Carbon Sequestration Estimates for Minor Exotic Softwood Species for Use in New Zealand’s Emissions Trading Scheme

Abstract

New Zealand’s Emissions Trading Scheme (ETS) allows growers to receive payments through the accumulation of carbon units for increased carbon stock. For forests < 100 ha, growers rely on pre-formulated lookup tables (LUTs) to estimate changes in carbon stock by age. Currently, minor exotic softwood species, which are predominantly redwood and cypresses, are covered by a general Exotic Softwoods LUT. However, this table has been found to significantly underestimate carbon sequestration for these species. Using a combination of growth models and productivity surfaces, the objective of this study was to provide draft updates for the Exotic Softwoods LUT based on redwood, and two key cypresses (Cupressus lusitanica and C. macrocarpa), at different scales (national, Island level, regional), and to identify the most appropriate scale for a revised LUT. For cypress species, carbon predictions were made using C. lusitanica for the North Island and C. macrocarpa for the South Island, as these are the preferred species for each island. Variation in redwood carbon among New Zealand’s nine regions ranged over two-fold at ages 30 (390–847 tonnes CO₂ ha⁻¹) and 50 (926–1956 tonnes CO₂ ha⁻¹) and carbon was much higher within the North Island than the South Island. Predicted carbon for cypresses was higher within the North Island than the South Island at all ages and varied across regions, by 38% at age 30 (610–840 tonnes CO₂ ha⁻¹) and 12% at age 50 (1019–1146 tonnes CO₂ ha⁻¹). These findings suggest that a separate LUT for redwood is warranted, and that cypress species could serve as the default species for a revised Exotic Softwoods LUT. They also suggest that regional tables should be considered for both redwood and cypresses. However, the government may consider factors other than the technical considerations outlined here when updating the LUTs.

1. Introduction

Carbon can now provide a significant source of revenue to forest growers [1] and the New Zealand Emissions Trading Scheme (ETS) is relatively unique from a global perspective in that it includes the forestry sector [2]. Forest growers registered in the ETS receive a New Zealand unit of carbon for each tonne of CO₂ sequestered (one tonne of CO₂ = 0.27 tonnes of carbon). Within the ETS, pre-1990 forests are defined as those established before 1 January 1990, and are recognized as part of New Zealand’s baseline carbon storage. Pre-1990 forests are ineligible for earning carbon credits in the New Zealand ETS. However, if they are deforested or converted to another land use, they are automatically enrolled in the New Zealand ETS, and units must be surrendered to the government to offset the loss of baseline carbon storage. Post-1989 forests, which are forests established after 31 December 1989, are regarded as new carbon sinks. Post-1989 forests can be voluntarily registered in the ETS to earn carbon units for their stored carbon.

Payment of carbon units within the ETS to growers is either through an averaging or stock change carbon accounting methodology, which are, respectively, suited to forests that are periodically clear-felled or permanently established [3]. Under the stock change method, the grower is allocated units as carbon accumulates over the lifetime of the stand but must surrender some when carbon is lost at harvest [3]. The stock change method is used for the permanent forest category, which came into effect within the ETS on 1 January 2023, and allows owners to earn carbon units from forests that are not intended to be harvested for at least 50 years after they are registered [3]. Under the averaging method, forests earn units during the first rotation up to a specified age representing the point at which the forest achieves its long-term average carbon stock. This is illustrated by the indicative carbon stock changes shown in Figure 1. Beyond this age, payments cease, which has the advantage that no repayment of units is required when the forest is harvested as long as it is replanted. For exotic softwoods, this long-term average age is currently set at 16 years for radiata pine (Pinus radiata D. Don), 26 years for Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) and 22 years for all other exotic softwood species.

Within the ETS, there are two ways in which the quantity of carbon units allocated to different species is measured. Participants who have registered 100 ha or more in the ETS must establish plots using the Field Measurement Approach (FMA) to determine carbon values. However, growers with less than 100 ha registered must use pre-defined look-up table (LUT) values, which describe annual changes in carbon for different species [4]. Currently, there are regional LUTs for radiata pine and a national LUT for Douglas-fir, these being the two most widely planted exotic softwoods in New Zealand. For all other exotic softwoods, a general Exotic Softwoods LUT is used. However, compared to data collected using the FMA method, this Exotic Softwoods LUT has been found to considerably underestimate sequestration, with FMA participant-specific tables having carbon values up to 130% higher than the LUT (based on 2013–17 reporting period). In particular, evidence suggests that significantly more carbon is stored in redwood (Sequoia sempervirens (Lamb. ex D. Don) Endl.) forests than is credited under the default Exotic Softwoods LUT.

Given the disparity in the current Exotic Softwoods LUT with available measurements, the development of a revised LUT more reflective of current data would provide more equitable outcomes for forest growers. Within the Exotic Softwoods class (i.e., exotic softwoods other than radiata pine and Douglas-fir), redwood covers a total of ca. 16,000 ha, including 10,000 ha established since 2000, while cypresses are the second most important group covering 9057 ha [5]. The area of all other exotic softwoods can be estimated from the above at around 9300 ha [5]. These other exotic softwoods comprise an assortment of pine species (other than radiata pine), larches, firs (other than Douglas-fir), Japanese cedar, and other species, all of which, as individual species, have been planted at a relatively small scale. Among the cypresses, the two most widely established species in New Zealand are Cupressus lusitanica Mill. (Mexican cedar) and C. macrocarpa Hartw. (Monterey cypress) [6]. As these two cypresses are widely planted and have a moderate growth rate, they represent a reasonable reference for the trees in the Exotic Softwoods category apart from redwood. Whether a separate LUT for redwood is required will depend on whether its carbon yield is sufficiently different to justify a separate forest type.

