1. Introduction
Redwood (
Sequoia sempervirens (Lamb. ex D. Don) Endl.) is a coniferous species indigenous to the western coast of the United States. This species exhibits shade tolerance and, within its native habitat, can develop stands of varying ages, with trees attaining heights of 115 m and ages surpassing 2200 years [
1,
2,
3]. Redwood forests store exceptionally high quantities of biomass and carbon within the decay-resistant heartwood of old-growth stands, and growth rates have been found to be highest in plantation forests [
1]. Generally, redwood stands are healthy [
4], being resistant against damage from wind [
3,
5] and fire [
6,
7,
8]. Consequently, the significant carbon reserves accumulated in these forests are relatively well-protected from threats such as pests, pathogens, and environmental stressors.
Redwood trees yield high-quality timber, which is primarily utilized for aesthetic purposes given its red-brown heartwood and pale sapwood, workability and dimensional stability [
5]. This timber is recognized for its lack of odor, attractive grain, and importantly, its naturally durable heartwood [
9]. The exceptional timber quality of mature redwood trees has led to rapid depletion of old-growth forests within their native range. Over the past 170 years, the area covered by these forests has decreased from 8900 km
2 to 457 km
2, with only seven remaining locations exceeding 4 km
2 in the United States [
10]. The timber demand has been partly met by plantations established within the native range, and there is potential for sourcing timber from plantations established outside the United States [
5].
Redwood can thrive under a wide array of environmental conditions, but optimal growth occurs in temperate climates characterized by mild temperatures and moderate to high rainfall. Successful plantations have been established in numerous countries with such climates, including France [
11], Germany [
12], and Chile [
13]. Redwood has proven to be particularly well-suited to New Zealand, where 8000 hectares were established between 2000 and 2018 [
5]. Afforestation rates saw a sharp increase after 2019, with approximately 2000 hectares established in 2023 alone [
14]. Redwood is currently the second most widely planted species in New Zealand, following radiata pine (
Pinus radiata D. Don) [
14]. Most of these plantations are in the North Island, where the climate is particularly favorable for this species [
5]. As a result, growth rates in these areas, which reach maximum periodic annual increments (PAI) of 87–90 m
3/ha/yr, surpass those in native-range plantations, where maximum PAI of 75.6 m
3/ha/yr have been observed [
15].
Data collected from a network of permanent sample plots (PSPs), along with predictive models and spatial data, have demonstrated the substantial potential of redwood in New Zealand for both timber volume and carbon sequestration [
15]. Within New Zealand redwood is a low-risk species to establish as it is very resistant to pests, diseases [
4,
16], wind [
17] and fire [
6,
7,
8]. Additionally, redwood has a very low risk of wilding spread [
18]. The species reproduces both vegetatively and through seedlings [
19], but cone production is irregular and seeds are characterized by very low viability.
Growth models that can predict carbon accumulation have been used to undertake a financial comparison of the net present value of redwood and radiata pine of New Zealand forests for which the only derived income was from carbon accumulation [
20]. This research showed that by age 100 years, the total carbon stored in redwood forests exceeded that of radiata pine in eight out of nine regions in New Zealand. Mean regional carbon for redwood exceeded that of radiata pine in the five North Island regions by on average 74%, with regional differences reaching up to 85% (5203 vs. 2808 t CO
2/ha at 100 years of age). The net present value of redwood was found to be closely comparable to that of radiata pine across a spectrum of carbon prices in four out of five regions in the North Island [
20].
Carbon now represents a major source of revenue for forest growers [
21]. Globally, compliance carbon markets play a central role in climate mitigation, with Emissions Trading Systems (ETSs) serving as key policy instruments. As of 2024, 36 ETSs are active worldwide, covering roughly 18% of global greenhouse gas emissions and generating a record USD 74 billion in revenue in 2023 [
22]. An additional 22 ETSs are currently in development or under consideration in countries such as Japan, Canada, Brazil, India, and Türkiye, highlighting growing international momentum toward carbon pricing as a climate strategy [
22].
The New Zealand Emissions Trading Scheme (ETS) stands out globally for its inclusion of the forestry sector [
22]. First introduced in 2008 under the Climate Change Response Act, the ETS remains one of the government’s primary tools for supporting both domestic and international efforts to reduce greenhouse gas (GHG) emissions. Under the scheme, registered forest growers earn one New Zealand Unit (NZU) for every tonne of CO
2 sequestered. Over time, the total number of NZUs is gradually reduced, placing a cap on allowable emissions and creating a financial incentive for businesses to adopt lower-emission practices. The ETS serves as a key policy tool for meeting New Zealand’s international obligations under the Paris Agreement, as well as its 2050 net-zero target and emissions budgets outlined in the Climate Change Response Act 2002 [
23].
Managing redwood under a continuous cover forestry (CCF) regime is likely to be highly profitable, particularly in the North Island of New Zealand. Initial carbon revenue can be leveraged to support the transition to a steady-state forest that generates a continuous stream of valuable timber with a very high rate of return [
24]. However, several challenges, important to forest practitioners, remain regarding CCF implementation–particularly around coupe size, green-up requirements and operational costs. Therefore, the aim of this study was to develop CCF tactical plans for two forest estates with dissected terrain, exploring the potential of CCF to produce high-quality timber while maintaining a constant carbon stock over time. Specifically, we aimed to test the hypothesis that harvesting small coupes while leaving all adjacent coupes unharvested for a defined green-up period is operationally and economically viable. We predict that some of the alleged drawbacks of continuous cover forestry are not insurmountable.
