pages about
today's forests


The Basic Bookshelf

50-year Harvest Statistics

Backlot Logging Equipment

Vulnerable Soils

Productivity Maps


pages about
forest history

The Presettlement Forest

Yellow Island Change

Lost & Found Prairie

Late 19th-Century Landcover

Bigfoot's Forest (incomplete)

Early Forest Composition



     For nearly a century limekilns in the San Juans burned fuelwood from local forests to convert quarried limestone into commercial quick lime. Local opinion holds that by the 1950s essentially all of the county's forests had been cut for this purpose. The caption of a 1920s photograph hanging in the Lime Kiln Café at Roche Harbor makes a typically exaggerated claim that each kiln consumed 192 cords of wood every day, without acknowledging the fact that so much cordwood would occupy a stack four feet wide, four feet high, and more than a quarter mile in length. That preposterous assertion is obviously incorrect, and a healthy skepticism suggests that other folklore estimates may also be wildly excessive. But, if so, what are the correct figures?

     This article offers the first realistic determination of how much fuel and timber were consumed by San Juan County's historic lime industry. The calculations are derived from hard data about how much lime was actually produced, the energy efficiency of the production process, and the quantity of standing timber that was available during the operation of the limeworks. The initial sections of this story summarize the main quantitative results regarding fuel and timber consumption, while deferring the analytical methods. Extended appendices contain background perspectives on: how the limeworks arose and functioned; how various operations differed in scale of production; how the present calculations were carried out; features of the source data behind the calculations; and discussions of accuracy and sources of error in the quantitative analysis.


     Only actual records and objective measures can shed light on how much limestone was quarried, how much quick lime was produced, and how much fuelwood was burned by the various limeworks throughout their ~80 years of operation. Fortunately, business records do exist for Roche Harbor Lime Co., which was by far the county's largest limework. A professional geologist documented the company's annual lime production over a period spanning 67 years (1890-1956) in a technical document commissioned by the state. That report also contains observations concerning efficiency that are key to the present calculations of fuel consumption:

"During the summer of 1951, six kilns were running, burning 3 1/2 cords of woods per kiln every 24 hrs. Four kilns were not running. The kilns processed 120 tons of limestone per day to produce 54 tons of quicklime, or 9 tons per kiln." (Danner, 1966, p. 88).

     These comments establish a direct quantitative relationship between fuelwood use and lime production. It is more direct and presumably more accurate than whimsical assertions about fuel consumption per kiln per day, which may incorporate biased perceptions or misleading best-case exceptions that fail to properly account for such details of kiln operation as down time for cooling, reloading, replacing firebrick linings, and maintenance. Danner's information about kiln efficiencies is independent of the time element in kiln operations; it may be restated more usefully as:

    • 2.22 tons of limestone yielded 1 ton of lime
    • 1 ton of lime was produced by burning 0.39 cords of kilnwood
    • 1 ton of limestone was converted by burning 0.175 cords of kilnwood.

     Because the direct coupling between the product and fuel (lime and wood) is now explicit, Roche Harbor's annual lime production (tabulated in Danner's report) and fuel consumption (calculated from the above relationship) can be plotted together by simply scaling the coordinate axes. Accordingly, cumulative values of product and fuel for the 67-year period are graphed together in Figure 1 (values for 1914-1918 were missing from Danner's report and have been interpolated). This exercise shows that by the end of the 67 years 955,100 tons of lime had been produced and 371,500 cords of kilnwood had been burned. Stated another way, lime production averaged about 14,300 tons per year, and wood consumption correspondingly averaged about 5,500 cords per year.

Figure 1. Cumulative production of lime and consumption of Douglas-fir kilnwood by Roche Harbor Lime Co. during 67 years of operation.

     So, with the quantity of fuel consumption now in hand, we may next ask how much acreage of average forestland in San Juan County was cut in order to produce that quantity. Luckily, that value can also be computed, thanks to a little-known timber inventory (of sorts) that dates from a mere two years after the end of the Pig War.


     The first land survey of San Juan County, conducted by the U.S. General Land Office (GLO) in 1874, included detailed information for more than two thousand "witness trees" from the forests of that time (U.S. General Land Office, 1874). Those data constitute a primitive kind of timber inventory and from them a quantitative description of the county's forest in 1874 can be computed using methods elaborated in Appendix 6. Because the present study is concerned with kilnwood, the full GLO database of witness trees of all species was reduced to include only Douglas-fir trees, because it was the most suitable for fuelwood. Douglas-firs were also the most abundant trees, as they are today. Simply stated, the complex computations show that countywide the average volume density of Douglas-fir wood in 1874 was equivalent to 38.3 cords per acre.


     Now, at last, straightforward arithmetic can derive the land area of forest that was cut to generate kilnwood for the limeworks. Because Roche Harbor Lime Co. consumed 371,500 cords of kilnwood from 1890 to 1956, it follows that 9700 acres (or 15.2 square miles) of timberland were cut to supply them. This amount of forest represents 8.7% of the county's land area, as shown graphically in Figure 2.

Figure 2. Douglas-firs were distributed unevenly throughout the county, as computed from the 1874 GLO witness tree data (left). Countywide, the overall average volume density of Douglas-fir in 1874 was 38.3 cords per acre (right). Roche Harbor Lime Co.'s total fuelwood consumption during its 67 years of operation was 371,000 cords, or the equivalent to 9700 acres of forest of average Douglas-fir density, or 15.2 sq. miles of average forest. Whereas the amount of wood consumption was considerable, it was but a fraction of the county's available timberland, contrary to widespread folklore.

     Of course, San Juan County had many limeworks besides Roche Harbor Lime Co., so total kilnwood consumed would have been greater. However, Roche Harbor's operation was very much larger than all the others put together (see Appendix 2) and a reasonable estimate of its share of overall lime production and fuel consumption is 80%. Therefore, kilnwood consumption for the entire lime industry must have approached one-half million cords, which represents timber from about 10.9% of the county's land area. This amount obviously does not support the common notion that "all of the county's forests were cut to feed the kilns." Indeed, 91% of the county's forest was not cut to feed the kilns.

