LONG-AGO LANDCOVER REVEALED
IN GILBERT'S EXQUISITE T-SHEETS

 

Abstract. This article summarizes the landcover information contained in a set of unpublished, late-19th century topographic maps (T-sheets) of San Juan County that have recently resurfaced after being "lost" for a hundred years. This webpage distills the landcover contents in two ways: as a derived map and as bar charts that quantify the several distinctive landcover types.

The original, large-format documents occupy over 200 square feet, so they cannot be reproduced here. However, links to online images of the originals are provided.

Information contained in the T-sheets helps us to understand the appearance of the landscape long ago ­ specifically to conditions before Euro-American settlement. They also allow us to draw comparisons with today's landcover.

Appendices explain more about the T-sheets: their creator, cartographer John J. Gilbert; how they were made; how they became "lost" for most of a century and were recently rediscovered; and how they have been analyzed for this summary.

  A broad understanding of the vegetation of San Juan County, historically and at present, opens the possibility of comprehending how the natural landscape is affected by the people who occupy it. Temporal changes in landscape patterns (distribution and sizes of landcover types) can be examined if data from different times can be compared. Nowadays landcover data can be acquired rapidly by remote, earth-orbiting imaging satellites (an example is given at the end of this article), but what has been lacking are acceptable data from earlier times to form the basis for comparison.

   This article announces that high-definition landcover information was collected more than a hundred years ago using 19th-century techniques. The early information has remained unknown because very soon after it was collected it was filed out of sight in a remote archive. Recently, this remarkable database has resurfaced and is now available for study. As a contribution to the environmental history of the San Juans, this article summarizes the century-old information.

   Near the end of the 19th century, when the county was very young, fifteen highly detailed landcover maps of the San Juans were hand drawn by the flamboyantly mustachioed and technically gifted cartographer John J. Gilbert of the U.S. Coast and Geodetic Survey (see Appendix). Completed between 1888 and 1897 (Figure 1), the detailed maps document the patterns of vegetation, land use, and human occupancy at a critical stage in the county's history, namely during the very first generation of Euro-American settlers. It may be recalled that homesteaders began arriving rather late, only after 1874, a scant 14 to 23 years before the maps were executed; settlement (and its associated landscape transformations) began only after the GLO survey of 1874, which itself occurred upon resolution of the "Pig War" in 1872. Given the human and technological limitations of those early years of settlement, it is plausible to assume that the extent of wholesale land conversion was very limited between the onset of settlement and completion of the maps, so the landcover depicted in Gilbert's maps may significantly echo the late pre-settlement landscape, at least in the general ratio between forested and open land.

   A primary objective of the U.S. Coast and Geodetic Survey, for which Gilbert worked, was to characterize land and sea features that might benefit navigation. Thus, Gilbert's topographic sheets or T-sheets in the Pacific Northwest documented topographic details and cultural constructions near the shore while another team of surveyors created hydrographic sheets (H-sheets) of the ocean floor. For unknown reasons, but fortunately for later historians, Gilbert's surveys of the San Juans extended right across the islands (except for an omitted sliver of central Orcas Island). The original T-sheets were at such a large scale (1:10,000), as illustrated in Figure 2, that they cannot be conveniently reproduced in this report; they are, however, available for viewing online (see Appendices). Instead, this article extracts, illustrates and quantifies the landcover content represented in the T-sheets; in this analytical process, many details of cultural geography are regretably omitted, although their existence will be mentioned in passing.

    In the following map (Figure 3), landcover information in Gilbert's large-format T-sheets has been reduced and rendered into a single, accessible small-scale graphic map that covers the entire county. This map captures the geographical distribution and spatial relationships of all of the landcover units that Gilbert identified in the late 19th century. It shows that the landscape was dominated by dense forest interspersed with pastures, fields, and orchards. In very broad terms, the same distribution of forest and open land persists to this day, as we shall see, doubtless because surficial geology, topography and drainage are such strong governors of landcover.

   The surveying methods that Gilbert used for delineating inland landcover were not documented, but the results indicate a high level of accuracy. The consistent clarity and competence of his hand is apparent throughout the T-sheets and lend confidence that the location and the sizes of the units are also reliable and precise. Accepting that the unit areas are accurately conveyed, they can be analyzed quantitatively. Accordingly, the following bar charts (Figure 4) complement the preceding map.