2. Materials and Methods
2.1. Forest Estates Under Study
The case study simulates the management of two forest estates: Blue Mountain, Taranaki (109 ha) and Spring Creek, Manawatu-Whanganui (467 ha) both of which are located in the North Island of New Zealand (
Figure 1). For modeling purposes, both estates were assumed to be fully planted with redwood on bare land at the beginning of the planning horizon (year 0). The strategic planning horizon was 300 years while the tactical horizon was 60 years (1/5 of the strategic horizon). Time was aggregated into 5-year periods.
Continuous cover forestry was defined within this study as the retention of at least 30% of each hectare in forest with an age ≥ 5 years. In practice this was achieved by dividing both forest estates into 0.7 ha hexagonal units, with the condition that once a unit is harvested the neighboring units could not be harvested for a period of at least 5 years.
The methods used for this theoretical study were developed in conjunction with four senior forest managers who had an advanced understanding of harvesting and redwood management and silviculture. An initial draft of the manuscript was circulated to managers and then modified to incorporate their feedback. This process was repeated two more times until the managers were satisfied that the developed methods could be practically implemented.
Terrain across the two properties differed substantially. Blue Mountain had a gentler landscape, with an average slope of 25%, while Spring Creek was more rugged, averaging 42%. Slopes between 30% and 40% are generally considered the upper operational limit for ground-based harvesting systems, whereas cable systems can be used on steeper terrain [
25,
26,
27]. At Blue Mountain, 65% of the area had slopes below 30%, and 77% below 40% (
Table 1), suggesting that ground-based harvesting systems are likely to be feasible across most of the estate. In contrast, 77% of Spring Creek exceeded a 30% slope, indicating that cable-based harvesting systems would be required—an assumption supported by the highly dissected terrain (
Appendix A).
The silvicultural regime consists of planting 625 stems/ha, prune 450 stems/ha at ages 6 (8.2–9.9 cm dbh, 4.1–4.9 m height), 8 (10.6–14 cm, 6.1–7.4 m) and 10 (20.8–23.5 cm, 8.2–9.8 m) to a height of 6.5 m, and thin at age 11 (24.5–27.6 cm, 8.2–11.1 m), to 450 stems/ha. This silvicultural regime was selected to maximize the yield of high-value clear heartwood, which commands a substantial market premium over grades containing sapwood or unpruned “tight knot” timber—both of which are still more valuable than timber with bark-encased knots [
28]. Harvesting will be undertaken on stands aged 35 years and older, and natural regeneration (coppice) will be established after felling the first rotation. The coppice will be thinned at ages 2 to three–five dominant shoots/stump, and at age 5 to a single dominant shoot/stump, maintaining the same pruning regime, outlined above, at ages 6, 8 and 10 years old. Assumptions are that yield tables will not be changed by plant source (seedling vs. coppice) or by re-establishing trees in 0.7 ha coupes surrounded by well-established stands.
2.2. Overall Modeling Framework
The modeling framework comprised the six stages depicted in
Figure 2. The first stage consisted of predicting carbon and timber yield over ages 0–100 years for both properties. We consider 100 years to be the maximum harvesting age based on piece size. The second stage consisted of estimating realistic carbon and timber prices achievable for the region where both properties are located. During the third stage we estimated likely silvicultural, management, harvesting and transport costs. The fourth stage consisted of planning potential roads to access all 0.7 ha units to estimate the total fixed roading cost. Once all data needed was collected, we ran a strategic harvest scheduling model based on linear programming in order to derive a general strategy for the management of redwood for both properties. The sixth and last stage, which is the focal point of this study, was to run a tactical spatially explicit harvest scheduling model, based on integer (binary) programming, analyzing the feasibility of implementing a continuous cover forestry regime for redwood. The tactical level uses the outputs produced at the strategic level, such as timber flow, applying them over shorter periods of time that are typically less than a rotation [
29]. All these stages are described in greater detail in the following sections.
2.3. Estimates of Carbon and Total Recoverable Stem Volume
Regional estimates of volume and carbon were based on a redwood growth model which is fully described in [
30,
31]. In brief, the redwood growth model, which was based on the 300 Index methodology, was developed using a comprehensive set of plot data distributed throughout New Zealand. The 300 Index approach incorporates both site index and the volume-based 300 Index as key inputs for the growth model. Using these inputs, the model estimates annual changes in stem volume, carbon accumulation, and log out-turn through an integrated framework of growth equations, allometric relationships, and functions for carbon partitioning and basic wood density. The redwood growth model is implemented in the freely available multi-species carbon calculator (Version 1.2) available at:
https://fgr.nz/tools/multi-species%20carbon-calculator/ (accessed on the 20 February 2025).
Estimation of the two key productivity metrics (site index and 300 Index) forms the foundation of volume and carbon predictions. For redwood, site index is defined as the mean height of the 100 largest-diameter trees (mean top height, MTH) at a reference age of 30 years. The 300 Index, a normalized volume measure developed to address the limitations of site index, represents the mean annual stem volume increment of a stand stocked at 300 stems per hectare at age 30. These two productivity indices, which are used as input to the growth model, can be extracted from plot data. Alternatively, if no plot data are available these two productivity indices can be extracted from national spatial predictions, described in detail in [
31], which are freely available at:
https://koordinates.com/ (accessed on the 14 January 2025).