Redwood is a fast-growing evergreen tree species native to a coastal strip in the western United States where it produces among the tallest and longest-lived trees on earth. Within its native range, redwood can develop mixed-aged forests with individuals reaching heights of 115 m and ages surpassing 2200 years [7,8,9]. Redwoods have been found to store exceptionally high quantities of biomass and carbon within decay-resistant heartwood in primary (old growth) stands, but the rate of carbon sequestration is greatest in secondary or plantation forests [9]. As redwood stands are typically very healthy [10,11] and resistant to damage from wind [12,13] and fire [14,15,16], the large quantities of carbon that accumulate are relatively safe from damage caused by pests, pathogens and abiotic factors. Redwood also yields premium, dimensionally stable lumber characterized by its naturally durable heartwood and appealing grain pattern [17], primarily utilized for appearance grade applications [13].

Growth rates of redwood are optimal in temperate climates that are mild with moderate to high rainfall. Plantations have been successfully established in many countries with these climates [18,19,20]. Redwood is very well suited to the New Zealand climate. The existing redwood resource is relatively small, comprising approximately 16,000 ha, with 10,000 ha established between 2000 and 2018 [13]. However, the species has recently increased in popularity and, after radiata pine, was the second most sold nursery species in New Zealand during 2023 [21]. The majority of redwood plantings have been in North Island locations where the climate suits the species [13], and suitable sites provide very high growth rates exceeding those of plantations within the native range [22]. Redwood growth and yield models have recently been developed based on data from a nationwide network of permanent sample plots (PSPs), which are permanently marked and regularly measured forest plots typically 0.04–0.06 ha in area. These models, along with associated spatial layers depicting volume and carbon, clearly demonstrate the exceptional growth potential of redwood in New Zealand’s environmental conditions [22,23,24].

Cypress species are also widely planted exotic softwoods within New Zealand. Cypresses are evergreen conifers that belong to the Cupressus genera and are native to warm temperate regions in the northern hemisphere. Two important species within this genus are C. macrocarpa and C. lusitanica. Cupressus macrocarpa is native to the Californian coast [25,26] and has been planted in western Europe (Great Britain, France, Ireland, Greece, Italy and Portugal), Australia, Kenya and South Africa. Plantings of the species have naturalized in New Zealand, where it is grown commercially. Cupressus lusitanica naturally occurs in central America and Mexico [27,28,29,30] and has been established commercially in Bolivia, Colombia, South Africa, eastern Africa and New Zealand. These two species have been widely planted, as they are relatively fast-growing and have dimensionally stable timber, with an attractive grain that is naturally durable and suited to many high-value appearance grade purposes [31].

The objective of this study was to estimate annual carbon stocks for redwood and cypress plantations in New Zealand at a range of scales (national, Island level, regional) in order to identify the most appropriate scale for a national Exotic Softwoods LUT, and to determine whether a separate LUT is required for redwood. Annual carbon stocks were determined for pre-1990 forests to determine liabilities following clear-felling, and for post-1989 forests to characterize carbon sequestration following establishment of these predominantly new forests. The study provided draft sets of LUTs for both types of forests at different scales.

2. Materials and Methods

2.1. Growth Models

The growth models used in this study are implemented in the freely available multi-species carbon calculator (Version 1.2), an updated Excel model which uses 300 Index and site index as inputs to account for site productivity (https://fgr.nz/tools/multi-species-carbon-calculator/ accessed on the 15 September 2024). The redwood growth model is based on a model described in [24,32], updated using recently re-measured PSP data, and fully described in Supplementary Material S1. The C. macrocarpa and C. lusitanica growth models are described in [33]. All three models use the 300 Index approach to account for variation in site productivity and silviculture. Traditional growth and yield models use site index to account for variation in site productivity, which is defined in New Zealand for all three species as the mean top height (MTH, or the average height of the 100 largest diameter trees per hectare) at age 30. The 300 Index was introduced as a volume productivity measure to address the shortcomings of site index for estimating site quality in growth models. There is often significant variation in volume in stands of identical height that is not captured by site index [34,35,36,37], with this variation sometimes reaching ±30% [38]. While stand volume provides a more robust assessment of site quality compared to height, it is influenced by stocking and silvicultural practices. The 300 Index was formulated to account for these two factors [24,32], and represents the mean annual stem volume increment at 30 years of age for a reference stand with 300 stems ha⁻¹. A significant benefit of the 300 Index, and the accompanying growth models, is that once the 300 Index and site index have been defined, volume and carbon can be predicted over varying stand ages, densities, and silvicultural regimes. Furthermore, the models can be used in reverse to estimate site index and 300 Index from a plot measurement consisting of stand density, basal area, MTH and thinning history [24,32].

