Policy interventions and competing management paradigms shape the long-term distribution of forest harvesting across the landscape

Significance

Humans have dramatically affected ecosystems and stocks of natural resources worldwide, often by sequentially harvesting the most profitable components of the environment. Although research has examined historical harvesting patterns in forests and broad issues of resource depletion, little has investigated how “high grading” along gradients of productivity and accessibility has changed the forest landscape over time. Analysis of a dataset from coastal British Columbia, Canada shows a tendency to log down the value chain when management choices are unconstrained and the potential of government policy to impose a greater stewardship ethic on harvesting behavior. This history of serial depletion of profitable environmental components has left a legacy of reduced value in the forest, with implications for intergenerational and intercommunity equity.

Abstract

Industrial economic models of natural resource management often incentivize the sequential harvesting of resources based on profitability, disproportionately targeting the higher-value elements of the environment. In fisheries, this issue is framed as a problem of “fishing down the food chain” when these elements represent different trophic levels or sequential depletion more generally. Harvesting that focuses on high grading the most profitable, productive, and accessible components of environmental gradients is also thought to occur in the forestry sector. Such a paradigm is inconsistent with a stewardship ethic, entrenched in the forestry literature, that seeks to maintain or enhance forest condition over time. We ask 1) how these conflicting paradigms have influenced patterns of forest harvesting over time and 2) whether more recent conservation-oriented policies influenced these historical harvesting patterns. We use detailed harvest data over a 47-y period and aggregated time series data that span over a century on the central coast of British Columbia, Canada to assess temporal changes in how logging is distributed among various classes of site productivity and terrain accessibility, corresponding to timber value. Most of this record shows a distinct trend of harvesting shifting over time to less productive stands, with some evidence of harvesting occurring in increasingly less accessible forests. However, stewardship-oriented policy changes enacted in the mid-1990s appear to have strongly affected these trends. This illustrates both a profit-maximizing tendency to log down the value chain when choices are unconstrained and the potential of policy choices to impose a greater stewardship ethic on harvesting behavior.

Humans have dramatically affected ecosystems and stocks of natural resources worldwide in both aquatic and terrestrial systems. For example, in the marine environment, much attention has been given to the issue of “fishing down the food chain” (1) or “fishing through food webs” (2); as valuable and preferred, larger, high trophic-level species are depleted, harvesting increasingly targets lower trophic levels. Serial depletion of fisheries stocks is a global issue (3) and is often highly correlated with spatial patterns of resource accessibility and profitability (4). There is also evidence that preferential harvesting of large land animals for human consumption is imperiling megafauna around the world (5) and that the global decline of large, old trees is affecting diverse ecosystem services and functions (6–8). Similarly, after depleting conventional sources with higher value and lower production costs, scarcity is causing global energy production to undergo a transition toward higher-cost methods of extraction and alternative sources (9). Among other issues, these examples of systematic erosion of high-value components of the environment raise concerns about sustainability and intergenerational access to natural resources.

Disproportionately altering one portion of an environmental gradient can cause cascading effects across ecosystems (10) and can affect ecosystem services and functions, which are maintained through a portfolio of diverse ecological structures and functions (11). Historical high grading of large old trees, for instance, is thought to have restructured forest demography, gap dynamics, and fire behavior in dry conifer ecosystems, such as ponderosa pine (Pinus ponderosa) forests (12). These types of ecological changes can also catalyze shifts in natural resource policy—a prominent example being the overharvesting of old-growth forests in the Pacific Northwest of the United States, resulting in the loss of habitat for the threatened northern spotted owl (Strix occidentalis caurina) and the subsequent development of the conservation-based Northwest Forest Plan (13).

Many factors influence harvesting patterns of resources. Complex land-use histories reflect shifts in societal views over time, altering ecosystem services and patterns of conversion of regional land cover (14, 15). In forestry, the traditional economic paradigm associated with the industrial forest sector prioritizes timber stands that maximize profits, usually linked to accessible areas, such as flat terrain, or with short transportation distances to log markets, as well as productive locations that promote fast tree growth and stands with large, high-value trees. Analyses of past harvesting activity in Alaska and coastal British Columbia (BC) demonstrate a forest sector preference for logging large-stature forests on valley bottoms—a classic example of accessible and productive high-value forests (16–18). There is also evidence that valuable, rare tree species, such as mahogany (Swietenia macrophylla) in tropical rainforests (19) and large, old, high-quality Douglas-fir and western redcedar (Thuja plicata) in temperate rainforests (12, 20, 21), have been disproportionately logged over recent decades.