The spatial predictions of productivity indices were used to determine 300 Index and site index shown in
Figure 1 for Blue Mountain while existing redwood plot data was used in the Multi Species Carbon Calculator to estimate both productivity metrics for Spring Creek. Once the 300 Index and site index were defined, the Multispecies Carbon Calculator was used to predict redwood volume, carbon and log grade out-turn. Compared to Spring Creek, Blue Mountain exhibited lower values of both 300-Index (21.5 vs. 29.3 m
3 ha
−1 yr
−1) and site index (27.9 vs. 32.7 m). Consequently, timber and carbon yield were lower in Blue Mountain than Spring Creek (
Figure 3).
For strategic and tactical harvest scheduling purposes, we used a fitted model to interpolate the data presented in
Figure 3: Blue Mountain: TRV = 5844 (1−1/e
(0.01945 age))
2.6563; CAR = 6841 (1−1/e
(0.01916 age))
2.5488; Spring Creek: TRV = 6663 (1−1/e
(0.02241 age))
2.7591; CAR = 7726 (1−1/e
(0.022186 age))
2.65939, where TRV is total recoverable volume (m
3/ha), CAR is carbon stock (t CO
2/ha) and age is expressed in years since planting.
2.4. Timber and Carbon Value
Redwood log values were estimated using two complementary approaches that included a comparative price analysis and lumber recovery, or ‘millback’, analysis [
32]. These at mill gate (AMG) values, by log diameter and pruned status, are shown in
Table A2. These values were integrated with predicted timber grade recoveries generated by the Excel-based redwood growth model, which estimates under-bark stem volume by pruned butt logs and unpruned upper logs across small-end diameter (SED) classes. The resulting mean AMG log values per cubic meter by stand age are shown in
Table 2 for both Blue Mountain and Spring Creek. These AMG values increased markedly with harvest age by 38.2% from ages 35 to 100 years (
Table 2).
A base carbon price value of NZD 80/t CO
2 was assumed for this study, according to current projections by the New Zealand Ministry for the Environment [
23]. A sensitivity analysis was conducted using selected carbon price points to represent different scenarios that included: a timber-only regime (NZD 0/t CO
2), the current minimum auction price under the ETS (NZD 35/t CO
2), and an upper bound of NZD 160/t CO
2, reflecting the Climate Change Commission’s projected carbon price for 2035 [
33].
For strategic and tactical harvest scheduling purposes, we used a fitted model to interpolate the data presented in
Table 2:
y = 213.39775910 + 3.23221289
x − 0.01023076
x2, where
y is price AMG (NZD/m
3) and
x is age (years).
2.5. Establishment, Silviculture, Harvesting, Roading and Management Costs
All costs related to establishment, silviculture, roading, and management were estimated based on a combination of industry consultation and published sources (see
Table 3). Initial establishment expenses in year 0 were NZD 3159/ha and covered activities such as land preparation, planting, release spraying, and mapping. Since redwood regenerates via coppicing after harvest, these upfront costs were not incurred in subsequent rotations. During the first rotation, pruning was carried out to a height of 6.5 m in years 6, 8, and 10, with each lift costing NZD 2329/ha. Thinning was performed at age 11 to reduce stocking to 450 stems/ha at a cost of NZD 900/ha. For second and later rotations, the same cost for the three pruning operations from the first rotation were used. However, there was a higher total cost for thinning the coppicing redwood sprouts (NZD 2000/ha) at age 2 (selecting 3–5 dominant shoots per stump), and age 5 (selecting a single dominant stem per stump).
Periodic costs included carbon-related administration, upkeep of infrastructure such as roads, tracks, and fencing, as well as animal pest control. Participation in the ETS was assumed to incur a cost of NZD 71/ha, with assessments scheduled every five years which is the minimum interval allowed under current regulations [
32]. These periodic expenses were incurred up to 35–40 years after establishment. Annual operating costs covered forest management, local rates, insurance, and general administration. A forest management cost of NZD 60/ha/year was applied to reflect the greater complexity involved in managing continuous-cover forestry (see
Table 3).
Ground-based harvesting was estimated at a cost of 60 NZD/m
3 through consultation with industry experts. As usual in NZ forestry, we referred to felling by a feller-buncher and extraction of the stems by a grapple skidder. Cable based harvesting (90 NZD/m
3) was estimated to cost 50% more than ground-based harvesting based on figures by [
34]. Considering an extraction distance in the range of 300–400 m, we selected a guyless excavator-based yarder (shovel yarder) equipped with a grapple carriage. For the felling operation, a tethered feller-buncher was chosen. This machinery enables significantly faster relocation times, between 1 and 1.5 h to establish a new line, compared to conventional tower or swing yarders [
35].
It is useful to define the theoretical layout of the cable harvesting lines that we have tentatively considered for tactical planning. This also helps assess whether there is sufficient harvestable material to justify the use of this extraction system. The proposed layout involves placing the shovel yarder at the edge of the patch, from which approximately three lines radiate to cover the area. Three lines are sufficient, given the use of a feller-buncher, which produces stacks accessible by the grapple carriage, and considering that the coupes are about 94 m wide at their broadest point.
To determine whether cable harvesting is justified, we can refer to a common rule of thumb in forest engineering: at least one cubic meter of harvestable timber is needed per meter of cable line. Given a 7000 m2 hexagon, the maximum width is approximately 94 m. Assuming an average extraction distance of 350 m and three lines required to extract timber from the entire patch, the total line length amounts to 1050 m, which implies a need for 1050 m3 of timber. This number will be reached and far surpassed once trees are 40 and 50 years old in Spring Creek and Blue Mountain, respectively. Furthermore, the rule of thumb was established for conventional tower yarders. Through using shovel yarders the relocation costs are lower, and this will further reduce the amount of harvestable volume required for making a given cable line profitable.