2.2. Growth Data

Growth data were sourced from the Scion Permanent Sample plot (PSP) database and from an FMA dataset supplied by New Zealand’s Ministry for Primary Industries (MPI). Redwood plots that were ≤5 years old were not used, as a key productivity index (300 Index) cannot be reliably estimated in very young stands. Three redwood plots that could not be matched with aspect, which was a key independent variable in the modelling, were also excluded. Following these exclusions, there were, respectively, 428, 405 and 569 plots, for cypresses (i.e., C. lusitanica, C. macrocarpa) and redwood (

Table 1. Variation in the number of plots at the Island, national and regional levels for the two cypress species (Cupressus lusitanica and Cupressus macrocarpa) and redwood (Sequoia sempervirens). Also shown is the partitioning of plot numbers by data source (PSP and FMA data) and the period of forest establishment (pre-1990 and post-1989 forests).

Region	Cypresses		Redwood
	C. lusitanica	C. macrocarpa	S. sempervirens
All data
North Island	324	92	435
South Island	104	313	134
New Zealand	428	405	569
PSP data
North Island	159	53	147
South Island	21	106	44
New Zealand	180	159	191
FMA data
North Island	165	39	288
South Island	83	207	90
New Zealand	248	246	378
Pre-1990 forests
North Island	68	27	61
South Island	8	81	10
New Zealand	76	108	71
Post-1989 forests
North Island	256	65	374
South Island	96	232	124
New Zealand	352	297	498
Regions
Auckland	52	7	10
Waikato/Taupo	65	22	157
Bay of Plenty	115	26	39
Gisborne	41	13	76
Hawke’s Bay/Southern Nth Island	51	24	153
Nelson/Marlborough	20	28	6
Canterbury/West Coast	83	151	115
Otago	1	119	12
Southland	0	15	1

).

The location of cypress plots differed by Island with the majority of C. lusitanica (76%) located in the North Island, while the majority of C. macrocarpa plots (77%) were in the South Island (Table 1; Figure 2). Most redwood plots (76%) were located in the North Island. Although most data were from post-1989 forests, the split between pre-1990 and post-1989 forests varied slightly by species (Table 1). The proportion of plots for post-1989 cypress forests ranged from 73% for C. macrocarpa to 82% for C. lusitanica and 88% for redwood. In total, the FMA plots constituted 58, 61 and 66% of the plots, respectively, for C. lusitanica, C. macrocarpa and redwood and the breakdown by data source (FMA or PSP data) is shown in Table 1.

2.3. Prediction of Productivity Indices

2.3.1. Extraction of Productivity Indices and Predictor Variables

National maps for both productivity metrics were generated for redwood and the two cypress species using the growth models and growth data described above. These surfaces are updates of surfaces previously developed for redwood [23] and the cypresses [39] using the more extensive set of plot measurements available for this study.

For each plot, the site index and 300 Index was estimated from the final plot measurement using the growth models described in Section 2.1. These were matched with data describing climatic, edaphic, topographic, and landform features which are fully described in Appendix A 

Table A1. Continuous variables included in the analysis. For each variable, the units, abbreviation and reference for the variable are given.

Variable	Units	Abbreviation	Reference
Climatic variables
Air temperature—mean, max, min.	°C	Tmean, Tmin, Tmax	[41]
Degree ground frost (frosts/month)	days	DGF	[40]
Growing degree days	Degree days	GDD	[40]
Rainfall	mm	Rain	[41]
Rainfall days	days year−1	Rain days	[41]
Relative Humidity at 9 am	%	RH	[41]
Solar radiation	MJ m−2 day−1	S Rad	[41]
Sunshine hours	hours	Sun hours	[41]
Windspeed	km hr−1	Wind	[41]
Water balance
Annual water deficit	mm	AWD	[45,70]
Drainage	mm	Drain	[45,70]
Gravel content	%	GC	[44]
Percentage root zone water balance	%	Wbal	[42]
Plant rooting depth	mm	PRD	[44]
Profile available water content	mm	PAW	[44]
Profile readily available water	mm	PRAW	[44]
Rainfall/Potential evapotranspiration		RPet	[45,70]
Topographic wetness index		TWI	[71]
Edaphic variables
Carbon: Nitrogen ratio		CN ratio	[43]
Macroporosity	%	Macro	[44]
Particle size			[45,70]
pH		pH	[44]
Phosphorus retention	%	P retention	[44]
Topographic and landscape
Aspect	°	Aspect	[45,70]
Distance to stream	m	DTS	[46]
Elevation	m	Elev	[47]
Multiresolution ridge top flatness		MrRFT	[72]
Multiresolution valley bottom flatness		MrVBF	[72]
Profile curvature		Profc	[72]
Slope	°	Slope	[46]

. The environmental variables were extracted from 25 m resolution surfaces covering the spatial extent of New Zealand. The climate variables included total rainfall, vapor pressure deficit, solar radiation, sunshine hours, windspeed, mean, and minimum and maximum air temperature [40], which were summarized at different temporal scales (yearly, seasonal, monthly). Also extracted were growing degree days and the number of days with rainfall and ground frosts [40,41]. Variables related to soil water balance included mean seasonal and monthly root-zone water balance with values expressed both in millimeters and as a percentage of the available root-zone water storage [42]. All variables described above were derived from climate data collected over a 30-year period from 1971 to 2000 and as such describe average spatial variation in climate and water balance.