   As portrayed in Gilbert's elegant T-sheets and represented here in charts and a derived map, in the late 19th century 83% of the county's total area of 112,000 acres, or 96,000 acres, was occupied by dense forest. Another 2,100 acres were sparsely forested, either because they naturally existed so (such as occurs on thin-soiled, rocky highlands) or because patches of dense forest had been cut or burned prior to mapping. Perhaps it goes without saying that the depiction of forestland as "dense" does not allow further inferences about the forest's actual composition or condition. Since Gilbert's time most of the forests have been cut and have subsequently regrown, but one may justifiably assume that today's forests differ in character from those of the 1890s. Some of the details of those differences are discussed in The Presettlement Forest and Early Forest Composition of this website.

   Non-forested open areas in Gilbert's T-sheets (17% of the total) were unevenly distributed. Some islands were thoroughly forested, yet 25% of San Juan Island was open, primarily due to grasslands. Throughout the county grasslands accounted for 9,100 acres (8% of the total), half of which were on San Juan. These areas would have been "natural" grazing lands or livestock pastures that had never have been plowed, and they undoubtedly represented the coastal prairies and other savannahs as described by the earliest explorers, Hudson's Bay Company employees, the U.S. Boundary Commission, and the federal surveyors of 1874 .

   Open land categorized as cultivated fields had been plowed and were probably sown with grains such as oats. A few of these fields had likely been reclaimed for farming by draining pre-existing wetlands but most were probably converted from natural, presettlement prairies. The 182 plots and 3,355 acres under cultivation in the T-sheets were primarily located in San Juan Valley and other low-lying areas. Plowing in that era would have been accomplished by small teams of draft animals, mainly horses.

   Orchards that produced apples, plums and pears for urban markets on the mainland were extremely important to the early agricultural economy of San Juan County, although most of them enjoyed only brief prosperity (see Bigfoot's Forest in this website). Gilbert's T-sheets recorded 309 separate orchards, nearly all of which were family enterprises of less than 10 acres situated close to houses. The magnitude of this early orchard industry can be understood with a simple calculation; if fruit trees were planted 24 feet apart throughout the 1341 acres of orchards countywide, the total number of trees would have been about 100,000, which comports well with the number of orchard trees documented for 1899 by N. S. Hayner (Publ. Amer. Sociol. Soc., 1929, Vol. 23, pp 81-92).

  Marshes or wetlands constituted a minor landcover type that would have been recognized and delineated by rushes, sedges and other diagnostic plants, as well as shallow surface water. Altogether, 1,260 acres were classified in this way and the majority were located on San Juan.

  Besides landcover features, Gilbert's T-sheets separately identified many human developments, including fences, tracks and roads, trails, ditches, residential houses, barns, farm outbuildings, schoolhouses, a few rural churches, post offices, lime kilns, lighthouses, and some commercial farms. Several fish traps projecting from the south shore of Cattle Point extended for an aggregate total of 3.4 miles. Among details that were not depicted were dwellings in towns of Friday Harbor and Eastsound and some hamlets. Nowhere were land-ownership boundaries shown (T-sheets were not cadastral maps). Gilbert documented landforms with 20-foot contour lines. Based on notes that appear on twelve of the fifteen T-sheets (three others failed to include such data), Gilbert claimed to have recorded a total of 340 miles of shoreline and 193 miles of roads and tracks in the San Juans.

  In 471 instances landowners were identified by name next to their houses, and many other houses are indicated but lack names. These strictly rural residences were usually situated close to cultivated fields, orchards, and fenced grasslands. There were 162 named farmsteads on San Juan, 147 on Orcas, 96 on Lopez, and 76 distributed on other islands. The identifications were very thorough, if not exhaustive, because the territorial census for the county in 1885 (before the T-sheets) showed a total population of 1,131 (including all members of families), and the U.S. Census of 1900 (after the T-sheets were completed) listed 686 families and a population of 2,991. This amazing thoroughness displayed in the T-sheets is another reason we should have confidence in other portrayed details, such as the sizes of landcover units. In some instances Gilbert labeled homesites as "deserted," by which Gilbert meant that "their owners find it necessary to seek employment elsewhere in order to make a living," in other words the owners were temporarily absent at the time of the survey (Descriptive Report for T-2231, 1895, p 9).