Transport costs to a mill were estimated at 19 NZD/m
3, assuming a cart distance of 100 km and transport carried out by a log truck [
36]. Other costs included road maintenance (3.75 NZD/m
3), post-harvest ancillary costs (1.88 NZD/m
3), a harvest management fee (5.63 NZD/ha) and the forest growers levy (0.33 NZD/m
3). These values totaled NZD 91/m
3 for ground-based and 121 NZD/m
3 for cable-based harvesting. These total costs were within or marginally above the range provided by [
37] (total cost range 42.2–119.0 NZD/m
3, average 68.6 NZD/m
3, corrected by CPI Q4 2017–Q4 2024) for small-scale owners of the Canterbury, Otago, and Southland regions. Harvesting and transport costs were subtracted from the AMG values in
Table 2 estimates to calculate net harvest returns. These net figures were then multiplied by the associated total recoverable volumes to estimate the value of timber extracted. Based on slopes (
Table 1), we assumed that harvesting will be ground-based in Blue Mountain and cable-based in Spring Creek. Although this is an oversimplification this could be readily refined during operational planning.
Using historical road construction costs (pers. comm. Robin Webster, 7 February 2025) corrected for inflation we arrived at a value of 137,091 NZD/km for primary roads, which ranged across forests from NZD 105,600 to NZD 166,320. The preceding was equivalent to a unitary cost per cubic meter of harvested timber ranging from 7.71 to 12.80 NZD/m
3 (average 9.01 NZD/m
3). A similar range was estimated by [
37] (7.7–19.2 NZD/m
3, average 12.4 NZD/m
3, corrected by CPI Q4 2017–Q4 2024) for small-scale owners in Canterbury, Otago and Southland. It was assumed that all existing roads in Blue Mountain and Spring Creek are primary roads while all new roads that need to be constructed to access coupes will be spur roads (single lane). Secondary roads construction costs were assumed to be half the cost of primary roads at 68,546 NZD/km. These values are similar to values reported by [
38] for spur roads (NZD 72,000/km) and higher standard secondary roads (NZD 90,000/km) in New Zealand.
2.6. Current and Potential Road Network
The current road network comprises 10.517 km and 20.446 km, which equates to a road density of 96.6 m/ha and 43.8 m/ha, for Blue Mountain and Spring Creek, respectively.
The tactical planning required an estimation of the size of the current and potential road network. The current road network is assumed to be primary roads, but secondary roads and skid tracks need to be planned. This is necessary because harvesting small, dispersed 0.7 ha hexagons across the landscape presents both logistical and cost-related challenges. To estimate an efficient access network, we applied the minimum spanning tree algorithm [
39], which identifies the shortest path that connects all nodes in a network. In the simplest case the weights or penalties are the distances between the nodes (the centroid within each coupe) although using construction cost as penalties are more common. The algorithm searches for the route connecting all nodes with the minimum sum of weights or penalties.
In this study, each arc connecting two nodes was assigned a penalty. Lower penalties were applied to arcs within a 25 m buffer of internal roads and ridgelines (for Spring Creek only), while penalties increased with slope for all other arcs. While this is a simplified approximation, and more detailed operational planning is needed to determine harvesting systems, landing locations, and spur road construction, the primary objective of this analysis was to estimate the potential road network length for calculating fixed costs.
The total road network corresponds to estimated fixed costs of NZD 2,109,494 for Blue Mountain—comprising NZD 1,441,791 for primary roads and NZD 667,704 for secondary roads—and NZD 6,372,969 for Spring Creek, including NZD 2,802,971 for primary and NZD 3,569,998 for secondary roads (
Appendix A). Although primary roads exist in reality, we conservatively assumed that this incurred cost should be included in our financial analysis.
2.7. Strategic Planning Model
Strategic planning in this study was based on a simplified version of Model III. This model, implemented through FOLPI (Forestry-Oriented Linear Programming Interpreter), was originally designed to optimize the timing and spatial distribution of harvesting activities across extensive forested landscapes [
40]. By solving this problem, we define the management strategy that should be adopted to guide the forest from its current condition to its ideal condition. Although strategic planning is not spatially explicit, this process provides key information (i.e., allowable cut, harvesting strategy, steady-state condition, rotation length) for the tactical planning which will be spatially explicit. Strategic planning is typically undertaken over relatively long periods of time that comprise several rotations.
In our particular case, the model identifies the optimal harvest schedule that maximizes net present value (NPV), subject to a series of management and financial constraints. The strategic planning horizon runs from year 0 to 300, and is split as a period of transition (0–150 years) and a steady-state period (150–300 years). Time was aggregated into 5-years periods. Trees are eligible for harvest at ages 35 and older, and from year 40 onward, at least 50% of the forest estate must consist of stands older than 20 years. Additionally, a non-declining carbon stock constraint is imposed to ensure that carbon income remains zero or positive throughout the planning horizon—thereby avoiding scenarios where carbon credits would need to be repaid. There are no restrictions on how many times a given hectare can be harvested over the planning horizon, aside from the constraints specified earlier. This modeling framework was applied to Blue Mountain and Spring Creek under the CCF regime.