In contrast and sometimes, in conflict, with these traditional economic approaches to timber production, forestry and foresters also have a long tradition of promoting stewardship. Such ethics date back to the foundational views espoused by prominent figures, such as Aldo Leopold and Gifford Pinchot (22, 23), and continue to be entrenched in the legal obligations of professional foresters who are tasked with upholding the public interest (e.g., https://abcfp.ca/web). Although the tradition of forest stewardship ethics has prioritized economic, ecological, and social objectives to varying degrees over time, it generally emphasizes that different types of forest values, including economic values based on timber, should be maintained on the landscape to provide benefits across generations (24–26). Under such an approach, the net effects of management should not decrease the average value of the forest estate.

One important idea that arises from this ethic is that, over time, forest operations should “log the profile” of the forest—harvesting ecosystem types, species mixes, and productivity classes roughly in proportion to their presence and condition on the landscape. In many ways, the principal of logging the profile in forest management is analogous to the principal of ecological representation in conservation planning, with obvious connections to the stewardship of ecological values. In the forest industry, however, “logging the profile” is generally associated with and discussed in the context of the long-term stewardship of timber values. Logging the profile is often contrasted with the notion of “high grading,” where the higher-valued components of a stand or landscape are targeted preferentially, decreasing its residual value over time.

Changes in forest policy can influence the decision-making of resource managers as they navigate trade-offs between these types of conflicting objectives. Policy changes over recent decades in jurisdictions around the world have addressed many issues that affect the spatial and temporal patterns of harvesting (27–30)—often catalyzed by shifting public values about stewardship, economics, and the allocation of benefits from harvesting. Prominent examples of such issues in the Pacific Northwest include the size and distribution of cut blocks (13, 27); whether clear-cut logging should be replaced by alternatives, such as retention silvicultural systems (6); protection for ecological features, such as riparian areas, unstable steep slopes, and wildlife habitat, particularly for threatened and endangered species (31); and concern over the loss and protection of high-value, productive, old-growth stands (13, 32).

Each of these paradigms—focused on short-term profits vs. stewardship—has clear implications for how logging will be distributed across space and over time (Fig. 1). If the industrial economic paradigm of maximizing short-term profits is primarily dictating timber-harvesting patterns, logging should initially focus on accessible, high-value stands and be allocated over time to sequentially less valuable, productive, and accessible stands. Consequently, if the highest-value components of the environment are targeted first, over time the areas being logged should increasingly represent the productivity and accessibility of the overall landscape (i.e., be inclusive of lower-profitability locations) until the time when average conditions are reached. Alternatively, if the stewardship paradigm of maintaining forest condition is primarily dictating timber-harvesting patterns, logging initially and over time should occur in stands that are broadly representative of the productivity and accessibility of the overall landscape and that do not trend downward over time with respect to these variables.

Fig. 1.

Conceptual hypotheses related to the environmental gradients of harvested areas over time. A traditional economic paradigm associated with industrial forestry would suggest that the most accessible and productive components of the landscape are sequentially targeted (A and B). Alternatively, a stewardship paradigm, focused on the goal to maintain forest conditions and intergenerational access to timber and nontimber values, would suggest that the types of forests being logged are similar over time (C and D). A policy intervention focused on shifting the pattern shown in A and B toward increasing stewardship might result in the patterns in E and F. Trend lines could level off if harvesting becomes more representative of the productivity and accessibility of the overall land base (E, ii and F, ii), shift toward lower average variable values if productive and accessible parts of the landscape (e.g., riparian areas) become unavailable to harvest (E, iii and F, iii), or shift toward higher average variable values if unproductive and inaccessible parts of the landscape (e.g., steep, unstable terrain) become unavailable to harvest (E, i and F, i). Image credit: Cécile Lienaux.

Here, we investigate the influence of economics, stewardship, and government policy intervention on harvesting behavior by assessing the spatial distribution of logging over a half century on the central coast of BC, Canada in relation to environmental gradients associated with productivity and accessibility. Specifically, we assess which paradigm best predicts spatial harvesting patterns and whether this has changed over time in response to a major policy change. Although excellent research has examined historical harvesting patterns in forests (e.g., refs. 16 and 17) or broad issues of depletion and risk (18), our study focuses on an analysis of harvesting behavior at scales that span differing policy regimes and reflect underlying environmental gradients. This allows us to assess both issues of serial depletion of valued ecosystem components and the management paradigm most consistent with actual industrial behavior.

Methods

Study Area.