     So much for "the answer" in a nutshell. The following sections expand upon and explain the quantitative analysis, which is admittedly rather complex in its basis and methodology.


  1. Geologic origin of the limestone deposits
  2. Locations and operational scales of lime production
  3. How lime was produced
  4. Timber "inventory" of 1874 and results of the analysis
  5. Verification of the calculated timber volume
  6. Definitions and the math equations used in analyses
  7. Variables that could affect the estimated acreage of cut forest
  8. Which forests were cut for kilnwood?
  9. Exaggerations explained
10. Photo gallery

Appendix 1. Geologic origin of the limestone deposits

     Limestone deposits are uncommon in the Pacific Northwest. Those in the San Juans derive from ancient geologic events involving tectonic convergence between a submarine San Juan Plate (which is largely calcareous having been formed the exoskeletons of small marine animals), and the relatively bouyant North American Plate (a continental landmass largely granitic in composition and devoid of limestone except in regions east of the Mississippi that were transiently inundated by ocean). The San Juans are complex amalgams of geologic materials derived from both of these tectonic plates due to subduction, a process which is now nearly complete.

     As geologic fragments were trapped in the turbulent subduction zone, they were first dragged many miles beneath the Earth's surface; calcareous rocks partially metamorphosed under the high pressure, thus obliterating most of the fossil exoskeletons. Some fragments were subsequently thrust back to the surface and in a series of accretional events formed the heterogeneous jumble that is now called San Juan Archipelago. The common assertion that the San Juan Islands represent "the tops of a submerged mountain chain" is a romantic but discredited myth. The notion originated with a pre-tectonic plate treatise on island geology by McLellan (1927) that persists thanks to uncritical merchandisers (Real Estate, n.d.) and others who should know better (U.S. National Park Service).

     As a result of these powerful geologic events, current limestone bodies in the San Juans are intimately fused with rocks of quite separate origin, including basalt, comprising a geological oddity. Deposits include surface outcroppings and subterranean masses.

Appendix 2. Locations and operational scales of lime production

     Because lime was an essential ingredient of commercial cement and concrete and was also used in agriculture and paper-making, quick lime was an important and highly profitable export product in the early history of San Juan County. Deposits that were most suitable for exploitation were close to forests (as the source of fuel) and the seashore (for shipping). Moreover, Seattle, San Francisco, and other fast-growing cities represented almost unlimited markets.

     The earliest Europeans in the region noticed the scattered deposits of limestone, although quarrying prior to 1870 was minuscule. During the joint U.S.-British military occupation of San Juan Island (1859-72) Lt. Roche of the Royal Marines noticed limestone outcrops above the shores of a protected bay that bears his name today (see Appendix 10, Figure 7). After the occupation, in 1874 meander surveys by the GLO recorded numerous "bars or ledges of limestone" throughout the islands. Not all were equal in quality.

     Limeworks can be roughly sorted into small-, medium- and large-scale operations with only one member in each the last two classes. Although about seventeen small-scale limeworks sprang up in 1860s and 1870s (Danner, 1966), they were primitive and short-lived. According to Richardson (1971, p. 203), "Lime quarries on the island (Orcas) had proved second-rate." The most successful of the small operations was located at "Eureka" on the northeast shore of San Juan Island. It began rudimentarily around 1860, discontinued for several years, reopened as the Eureka Lime Co. in about 1874, and continued intermittently for another ten years. Gilbert (1895) described its remains: "This limekiln has been abandoned and buildings are fast falling to decay. It was one of the largest of the many plants at one time in operation among the islands: - there are two substantial kilns and a number of vacant houses beside a good wharf andn large warehouse." In its heyday Eureka employed and housed several workers and briefly sported a post office, but today all that remains are a few overgrown, shallow quarry pits. Limestone blocks from the kilns have been used in nearby home construction. In terms of limestone quarrying and timber consumption, the impacts of Eureka and the even lesser enterprises were truly negligible when compared to the two much larger limeworks, namely Cowell Lime Co. and Roche Harbor Lime Co.

     A massive limestone deposit located on the steep western shore of San Juan Island was first mined in the early 1870s. Around 1880, when it produced about 2000 tons annually (Danner, 1966, p. 109), it was bought up by Henry Cowell, who already owned substantial limeworks in Santa Cruz County, California (Kelly et al., 1990). The Cowell Lime Co. eventually included extensive acreage of upland timberland and continued in operation for several decades. The enterprise was perched precariously near the shore with the quarry face looming hazardously above the buildings. It comprised three kilns and a functional but unremarkable infrastructure that included a rail system for transporting materials by handcart down the steep slopes, a warehouse for filling and storing barrels, and an inadequate wharf that was compromised by exposure to deep water and active currents. Cowell's on San Juan Island can be classed as medium-scale.

Figure 3. Locations on San Juan Island of the county's principal limeworks: Roche Harbor Lime Co. was the most productive; its main rival was Cowell Lime Co.; and Eureka Lime Co. was briefly active but small in scale. Historic land-cover and topographic maps of each site are from Gilbert (1894-5).

     After Henry Cowell died in 1903 his sons kept the company operating until 1940. Production may have peaked during the rebuilding of San Francisco after the 1906 earthquake and fires, but period photographs and the great mass of unexploited limestone that remains today suggest that output levels were never very large and that closure was not due to exhausted quarries. There may have been little reinvestment; one of the limekilns was particularly rudimentary and the other two large, stone-block kilns were merely satisfactory even from a 19th century prespective. Unless the company's business records are discovered, the precise amount of lime produced cannot be ascertained; it certainly was but a small fraction of its competitor 's at Roche Harbor.

Figure 4. The Cowell Lime Co. (~1920). Despite its large deposit of limestone, the operation was hindered by steep terrain and by an excessively exposed moorage.

     Roche Harbor's lime production also began early in a succession of small operations, until a classic, autocratic capitalist bought into the enterprise in 1886. John S. McMillin, previously of Tacoma, expanded and reorganized the quarrying and processing into the prodigiously successful Roche Harbor Lime Co. Fortunately for the present story, the company's detailed records of lime production from 1890 to 1956 are available for analysis. Fuelwood consumption was calculated from them (Figure 1).

