Evolution of Forest Mensuration in the West
Where have we come? Where are we? Where to from here?
By James D. Arney, Ph.D.[1]
Forest Mensuration in the 1970’s
By 1970 other industrial forest owners became interested in pursuing research into more aggressive forestry practices. The USFS Experiment Stations and BCFS Research Branch were active in thinning and spacing field trials; but the industry desired to expand into fertilization, soil nutrition, nursery practices and non-tree vegetation control in new plantations. The regional forestry schools were identified as logical host institutions for a new type of research organization – the research cooperative. This structure provided forest ownerships, too small to establish their own research teams, the opportunity to play an active role in prioritizing and funding forestry research of direct and applied interest to their own forests. It also provided a means to share investment in research without disclosing proprietary forest information between owners. The forestry schools gained additional funding by sponsoring these research cooperatives and the US Forest Service found popular support for their research through broad combinations of collaboration and team projects.
The early 1970’s also brought on the expanded opportunities of computer processing. Forest resources cover large acreages of landscape encompassing wide variations in topography, soils, species composition, age classes and stand densities. Almost every aspect of developing an inventory, projecting future yield and determining a harvest level required laborious and time-consuming calculations. These calculations were complicated by layers of statistical inferences due to the nature of measuring the forest through varying sampling designs on many subsets of the total landscape.
This new computer processing capacity expanded forestry research into a new direction – computer-based models for growth projection and harvest planning. The traditional Austrian formula and others were quickly replaced with Timber RAM, TREES and FORPLAN harvest scheduling computer models (among others). It became possible to evaluate alternative silvicultural investment strategies, species compositions and management intensities in addition to traditional ranges of site class and age class acreage distributions over time. These types of research investigations were more demanding of available computer resources than from access to an Experimental Forest. Thus, the university forestry school became the default host to launching new approaches to forest harvest planning. This also facilitated changes in the Federal and Provincial research organizations to establish new teams investigating and applying computer-based forest planning systems. Private industry paralleled these developments with groups labeled as “Operations Research” at their corporate offices where these new computer planning models were put to work.
Significant investment and research in field trials had evolved by 1972. This contributed to the opinion that computerized yield tables were now the weak link in the (inventory, growth and planning) infrastructure of forest management. There were ample numbers of permanent-plot field trials that had been established over the past 5 – 10 years to document stand volumes at various ages and densities on most sites in the Northwest. These field trials provided Chuck Chambers and Frank Wilson (1972) with 356 permanent plots to build regression models to predict Douglas-fir stand volume from site index, stand age and stand basal area per acre. Later in the same year they also published yield predictions for Western Hemlock using the same input parameters on 232 permanent plots. Both of these yield projection reports included applications of the tarif system from Turnbull et al (1963) to derive board-foot volume.
Now the forest planner had an inventory basis of total acres by site and age, regression models to project those inventories forward in time; and, a computerized harvest planning system to schedule the harvest level and capacity of their ownership. Forest management planning became dynamic with these analytic tools available. Each year it was now possible to reduce the total inventory for actual harvest removals for that year and then re-run the harvest capacity analysis on the whole ownership. Forest managers could begin to visualize a working forest over time, its species composition and stocking levels under various kinds and intensities of forest management.
In this environment of dynamic forest management options, observations of thinning and fertilizer effects and unprecedented computer analytical capacity, the expansion of research into forest simulation (growth & yield) models began. In 1969 with financial support from the computer service center at Oregon State University, Jim Arney (1971) began building a tree and stand growth simulator for Douglas-fir that could invoke multiple thinning and fertilization treatments through time. This simulator could emulate any number of trees per acre on any site index level. Using King’s (1966) site curves and Grosenbaugh’s (1954) height-accumulation tree volume methods, he could forecast total stand volume in cubic and board-foot volume for any time in the future up to approximately eighty years of age. Crown Zellerbach was already investigating similar computerized growth models (Lin, 1970) and provided permanent plot data and advice to assist in Arney’s developmental efforts.
Also in early 1972, the Canadian Forestry Service (CFS) held a meeting at the Petawawa Forest Experiment Station in Ontario of forest mensurationists from across Canada to discuss the tree growth simulation program within the CFS (Honer, 1972). A six-member working group (Brian Armitage, Jim Arney, Imre Bella, Jim Cayford, Frank Hegyi and Terry Honer) was appointed to develop recommendations for tree growth simulation research within the CFS (Honer, 1973). Their report includes the following statements:
“If the extensive forest management practices in Canada were to continue, yield table methods would probably provide results significant for planning purposes; however, if we assume that forest management will intensify in the future, we will need methods that can used to forecast the outcome of a range of alternative silvicultural strategies. The impact on tree and stand growth of spacing, density, site, defoliation, and fertilization will have to be considered. Finding solutions to these problems through thinning, spacing and fertilization studies, and then fitting this fragmented information together cannot be done effectively with the framework of yield tables.”
“… we need a new kind of framework, a model, that is complex enough to combine and integrate the effects of all these factors. What is required is an operational tree growth simulator capable of providing yield estimates for natural and managed forest stands.”