We focus our analysis on a portion of the central coast of BC, Canada (Fig. 2). This region encompasses diverse landscapes that are home to iconic wildlife, such as grizzly bears (Ursus arctos horribilis); productive salmon runs; and large, long-lived trees, such as western redcedar (32–34). Due to the wet and heterogenous biophysical environment of the central coast, the forests vary greatly from small-stature, low-productivity hypermaritime bog forests to tall, massive, highly productive floodplain forests (35, 36). The region is characteristic of temperate rainforest landscapes, where large disturbances, such as fire, occur very rarely, allowing a broad dominance of old forests on the landscape (37, 38). Indigenous populations have occupied the area continuously for over 10,000 y (39, 40), including the Heiltsuk First Nation, whose traditional territories cover much of our study area (https://heiltsuknation.ca/). This is a landscape of steep-walled mainland fjords and outer coastal islands, with significant portions of the area in nonforested high-elevation ecosystems on the mainland or lower-productivity forests on the outer coast. Our spatial analysis focuses only on the subset of this area that is forested (i.e., the green areas in Fig. 2, representing 855,133 ha of a total terrestrial land base of 1,484,298 ha).

Fig. 2.

Map of the study area (green) and forest harvesting within the central coast of BC, Canada. Due to extensive landscapes of relatively unproductive soil (e.g., outer islands in the study area), forests that are economically feasible to log represent a much smaller portion of the land base compared with forests in more southern parts of the province. The inset map shows the general location of the study area within Canada.

The central coast of BC forms a major part of the global distribution of coastal temperate rainforest and is often referred to as the Great Bear Rainforest (GBR). While there is a long history of Indigenous use of the land, the level of forest harvesting has increased dramatically over the past one and a half centuries due to the growth of industrial logging, mostly by non-Indigenous corporate entities, focusing on species, such as western redcedar, Sitka spruce (Picea sitchensis), western hemlock (Tsuga heterophyla), and amabilis fir (Abies amabilis), with Douglas-fir becoming common in the more southern portion of the region (41). In the 1990s and 2000s, these forestry activities were the catalyst for internationally significant land-use disputes, leading to new agreements, regulations, and policies to address sustainability concerns (27, 42, 43).

Because roughly 95% of BC’s forests occur on public land, with rights and responsibilities conferred to private interests through a tenure system, provincial policies have a strong influence on behavior in the forest industry. A major policy shift toward stronger stewardship objectives occurred in 1995 with the introduction of the Forest Practices Code (FPC) (44). The highly prescriptive regulatory environment of the FPC influenced many dimensions of forest management that previously had received little attention in regulations, such as new rules for cut-block configuration, size, and adjacency and increased protection for fish-bearing streams and unstable terrain. Subsequently, in the GBR specifically, a new regime of ecosystem-based management (EBM) was implemented in stages through various agreements and regulations between 2000 and 2016 (45). These changes led to new objectives for conservation and cultural heritage at the stand and landscape levels that further constrained logging in many parts of the GBR, including over entire watersheds. Negotiations around EBM also helped foster a new working arrangement between the forest industry and environmental organizations and a new government to government relationship between Indigenous and provincial governments (42).

Forest Harvest Datasets.

To determine the spatial extent of logging in each year between 1970 and 2016, we accessed the Geographic Information System (GIS)-layer Harvested Areas of BC from the BC provincial government (46). This layer combines harvest data from government sources, reporting from forestry licensees, and a gap analysis to estimate unreported harvesting based on remote sensing. The resulting spatial polygons represent the outer boundary of each cut block and typically encompass any in-stand reserved area (provincial policies currently require a minimum of 7% stand-level retention). We used polygons representing the area harvested each year—based on the Harvest Completion Date field. We noticed some missing spatial cut-block information in the 2018 dataset (verified using satellite imagery), which had appeared in a Harvested Areas of BC file that we previously downloaded in 2015, so we merged these two spatial datasets to create a final layer for analysis.

