wilson, Kennedy and Bouchard (1992:46) have discussed similar increases in site density on Quadra Island in an area thought to be well surveyed. Mackie analyses the Ohiaht Project in more detail and finds that:
Of the 22 previously recorded sites 73% [N=16] were larger than 2,OOO m2; (23% [N=6] larger than 10,000 m2). Ninety two percent (N=55) of those 160] newly recorded were smaller than 2,000 m2. In fact, 47% (N=23) were smaller than 200m2. (Ibid:3).
The complexity of midden deposits was also found to vary, with previously recorded sites being predominantly more complex, and the newer discoveries more frequently simple. There were also differences in some of the locational data associated with sites. A very important point for predictive model building relates to the applicability of ethnographic data to the archaeological record. It was found that a very high proportion (82%) of older recorded sites were still associated with an ethnographic place name, while the inverse proportion (86%) of newly recorded sites were without a remembered name. Mackie concludes that a researcher using the older data would find that the archaeological data "correlated well with ethnographic knowledge" (Ibid:5). Using ethnographic place-name data would obviously contribute to a model which predicts the larger, more complex sites, but equally would do little to predict the large majority of sites.
It is clear that the older survey results are of little use for applications such as settlement pattern or population reconstructions, resource management, ecological analyses, and so on. However, they will be of some use in finding out more about the nature of ethnographically known places. They may be helpful too in locating the focal points of regional activities.
Unfortunately, if soil probes have not been used to determine site boundaries, the sites will be so inadequately mapped and the dimensions so poorly known, that the records will prove of little use for more than locational data.
It is a good bet that similar biases, modified slightly for local cultural differences, will be found in most areas of coastal British Columbia. (One exception may be the Gulf of Georgia where sites are eroding and thus more visible.) (Ibid:S)
In other parts of the province, many surveys of varying degrees of systematization and intensity have taken place, predominantly along river valleys. This report includes those systematic surveys with data which appeared to have a reasonable degree of reliability, especially those in which areas without sites were identified. Many other surveys may have contained equivalent data quality without it being apparent in the Annotated Bibliography.