These recommendations were accepted by the Programs Operations Directorate (CFS) early in 1973, and a program of testing and evaluating was initiated the same year at the Pacific Forest Research Centre (Victoria, BC), the Forest Management Institute (Ontario) and the Great Lakes Forest Research Centre on Douglas-fir, White Spruce and Jack Pine, respectively.
Arney sent a personal letter in summer of 1972 to mensurationists in the West with the following invitation:
“Because there is a high level of interest among mensurationists, an informal workshop will be convened on September 6 and 7, 1972 at the U.S. Forest Service, Forest Sciences Laboratory near Olympia, Washington. Interested Mensurationists are invited to participate. No formal papers will be presented so that maximum advantage can be taken of this opportunity to discuss common problems and evaluate various approaches to tree growth modeling. Enclosed is a list of scientists who have expressed an interest to participate.”
Everyone on the list accepted the invitation.
Participant list of historical interest
| Weyerhaeuser Research | Centralia, WA | Dave Bower, Dave Lewis, Jim Woodman, Dale Shaw |
| Crown Zellerbach Research | Camas, WA | Bob Strand, Jim Lin |
| MacMillan-Bloedel | Nanaimo, BC | Don Reimer |
| PNW Forest Expt. Station | Olympia, WA | Dick Miller, Don Reukema, Dick Williamson, Bob Curtis, Dave Bruce, Don DeMars |
| Depart. Natural Resources | Olympia, WA | Gerry Hoyer |
| Faculty of Forestry, UBC | Vancouver, BC | Don Monro |
| Canadian Forestry Research | Victoria, BC | Jim Arney, Jim Lee |
| INT Forest Expt. Station | Moscow, ID | Al Stage |
| Rocky Mt. For. Expt. Sta. | Fort Collins, CO | Cliff Myers |
| NE Forest Expt. Station | Columbus, OH | Sam Gingrich |
| PSW Forest Expt. Station | Berkeley, CA | Dave Sharpnack |
| PNW Silviculture Lab | Bend, OR | Walt Dahms, Jim Barrett |
| BC Forest Service | Victoria, BC | Al Fraser, Nick Kovac |
Within two years an informal correspondence among a wide geographic array of growth modelers developed. This correspondence was mostly by personal letter or phone conversation among modelers. The correspondence grew to include Al Ek (Wisconsin), Lee Wensel (California) and Harold Burkhart (Virginia). The discussions and comments were highly technical and dedicated to specific model building questions and alternatives. This was a new science and there was ample room to share ideas without treading on anyone else’s developmental efforts. As a result, almost none of this is documented in the retrievable literature. Building on this series of correspondence, various forest modelers began documenting their progress. Ken Mitchell (1973) began developing the Tree and Stand Simulator (TASS) at the Research Branch of the British Columbia Ministry of Forests in Victoria. Al Stage (1975) began building the Prognosis growth model at the Intermountain Forest Experiment Station in Moscow, Idaho. Lee Wensel (1977) established a redwood growth and yield cooperative among Northern California forest land owners to build the CRYPTOS and CACTOS growth models. Al Ek and Rolfe Leary (1975) developed the STEMS growth model in the Lake States Experiment Station. Each of these modelers was in contact with the others and aware of the relative success and approaches being attempted elsewhere. There were actual pieces of Fortran source code sent among modelers where similar problems were encountered and someone found an efficient solution.
A IUFRO Meeting was held in August, 1973, in Vancouver, British Columbia. At this meeting Don Monro (1973) presented his paper on structure and approaches to growth and yield models. His group labels used to identify types of growth models are still used 35 years later (i.e., Whole-stand; Tree-list, distant-independent; Tree-list, distant-dependent).
Each of these growth and yield simulators handled individual tree lists containing at least diameter at breast height, total height and the number of trees per acre represented by the tree in the list. The growth projection would cycle through a number of periods resulting in diameter increment, height increment and mortality rates which would differ by varying degrees and kinds of species, site, silviculture, size and stocking conditions. These growth models provided extensive versatility to the inputs for the forest harvest planning models. This functionality became an immediate priority to many forestry companies to incorporate into their internal inventory designs, methods and planning analyses.
Perhaps it is not surprising that in 1973 Weyerhaeuser Director George Staebler and USFS Research Station Director Bob Buckman decided to co-host the development of a new managed-stand growth model for Douglas-fir in the Pacific Northwest. Staebler also wished to verify the yield projections for their internal “Target Forest” numerical thinning and fertilization tables. Staebler hired Jim Arney to join with Bob Curtis (USFS) as co-leaders of a two-year project to develop this new growth model. The cooperative project was announced in the Seattle Post Intelligencer and Portland Oregonian in October 1973.
The first action by Arney and Curtis was to enlist the participation of other organizations with permanent plot research trials established for four or more years. None of these field installations had been shared outside of their host organizations in the past. The invitation to participate was accepted by fourteen federal, state, provincial and private forestry organizations. This created the most extensive database ever combined in one yield development analysis in the West. The geographic range was west of the Cascade Mountains from Medford, Oregon to Campbell River, British Columbia. Interest in retiring Bulletin #201 for a managed-stand growth & yield model was high. Stand dynamics through thinning and fertilization was apparent in field observations. A tree-list growth model capable of forecasting these silvicultural effects was highly desired. Inventory methods and harvest planning technologies were being redesigned and developed to take advantage of this more complex and flexible yield forecasting structure.