To understand harvesting over a longer time period, we also analyzed historical cut-block data for 1860 to 1969 using a dataset that was prepared for an earlier analysis in the study area in ref. 16. This analysis used historical orthophotos to classify forest disturbances, including a category that reflects logging (ref. 16 and the corresponding author have more details and access to the dataset). Because this dataset did not list information about the harvest completion date, we instead used current stand age as a proxy to estimate harvest date. We obtained current stand age information from overlapping forest cover data found in the Vegetation Resource Inventory (https://www2.gov.bc.ca/gov/content/industry/forestry/managing-our-forest-resources/forest-inventory), a spatial data layer that characterizes stands according to various attributes of land cover and vegetation based on orthophoto interpretation. We categorized harvesting according to the data source (the historical orthophoto covering 1860 to 1969 vs. the modern provincial dataset covering 1970 to 2016) and the type of forest being harvested (younger second-growth forests on sites being harvested for the second time vs. old-growth forests harvested for the first time (the amount of naturally occurring younger forest being harvested in the dataset is negligible). We then combined these categories to create three datasets for further analysis: 1) cut blocks representing historical old-growth logging, 2) cut blocks representing modern old-growth logging, and 3) cut blocks representing modern second-growth logging. We determined the locations of second-growth logging by assessing areas of overlap between the historical and modern logging datasets and determined old-growth logging by areas that did not overlap. All three cut-block datasets were then converted to raster files for further analysis (25-m grid cells, BC Albers projection). Our analyses of year to year trends focus only on the modern old-growth logging dataset because of the higher temporal uncertainty in the historical logging dataset and the small sample size in the modern second-growth logging dataset. We, therefore, only report results for these latter two datasets aggregated across all applicable years.

Data Analysis.

We assessed various spatial factors associated with logging to examine the extent to which harvesting is targeting specific environmental gradients. We first used the Zonal Statistics tool in ArcGIS 10.5 (47) to calculate the mean values within each harvested area of three variables associated with site productivity and terrain accessibility: Slope, Site Index, and Distance to Large Stream (Table 1). To understand trends at the scale of individual cut blocks, we derived these values for each of our logging datasets by setting the zone of analysis to the year the area was harvested, resulting in mean variable values aggregated across raster cells for each discrete area logged (i.e., cut block) in each year (Datasets S1–S3). We then calculated the mean variable values of all logging in each year weighted by area using the plyr package (48) in the statistical program R (49) and created graphs using the ggplot package. Finally, for each variable, we used a linear regression model to examine statistical differences between the mean variable values for areas overlapping old-growth logging in the pre-FPC period (1970 to 1994) and the mean variable values for areas overlapping old-growth logging in the post-FPC period (1995 to 2016).

Table 1.

Environmental variables related to site productivity and terrain accessibility used as predictors in our model

Variable name	Source	Description
Slope	TRIM*	Slope (%) is related to terrain accessibility. We interpret increasing values as decreasing accessibility
Site Index	VRI†	Site Index is a measure of site productivity. Site Index represents the potential height (meters) of dominant trees within a forest stand at age 50, measured from breast height. Site Index is calculated using a slightly different method in young vs. old forests; so, to avoid biases, our analysis only includes comparisons across similar ages of forests. We interpret increasing values as increasing site productivity
Distance to Large Stream	MFLNRO‡	Distance to Large Stream is related to site productivity and accessibility. Increasing values represent further distances from rivers of major watersheds, which we interpret as decreasing site productivity and decreasing accessibility at a landscape scale (not necessarily reflecting operational-scale policies for riparian buffers)

In general, more productive and accessible forests are associated with more profitable harvesting.

^*TRIM indicates Terrain Resource Information Management (https://www2.gov.bc.ca/gov/content/data/geographic-data-services/topographic-data/terrain).

^†VRI indicated the Vegetation Resource Inventory (https://www2.gov.bc.ca/gov/content/industry/forestry/managing-our-forest-resources/forest-inventory).

^‡MFLNRO indicates the Ministry of Forests Lands and Natural Resource Operations High Value Fish Habitat spatial layer (the query is based on selecting large streams within higher-order watersheds).

We also predicted the distribution over time of modern old-growth logging (1970 to 2016) to examine the extent to which harvesting occurred disproportionately along environmental gradients reflecting productivity and accessibility. We used the machine learning algorithm Maxent (50), accessed through the dismo package (51) in R. Maxent is a robust method for predicting species distributions that quantifies relationships between occurrence data and environmental predictors (52). To understand trends over time, we created separate spatial models for each year of harvesting. To characterize raster cells representing harvested forests (occurrences or “presence” locations in the model), we extracted variable values for Slope, Site Index, and Distance to Large Stream from 1,000 random cells in each year of logging, with a minimum distance of 50 m between random points to avoid duplicate selections. The spatial extent of logging in some years led to fewer than 1,000 random points fitting within these constraints, so the total number of presence points across all 47 models was 30,347 (Dataset S4). To characterize raster cells representing unharvested forests (“absence” locations in the model), we used these same modeling rules within the unlogged portion of the study area. We used the default settings in Maxent and k-fold cross-validation (k = 5) for each dataset so that model training used 80% of the occurrences and model testing used the remaining 20%.