   In general, the broad landcover separations between forested and non-forested open land have not changed radically from pre-settlement times the present. The details, however, are considerably altered. Some of these changes are noted in Figure 6, which pictorially compares landcover on San Juan Island (similar changes occurred on other islands, although less intensely).

   Two years after the end of the "Pig War," the General Land Office in 1874 established the gridwork of section lines for ranges, townships, and sections across the county and surveyed the land features along the section lines. The distance and location of nearest witness trees at section corners and at quarter-section points (halfway between corners) were recorded; those data have been mathematically converted into tree densities at those points, as shown in Figure 6, upper left. A distinction between forests (7 or more trees per acre) and open land (less than 7 trees per acre, therefore indicating savannah and grassland) was arbitrarily determined. As explained elsewhere (see Early Forest Composition in this website) the GLO did not collect data between section lines, so the resulting database is very sparse (and therefore of only marginally validity) and the grain of the resulting map is very coarse. Open areas that existed in 1874 had, for some years before, been used by Hudson's Bay Company as sheep grazing grounds.

   Compared to the survey of 1874, Gilbert's T-sheets greatly improved the "grain size" of landcover mapping. Probably little overt conversion (from forest to agricultural open land) or even forest cutting had yet occurred by the time they were produced, so the matrix of dense forest was likely similar in distribution and condition to what it had been before settlement. Preliminary evidence suggests that the proportion of open land in the San Juans prior to 1750 was greater -- perhaps much greater -- than it has been in historic times; rediscovery of one ancient prairie now overrun by forest is described in Lost & Found Prarie of this website.

   Strict comparisons between landcover maps from different times are full of pitfalls because mapping methodologies were not only different but also undocumented. Unequal attention to details would have generated differences in the "grain" of the results. Nevertheless, some major themes of change are evident.

   By 1952 most of the pre-settlement forest had been cut to fuel commercial limeworks (see Impact of the Lime Industry, this website), for residential uses, and for milling into lumber; on the other hand, much of the cut forest was probably in the early or mid stages of regrowing. However, compared to the 1890s, farmland by 1952 was considerably expanded in San Juan Valley and elsewhere at the expense of forested land. From other records we know that certain crops such as grains and peas were briefly quite profitable. In general, however, by 1952 the economics of farming in the San Juans was in decline, and tourism, retirement and other non-farming pressures had not yet really begun.

   Between 1952 and the present the San Juans have experienced an explosion of rural residential occupancy; parcels of land have been repeatedly subdivided. Many new landowners have created small forest openings but they have not engaged in commercial-scale farming, grazing or forestry. By such means, contiguous blocks of dense forest are now reduced in size and the amount of sparse forest has increased. Overall, forest lands have become severely fragmented by a large number of small grassland, scrubland and transitional landcover units. Of course, the qualities of forest in different units and probable declines over time are not conveyed in any of the maps shown here.

APPENDIX I. What is the form of the original T-sheets and where may they be observed?

   Each of the fifteen original T-sheets depicts a different portion of the county on a large sheet of paper or collodion-coated tracing fabric measuring about 3 x 5 feet. In most cases two identical copies exist, although all are manuscript versions, signifying that they were executed by hand rather than by any mechanical printing process. Landcover features were depicted in black ink throughout, but contour lines on some copies were marked in red ink. Half of the T-sheets were executed on coated linen and half on very thick paper card. Each finished map contains a clear title block that is signed by J.J. Gilbest, Ass't. The T-sheets are stored at the National Archives and Records Administration in College Park, Maryland, Cartographic Division, Research Group 23 and can be examined on-site upon request. Large-format photocopies may be acquired for a fee.

   The scale of the T-sheets is 1:10,000, which means that one mile is represented by 6.3 inches (1 inch = 833 feet). Thus, when all of the full-scale T-sheets are fitted together, San Juan County occupies an area of about 12 by 11 feet. On the other hand, details are so small that individual cartographic symbols are about 1/16th of an inch or less in size.