Edaphic variables that were included in analyses were carbon–nitrogen ratio (CN ratio) from 0 to 10 cm [43], pH, macro-porosity, phosphorus retention from 0 to 20 cm [44], and particle size [45]. Water balance-related soil properties included plant rooting depth, gravel content, profile available water content, and readily available water content [44]. Topographic and landform variables included aspect [45], slope, distance to the nearest stream [46], and elevation [47], amongst others (see Appendix A Table A1).

2.3.2. Modelling Approach

Random forest was used to predict the two productivity indices for redwoods and cypresses using scikit learn, version 0.23.2. [48], which was implemented in Python, version 3.9.18. Random forest was used, as all datasets were relatively large and covered almost the entire environmental range over which the three species occur in New Zealand, which reduced the possibility of erroneous extrapolation to unseen conditions. Random forest is one of the most widely used machine learning methods [49,50], as it is capable of describing non-linear relationships, is easy to implement and handles collinearity and high-dimensional data very well. Random forest is a tree-based method that creates a large number of decision trees with the final predictions constituting the average predictions from individual trees. The term ‘random’ within the algorithm name is derived from the random sampling, with replacement (i.e., bootstrapping), of training observations for each learner (individual tree) within the forest and the use of a random subset of predictors at each split (node) within the tree. This approach enhances diversity among the trees, which reduces overfitting and improves generalization to unseen data.

Recursive feature elimination (RFE) was used prior to model fitting to subset the environmental predictors to the most important variables. The RFE undertook a 10-fold cross-validation and tested the accuracy of models with between 1 and 30 variables, with each of the 30 sets constituting the most important variables within the entire dataset. Each of the 30 datasets were further constrained, (through code integrated into the RFE) so that the set of selected variables had absolute correlation coefficients (R) < 0.9. The least important of each highly correlated pair was eliminated to meet this condition, which reduced the possibility of overfitting. The RFE process reduced the number of variables to 13 for redwood 300 Index and site index, 20 and 21, respectively, for C. lusitanica 300 Index and site index, and 15 and 8, respectively, for C. macrocarpa 300 Index and site index.

Following variable selection for each model, the dataset was divided into a training dataset comprising 80% of the observations with the remaining 20% retained as a test dataset. The random forest model was fitted to the training dataset using a 10-fold cross-validation with five repeats, and predictions from this fitted model were made on the independent test dataset. This process was repeated 49 times using a different train/test split during each iteration, and performance statistics were averaged over predictions on all 50 test datasets. This approach removed any potential bias that may have resulted from an unbalanced choice of test dataset using a single train/test split, providing a more rigorous evaluation of true model performance.

Model performance was determined on the test dataset using the root mean square error (RMSE) and the coefficient of determination (R²), which were calculated, respectively, as

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{\sum _{i=1}^n({\widehat{y}}_i-{y_i)}^2}{n}}}$$

(1)

$$R^2=1-{\displaystyle \frac{\sum _{i=1}^n(y_i-{{\widehat{y}}_i)}^2}{\sum _{i=1}^n({y_i-\overline{y})}^2}}$$

(2)

where $y_i$ is the observed value and ${\widehat{y}}_i$ the predicted value in plot $i$, $\overline{y}$ is the average of the observed values, and $n$ is the number of plots.

The importance score of the six most important variables in each model was obtained. Residuals for each model were plotted against predicted values, all key variables in each model and important variables that were not included in the model to identify any bias.

2.3.3. Extraction of 300 Index and Site Index for Spatial Predictions of Carbon

Spatial predictions of 300 Index and site index covering New Zealand were made for all three species using the models described above and the relevant environmental surfaces as input. A number of exclusions were made to limit the predictions to areas where each species is likely to be established. Regions with a mean annual air temperature (MAT) below 9 °C were excluded for redwood, and regions with a MAT below 8 °C were excluded for the two cypress species, as it is unlikely these species would be established in these colder/predominantly high-elevation areas. These cutoffs were strongly aligned with the data, which showed that the minimum MAT was 9.27 °C for redwood plots and 8.03 °C for the cypress species, C. macrocarpa, used for South Island carbon predictions. All lakes larger than the minimum mappable unit (0.0625 ha) were excluded. Contiguous areas of natural forest above 10 ha (LUCAS LUM), urban areas (LUCAS LUM), and protected areas (LINZ Protected Areas) were excluded. Major roads (LCDB5) that intersected urban areas were also removed. The Land Use Capability (LUC) layer was used to further exclude very productive land classes. The LUC categorizes land into eight classes according to its long-term capability to sustain one or more productive uses based on physical limitations and site-specific management needs. The LUC classes 1 and 2 were excluded, as forestry is unlikely to be established on these productive land classes.

Values of 300 Index and site index for the constructed rasters were averaged across each of the nine regions displayed in Figure 3 and Figure 4, as these regional boundaries have been previously used by MPI for LUT construction. The two productivity indices were also averaged to the Island (North Island, South Island) and national level using area-weighted regional values.