The strategic model is described by four components: decision variables, parameters (or known values), an objective function and constraints which are described in more detail, below.
2.7.1. Decision Variables at the Strategic Level
As this is a strategic model, decision variables are not spatially explicit and represent the area being harvested from each age class in each period.
where
t is an index of time periods and
j is an index of age classes.
2.7.2. Parameters Used at the Strategic Level
The parameters of the strategic model refer to specific values that characterize the forest and its management over the long run. They comprise area by age class, carbon stock and recoverable volume along the planning horizon, costs and prices, among others. A list of these parameters is given below:
T | total number of periods; |
Pj | Net price per cubic meter (NZD/m3) of timber harvested from age class j; |
ϕ | Net price per tonne of CO2 stored (NZD/t CO2); |
Vj | Realizable (merchantable) yield per hectare (m3/ha) in age class j; |
Sj | Carbon stock (t CO2/ha) in age class j; |
r | Discount rate (as decimal number); |
aj | area (ha) of age class j at the beginning of the planning horizon; |
E | variable cost of establishing and tending 1 ha that was harvested. |
2.7.3. Auxiliary Variables at the Strategic Level
Auxiliary variables do not influence the optimization outcome but are included to aid in interpretation of the results. These variables include the following:
where
ztj is the area remaining in age class
j at period
t,
Ht is the total merchantable volume (m
3) harvested in period
t, and
Kt is the total carbon stock (t CO
2) in period
t.
2.7.4. Objective Function at the Strategic Level
The objective function, which is given below, consists of maximizing the net present value over a 300-year planning horizon. The base discount rate was set to 6% p.a. The revenues were derived from two sources: harvesting and carbon sequestration.
2.7.5. Constraints at the Strategic Level
- (a)
Area constraints
Three groups of constraints were implemented in order to ensure the conservation of area:
- (i)
The area harvested must be immediately replanted (Rt).
- (ii)
The area planted in any period (Rt) must be subsequently harvested.
- (iii)
The area at the beginning of the planning horizon in each age class must be subsequently cut.
But (i) and (ii) can be equal as
Rt is the same for both constraints, thus,
- (b)
Minimum and maximum harvest ages: harvesting is permitted for trees aged between 35 and 100 years, i.e.,
where
j− and
j+ are the age classes below and above which trees cannot be harvested.
- (c)
Minimum canopy cover. Ensure that 50% of the forest estate should be age 20 years or older after year 35 along the planning horizon. That is,
where
fa is the fraction of the forest estate that should be over a given age (
fa = 0.5,
j· = 5 [20–25 years]) and
A is the total area of the forest estate.
- (d)
Non-declining carbon stock. This constraint ensures that carbon stock does not decline so that revenue from carbon is received but does not have to be returned.
- (e)
Timber flow constraints. These are expressed as relationships between consecutive time periods, specifying the allowable maximum increase (α) or decrease (β) in harvest volume. Both α and β are defined as proportions, ensuring that the volume harvested in any period does not deviate beyond these limits relative to the previous period.
Timber flow constraints were introduced beginning in period 8 (years 35–40), reflecting the assumed minimum harvest age of 35 years. The parameters α and β were both set at 0.1, limiting harvest volume fluctuations between periods to a maximum of ±10%.
2.8. Tactical Planning Model
At the tactical level, numerous management models exist, with those incorporating spatial and adjacency restrictions gaining prominence [
41,
42]. These models schedule harvests in a way that ensures adjacent units are not harvested simultaneously, maintaining compliance with clearcut size regulations and policy limits [
29].
From a modeling perspective, representing spatial constraints can be challenging. There are two potential models that can be applied in tactical forest management planning. The Unit Restriction Model (URM) evaluates harvest units so that they do not exceed a maximum permitted area. A mathematical model is then applied to ensure that adjacent units are not simultaneously harvested. In other words, the URM harvests previously defined spatial units of a size less than the area restriction, so that simultaneously harvesting two adjacent units would violate the spatial constraint [
43]. In contrast, the Area Restriction Model (ARM) allows adjacent smaller units to be harvested simultaneously, provided the combined contiguous area remains within the upper limit [
43].
In this study, spatial restrictions were imposed using the URM. A grid of 0.7 ha hexagons was overlaid across both Blue Mountain and Spring Creek. This unit size was selected to ensure that, under adjacency constraints, at least 30% of any given hectare retained forest cover at all times. A key consideration when applying adjacency rules is the minimum interval before any other neighboring unit can be harvested. Based on practices in Californian redwood forests, where common time lags for harvesting adjacent clear cuts are around four years [
29], a five-year interval was adopted in our analysis. Under these conditions several questions arise about coupe size, green-up, and the operational costs of CCF. For example, managers may question whether using a coupe size of 0.7 ha is feasible or whether it is viable to maintain unharvested stands of a minimum age around a harvested stand for a designated green-up period. Under current legislation, a conservative minimum age for green-up would be at least five years.
The proposed tactical spatially explicit model aims to determine a harvest schedule for a set of stands that maximizes profit while adhering to adjacency constraints and maintains a relatively even flow of timber harvested and carbon stock. This model was applied to both Blue Mountain and Spring Creek forest estates.
The tactical model is described by four components: decision variables, parameters (or known values), an objective function and constraints which are described in more detail, below.