To quantify model fit, we calculated the area under the receiver operator curve (AUC) statistic (53), which shows how well a model can discriminate between logged areas and unlogged areas. An AUC above 0.5 would suggest that logged areas are predictable based on their environmental conditions because they are sufficiently different from a random sample of conditions in the study area. We also evaluated statistical relationships by examining variable contributions to model performance as well as individual variable response curves with covariates held at their mean value. This overall approach allowed us to understand factors influencing historical harvesting patterns based on the presence and absence of logging in the study area through time.

Results

Trends in Productivity and Accessibility over Time.

The total area harvested each year generally increased through the 1970s and 1980s until the mid-1990s, at which time there was a clear transition to a long-term trend of decreasing area logged per year (Fig. 3A). The time of this transition roughly corresponds with the introduction and implementation of major changes in the provincial regulatory policy for forest stewardship in BC under the new FPC (as discussed above). The area logged in the study area over the past 156 y is 56,811 ha, ∼7% of the total forested area. Of this logged area, historical logging (prior to 1970) represents 12%, modern old-growth logging (since 1970) represents 87% (Fig. 3), and modern second-growth logging accounts for less than 1%. The mean cut-block size for modern old-growth logging is 5.0 ha (range = 0.1 to 433 ha, although data for cut-block size are somewhat uncertain because our harvest data may not precisely match the discrete cut-block boundaries that were engineered in the field each year).

Fig. 3.

Average (A) Area Harvested, (B) Site Index, (C) Slope, and (D) Distance to Large Stream between 1970 and 2016. The productivity and accessibility of the areas being logged decline over the study period, and the policy changes associated with the FPC are correlated with distinct changes to harvesting patterns. The blue and red lines represent best-fit regression lines, and the shaded bands represent 95% CIs.

All three logging datasets show clear changes over time in the spatial distribution of cut blocks across the landscape (Fig. 4). For instance, through the 1970s and 1980s, logging activity begins closer to the ocean and moves up valley bottoms. In the 1990s and 2000s, logging occurs in much smaller discrete areas that are dispersed more broadly across the landscape. Starting in 2008, there was a major forest sector downturn tied to a global recession that, combined with agreements and policies to implement EBM, led to reduced harvesting activity in coastal BC—73% less area was logged in the study area during the period 2008 to 2016 inclusive relative to the previous 9-y period.

Fig. 4.

Map showing the spatial distribution of cut blocks in different time periods. The highest rate of logging occurred in the 1980s to 2000s. These panels are zoomed into the landscape around the Kwatna River watershed, which represent less than 10% of the overall study area.

Notwithstanding variability and nonlinear relationships, logging clearly began in accessible, productive locations and over time, shifted to forests that were progressively less productive and accessible. Specifically, between 1970 and 2016, there is a downward trend in the mean Site Index of old-growth logging (Fig. 3B), a slight upward trend in the mean Slope (Fig. 3C), and an upward trend in the Distance to Large Stream (Fig. 3D). Over the most recent 9 y in the sample, there is greater variability among years in mean variable values, corresponding to the period with a downturn in the forestry sector and less harvesting in the study area. The mean values for all historical logging are similar to the mean values for second-growth logging (historical: Site Index = 19.1, Slope = 42.5%, Distance to Large Stream = 4,017 m; second growth: Site Index = 20.7, Slope = 30.1%, Distance to Large Stream = 3,938 m). Since the second-growth logging is taking place entirely in stands that were logged historically, this alignment makes sense.

For the variables Site Index, Slope, Distance to Large Stream, and Annual Area Harvested, there is a change in the regression line in the mid-1990s, a period associated with the FPC policy changes (Fig. 3). However, the trends in variable values exhibit different types of changes around this time. The regression line for Site Index shifts and becomes relatively steeper, reflecting a more rapid transition toward lower-productivity stands. The regression line for Slope becomes relatively shallower, with a shift toward more accessible forests on less steep slopes. The regression line for Distance to Large Stream does not shift but becomes relatively shallower, roughly reflecting the mean value for the study area.