2.4. Predictions of Carbon

2.4.1. Overview

Regional estimates of volume and carbon were based on the 300 Index methodology, which is summarized above (see Section 2.1). Predictions of carbon for all three species were based on regional, island and national averages of 300 Index and site derived from the spatial models of the productivity indices (see Section 2.3.3). Carbon was predicted from these productivity indices using the multi-species carbon calculator, which uses methods described for redwood in [22] and for the cypresses in [39]. Carbon was predicted using a combination of growth models, allometric equations and functions describing carbon partitioning and basic density.

Annual predictions of carbon were made using standard silvicultural regimes for each species designed to optimize production of high-value timber but also allowing accumulation of carbon at a moderate rate. For Cypress spp., the regime used an initial planting density of 1000 stems ha⁻¹, with 450 stems ha⁻¹ pruned to 6.5 m over three stages at ages 6, 8, and 10, and a thinning to a residual stocking of 450 stems ha⁻¹ at age 11 years. An identical regime was used for redwood, except the stand was established at 815 stems ha⁻¹, as the cost of planting stock discourages higher initial stockings.

Although carbon predictions were made for both cypress species in both Islands, the final carbon predictions for cypresses used C. macrocarpa for the South Island and C. lusitanica for the North Island, as these are the preferred species for each island. As there was also a markedly larger proportion of plots for C. macrocarpa in the South Island and C. lusitanica plots in the North Island, this allocation also provided more robust productivity estimates for the cypresses. The Island and national level mean values of carbon documented for redwood and cypresses were area-weighted averages determined from the regional estimates.

2.4.2. Redwood Carbon Predictions

Carbon estimates for redwood provided by the multi-species carbon calculator were made using methods that have been described previously in detail [22]. The updated growth model and accompanying figures are given in Supplementary Material S1. Utilizing the 300 Index and site index estimates described for the three scales above, the redwood 300 Index growth model was used to predict annual changes in the mean top height and basal area, which were used to predict stem volume. Basic stem wood density for redwood was assumed to average 318 kg m⁻³, which was the average for 21 stands dispersed throughout New Zealand from the Canterbury region in the south to the Auckland region in the north. Stem wood biomass was derived as the product of basic density and volume. Bark volume was assumed to be 18% of stem wood volume [51] with density of 437 kg m⁻³ [52]. Biomass in redwood branches and foliage were estimated from diameter at breast height using allometric models [53] and roots were predicted using the IPCC default root–shoot ratio for coniferous forest of 0.23 [54]. Carbon inputs into the dead wood and litter pools were estimated using the growth model mortality functions and estimates of litter fall, and annual losses from these pools due to decay estimated as described in [22,54,55,56].

Four carbon pools were modelled, namely above-ground biomass (consisting of stem wood, bark, branches and needles of live trees), below-ground biomass (roots of live trees), dead wood, and litter. The litter pool was modelled assuming a constant litter fall biomass of 4.123 t ha⁻¹ yr⁻¹, which is the mean of three redwood forest sites [55] capped to a maximum of 1/3 needle biomass. Other additions to the litter pool consisted of foliage and branches from dying trees, all above-ground biomass from thinning events, and logging slash at harvest. Inputs into the dead wood pool consisted of all below-ground biomass from dying trees and thinned or harvested trees, and stem and bark material from dying trees. The carbon losses from the decay of dead wood from mortality, thinned and felled trees were modelled using an exponential decay function, with a decay half-life of 15 years [22,56]. Carbon losses from decay in the litter pool were modelled using the exponential decay function used in the C_Change model for softwood litter [57].

To convert biomass into carbon, we used carbon fractions of 0.495 for foliage [58] and 0.520 for bark [59]. We used a carbon fraction of 0.530 for wood, which is the documented value for sapwood within secondary forest and plantation stands of redwood growing in the United States [60]. We did not have sufficient data to characterize the percentage of heartwood within the model. However, carbon fractions for heartwood and sapwood are documented as being very similar in Jones and O’Hara [60] (0.529 vs. 0.539), which reduces error from this source on the predictions of carbon. For dead wood, we used the IPCC default carbon fraction of 0.500. As the supporting PSP data included many stands older than 50 years (see Supplementary Material Figures S1–S4), carbon predictions for redwoods were made to 100 years.

2.4.3. Cypress spp. Carbon Predictions

Predictions of carbon sequestration (in above- and below-ground biomass, dead wood, and litter pools) for both cypress species provided by the multi-species carbon calculator were made using the C_Change model; see [57] for a detailed description. This model requires as input a yield table containing annual under-bark stem volumes, volumes losses from mortality (and thinning events if any), and basic density of the stem wood. The stem volume yield tables were generated using the 300 Index growth models and stand volume function developed for both cypress species described in detail in Kimberley and Watt [33]. Basic density of stem wood was assumed to be 404 kg m⁻³ at all ages, this being the mean whole-tree stemwood density of 107 New Zealand-grown cypress trees [61]. Stem volume and basic density are used by C_Change to estimate stemwood dry matter, which is then used to predict biomass in other components including roots, bark, branches and foliage using growth partitioning functions [57] adjusted using modifiers for cypress species derived from biomass studies, as documented in Beets, et al. [62].