     McMillin's Roche Harbor Lime Co. overshadowed all of its competitors by a combination of natural advantages, bold business management, and lucrative political stratagems. In the first category, limestone deposits at Roche Harbor were extensive yet discretely separated as several quarries; the landscape was generally low and manageable; and the low-bank harbor offered exceptional moorage. Nearby timberlands were well-stocked and productive.

     It was also important that the owner resided adjacent to his operation and provided on-site supervision, unlike Cowell who remained in California. Owing to an overbearing management style, McMillin built a quintessential 19th-century capitalist "company town" complete with workers' housing, wages paid in private scrip redeemable only at the company's general store, a post office, and even a medical facility, all of which stabilized a large workforce.

Figure 5. McMillin's technologically advanced Roche Harbor Lime Co. (~1920). Eight efficient steel kilns (right, behind two-story building) complemented older stone-block kilns (left). Facilities for lime processing, warehouse storage, shipping fleet and moorage were unmatched by other lime operations west of the Mississippi.

     McMillin aggressively invested company capital to expand and improve the technological facilities for quarrying limestone and producing and marketing lime. The capacity and through-put of the limekilns was increased by augmenting less efficient stone-block kilns with a battery of eight advanced steel kilns; four stone kilns are visible in Figure 3 (dated 1894) but only two remain in Figure 5 (~1920). Limestone and lime were conveyed by progressively more mechanized means (horses, steam locomotives, and finally trucks); large diesel generators produced electricity to operate crushers and other devices; fuelwood was collected from remote producers by water using one or more 100-cord scows, and thousands of cords were stored for drying; expansive warehouses stored an immense, 20,000-barrel inventory of lime ready for shipment; and the wharf was improved to handle a steady traffic of shipping vessels, including the company's own fleet, and visitors.

     McMillin even controlled his own supply of lime kegs by creating a subsidiary barrel-making company on the north shore of Roche Harbor. The airtight barrels were manufactured from peeled wood (a thick veneer) instead of more labor-intensive curved staves. Fifty factory workers could produce a few thousand barrels per day. A barrel accommodated 2 cubic feet, or 200 pounds of lime (ten barrels per ton), measured about 15" by 26" and could be manipulated by one or two men. The company sold a barrel of lime for about $2 (equivalent to $25 today).

     Ever the capitalist tycoon, McMillin excelled in the political arena. As boss of the state Republican Party, he cultivated powerful associates and customers far and wide, including President Theodore Roosevelt, by entertaining them at a sumptuous hotel built on the grounds of the lime operation. His friend and business associate Victor J. Capron lived nearby (Appendix 10, Figure 9), was the company's resident physician, served in the state legislature and as mayor of Friday Harbor, and owned over a thousand acres of forestland adjacent to the much larger holdings of the Roche Harbor Lime Co. (see Appendix 8).

     This company flourished for several decades until the quality of limestone began to decline, whereupon in 1956 everything was sold. Today many components of the original enterprise have been converted to resort amenities, tourist attractions, condominiums and other residential areas, and timber resource lands. Surviving records of Roche Harbor's limestone quarrying and lime production detail 62 years of operation (1890-1913 and 1919-1956); after interpolating the missing five years, these records permit fuelwood consumption to be calculated for 67 years, which is nearly the entire period of operation.

     This company vastly out-produced all other limeworks in San Juan County, as indicated by its unmatched scale, huge success, and longevity of operation, but absolute comparisons are not possible since it is the only one whose production records have been discovered. However, it seems reasonable (if also arguably speculative) that McMillin's operation produced about 80% of all lime ever generated in San Juan County. It would have consumed an equivalent proportion of all fuelwood ascribable to the entire lime industry. Cowell Lime Co. accounted for most of the remaining 20%, whereas the pooled share of the Eureka Lime Co. and other small producers was relatively negligible. This assumed figure of 80% allows total county kilnwood consumption to be estimated at 1.25 times that Roche Harbor Lime Co.

Appendix 3. How lime was produced

     When limestone is heated to 1648 degrees Fahrenheit its calcium carbonate emits carbon dioxide and water vapor and leaves calcium oxide or quick lime as a solid residue. Quick lime may remain as lumps or be crushed to powder, but if it is exposed to water or air it soon reverts to calcium dioxide and eventually to calcium carbonate, which accounts for the tight shipping barrels.

     In order to "burn" limestone in a pot kiln, alternate layers of coarsely crushed limestone and fuelwood were placed in the enclosure and the wood was ignited. Eventually, quick lime was "drawn" from the bottom of the kiln's chamber. Simple kilns needed to be cooled between loads, but more modern, higher through-put designs operated continuously as segregated materials were added from above and withdrawn from below.

     In the San Juans local timber provided the fuelwood and it was almost exclusively from Douglas-fir trees. This wood is very dense and has the highest heat value of any conifer in the islands (lodgepole pine also has high heat value but is much less abundant). Just as today, the historic forests contained many species, but Douglas-firs were by far the most abundant (countywide relative frequency of 0.56); and even though growth in the San Juans is slow, Douglas-firs are long-lived and can eventually grow to large size.

     Fuelwood used in limekilns was typically cut to 4-ft lengths and coarsely split into large wedges and allowed to dry for about six months. Such "kilnwood" was produced by laborers attached to the limeworks as well as by independent suppliers. The standard measure of kilnwood was the cord, which is an orderly stack measuring 4 x 4 x 8 feet or 128 cubic feet. A cord of firewood for domestic use, which is cut shorter and split into smaller pieces, customarily includes about 25% air space and only about 96 cubic feet of actual wood; in contrast a cord of kilnwood contains about 10% air space and 115 cubic feet of wood. In 1956, when Roche Harbor Lime Co. was discontinued (Anonymous, n.d.), 4000 cords of kilnwood were left behind in the drying sheds. That was enough to supply active kilns for about a month.

     During the decades of lime production, logging was carried out exclusively by hand (the chainsaw became commonplace only in the 1950s). One- and two-man "misery whip" saws were quite efficient when kept sharp and when the kerf was kept open with wedges; they were employed both for felling and bucking. Four-foot logs were split by maul and wedges. Splitting was possible as far up a tree trunk as the absence of large branches (knots) allowed, which was usually more than halfway; tops were frequently left on the ground. Kilnwood was typically transported overland by horse-drawn cart (Appendix 10, Figure 10). In situations where a high bank overlooked the water, simple chutes conveyed kilnwood directly onto a beach or a buyer's scow.