APPENDIX II. Can the T-sheets be accessed online?

   Digital versions of the T-sheets of the San Juans may be viewed at http://www.wsulibs.wsu.edu/Holland/masc/xmaps.html (follow the link: "San Juan Islands T-Sheet Index"). This site provides an index to the cartographic symbols. Some exploration is required to zoom the T-sheets to their original scale. If entire maps cannot be downloaded, screen images can be captured using a "screen grab" tool.

APPENDIX III. Who was John J. Gilbert and how did he execute the T-sheets?

   Gilbert was a lifelong employee of the U.S. Coast and Geodetic Survey (previously called the Coast Survey until 1878 and known as the National Geodetic Survey since 1970). He spent 22 years (late 1880s to early 1900s) producing maps in the Georgia Straits region of western Washington state. During that time his title was Assistant. Judging by his work, Gilbert was a masterful surveyor and cartographer. Notebooks in his lucid handwriting illustrate his qualities of thoughtful observation and decisive accuracy. Photographs of Gilbert portray a somber individual with prominent facial hair (Figure 7). His official obituary upon his death in 1929, at age 84, stated:

"His genial disposition endeared him to all. The unparalleled record for continuous surveying operations in one locality was held by Captain Gilbert.... His original records are as current now as then.... Although some of the modern-day methods and instruments were lacking, he was able to produce substantially as accurate and thorough results as are possible today." (www.history.noaa.gov/cgsbios/biog6.html)

   According to a 1909 story in the New York Times, by which time he was the agency's Inspector of Hydrography and Topography working in Washington, D.C., Gilbert was reprimanded for on-going friction with his boss, a man he had engaged in a workplace fistfight two years before.

   Executing a T-sheet of a coastal area required a land-based headquarters for drafting, a launch and cutter for transport, a crew of supporting assistants and equipment. Most details of the surveying process and the number of helpers on Gilbert's team are not recorded. When commencing work on an island, scores of triangulation stations were established along its shore. Survey data were recorded by plane table and the final product was completed at the headquarters office, usually in the winter. Given limitations implicit in line-of-sight surveying within densely forested landscapes, it is not immediately evident how many of the nearly invisible inland details were collected, let alone with such remarkable accuracy.

APPENDIX IV. What purpose did the T-sheets serve?

   The nation's first scientific agency, the U.S. Coast Survey (later Coast and Geodetic Survey) set a goal to produce detailed T-sheets for the entire coastline of United States. It was an ambitious and expensive project, and the T-sheets became the centralized and standardized tools for understanding coastal resources and for guiding development. In most cases the geographical coverage in T-sheets was limited to precise delineations of the contours of the coastline, the character of the immediate uplands, and the locations of a few key landward features. In most locales, T-sheets represent the earliest and most reliable geographical records of coastal natural features.

   Gilbert's T-sheets always included landcover and landform details farther inland, and as a result they have become valuable historical documents. For mainland portions of Whatcom and Skagit Counties, he studiously included about one half mile of inland landcover and cultural features. However, for the islands of those counties, and of course for the San Juans, his T-sheets thoroughly included all landcover and cultural features right across the landscapes.

   In the end, T-sheets were never distributed or circulated through publication, so they received little attention outside the bureaucracy of the Coast and Geodetic Survey. Some of their information no doubt filtered back to urbanizing centers and ports, before they were simply filed away and basically forgotten. If, at the time they were commissioned, the T-sheets of the San Juans were conceived for some further specific purpose, it is not apparent today nor has it ever materialized. Nowadays T-sheets of some sensitive sites are being rediscovered as historical tools for projects in ecological restoration, particularly of estuaries (Grossinger, R. 2001. in The Historical Ecology Handbook, D. Egan and E.A. Howell, eds. Island Press, Washington, pp 425-442).

APPENDIX V. How were San Juan County's T-sheets rediscovered?

   In the 1990s this author (Tom Schroeder) began an amateur study of the forests of San Juan County. From the beginning, a question arose: What was the condition and distribution of the forests before Euro-American settlement (or immediately following)? Very little reliable information was initially available. The notes of the General Land Office survey of 1874 (which themselves had to be retrieved from near-obscurity) offered some insights (see Figure 7), but much remained unclear. At first, I knew nothing of USC&G's T-sheets.