Overall, the trend showing a decline in the productivity of harvested forests becomes more pronounced after the mid-90s policy change, and the trend showing a decline in the accessibility of harvested forests becomes less pronounced after this period (although still trending toward decreasing accessibility overall). When examining the mean coefficient values from the linear regression model for all years in the post-FPC period, Site Index is 82% of the pre-FPC period (P > 0.005), Slope is essentially unchanged relative to the pre-FPC period (P > 0.695), and Distance to Large Stream is 126% of the pre-FPC period (P > 0.005). Although the differences across policy regimes are partly explained by the overall directional trends across the entire study period, these results show that the FPC is correlated with changes to the areas selected for harvest. The dramatic change in the annual area harvested around this period also strengthens the hypothesis that the FPC strongly influenced harvesting activity.

Concordance between Logged Areas and the Broader Landscape.

Over the course of the study period, logged areas became more representative of the broader environmental conditions across the study area. For example, comparing models representing each year of modern old-growth logging shows a temporal trend of declining AUC values, which suggests that our model has progressively less ability to discriminate locations with cut blocks from locations outside of cut blocks, particularly when we aggregate data for the period after the forestry downturn between 2008 and 2016 (Fig. 5). When using all data for modern old-growth logging between 1970 and 2016, the model performs well (AUC = 0.87), indicating that over most of the period, logging is not well distributed across the range of environmental conditions existing within the study area. Across all years (all models), Site Index is the most important variable responsible for this relationship, contributing to 85% of the model’s predictive performance, although in any individual year, the relative importance of the three variables and their specific interactions changes. Hence, logged areas are biased toward higher-productivity locations on the landscape. Consistent with the trend highlighted in the AUC analysis, all three of our variables have their average values within cut blocks moving toward, reaching, or moving past their respective mean study area values over time (Fig. 3). Overall, the response curves show that forests with gentler slopes, with higher productivity, and closer to large streams have had a higher probability of logging activity compared with unlogged forests.

Fig. 5.

A scatterplot showing AUC values by year. Over time, the model has less of an ability to discriminate between logged and unlogged forests, suggesting that harvesting is becoming more representative of the broader environmental conditions of the landscape (average conditions occur when AUC = 0.5). The AUC values between 2008 and 2016 show much more variability than previous years, largely because the economic downturn led to less logging in those years (about 25% of the pre-2008 logging rates) and thus, a paucity of data. The overall trend using all of the data is approximated by the dashed red lowess smoother. Because of the reduced area logged between 2008 and 2016, we also considered a model that aggregated data across this period into one point, represented by the solid black circle. The solid black lowess smoother shows the trend for the entire study period when the years 2008 to 2016 are represented by that single aggregated point.

Discussion

A Changing Landscape.

Our findings reveal serial depletion of valuable landscape elements, leading to a long-term trend of “logging down the value chain.” Specifically, harvesting over the past half century in our study area on the central coast of BC generally started in higher-productivity and more accessible forests and proceeded to lower-productivity and less accessible forests. This pattern has led over time to the forests harvested being more representative of the environmental conditions of the broader landscape because the most profitable components were targeted first, with logging forced into increasingly average conditions over time. Over the period of our dataset, it becomes increasingly difficult to discriminate between the underlying environmental conditions of logged areas vs. unlogged areas. This convergence over time is mostly driven by the relationship between the spatial distribution of harvesting and Site Index, a metric of site productivity and the most significant variable in our model. Such a trend suggests that as time progressed across the study period, the forestry sector had fewer opportunities to harvest the types of productive stands that produce the largest trees, which are relatively rare in a regional context. Further, although not a panacea for stewardship, conservation-oriented regulations and policies implemented over the second half of the study period appear to have altered harvesting behavior by constraining the land base available for logging.

Our quantitative analysis brings empirical support to a long-standing narrative in the forestry sector about high grading and the ability of policy choices to influence stewardship. These results are consistent with other assessments of harvesting patterns in the coastal temperate rainforest, which show disproportionate logging activity in the most productive landforms in Alaska (17), in valley bottoms on the central coast of BC (16), and in forests of BC more broadly (18). Our study also reinforces the benefits, espoused by others (54), of using historical datasets to understand land-use and landscape change over time.

Competing Paradigms.

Two contrasting paradigms have shaped the development of forestry practice: one founded in a traditional approach to forest economics and one in forest stewardship. The trends we observe align more closely with the expectations of an economic paradigm focused on prioritizing profitability than with those of a stewardship paradigm focused on maintaining forest condition. This is not surprising given that in the first half of the period of analysis, government priorities and policies largely focused on exploiting the timber resource, maximizing short-term revenues, and promoting economic development through the granting of timber tenures to large corporations (41). Although maximizing profits to shareholders remains a core mandate for corporations (55), societal expectations for companies to manage forests sustainably have risen dramatically over recent decades. Furthermore, individual forestry professionals who work for these companies often have ethical and legal responsibilities, independent of their employers, to uphold the public interest and promote forest stewardship.