     During a long day of work, a skilled woodsman could produce a cord and a half of kilnwood per day; at that rate, all of Roche Harbor Lime Co.'s kilnwood would have required nearly 800 man-years of labor! A cord sold for $1.25 to $1.75, which was about a laborer's average daily wage. Assuming that Roche Harbor Lime Co. purchased or produced kilnwood at a cost of $1.50 per cord, their outlay for 371,500 cords over a period of 67 years was $557,250, or an average of $8317 per year. For the sake of comparison, income from lime sales over the same period was $19 million, or about $285,000 per year.

Appendix 4. Timber "inventory" of 1874 and results of the analysis

     The amount of forest that was cut and burned to produce lime reflected the number, size and species of the trees that were available during the relevant historical period. In 1874 the federal GLO commissioned three teams of surveyors to lay out a one-mile grid of N-S and E-W section lines (Appendix 10, Figure 11). Points of line intersection, called section corners, fell randomly with respect to a landscape that was primarily, but not entirely forested. The surveyors erected posts at each corner and then documented the nearest trees, one witness tree for each compass quadrant (Appendix 10, Figure 14). Each corner tree was identified by species, diameter in inches, compass bearing, and distance from the reference point in links (1 link = 7.92 inches; see Appendix 10, Figure 12). Halfway between section corners, at the so-called quarter corners, two additional witness trees were similarly identified, one on each side of the line. The present study relies on these two categories of witness trees, however GLO surveyors also recorded additional witness trees: "line trees" that lay directly along section lines, and "terminal pairs" where a section line was interrupted, as at a shoreline).

     Four considerations present themselves in connection with the GLO data: 1) the relation of surveyors' reference points to landscape features such as forest stands is independent and systematic, though not rigorously random; 2) the idea of a "nearest tree" was susceptible to personal bias, even though the GLO instigated rules concerning these matters; 3) some surveyors clearly resorted to crude estimation of tree diameters, instead of actual measurement; and, finally, 4) a one-half mile sampling grid is a lamentably dilute way to evaluate the county's 175 square miles of land area (the main goal of the survey was to accurately lay out sections, not to scientifically sample the landscape). Despite its limitations, the GLO database offers an unprecedented, quantitative glimpse of the forests as they were in 1874, which was merely two years after the islands were incorporated into the U.S. and as the lime industry was just starting up. The GLO field notes mention 2074 trees of all types and situations, but of those only 693 were appropriate for calculating kilnwood consumption (Douglas-firs adjacent to section corners and quarter corners, which amounted to 490 reference points).

     Generalizations about forest conditions are always fraught with complexities. Any statistical assessment of San Juan County's forests would fail to capture their inherently high degree of patchiness, which is due to the landscape's wild irregularities of terrain, diverse hydrology, and varied natural history. Even today these forests are poorly characterized, and no scientifically valid, countywide inventory of the current forests exists. Seen in this light, the timely GLO "inventory" is a fortuitous miracle, even if its sampling is woefully sparse and problematic.

     The amount of wood in the forests in 1874 can be calculated from the GLO's raw data for each island, as described in Appendices 4 and 6. Since the survey directly recorded tree species and diameters, proportional species composition (relative frequencies), and average diameters (either arithmetic or quadratic mean diameters) are easily calculated. Tree densities (in trees per acre) can be arithmetically derived from the distances that separated witness trees and their reference points, and, in turn, tree densities lead to total numbers of trees per island and average tree-to-tree distances. Tree densities combine with diameters to derive "basal area," which expresses the cross-sectional area of wood per area of land (in square feet per acre). Simple assumptions about average tree height then combine arithmetically with basal area to yield timber volume densities (here computed in terms of cords per acre in order to relate directly to the central question of the quantity of potential kilnwood that was available to the limeworks). These calculated values are shown in Table 1; see Appendix 6 for the methodological equations, Appendix 5 for validating cross-checks, and Appendix 7 for factors that could alter the estimates.

               Table 1. Principal metrics of kilnwood as calculated from the GLO forest database of 1874. Values are derived from and relate exclusively to the Douglas-fir component of the forests.

     Statistical analysis of the GLO database (Table 1) reveals that in 1874 San Juan County contained nearly 5 million Douglas-fir trees and over 4 million cords of potential kilnwood throughout its 175 square miles of land area. The primary consumer of kilnwood, Roche Harbor Lime Co., utilized 371,500 cords during operations from 1890 to 1956, or 8.7% of all available Douglas-fir timber.

      Notes concerning Table 1:

    • The table refers exclusively to Douglas-fir trees, which were the primarily source of kilnwood. Other species of trees were filtered out of the calculations.
    • The unit labeled San Juan includes San Juan Is. and nearby Henry Is.
    • GLO data was examined for the seven largest islands in the archipelago, whose combined land area is 93.7% of the total area of the county. Altogether 693 witness trees were analysed (332 corner trees and 361 quarter-corner trees).
    • Tree Density on San Juan Is. is depressed relative to Orcas in part by the diluting effects of a few thousands of acres of treeless prairie in the southern half of the island. Even though the two islands are about the same size, total tree volume is much less on San Juan due to Douglas-firs that are substantially smaller as well as somewhat fewer.
    • Tree Density on Lopez Is. is depressed because a) some areas were "much injured by fire," b) "near its center is a prairie of nearly a square mile in extent also a small one near its northerly extremity," and c) other parts were "bottom land covered with fern & alder," all conditions that would have excluded Douglas-firs (U.S. Boundary Commission, 1867, pp. 118, 142).
    • Tree Density on Shaw Is. is depressed by wetlands that were drained after settlement. According to U.S. Boundary Commission (1867, p. 150), "the timber, consisting of fir and cedar, is small and scattered. Valleys are small and generally very swampy, and are rendered almost impassable by thorny bushes everywhere heaped up in tangled masses."
    • Height data were not collected by the GLO surveyors, so in order to derive volumes this analysis assumes a standard height of 100 feet for all trees. This estimated height is approximately correct as an average for trees of age 100 years or more under the slow-growing conditions in the San Juans.
    • Average Spacing between trees refers only to Douglas-firs; relative frequency (of Douglas-firs) suggests how many trees of other species might be interspersed. An historic photo possibly captures the condition of a pre-settlement forest that accords well with the metrics in Table 1 (Appendix 10, Figure 14); it shows large, widely separated trees, presumably mostly Douglas-firs, on Blakely Island before logging of any kind had begun.