   In late August 1999 a window into the county's landscape history serendipidously opened for me. The 1999 San Juan County Fair organized a special exhibit in a former poultry shed; the show featured mementos of the county's maritime and fishing history. Alongside historical photographs and diverse memorabilia exhibited by local fishers and sailors, a framed photocopy of an old map featured Cattle Point on San Juan Island; the map depicted several long fish traps as they existed a century before, in the waters where the map's current owner still gillnetted salmon. Although I was only mildly interested in fishing history as such, the map caught my eye because of the upland portions of the old map -- the land was thoroughly filled in with cartographic symbols that denoted grasslands, forests, roads, fencelines and dwellings. More landcover details were crammed into that map than any other I had seen for the San Juans, even including some aerial photographs of a much later era. The map's hand-lettered title box indicated a date of 1897, it credited the cartographer as J.J. Gilbert of the U.S. Coast and Geodetic Survey, and it identified the product -- now known as T-sheet #2301 -- as "Sheet No. 12, Register No. 2301." These latter details signaled to me that the framed map of Cattle Point was probably one of a larger series. Did the others of the series illustrate additional portions of the San Juans? If so, where were they located, and why had they never been mentioned by local historians? From that moment forward, I needed to find answers.

   The owner of the exhibited map knew nothing about its ultimate provenance or the existence of additional maps. He had received the copy as a gift from a fisher friend in San Francisco because it portrayed the familar old fish traps off Cattle Point.

   After several communications I discovered that records of the Coast and Geodetic Survey were stored at the National Archives and Records Administration (NARA) in College Park, Maryland, and that the Cartographic Division possesses several thousand original T-sheets from all around the U.S., filed under Research Group 23. Fifteen of NARA's T-sheets (plus copies) dealt with San Juan County, and "No. 12" from the County Fair exhibit was among them. I then learned that visitors could present themselves to NARA to view the maps but that the agency was unable to provide and mail photocopies to remote researchers. I also learned that the original T-sheets were large in format and sometimes physically quite fragile. I realized that a personal visit to NARA had to be considered.

   In the spring of 2000 an acquaintance who lived in Maryland (Cathy Lundmark) agreed to visit NARA on my behalf as an associate investigator to track down the T-sheets, to verify their geographical coverage and condition, to obtain photocopies at full-scale and at half-scale, and to mail them back to me. During this assignment Lundmark maintained moment-by-moment contact with me by email and tenaciously undertook these tasks, for which I am deeply grateful, thereby unearthing and "repatriating" San Juan's T-sheets. Only in 2001 and 2002 was I personally able to visit NARA to investigate and collect additional materials related to Gilbert's T-sheets.

APPENDIX VI. How were the T-sheets posted online?

   The unwieldy, full-scale photocopies of the T-sheets were reserved for detailed study. My original goal with the half-scale photocopies was to reproduce sets of T-sheets, organized as atlases, for donation to the three libraries in San Juan County, because the public also deserved to see and study these remarkable documents. Unfortunately, in the reduced versions details were only barely legible because they were so minute at half-scale. Consequently, in April 2000 I abandoned plans to donate reduced, paper versions and embarked on a project to post the fifteen T-sheets online in electronic form at full scale. This required more technical proficiency than I possessed. When I approached University of Washington's Map Library with the T-sheet photocopies, the team of specialized librarians enthusiastically endorsed my project to post them electronically, and they agreed to scan the photocopies digitally. However, some weeks later the UW librarians backtracked; UW map librarians wanted to defer digitization for several months in order to leverage the T-sheet project into a grander enterprise to enhance their entire digital technology infrastructure, which would have necessitated a long delay (to prepare a grant application, allow time for grant review, set up a budget, purchase and shake down new equipment, and realign library staff). So, in order to keep the momentum for web-posting the T-sheets, I decided to take my project and paper T-sheets elsewhere.