Each of these developments in the 1960s and early 1970s cycled back into the inventory methods both in the cruising designs and information retrieval. Since the growth models were requiring tree lists of diameter, height and numbers of trees per acre, then the cruise design began to collect diameter at breast height in addition to tree count on a point sample that previously only required a tree count to determine an estimate of total basal area per acre. This provided stand structure detail for silvicultural regime building.
In the 1970s, point sample cruises began recording all trees (all commercial species) to the nearest 1-inch which were greater than 5-inches diameter at breast height. This provided a more complete estimate of the diameter distribution found in each cruised stand.
Meanwhile, ownerships with interest in moving forward into new silvicultural regimes of young stands quickly noted limitations in their CFI approach to inventory. It provided good information about past silvicultural treatments; but it provided no basis for evaluating future regimes. In the early 1970s and 1980s, these ownerships switched to more intensive temporary plot designs with fixed-area sub-plots for small trees (less than 5-inches diameter); crown ratio, upper-stem taper, tree age and tree damage measurements. This degree of measurement intensity provided detail about species composition, log size distributions, total basal area and stand height/age indices to site index classification. The permanent plot inventory grid was replaced with a stand-based sample of temporary plots.
However, the size of the database required to handle this detail for each stand in the inventory became a concern. This brought on the development of diameter distribution functions to estimate the size and number of trees in a stand given only a few parameters. The Weibull function (Bailey and Dell, 1973) was the most popular method of generating diameter distributions. It used only the number of trees per acre and average diameter of the stand to generate normal distributions of trees. This allowed for much smaller demands on the size of inventory databases for an entire ownership.
As interest evolved into higher resolution inventory methods, so did the database structures and database handling methods. Initially in the early 1970s most ownerships were obtaining an aerial photo coverage of their forested lands on a periodic basis. This aerial photo coverage was then “photo-typed” into polygons of 10 to 200 acres in size with labels identifying the species, size and density of that polygon. Typically, these were not more than 4 to 6 species groups, 4 size classes (seedling, pole, small saw and large saw), and 3 density classes (poor, medium, well-stocked). A single interpreter would classify an entire ownership and transfer these polygons and labels to an inventory map in a matter of a few weeks. This map became the new inventory base with acreage computed for each polygon.
In 1970 this inventory map was used in conjunction with field-cruised information to assign average volume to each “photo-type” label. As the computer capacity expanded during the 1970s, the map was replaced with a computer database where the acreage and volume could be stored for each polygon. Initially only the acreage changed from polygon to polygon within a common label since all field cruise plots were being combined within each stratum (species, size, stocking) prior to compiling the volume estimates. There was only one volume level per stratum within each ownership.
By 1975 more aggressive organizations began tracking the cruise plots from individual polygons (stands) in order to compile each polygon with observations only from that polygon. This provided the forester with much more confidence that what was contained in the inventory database could be actually found in the field. By late 1970s, average volume by “photo-type” label was quickly becoming recognized as not sufficiently precise when the forester went to the field to develop silvicultural plans. This began the transition from one volume class per stratum to a combination of individual compiled volumes per stratum for cruised stands with an average volume per stratum for the remaining un-cruised stands within each stratum. In the late 1970s there was typically about 10 to 25 percent of the acreage in each stratum which was cruised and recorded as individual volumes by stand polygon. The remaining 75 to 90 percent of the acreage in those strata were characterized with the average volume found from the cruised stands within each stratum. The next year additional un-cruised stands were cruised, compiled and recorded. This caused a continuation of the transition to almost every stand (polygon) within a stratum being cruised at least once by the early 1990s. This transition occurred over about twenty years (1975 – 1995). By the mid 1990s most forest ownerships were maintaining inventory databases which contained eighty percent or greater cruised acreages among all stands and strata.
These stand-based inventories were being recognized as much more robust and site-specific for forest planning and management. The typical application was to directly cruise each stand (polygon), compile it and load the forest inventory database. The remaining un-cruised stands could then be updated with the weighted-by-acres average of all cruised stands within each stratum. These remaining un-cruised stands were gradually becoming a smaller and smaller subset of each stratum as the years progressed.
This progression to a much higher proportion of the forest inventory with direct field measurements resulted in a significant decline in the number of contracted projects for aerial photo-classification and mapping by the mid 1990s. It was common practice in the 1970s to fly a complete set of aerial photography to generate a timber-type photo classification. This was the standard approach to develop a forest-wide inventory. However, once the complete photo-type coverage was available, the inventory forester began cruising stands to convert from a dependence on photo-type labels to a directly measured stand-by-stand forest inventory. The era of aerial and satellite vegetation mapping for forest resources began and ended in a matter of twenty years (1970 – 1990).
[1] Submitted 2009-June-30, Final 2009-September-04. Senior Biometrician, Forest Biometrics Research Institute, Philomath, Oregon.