2.8.1. Decision Variables at the Tactical Level
The decision variable selected is a integer binary variable, which means that could take either a value of 0 or 1 but not a fractional value. There are as many variables as stands and periods.
xit | A binary variable that indicates whether stand i is harvested in period t (1) or not (0) |
where
i ∈ A (stand, A is the set of stands);
t: 1, 2, ……, m (time period).
2.8.2. Auxiliary Variables at the Tactical Level
Auxiliary variables are a combination of decision variables and parameters. They do not influence the final solution but document and keep track of management quantities relevant for decision making. For the purposes of the exercise there are two auxiliary variables:
Ht | An auxiliary variable to account for the total volume harvested in period t; |
Ct | An auxiliary variable to account for total carbon stock in period t. |
2.8.3. Parameters at the Tactical Level
The parameters of the model refer to specific values that characterize the forest and its management at the tactical level. Parameters describing the forest estate include the stand areas, recoverable volume and carbon stock along the planning horizon, among others. A list of these parameters is given below:
A | Set or list of all stands that are within a forest estate. |
Bi | Set or list of stands adjacent to a particular stand i. For each stand in list A there is a list of stands which are neighbors of stand i, e.g., B[1] = {2,4,6,7,9}, meaning that the neighbors of stand 1 are stands 2,4,6,7 and 9. |
m | Number of periods, e.g., for 12 periods of 5 years each, m = 12. |
ai | Area of stand i (ha). |
vit | Total recoverable volume (m3/ha) that could be harvested from stand i in period t (period t goes from 1 to m). |
npvit | Net present value (NZD/ha) from harvesting stand i in period t. |
Hmin | Minimum volume to be harvested per period (it comes from the results of the strategic planning). |
Hmax | Maximum volume to be harvested per period (it comes from the results of the strategic planning). |
The
npvit is an auxiliary parameter derived from harvesting 1 ha of spatial unit
i in period
t. Therefore,
where
pit is the net price per cubic meter (at mill gate price from
Table 2 minus the harvesting cost, i.e., 91 NZD/m
3 for ground-based and 121 NZD/m
3 for cable-based harvesting) harvested from unit
i in period
t,
Cs is the discounted silvicultural cost of pruning and thinning after harvesting, and
r is the interest rate.
2.8.4. Objective Function at the Tactical Level
The objective function aims to maximize the net present value from harvesting the stands over the planning horizon. This is represented by:
2.8.5. Constraints at the Tactical Level
- (a)
Area Constraints: Each stand can only be harvested once along the tactical planning horizon (60 years) or left unharvested:
- (b)
Adjacency Constraints: Two adjacent stands cannot be harvested in the same period:
- (c)
Total volume harvested per period: The total volume harvested in each period is calculated as:
- (d)
Total volume harvested per period within a range: The total harvest volume in each period should stay relatively constant within a range given by the results of the strategic planning (Hmin, Hmax):
2.9. Optimization Algorithm
The strategic model was developed in Microsoft Excel (version 16.54, 2021) and solved using the OpenSolver add-in [
44]. OpenSolver is an open-source extension for Excel that enables the formulation and solution of linear and integer programming problems. Optimization was performed using the COIN-OR CBC solver, which is integrated within the OpenSolver platform.
The tactical problem was programmed using the Operation Research’s language MATHPROG (a subset of AMPL) and solved using packages Rglpk (version 0.6-5.1) and Rsymphony (version 0.1-33) in the R system for statistical computing [
45]. Both R packages provide a high-level interface to R for solving large-scale linear programming (LP) and mixed integer linear programming (MILP) problems based on GLPK and COIN-OR SYMPHONY solvers.
2.10. Sensitivity Analyses
Sensitivity analyses were carried out for each forest estate by changing interest rates (4, 6, 8% p.a.), yields (±50%, ±20% base condition), timber prices (±20% base condition), carbon prices (NZD 0, 35, 80, 160/t CO
2), harvesting costs (1× and 2× those in
Table 3), roading costs (1× and 2× those in
Table 3) and whether primary roads were constructed just before planting or when harvesting starts. We also considered delaying harvesting by 5 years from year 35 to 40. During the sensitivity analysis everything else was kept constant (
ceteris paribus), changing one value at the time and recording the effect on the optimization run.
4. Discussion
The management of redwood plantations under continuous-cover forestry (CCF) has been shown to be highly profitable particularly in the North Island of New Zealand [
24]. Our research hypothesis about the operational feasibility of managing redwood forests by harvesting 0.7 ha coupes under a CCF regime was confirmed. The IRRs were 9.16 and 10.40% p.a. for Blue Mountain and Spring Creek, respectively, which exceed those considered to be robust IRRs for radiata pine and other forest species (Douglas-fir, eucalypts and blackwood) in New Zealand [
47]. The same authors reported an average IRR of 9.3% for redwood plantations managed under clearcut regimes in New Zealand.
A key advantage of managing redwood plantations under a CCF regime is the ability for small-scale owners to generate substantial and continuous annual income by harvesting small 0.7 ha coupes starting from age 35. This contrasts with species such as radiata pine, which typically require the harvesting of larger stands at longer intervals. For instance, at Blue Mountain about 3673 m3 will be harvested annually, over ages 35–95 years, according to the tactical modeling. Considering a net price of about NZD 299/m3 (price 390 NZD/m3 at age 70, minus a variable cost of 91 NZD/m3) that would amount to about 1,098,227 NZD/year from only 108.9 ha. For Spring Creek, about 20,763 m3 will be harvested per year, over ages 35–95 years, which at a net price of NZD 269/m3 (price 390 NZD/m3 at age 70, minus variable cost of 121 NZD/m3) would amount for 5,585,247 NZD/year from only 467.2 ha. These figures highlight the strong financial returns achievable through the steady harvesting of small, high-value coupes under a redwood CCF system.