   The Map Library at Western Washington University in Bellingham made a point of maintaining resources that related to the nearby San Juans. It was staffed by a solitary, tireless, but overworked librarian (Janet Collins); although she expressed immediate enthusiasm for the project, her library had no facilities or experience with large-format scanning. Nevertheless, Collins requested that I entrust her with the full-scale T-sheets for a short time. Astonishingly, within two days, she triumphantly announced that the T-sheets had already been scanned at very high resolution and recorded electronically as large CD files that preserved all of the detail of the originals. As it happened, during the intervening two days Collins discovered that WWU's Physical Plant owned a large-format scanner that was used for architectural, construction, and infrastructure renderings, and she single-handedly negotiated the scanning of the T-sheet photocopies. Much can be said in favor of personalized commitment to a project.

   With the electronic files of the T-sheets in hand, I again contacted the UW Map Library. This time they agreed to compose the necessary meta data for the finished files and, in a timely fashion, to arrange for web-posting through Washington State University's server that accommodates regional historical image databases. In turn, I provided a brief webpage introduction. By midsummer 2000 Gilbert's T-sheets of the San Juans were available for the entire world to view, one hundred years after he had created them, thanks to a fruitful three-university collaboration. The files may be viewed at http://www.wsulibs.wsu.edu/Holland/masc/sanjuanindex.html, however extracting data from them is still inconvenient (follow "View Map Image (Zoom))." Because the files utilize the MrSid compression protocol, some perseverance is required in order to discover how the images may be zoomed to full scale, and, because of that format, after downloading they require specialized software to be opened; alternatively, small zoomed fields may be acquired by ordinary image-capturing tools (known as "screen shots" or "screen grabs").

   Subsequently and independently, UW's ambitious Puget Sound River History Project posted many additional regional T-sheets through WSU's server (see http://riverhistory.ess.washington.edu/tsheets/framedex.htm). Their additional T-sheets (however, not including the San Juans, presumably because they lack rivers and estuaries) are readily downloadable as medium-resolution TIF images; for San Juan County T-sheets links are provided to the MrSid files at the website mentioned in the preceding paragraph.

APPENDIX VII. Are there accessory materials associated with the T-sheets?

   A number of incidental field notebooks at NARA were produced in association with the T-sheets, but the most significant accessory materials are handwritten Descriptive Reports that accompanied each completed T-sheet as it was being registered by the Coast and Geodetic Survey. These DRs contain remarks exclusively concerning coastal features The fifteen DRs for the San Juans contain a total of 187 pages of text and 164 contemporaneous photographs! The Puget Sound River History Project, cited above, has posted these records online; the text is legible but the photographic images are obscured because they were scanned as line drawings instead of halftones. I have a digital collection of these images as halftones and have used a couple of images in this web page and elsewhere in this website (see Yellow Island Change).

APPENDIX VIII. How were data from the T-sheets transformed into the derived map and bar charts of the present webpage?

   Landcover units of the original T-sheets (identified by cartographic symbols) were converted into the color-coded map of Figure 4 using low-tech methods. That is, units in half-scale photocopies of T-sheets were first colored according to type. Using a light table, outlines of units were traced onto overlay sheets, which were then scanned and saved as digital image files. Using Adobe Photoshop the fifteen traced overlays were assembled and merged into a single image showing all landcover units throughout San Juan County. With the colored T-sheet photocopies as guides, each unit outline in the digital image was filled with a color chosen to represent that type.

   Likewise, the bar charts in Figure 5 were generated by the low-tech method of cutouts and paper-weighing. Using the half-scale tracing sheets mentioned above, a total of more than 1,200 landcover units were manually cut out with scissors and segregated into a hierarchy of piles: by island (4 piles) and by landcover types (7 piles). A final hierarchical tier was generated as the size class (0-10 acres, 10-20 acres, etc.) of each unit was determined by visually comparing the cutout against a 1/8 inch grid (at a scale of 1:20,000, which is that of the half-scale T-sheets, a 1/8 x 1/8 inch square represents almost exactly one acre). This created five more piles. Altogether, 140 hierarchical piles resulted, and the total area of each pile was determined by weighing it on a highly sensitive electronic balance and comparing it against the weight of a piece of paper of known area (acreage). Accordingly, the aggregate areas of landcover units of each size class for each island were compiled in a graphics program to construct bar charts.