Appendix 5. Verification of the calculated timber volume?

     Can the present analysis of the 1874 database be cross-checked for accuracy? Perhaps the only realistic possibility is to compare the analytical results against modern forest-growth expectations as elaborated by professional foresters. Published yield tables are designed to predict parameters of Douglas-fir trees (e.g. density, diameter, height, and volume) that should occur in regenerating forest stands on sites of specified capability and at specific ages. According to Atterbury (1990, p. 39), an idealized 100-year-old, fully-stocked stand of Douglas-fir on a site of "50-year site index 95" (which is somewhat higher than most sites in the San Juans) would be expected to have a basal area of 185 square feet and a gross volume of 6000 cubic feet (which converts to 52 cords/ac of kilnwood or 61 cords/ac of firewood). These measures are similar to those have been determined from the 1874 database (actually slightly higher, as is the site index of the Atterbury yield table).

     Another yield table (Curtis et al., 1982, Table 1A) does not agree so well. It claims that a 100-yr-old natural-regeneration forest (that is, one with no planting or management treatment) of pure Douglas-firs of 50-year site index 85 would be expected to yield a gross volume considerably greater than what was found in the analysis; it would exhibit more than twice the tree density but only half the average tree diameter. In other words, this yield table predicts a very different kind of forest.

     Perhaps the most validating conclusion one can make from these cross checks is that the 1874 values are at least of the same order of magnitude as predictions from yield tables. After all, the situations being compared are dissimilar; forestry yield tables are designed for quasi-monoculture, second-growth tree farms, whereas forests in the San Juans in 1874 were naturally mixed stands in which 44 trees out of every 100 were of species other than Douglas-fir (relative frequency of 0.56). The original forests had probably experienced environmental disturbances whose intensities, extents and frequencies can barely be estimated, including frequent low-intensity fires and localized wind damage. Finally, many of the surviving Douglas-firs in 1874 were large enough to be ~250 years old, unlike the 100-year endpoints of the yield tables.

Appendix 6. Definitions, equations, and methods used in analyses

    • Relative Frequency of Douglas-firs is the fraction of all witness trees of the unit that are Douglas-firs. For additional definitions of forestry terms, see Hanley et al. (1987).
    • Dominance is the proportion (as a percentage) of the Basal Area of a given species in a forest stand, relative to the Basal Area of all species present.
    • Tree Density (in trees per acre) relative to a point (i.e. section corner or quarter corner) is described by the equation:
                                       (Equation 1)
      where distances (in links) are measured from each witness tree to a reference point; n is the number of witness trees referenced to the point (maximum of 4 for section corners and 2 for quarter corners; and the coefficient C = 4 when n = 4, C = 2.43 when n = 3, C = 1.31 when n = 2, and C = 0.5 when n = 1. This important "point-centered quarter" equation is based on equations derived by Delcourt (1976), based upon prior studies by Morisita (1954) and Cottam and Curtis (1956). Tree densities at points, as such, are not displayed here but they underlie computations of Tree Densities for geographic units composed of many reference points.
    • Tree Density of a geographic unit, such as an island, is the arithmetic average of all Tree Densities at individual points in that unit.
    • Arithmetic Mean Diameter (in inches) of a set of trees is the sum of all individual diameters divided by the number of trees in the set. These values are not displayed in Table 1. This value may also be termed the mean observed diameter.
    • Quadratic Mean Diameter (in inches) of a set of trees is described by the equation:
                           (Equation 2)
      For deriving the amount of wood (Basal Area or Timber Volume), this value is more valid than the Arithmetic Mean Diameter.
    • Total Trees is the product of the unit's average Tree Density and the area of the unit (in acres).
    • Basal Area is the cross-sectional area of trees per unit of land area. Basal Area at a point (in square feet per acre) for the nearby stand is described by the equation:
                                  (Equation 3)
    • Average Basal Area for a geographic unit = (Quadratic Mean Diameter of all trees in the unit / 24)squared x 3.1416 x Tree Density of the unit. This is not the same as the arithmetic average Basal Area of all points in the unit.
    • Average Spacing for a geographic unit (in feet) = square root of (43560 / Tree Density of the unit). There are 43,560 square feet in an acre.
    • Volume Density of a geographic unit (in cords per acre) = Basal Area of the unit x 0.29. The dimensionless coefficient of 0.29 follows from assumptions a) that all trees were 100 feet in height, which is about average for local trees at age 100 and b) that entire trunks from base to apex were converted to kilnwood, according to the equation, BA x C = (BA x H / 3) / V, where BA is Basal Area (in square feet), C is the coefficient, (BA = H / 3) (in cubic feet) is the volume of the cone of base area BA and height (H) of 100 feet, and V is the volume of wood in a cord of kilnwood that includes 10% air space (that is, 128 x 0.9). When the equation is solved for C, BAs cancel and C = 100 / (3 x 115.2), or 0.29. See Appendix 7 for incidentals that could moderately increase or decrease these final results.
    • Total Volume of a geographic unit (in cords) is the product of the unit's Volume Density and land area (in acres).
    • Methods of calculation and the sequence of steps are important in order to maximize consistency and accuracy; in particular, steps must be taken to compute averages correctly, especially in cases where quadratic functions are involved (the sum of the squares of numbers is considerably smaller than the square of the sum of the numbers).
      • Organize witness tree data by section. GLO surveys present up to eight witness trees per section (including four corner trees and two pairs of quarter-corner trees, but not including line trees or pairs of trees that witness the ends of fractional section lines). In the present study a section's representative witness trees are the four trees from the section's northwest corner and the trees from quarter-section point to the east and south of the northwest corner. These eight (or fewer) trees form an array that stand for the section, even though they are offset from center toward the northwest.
      • Tree Density for a section is calculated according to Equation 1 above, but as the arithmetic average of three individual tree-density calculations for the three reference points (one section corner and two quarter-section points) for the reason that the factor C has been generated for no more than four trees.
      • Tree Density for a large geographic unit (larger than a section, for example an island or an assemblage of islands) is the average for all sectional values in that unit. Thus, the average Tree Density for San Juan County is not the average of each island Tree Density (in part because the islands vary in land area), rather it is the average of all sectiona Tree Densities.
      • The Quadratic Mean Diameter of a geographical unit is calculated after Equation 2 above in which the squares of all tree diameters in the unit are averaged. It is not accurate to take an average of Quadratic Mean Diameters of separately computed subunits; that is, the value for an island is otained by summing the squares of all witness trees throughout the island's sections, not by first computing the value for each section and secondarily averaging sectional values.
      • Basal Area per Acre. The above (correct) way of calculating Quadratic Mean Diameters for geographic units of different size will carry over to calculations of Basal Area of the same units using Equation 3, assuming that the Tree Density value is also appropriate to the geographic unit.