Biomass was converted into carbon using conversion factors specific to each biomass component. Carbon flows to the dead wood and litter pools were estimated and decay functions used to predict losses of carbon from these pools. The carbon sequestration predictions covered the summed carbon in the above- and below-ground biomass pools, the dead wood pool, and the litter pool but did not include carbon in the soil. Carbon predictions for cypresses were made to 50 years, which aligned with the maximum measurement ages for most PSP data, which underpin the growth model [33]. For this underpinning dataset, the age ranges for C. macrocarpa and C. lusitanica stands were, respectively, 2–77 years and 2–61 years [33]. However, as the vast majority of these data were less than 50 years old for both species [33], robust models could only be developed to an age of 50 years.

2.4.4. Predictions of Total Carbon and Residual Carbon for Redwood and Cypresses

Residual carbon remaining following harvest is defined as the carbon stock per hectare in above-ground residual wood and below-ground roots in a cleared plantation. This was modelled annually for redwood and cypresses by predicting the biomass of logs removed during a simulated clear-fell harvest using the multi-species carbon calculator. This was achieved by applying a log cutting routine to modelled trees which were assumed to follow a Weibull diameter distribution, using the taper functions for each species implemented in the calculator. The simulated harvested logs were cut up to a break height of 65% tree height using a minimum log small end diameter of 100 mm, and assuming a volume loss at harvest of 4%. The biomass of harvested logs was subtracted from this total above-ground biomass at the time of harvest with the remaining above-ground biomass representing logging slash assumed to enter the litter pool. All below-ground biomass of harvested trees was assumed to enter the dead wood pool. The residual carbon was calculated as the sum of carbon from logging slash and roots of harvested trees, plus the carbon in the existing dead wood and litter pools.

Predictions of total carbon for post-1989 forests assumed no previous afforestation and therefore did not include any residual carbon from a previous crop. However, the predictions of total carbon for pre-1990 forests assumed that the forest had a previous crop of the same species which was clear-felled at age 45 for both redwood and cypresses. The pre-1990 estimates of total carbon included the carbon accumulation plus the residual decaying biomass from the previous rotation (stumps, roots, logging slash, etc.) and were estimated annually from age 9. For the pre-1990 carbon tables, the residual biomass from the previous crop was decayed linearly over a 10-year period to zero. This method aligns with the other pre-1990 deforestation tables in the ETS. However, to better reflect the actual residual decay in these forests, the residual biomass was also decayed using an exponential decay function with a half-life of 15 years. This model had the following form: C(t) = C₀ exp(−λt), where C(t) and C₀ are carbon at age t and age 0, respectively, and λ = 0.0462, reflecting the 15-year half-life of the residual carbon.

3. Results

3.1. Productivity Models for Redwood

The productivity surface models for redwood had an RMSE of 9.57 m³ ha⁻¹ yr⁻¹ and R² of 0.48 for the 300 Index and RMSE of 4.94 m and R² of 0.61 for site index. In total, 13 variables were used to describe both 300 Index and site index with variable importance for the top six variables shown in

Table 2. Variable importance for the six most important variables included in the redwood 300 Index and site index models. Variables are sorted in descending order of importance.

300 Index		Site Index
Variable	Importance	Variable	Importance
CN ratio	0.172	Tavg March	0.394
Tmax Dec.	0.136	Tmax Dec.	0.105
Wind Summer	0.104	CN ratio	0.0641
Tmin June	0.0997	Drainage June	0.0524
No. frosts June	0.0862	Water balance Feb.	0.0519
No. frosts Feb.	0.0693	Wind Summer	0.0471

. There was marked similarity between the key variables that influenced growth (Table 2) for both productivity indices. Productivity increased markedly with increases in air temperature and reductions in the number of ground frosts and was higher on sites with lower carbon–nitrogen ratio (CN ratio) which is associated with greater soil fertility. Residuals from both models showed little apparent bias against predicted values, the key variables within the model, or a number of important variables that were not included in the model. Year of establishment, which was tested for inclusion in the model to account for improvements in genetics or management practices over time, was not selected by the recursive feature elimination for either productivity index and model residuals showed little pattern when plotted against establishment year.

3.2. Productivity Models for Cypresses

The productivity surface models for C. lusitanica had an RMSE of 4.16 m³ ha⁻¹ yr⁻¹ and R² of 0.60 for 300 Index and RMSE of 2.69 m and R² of 0.72 for site index. In total, 20 and 21 variables, respectively, were used to describe 300 Index and site index. There was marked similarity between the key variables that influenced growth for both productivity indices, with the most important six variables shown in

Table 3. Variable importance for the six most important variables included in the 300 Index and site index models for the two cypresses. Variables are sorted in descending order of importance.