An interesting finding of this study was that the optimal rotation length exceeded that typically used in New Zealand for redwood managed under clearcutting regimes. Previous research [
47] proposed a silvicultural regime for redwood involving an initial planting density of 800 stems/ha, followed by up to three pruning lifts and one thinning to a final stand density of 350 stems/ha, with a rotation length of 35 years. However, strategic and tactical modeling, accounting for both timber and carbon income, indicated an optimal rotation length of approximately 71 years. Adopting a shorter rotation (i.e., 35 years) would result in a lower steady-state carbon stock and potentially trigger the repayment of carbon credits. Consistent with our findings, [
48] suggest that the optimal rotation for second-growth redwood stands lies between 60 and 80 years, and these older ages are associated with superior wood quality, with higher heartwood content [
49].
The coupe size used in our study was generally consistent with those previously reported for redwood. Silvicultural systems for redwood in California are grouped into even-aged (clearcuts) and uneven-aged (selection cuts) with both producing heterogeneous landscape patterns, where the group selection system consists of small clearcuts of 0.04–1.2 ha [
48]. Similarly, our proposed system for redwood plantations in New Zealand consists of harvesting small 0.7 ha coupes which are spread across relatively small properties (100–500 ha). A previous study [
50] argues that smaller coupes of less than 0.25 ha should be used under continuous cover forestry, but when redwood experts were consulted, they suggested that such small coupes would not be operationally feasible for New Zealand. Additionally, although redwood trees continue to grow and develop rapidly even after crown closure [
48], the species grows most rapidly under large openings [
51], that have at least 50% of the above canopy light [
19], with this threshold easily achieved using a coupe size of 0.7 ha [
52].
The coupe shape did not show any influence on profitability, timber and carbon flows. For tactical planning, group clearcuts were defined using hexagonal units, which can be operationally approximated by harvesting a cluster of adjacent trees covering approximately 0.7 ha in a roughly circular shape. Following thinning at age 11, stand density is reduced to 450 trees/ha, resulting in an expected density of around 315–325 trees/ha by the end of the rotation (~71 years). At this density, a 0.7 ha unit would contain approximately 224 trees. Thus, knowing the UTM coordinates of the centroids of each 0.7 ha unit, should be sufficient to locate that point in the field and from that point draw a circle with an imaginary radius of 47.2 m from which 224 trees will be harvested. The trees harvested per coupe would have a recoverable volume of 630–869 m3 if harvested at age 35 and 1862–2450 m3 if harvested at age 70. The tactical model was also run using square (83.7 m sides) and rectangular (35 × 200 m) coupes, yielding results that were nearly identical to those obtained with hexagons. This indicates that alternative shapes—such as ridge-to-gully rectangular coupes—could be readily integrated into operational planning at no extra cost.
Adjacency constraints prevent contiguous units being harvested at the same time so that the maximum size of clearcuts does not exceed statutory or policy limits [
29]. These constraints are now being routinely used in harvest scheduling planning for erosion control, biodiversity conservation and reducing the negative impacts of storms on forests, among others [
29,
41,
42,
53]. We found that imposing green-up constraints on the continuous-cover management of redwood had a marginal effect on profitability (<0.4% in NPV). Similarly, Ref. [
54] using a 5-year exclusion period for a 500 ha forest in Japan, resulted in a small reduction in NPV (0.96%) compared to a no exclusion period. Consistent with our study, Ref. [
29] suggested that common time lags for harvesting adjacent clear cuts of redwood forests in California are 4 years. In our case we set the lag to 5 years, which is the time considered necessary for coppice to reach a height of 5 m, which conservatively aligns with New Zealand ETS legislation.
Harvesting small 0.7 ha coupes spread over the landscape might impose operational constraints dictated by the terrain. Slopes up to 30–40% are generally considered as feasible for ground-based systems while cable-based systems are able to work at greater slopes [
25,
26,
27]. However, ground-based systems have also been employed on steeper country while operating from off contour roads spaced up to 100 m apart [
50]. At Blue Mountain, 66% of the area had slopes below 30%, and 78% below 40%, suggesting that ground-based harvesting systems are likely to be feasible across most of the estate. In contrast, 77% of Spring Creek exceeded a 30% slope, indicating that cable-based harvesting systems would be required. Given the hexagonal shape of the coupes, ground-based harvesting methods typically involve dragging logs through the stand to the nearest road. In cases where ridges and spurs can be tracked, ground-based systems may be extended to steeper areas; otherwise, cable-based systems will be required [
50]. Skylines set at ridges are routinely used to harvest small coupes under CCF in broken terrain in Europe, although this method is generally limited to distances of 300–500 m [
55]. These ridge-to-ridge distances are well within reach for both Blue Mountain and, in particular, the more dissected terrain of Spring Creek. The above suggests that detailed operational planning would be needed to harvest small 0.7 ha coupes particularly in broken terrain although difficulties are not unsurmountable.