These trends suggest that viewing forest sector behavior solely through an economic or stewardship lens is overly simplistic because harvesting decisions increasingly emerge within complex social–ecological systems that are attempting to address multiple values and perspectives. Moreover, although a desire to avoid high grading is a core part of the forestry ethical canon, the idea of forcing harvesting operations to log low-value stands in order to match the profile on the landscape might be seen as an inappropriately narrow perspective on “stewardship.” An alternative approach that places a higher value on the ecosystem services provided by unlogged stands of lower economic value would be to remove them from the harvesting land base and reduce the overall cut accordingly.

Policy Interventions Can Change Harvesting Behavior.

Profit- or stewardship-driven decisions by forestry companies and foresters do not take place in isolation of other factors, such as the mediating effects of policy, changing technology, and dynamic log markets. The trends over time and space in our data show that major changes in provincial regulations and policies in the mid-1990s correlate strongly with changes to how forest harvesting was distributed across gradients of productivity and accessibility. However, the specific statistical changes to our three indicators show mixed results and do not consistently align with our simple a priori expectations. Over this time, regulations and policies associated with the FPC and EBM in the GBR limited timber harvesting in portions of the land base and placed constraints on the size and distribution of cut blocks to address biodiversity objectives and negative public perceptions about industrial forestry (27). These policy constraints catalyzed a transition in harvesting from large, concentrated clear cuts that progressed up major watershed drainages to smaller cut blocks dispersed across the broader landscape.

However, government policies are not the only factor that influenced harvesting trends over the past century. For instance, incremental technological innovations, such as grapple yarders and helicopters, enabled access to more isolated and logistically challenging terrain outside of the historically harvested valley bottoms. Forest composition relative to log markets has also driven logging activity and patterns. While we did not integrate tree species data into our models because the preharvest stand composition was unknown, western redcedar—an iconic species important to Indigenous communities—has been targeted disproportionately over recent decades due to strong market demand for its wood, leading to dramatic shifts in the distribution and abundance of these large old trees (20, 21). Finally, the operational logistics of accessing forests across the landscape have likely affected patterns of logging because features such as towns, highways, and river mouths, which are also spatially correlated with gradients of productivity and accessibility, are logical entry points for developing road networks across watersheds.

An Alternative State of Forest Value.

An economic paradigm focused on maximizing profits by sequentially targeting the highest-value components of an ecosystem leads to future conditions with degraded states of value. This has long-term implications for intergenerational access to timber resources. In principle, because forestry can be renewable and is intended to span multiple harvest rotations, second-growth logging can provide an opportunity to target the same environmental gradients as historical logging of old growth, and long-term stand development along these gradients provides an opportunity to exploit conservation opportunities missed in the first rotation. Indeed, the average values of the environmental variables used in our study area for second-growth and historical logging are similar. However, despite representing the same land base in terms of inherent productivity and accessibility, second-growth stands typically contain lower timber volumes and lower market value, depending on the demand for particular forest products. Because the trees are younger, smaller, and sometimes of a different species than was there previously and often have different wood characteristics, these stands typically do not represent the same distribution of economic values as the original old growth (56).

Similarly, because the preindustrial rainforest landscape often developed over many centuries without stand-replacing disturbances (37, 38, 57), these second-growth forests can also lack many of the ecological values and ecosystem services provided by the old-growth dominated predisturbance forest. These include high carbon stocks, distinct patterns of species diversity, complex stand structures that support specialist wildlife, regulation of water flow and quality, and diverse cultural services for Indigenous communities (58–64). At the scale of human generations, this shift to a less valuable resource on the same area of ground could be thought of as a persistent alternative state of lowered value.

Under the traditional model of industrial silviculture in the coastal temperate rainforest, this shift to an altered state of value would occur even in the absence of a pattern of high-grading valuable landscape elements. In these types of forests, timber harvest rotations are typically much shorter than intervals between large-scale natural disturbances (32, 37, 57), and most of the forested landscape outside of some form of designated protection will be converted to these younger forests. Thus, the future managed forest landscape will have a dramatically younger age structure than the preindustrial, old growth–dominated landscape. Although logging the forest profile ensures that gradients based on accessibility and site productivity are logged proportionately, once the managed land base is entirely logged and converted to second growth through the first rotation, the end result will be the same as landscapes developed through a serial depletion algorithm. This could be addressed through dramatically longer rotation lengths, but forest companies often view this an undesirable option because, among other economic factors, the rate of incremental wood added to stands each year typically starts to decline after the first century of growth. Such limitations are in part why one of the goals of ecosystem-based management, as developed for the GBR, is to maintain sufficient natural late seral forest outside the managed land base to support diverse ecosystem services and functions (41).