300 Index		Site Index
Variable	Importance	Variable	Importance
C. lusitanica		C. lusitanica
Tavg autumn	0.199	Growing degree days	0.219
Solar rad. January	0.110	Annual rain days	0.132
Solar rad. December	0.084	Tmin June	0.098
VPD spring	0.063	Solar rad. December	0.087
Tmin June	0.063	Annual solar rad.	0.055
Annual rain days	0.062	Solar rad. September	0.053
C. macrocarpa		C. macrocarpa
Solar rad. Autumn	0.122	Tmax January	0.259
Solar rad. June	0.111	Annual rainfall	0.131
Water balance June	0.103	Foliar Nitrogen	0.124
CN ratio	0.088	Establishment year	0.121
Aspect	0.072	Elevation	0.103
Profile curvature	0.064	Aspect	0.100

. Productivity generally increased to a maximum with both air temperature and solar radiation and was high on sites with low to moderate rain days but declined in areas with a very high number of rain days. The establishment year was not selected by the recursive feature elimination for either productivity index and model residuals showed little pattern when plotted against this variable.

The models for C. macrocarpa had an RMSE of 4.61 m³ ha⁻¹ yr⁻¹ and R² of 0.34 for 300 Index and RMSE of 3.01 m and R² of 0.50 for site index. In total, 15 and 8 variables, respectively, were used to describe 300 Index and site index and the 6 most important variables for each productivity index model are shown in Table 3. The 300 Index increased with solar radiation and water balance, declined with increases in CN ratio and was highest on north-facing slopes (Table 3). Site index increased with air temperature and rainfall to a maximum. Increases in site index were also noted with higher foliar nitrogen while north-to-east-facing slopes had the highest values of site index among all aspects. There was little change in site index with planting year until 1980. After this time, there were steady increases in site index from 1980 to the year 2018, which was the most recently measured plot.

3.3. Spatial Predictions of 300 Index and Site Index

Spatial predictions of 300 Index and site index made using the above models are shown for the three species in Figure 3 and Figure 4. As establishment year was an input variable in the C. macrocarpa model for site index, separate spatial predictions were made for construction of pre-1990 and post-1989 LUT’s for this model. This was performed by setting year of establishment to 1984 for pre-1990 forests and 2007 for the post-1989 forests, which represent, respectively, the 10th and 90th percentiles within the plot dataset for this species. Predictions for C. macrocarpa site index shown in Figure 4 use the post-1989 year of establishment, although differences with pre-1990 values were minor. The two productivity indices, for all three species, averaged to the Island (North Island, South Island) and national level, using area-weighted regional values, are shown in

Table 4. Variation in 300 Index and site index at the regional, Island and national scales for redwood and the cypresses. Values shown include the mean and the coefficient of variation (CV) (standard deviation/mean) with the CV expressed as a percentage. The Island and national values are area-weighted means determined from regional values.

Region	Redwood				Cypresses
	300 Index (m3 ha−1 yr−1)		Site Index (m)		300 Index (m3 ha−1 yr−1)		Site Index (m)
	Mean	CV	Mean	CV	Mean	CV	Mean	CV
Auckland	19.8	19.5	27.8	7.7	17.0	11.5	26.5	5.1
Waikato/Taupo	19.1	40.9	28.3	21.1	14.6	14.7	24.5	9.4
Bay of Plenty	19.4	31.8	29.7	17.8	14.6	15.4	25.6	10.3
Gisborne	18.7	31.7	27.0	18.2	17.1	19.6	26.7	12.8
Hawke’s Bay/Southern NI	20.0	35.4	27.9	17.4	13.8	15.7	23.9	10.2
Nelson/Marlborough	12.7	52.5	23.2	23.6	11.7	19.4	23.2	9.4
Canterbury/West Coast	11.8	57.2	20.6	25.4	11.3	16.8	21.9	8.7
Otago	13.7	56.6	19.0	13.6	12.3	14.7	22.2	9.2
Southland	9.0	51.8	19.3	10.2	13.1	14.0	23.4	5.7
North Island	19.7		28.1		14.9		24.9
South Island	11.9		20.3		12.0		22.4
New Zealand	16.6		25.0		13.6		23.8

.

Regional predictions for redwood show values of 300 Index and site index in the North Island that exceeded those of the South Island by, respectively, 66% (19.7 vs. 11.9 m³ ha⁻¹ yr⁻¹) and 38% (28.1 vs. 20.3 m). In general, 300 Index and site index values peaked around central North Island regions and the lowest values for both indices occurred in either Southland or Otago (Figure 3, Table 4). Within-region variation in the 300 Index was far greater within the South Island than the North Island with the coefficient of variation (CV) ranging from 52% in Southland to 57% in Canterbury/West Coast (Table 4). This higher variation, which is clearly evident in Figure 3, mainly reflected the greater within-region range in air temperature and rainfall within the South Island caused by the Southern Alps, with 300 Index values generally higher in coastal areas on both the east and west coasts. Within-region variation was not as marked for site index and greatest within the Canterbury/West Coast region.

For the cypresses, productivity indices in the North Island exceeded those in the South Island although to a lesser extent than redwood, by 24% for 300 Index (14.9 vs. 12.0 m³ ha⁻¹ yr⁻¹) and 11% for site index (24.9 vs. 22.4 m). Within the North Island, productivity indices were highest in the Auckland and Gisborne regions, while the lowest values in the South Island occurred in the Canterbury/West Coast, reflecting the strong influence of air temperature and rainfall on productivity (Figure 4, Table 4). Within-region variation in both productivity indices for cypresses was relatively uniform between regions and lower than that of redwood, reflecting the lower sensitivity of cypresses to site (Table 4). Values for the CV were highest in the Gisborne and Nelson/Marlborough regions for 300 Index and highest in the Gisborne region for site index (Table 4).