Appendix 7. Variables that could alter the estimated acreage of cut forest

Factors that would INCREASE acreage

    • Here fuel consumption has been calculated only for Roche Harbor Lime Co. production, with a 1.25-times extrapolation to the entire industry. If new evidence shows that Roche Harbor's proportion of total fuel consumption was less than 80% (signifying that the other limeworks were more productive than has been estimated), the acreage of cut timberland would increase.
    • If only the lower halves of trees (87.5% by volume) were cut for kilnwood, and the upper halves (12.5%) were wasted, the acreage cut would increase by 0.14 (i.e. multiply consumption by 1.14).
    • If only some Douglas-fir trees were selected for cutting, say by certain criteria of diameter or lack of low branches, more acreage would have been cut, though less severely.

Factors that would DECREASE acreage

    • If species other than Douglas-fir were used, acreage consumed must be decreased, but the amount would never be more than more than 10-15%, because Douglas-fir had a Dominance of about 87% in 1874. That is, Douglas-fir trees were not only more numerous than other species, but also generally larger in diameter (and therefore basal area).
    • If the yield tables of Curtis et al. (1982) unaccountably provide a more accurate estimate of forests in 1874 than the GLO survey, Basal Area would be much larger than presently calculated and trees would have been taller, thereby adding 50% more wood volume per acre. In other words, according to Curtis' prediction, less acreage would have been cut in order to extract the same amount of kilnwood.
    • The estimated wood volume was based on GLO data from a "snapshot" taken in 1874, yet kilnwood consumption spanned several decades beginning well after that date, during which time natural tree growth would have increased wood volume per acre. Whatever such incremental growth might have been, it would have significantly reduced the amount of forestland cut for the needed amount of kilnwood.

Appendix 8. Which forests were cut for kilnwood?

     The results of this analysis dispute the notion that timberlands in the San Juans were exhaustively cut over just to feed the limekilns. In fact less than 10% was cut for that purpose, and the cutting was spread over a period of about 80 years. Nevertheless, one might inquire where the cutting done, and a place to start this investigation is by examining the pattern of land ownership.

     Clearly, large landholdings of forest were consolidated by the large limeworks. By the time that competition had winnowed the active operations down to only two, both Roche Harbor and Cowell Line Co. were also the preeminent landowners on San Juan Island, as illustrated in Figure 6. The case with Roche Harbor was complicated by the participation of McMillin's business partner, Dr. Victor Capron, as another owner of large holding. For years after its quarrying had ceased the Eureka Lime Co., by then called the Hidalgo Lime Co., also owned nearby forestland. These four owners controlled 6410 acres or 10 square miles of forestland adjacent to their respective centers of lime production. Their holdings contained enough timber to have furnished their kilnwood needs. But were these company lands actually cut? 

Figure 6. Property-ownership map (Metsker, 1933) highlighted with the major landowners (three limeworks and V.J. Capron, a major business partner with McMillin at Roche Harbor Lime Co.) By this date the Eureka Lime Co. had been abandoned for decades, and its lands were owned by an inactive entity named Hidalgo Lime Co. The limeworks-associated lands contained sufficient timber for all of the needed kilnwood, but that fact does not establish that they were cut for that purpose.

     It is not at all evident how much of the forestlands owned by lime interests were actually cut. Were they cut first or, on the contrary, held in reserve? No hard evidence has been found on the subject. As discussed in Appendix 9, many arguments can be marshaled for and against the cutting of kilnwood on company lands. After all, there was abundant non-industry timber elsewhere in the county, including one million cords on San Juan Island alone (see Appendix 4, Table 1), which is three times what was ever used by all of the kilns. Oral histories suggest that Roche Harbor drew heavily from timber on Waldron, Henry, and Stuart Islands, as well as northern San Juan Is. away from its own property.

     Local folks who were contemporaneous with the limeworks were surely impressed with the seemingly endless cutting and splitting of timber for kilnwood. But memories could easily have conflated that activity with timber cutting for other purposes. How extensive were other forms of timber exploitation during the active limekiln period? Relatively how extensive was cutting for domestic firewood, pilings for salmon "traps," fueling steamboats that plied the county's waterways, and, of course, lumber? A brief review of each of these uses demonstrates how they pale in comparison with the extent of timber cutting for kilnwood.

     At the turn of the 20th century there were fewer than seven hundred households in San Juan County. If each family burned ten cords for cooking and heating, again presuming the firewood was Douglas-fir, their total consumption of 7000 cords per year could have been obtained from 0.3 square miles of average timberland. This is somewhat more than what was annually cut for kilnwood.

     Pilings for salmon traps were also Douglas-fir logs; they totaled a few thousand in the two-decade fishing heyday straddling the turn of the century. But pilings required more maintenance than outright replacement, so less than 50 acres could have furnished all of the logs. As for steamboats, regardless of their efficiency they operated only briefly before being replaced by diesel-powered craft in the 1910s and early 1920s and so their total fuelwood could not have been significant when compared to limekilns.