The large piece size resulting from the extended rotation age under the CCF regime may pose operational challenges. The tactical planning shows that all first rotation crop will be harvested within 60 years (ages 35–95), with regeneration by coppice. The mean diameter at breast height (dbh) size range over these ages ranges from 62 to 118 cm in Blue Mountain and 68–133 cm in Spring Creek. Experts argue that up to age 50 the piece sizes are easily manageable (77–84 cm in dbh) but that after that harvesting and logging would be more difficult although still feasible. Difficulties can arise specifically in the mechanized felling of trees. Fully mechanized tree harvesting is generally limited to trees with a dbh of 60–80 cm, depending on the machine type. Feller bunchers typically handle trees up to 60–65 cm dbh, with productivity and safety declining near the upper limit. Cut-to-length harvesters, particularly larger models used for final felling of conifers, can process trees up to 75–80 cm dbh, though performance and efficiency decrease at these extremes. Trees exceeding these thresholds often require manual felling or specialized equipment. However, in the short term, a sensitivity analysis of costs indicates that shifting to motor-manual felling is feasible. While this method is more expensive, it allows larger trees to be harvested. The analysis clearly shows that, despite higher harvesting costs, the economic viability of redwood plantations managed under a CCF regime remains unaffected.
Continuous cover forestry would generally require a higher road density than forests under rotational clearfell regimes [
50]. The road density in our case study was in the range 155–186 m/ha. As expected, these values are much higher than the average of 25.2 m/ha (range 0 to 52.8 m/ha) for small-scale clearfelled radiata pine plantation forests (woodlots) in New Zealand [
56]. However, in the alpine region of Europe, where CCF is more widely practiced, higher road densities of 38–51 m/ha are required to reduce extraction distances and minimize cable logging costs [
57,
58]. Despite the high density of our roading system the sensitivity analysis shows that the effect of a doubling of the roading cost was marginal on the profitability of redwood CCF. This insensitivity was attributable to the high timber value, high timber volume and the fact that the roading cost was discounted 35 years.
In tactical planning, prioritizing harvesting areas near existing roads is a common strategy to minimize road construction, transportation and streamline harvest operations [
59]. In our study we implemented a “roads-first-policy” by doubling the selection priority for units located within a 50 m buffer of primary roads, with additional emphasis placed on units scheduled for earlier harvest. As a result, most units harvested during the initial periods were clustered near existing roads, while those scheduled for later harvest were generally located farther away. The implementation of this policy incurred a marginal cost of less than 1.6% of the net present value (NPV) compared to the condition without it, which was deemed to be acceptable by redwood managers.
Several potential limitations of this study warrant further investigation. Firstly, although redwood is shade-tolerant [
7], there is uncertainty regarding whether coppice shoots will suffer from shading by surrounding mature trees. This could potentially reduce carbon accumulation and hinder regeneration across significant portions of the coupe, particularly the most shaded areas along the coupe edge. Nevertheless, past research generally supports our assumption that 0.7 ha coupes will not be unduly impacted by shade [
19,
52,
60], and future studies should examine the resprouting success of redwood coppices, and growth rate of young trees, within coupes bordered by trees with a range of heights. Second, tactical planning on steep terrain may require modifications to coupe size and shape to ensure cost-effective operation of cable-based harvesting systems. While the high timber volume of redwood suggests that even small 0.7 ha coupes could yield enough to justify cable-based extraction, this assumption should be validated in operational settings. Thirdly, the generation of a potential road network based on extensions from existing primary roads likely overestimates the required length and cost of secondary and spur roads. Our model assumed a road density which exceeds those used in clearfell regimes in New Zealand [
56] and even within European studies where CCF is more widely practiced [
57,
58]. However, higher road density may be justifiable under CCF systems [
50]. Importantly, sensitivity analysis showed that even doubling the roading cost had a limited effect on overall profitability, reducing the internal rate of return (IRR) only slightly—from 9.16% to 8.92%. Finally, all the simulations were made under the current climate, which we are aware is an oversimplification.
Land use change from agriculture to forestry is currently a topic of active debate in New Zealand, particularly due to its potential implications for emerging carbon market needs under the New Zealand Emissions Trading Scheme (ETS) [
61]. Estimates suggest there are approximately 1.2 million hectares of marginal or unproductive agricultural land suitable for afforestation [
62]. Future afforestation is expected to occur primarily on former sheep and beef farming land, especially in areas with low grass productivity or terrain too steep for dairy farming [
63]. Exotic forests are commonly established on Land Use Capability (LUC) classes 6 and 7—land also traditionally used for sheep and beef farming. However, afforestation must comply with the National Environmental Standards (NES) (
https://www.environmentguide.org.nz/rma/planning-documents-and-processes/national-environmental-standards/, accessed on 28 June 2025) and the Resource Management Act (1991) (
https://www.legislation.govt.nz/act/public/1991/0069/latest/DLM230265.html, accessed on 28 June 2025), which together form the core framework for sustainable land use planning in New Zealand. These regulations may constrain afforestation in certain areas to ensure the sustainable management of natural and physical resources [
64]. While sheep and beef farming may at times compete with forestry, robust institutional frameworks and legislation help balance competing land uses to achieve New Zealand’s broader social, environmental, and economic sustainability goals.
Our model, while currently theoretical and not yet field-tested, suggests that the proposed management approach is financially attractive for high-productivity, high-value timber species like redwood. We predict that harvesting small coupes (approximately 0.7 hectares), with a minimum five-year green-up period, would likely be more ecologically and socially acceptable than large clearcuts. This is supported by evidence that smaller harvest patch sizes enhance wildlife habitat, biodiversity, and landscape aesthetics [
65]. In addition, we anticipate this approach would be perceived as superior to other alternatives by a relevant segment of society [
66].