Increasing attention is being paid globally to the ecological and cultural roles of large old trees and the services they provide (6, 8, 65). Although our data focus on trends in site productivity rather than representing tree size per se, these factors are strongly correlated, with the largest, most iconic examples of trees in this biome typically associated with the most productive sites (20). The trajectory of landscape development we describe here is one where the large, old landscape elements were preferentially targeted by early resource development, producing a land-base depauperate in key historical components. The landscape legacies of this manifest, for instance, in a scarcity of culturally important species, such as monumental cedar trees that are critical to Indigenous carvers (20, 66), a condition that will take many hundreds of years to redress. As human perceptions and management systems accommodate to such degraded landscapes, we risk committing the fallacy of a perceived shifting baseline (67–70) in that our modern conception of the forest land base will have normalized a condition of degraded value economically, culturally, and ecologically.

Intergenerational Access to Forest Value.

Around the world, many jurisdictions are shifting the authority for land-use governance decisions from state and corporate actors to local communities and Indigenous groups (71, 72). This is the case for our study area in the GBR and for many other areas of the coastal temperate rainforest in northwestern North America. What are the consequences of this legacy of a state of reduced forest value for these communities? As in our study, the land base now being allocated to communities reflects the previous decades to centuries of extraction of the most profitable resources—a situation described as “managing the leftovers” (73). If the profits from originally extracting this natural capital had been invested locally in other forms of human and physical capital (e.g., schools, hospitals, water treatment, sustainable power generation, or business opportunities), then local communities might perceive some equity in past land-use decisions and harvesting practices. In many cases, however, this type of devolution of governance to the local level is occurring in isolated regions inhabited primarily by Indigenous populations and where much of the accumulated benefits from resource extraction were exported outside local communities (74).

In regions like our study area, the drawdown of natural capital from logging over the past century has been concentrated locally, but its conversion into physical and human capital has largely been dispersed outside the local area (41). As Solow (75) argues, “the cardinal sin is not [extracting]; it is consuming the rents from [extracting].” Capital substitution resulting from harvesting the most productive and accessible forests might be thought of as a societal benefit at a provincial scale but will not have necessarily maintained or increased well-being over time at the scale of local communities. This is particularly the case when local market externalities, such as degraded ecosystem services in community watersheds, are factored into the costs and benefits of past management decisions. Such an analysis emphasizes the pervasive, cascading social–ecological consequences of the systematic targeting of high-value, high-productivity landscape elements over many decades.

Conclusion

Overall, it is clear that forest management in coastal BC has resulted in logging down the value profile over the past century and that recent more conservation-motivated policies have affected land use in complex but distinct ways. Logging down the value profile is similar to trends observed in fisheries, described as fishing down or through food webs (1, 2) and as serial depletion of valued taxa (3, 4). In both systems, the most valuable components of the environment have been targeted first, with more attention focusing on sequentially less-valued components over time. Although the terrestrial case does not focus on differences across trophic levels, the landscape elements of varying economic value we examine also express varying ecological function across landscape-scale environmental gradients. These broad implications for ecological integrity, the risk of perceived shifting baselines, and intergenerational access to resources create a strong thematic connection between our work and the foundational marine research.

In our study system, the choices made by resource managers initially reflected the dominance of an industrial economic paradigm, and forest policies focused on short-term profitability, enabled by the widespread availability of high-value forests and the allocation of tenure rights to large, nonresident industrial forestry companies that could exploit those opportunities. As the industrial development of the land base progressed, this was eventually modified by depletion of accessible, productive, high-value locations on the landscape, imposing constraints on the choices available to forest managers. Finally, explicit changes in the policy environment intended to reflect broader goals of stewardship interacted with these historical constraints to produce a modern pattern of harvesting that differs substantially from that in the early study period. Thus, the history of logging on the central coast of BC illustrates both the tendency to harvest down the value chain when choices are unconstrained and the ability of policy choices to impose a greater stewardship ethic on harvesting behavior. Although occurring in the context of a degraded state of resource value, new opportunities are emerging to more equitably balance intergenerational access to resources and the environment through decision-making devolved to local communities and through EBM.

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.