     Finally, it might seem reasonable to assume that logging for lumber was more extensive than cutting for kilnwood. But, again the presumption is not justified for San Juan County, whose history of forest exploitation for lumber was very different from other parts of the state. Despite the existence around 1900 of a handful of small-scale lumber mills in the county, they were minuscule, short-lived and economically not competitive with the gigantic mainland mills. The market for lumber from mills on San Juan, Lopez and Orcas was strictly local, and therefore it was also small. Regarding exportation of island logs to mainland mills, for many decades it was neither expeditious nor profitable, a circumstance that only changed with the housing boom after World War II (see the webpage of this website entitled "50-yr harvest statistics" for the extent of post-war logging).

     The low quality of island-grown timber was another factor that limited logging for export to the mainland. According to Hayner (1929, p. 84): " Island timber was for the most part too small, wind-blown, and pitchy... but (it) was excellent for fuel in the lime kilns." So, in fact, by regional standards, there was extremely little logging for lumber before 1900 and for the subsequent decades when the islands suffered isolation and economic depression. During the time of the limeworks, then, timber cutting was predominantly for kilnwood, and as we have seen that use still left the majority of the county's forestlands untouched. Of course, the visual impact of severely cut tracts must have been astonishing then, as now. Eventually, during the 1940s and 1950s, logging for mainland limber mills did substantially increase, but by then the lime industry was in decline. So the full history of massive logging in the San Juans must include the episode of cutting for fuelwood that was followed sequentially by logging for lumber at out-of-county mills.

     One can conceive several scenarios that would have affected the geographic pattern of harvesting kilnwood:

    • Cutting timber that was closest to a kiln, such as that owned by the limeworks themselves, would have reduced transportation costs.
    • Conversely, timberland owned by the limeworks could have been more valuable if held in reserve and used to depress the selling price of kilnwood derived from independent suppliers.
    • Kilnwood production was a widespread cottage industry. It was produced and sold, without regard for location, by whomever could deliver it cost-effectively.
    • Timber was principally cut from properties near shorelines, even if high-banked, where simple chutes conveyed wood onto a beach, into the sea, or directly onto a buyer's scow (such chutes are frequently mentioned in oral histories throughout the San Juans and were built by binding pairs of logs together).
    • Kilnwood that was transported overland by cart was cut mostly from lands close to tracks and roads, not remote inlands.
    • Although crews of woodsmen were retained by the larger limeworks, it is unknown whether they operated exclusively on company-owned timberlands or if they also contracted out to cut on independently owned lots.
    • Independent landowners throughout the county, such as farmers and ranchers, routinely cut parts of their forestlands for homesites, fields, or orchards, and they would have sold kilnwood opportunistically.

Appendix 9. Exaggerations explained

     In the absence of accurate records and critical analysis, any technical issue that should be resolved objectively and quantitatively succumbs to subjective distortion. The genealogy of exaggerated impacts of the lime industry upon the forested landscape of the San Juans began in the early 20th century with hyperbolic "booster" literature. Regional newspapers circulated stories (probably all derived from the same source) that glorified the commercial prospects of San Juans, including the limeworks at Roche Harbor; the Everett Morning Star and San Juan Island's Islander newspaper (June 22, 1907) reported that each of Roche Harbor's ten limekilns produced 100-150 tons of lime per day, i.e. a company output of 1000-1500 tons per day. This amount is 12- to 19-fold greater than annual production reported by Danner (1966), which of course is the basis of the present analysis. What explains the discrepancy?

     The newspaper reporter in 1907 would have obtained his figures from someone at the company, but then he apparently garbled the story. While it is perhaps true that an entire battery of advanced kilns could process 100-150 tons of limestone in a day (Danner reported 120 tons by six kilns), this amount refers to the starting quarried material (limestone), not the end product (lime), which weighed less than half as much. More significantly, this quantity relates to multiple kilns, not each one. Did the reporter fail to distinguish limestone from lime and then compound his error by confusing the productivity of the entire company with the output of a single kiln? Probably.

     It should be reiterated that the present quantitative analysis relies on published lime-production records and a published fuel-to-lime conversion factor; it is independent of 1) all estimates or claims concerning the optimized productivity of single kilns (whether documented or imaginary) and 2) the number of kiln-days that may have been involved in the company-wide output.

     Decades after the era of boosterism, Evans and Burley (1971) reported ­ without documentation ­ that each kiln at Roche Harbor consumed 64 cords of fuelwood per shift (another measurement left unclarified). At some later date, an historic photograph illustrating limekilns belching smoke and steam was hung in the Lime Kiln Café at Roche Harbor Resort; it was captioned with the statement that each kiln consumed 192 cords per day. Evidently, the caption's author surmised from Evans and Burley that three shifts must have kept the kilns in operation around the clock and so he multiplied their number by three without considering that he was embellishing a myth that was already inflated. Such is the genealogy of the exaggerated claim that is cited in the introduction of this webpage. In contrast, the present analysis shows that about 15 cords were consumed by each kiln each day, a 13-fold difference from 192.

     Unidentified "historians" continued the tradition of altering and expanding upon the facts. Without the slightest substantiation, Orcas Island Historical Museum's current website claims:

"Historians believe that most of the 35 kilns ran 24 hours a day, seven days a week, eleven months a year. Each kiln required at least four cords of wood a day to attain and remain at the required 1600 degrees F. In only fifty years, these kilns burned an estimated 2,352,000 cords of wood!"

Although the daily fuel consumption per kiln in this quotation is close to what has been calculated here, the claim that as many as 35 kilns operated ceaselessly for 50 years vastly exaggerates (about ten-fold) the actual scale of the lime industry in the county. The museum's historian evidently credited every rudimentary backyard pot kiln (even those left in ruin after a single season) with equal productivity and longevity of operation as McMillin's advanced, long-lasting, and high through-put steel kilns. This assertion is obviously nonsensical. As a consequence, we witness an educational institution that carelessly promotes the unauthenticated claim that mostly phantom limekilns burned up more than half of the county's entire timber volume (2.4 million cords out of 4.3 million cords calculated in the present report), rather than 10.9% as calculated here. As legend-mongering would have it, yet another amateur historian in a 1997 audiotape at the Friday Harbor Historical Museum reiterated the same flawed calculations and then unabashedly rounded the total kilnwood consumption upward from 2.4 million to 3 million cords, or 70% of the county's timber! At that inflated rate, it is not hard to understand why so many people believe that all of the timber was indeed burned up as kilnlwood -- they just keep rounding up exaggeration after exaggeration. Present calculations demonstrate that such a conclusion is untenable.

     In summary, it is well to remember that the urge to understand our surroundings requires that we vigilantly discriminate verifiable facts from invented fictions, what is knowable from what is not, and irrelevancies from answers that appropriately address the questions that are asked. Much confusion arises when details are overlooked, omitted, or fabricated out of impatience or bias. In San Juan County there is much that remains mysterious, which is both enchanting and frustrating which the confusion lies with poor scholarship. For example, the is even an unresolved discourse regarding the number of islands that comprise the archipelago. Answers range from 457 to only four. Up[on examination there are indeed 457 emergent rocks, but only about 172 remain above ordinary high tide and are actually named as islands. Of these, only a few dozen support vegetation and fewer still are occupied by people. Only four are served by Washington State Ferries. Obviously, the number of islands depends on one's criteria and agenda, and the example serves to illustrate how the misguided use of information can distort understanding. With regard to kilnwood consumption by the historic lime industry in San Juan County, and the cutting of the forested landscape to supply that kilnwood, the present study has attempted to follow the facts with a reasoned analysis.

Appendix 10. Photo gallery

Figure 7. Lt. Richard Roche of the Royal Marines, who in the 1860s discovered a large deposit of limestone above the bay that bears his name.   Figure 8. John S. McMillin, boss of Roche Harbor Lime Co. and of political life on San Juan Is. (1920s).

Figure 9. Victor J. Capron (physician, McMillin's partner, owner of large timberlands, mayor of Friday Harbor, and state legislator) and the remains of the Victorian-era house near Roche Harbor that he built in the 1890s.

Figure 10. A cord of split timber being transported out of the forest (~1910). Timber destined for lime kilns was typically Douglas-fir cut into 4-foot lengths, split into large pieces, and dried. This load is 4 feet wide, 3 feet high, and 12 feet in length, or 128 cubic feet in volume. (The wood in this photo resembles redcedar more than Douglas-fir and may have been destined for roof shakes rather than kilnwood).

  Figure 12. A GLO surveyor's chain was 66 feet in length   (1/80th of a mile) and had 100 links. Its adjustable length was   calibrated daily.
  Figure 11. A typical GLO survey team (~1870).    

                                  Figure 13. The GLO Field Notes for San Juan County. Suveyors in 1874 recorded their                                   witness-tree information here.

Figure 14. View northeast from King Hill in Friday Harbor across Brown and Shaw Islands to Blakely Is. and beyond to Mt. Baker (~1904). Blakely's timberland seems to exhibit sparse underbrush and generous spacing between trees, conditions that were more common in pre-settlement forests due, preumably, to frequent, low-intensity fires. A fire on Blakely as late as 1895 is reported by Richardson (1971, p. 219). In 1874, according to the GLO survey, nearly three-quarters of Blakely's trees were large Douglas-firs spaced 33 feet apart. (From photo collection of J.M. McCormick, highly magnified).


Anonymous (no date). A Walking Tour of Historic Roche Harbor. Roche Harbor Resort brochure.

Atterbury Consultants, Inc. (1990). A Compilation of Yield Tables for Northwest Species. Published privately, Beaverton, Oreg. 255 pp.

Cottam, G. and J. T. Curtis (1956). The use of distance measures in phytosociological sampling. Ecology 37:451-460

Curtis, R. O. et al. (1982). Yield Tables for Managed Stands of Coast Douglas-fir. GTR-135, PNW Forest and Range Exp. Sta. FS/USDA, Portland, Oreg. 182 p. Also available online at

Danner, W. R. (1966). Limestone Resources of Western Washington. Bulletin 52. Wash. Division of Mines and Geology, Olympia. 477 pp.

Delcourt, H. R. (1976). Presettlement vegetation of the North of Red River District, Louisiana. Castanea 41:122-139.

Evans, L. and G. Burley (1972). Roche Harbor. A Saga in the San Juans. B & E Enterprises, Everett, Wash.

Gilbert, J. J. (1894-1895). Topographic Sheets T-2194 and T-2231, U.S. Coast and Geodetic Survey, RG 23, National Archives and Records Administration, College Park, Md. Also available online at .

Gilbert, J. J. (1895) Descriptive Report for Topographic Sheet T-2231, page 1, U.S. Coast and Geodetic Survey, RG 23, National Archives and Records Administration, College Park, Md. Also available online at .

Hanley, D. P. et al. (1987). Terminology for Forest Landowners. Washington State University Cooperative Extension, Pullman, Wash. 37 pp.

Hayner, N. S. (1929). Ecological succession in the San Juan Islands. Publ. Amer. Sociological Soc. 23: 81-92.

McLellan, R. D. (1927) The Geology of the San Juan Islands. University of Washington, Seattle, Wash. May be seen online at

Metsker, C. F. (1933). Map of San Juan County, State of Washington. Tacoma, Wash.

Morisita, M. (1954). Estimation of population density by spacing method. Mem. Fac. Sci. Kyushu Univ., series E., 1:187-197.

Orcas Island Historical Museum:

Perry, F. A. et al. (2007). Lime Kiln Legacies: The History of the Lime Industry in Santa Cruz County. 253 p.

Real Estate (no date):

Richardson, D. (1990). Pig War Islands. Orcas Publishing Co. Eastsound, Wash. 362 p.

U.S. Boundary Commission Report, U.S. Congress (1867). The Island of San Juan. Executive Document No. 29. Senate, 40th Congress, 2nd Session. Report of the Secretary of State. Washington, DC.

U.S. General Land Office, Field Notes, San Juan County, 1874. (Manuscript and typescript copies at San Juan County Department of Public Works, Friday Harbor, Wash.)


Return to fOREST iNFO Home

Site created by Tom Schroeder