Predictions of carbon were based on the spatial predictions of 300 Index and site index. These spatial predictions of 300 Index and site index, which were averaged to the regional, Island and national level (Table 4), were used as inputs to the carbon models, which were used to predict carbon as described in the next section.

3.4. Differences Between Species and Islands in Post-1989 Carbon

A comparison between redwood and cypresses of carbon sequestration in post-1989 forests on the same scale, made over a 50-year period, is shown in Figure 5. Regional means from these data at ages 30 and 50 are shown in

Table 5. Variation in predicted total carbon at ages 30 and 50 for post-1989 forests, averaged to the regional, Island and national levels. Values for cypresses show carbon for Cupressus macrocarpa in the South Island and Cupressus lusitanica in the North Island, as these are the preferred regions for these species. For reference, carbon predictions from the MPI lookup tables (LUT’s) are displayed. The percentage differences in carbon predictions between Islands are shown. The percentage difference between predictions and values from the MPI LUT for exotic softwoods are shown and the percentage regional range between regions with the lowest and highest carbon is also displayed.

Region	Redwood		Cypresses
	Carbon (Tonnes CO2 ha−1)		Carbon (Tonnes CO2 ha−1)
	Age 30	Age 50	Age 30	Age 50
Island/national means
North Island	831	1920	755	1068
South Island	504	1198	638	1052
New Zealand	701	1633	702	1061
% Island differences
Nth Island/Sth Island	64.8	60.3	18.3	1.5
Regional means
Auckland	838	1936	838	1144
Waikato/Taupo	810	1868	744	1059
Bay of Plenty	829	1905	747	1060
Gisborne	792	1836	840	1146
Hawke’s Bay/SN Isl.	847	1956	712	1028
Nelson/Marlborough	546	1278	632	1047
Canterbury/West Coast	502	1191	610	1019
Otago	566	1360	650	1065
Southland	390	926	689	1111
Range regional diff (%)
North Island	6.9	6.6	18.1	11.4
South Island	45.3	46.9	12.9	9.1
MPI LUT	400	641	400	641
% Estimates/MPI LUT
North Island	108	200	89	67
South Island	26	87	60	64
New Zealand	75	155	76	65

. Carbon sequestration was predicted to be greater in the North Island for both redwood and cypresses, but to a far greater extent for redwood (Figure 5, Table 5). Predicted redwood carbon at ages 30 and 50 within the North Island exceeded that within the South Island by, respectively, 65% (831 vs. 504 tonnes CO₂ ha⁻¹) and 60% (1920 vs. 1198 tonnes CO₂ ha⁻¹). In comparison, predicted values of carbon within the North Island for cypresses, at ages 30 and 50, exceeded values in the South Island by, respectively, 18% (755 vs. 638 tonnes CO₂ ha⁻¹) and 1.5% (1068 vs. 1052 tonnes CO₂ ha⁻¹).

Within the North Island, mean carbon values for redwood exceeded values for cypresses by 10% (831 vs. 755 tonnes CO₂ ha⁻¹) and 80% (1920 vs. 1068 tonnes CO₂ ha⁻¹), respectively, at ages 30 and 50 years, with this divergence over time reflecting the differing growth trajectory between species (Figure 5). Species variation within the South Island was less marked, and redwood carbon at age 30 was 21% lower than that of cypresses (504 vs. 638 tonnes CO₂ ha⁻¹) but 14% higher at age 50 (1198 vs. 1052 tonnes CO₂ ha⁻¹).

Predictions of redwood could be made from 50 to 100 years, as the underlying plot data supported this. These showed carbon increment to continue in a linear manner beyond 50 years of age for both Islands and at the national level, with a slight reduction in growth rate beginning at ca. 80 years. Values of carbon at age 100 were 4652, 3048 and 4015 tonnes CO₂ ha⁻¹, respectively, for the North Island, South Island and New Zealand (Figure 6).

3.5. Comparisons of Post-1989 Carbon with Existing Lookup Tables

Differences between the existing Exotic Softwoods LUT values (green lines, Figure 5; Table 5) and our predicted carbon values were most pronounced for redwood within the North Island. Predictions of carbon at ages 30 and 50 for redwood in the North Island exceeded the existing LUT by 108% and 200%, respectively (Table 5). In contrast, predicted values of carbon within the South Island exceeded those of the LUT values at ages 30 and 50 by 26% and 87%, respectively. At age 30, predicted carbon values for all South Island regions, apart from Southland, were higher than the LUT, and by age 50, all regions exceeded the LUT values by quite a margin. Overall, at the national level, predicted values exceeded those of the LUT by 75% and 155% for ages 30 and 50.

Predicted carbon values for cypresses exceeded those from the existing LUT by the largest margin within the North Island, particularly at age 30 years (Figure 5; Table 5). For post-1989 plantings, predictions in the North Island exceeded those of the existing LUT by 89% and 67%, respectively, at ages 30 and 50. In contrast, differences were less marked between predictions and LUT for the South Island with predicted values exceeding those of the LUT by 60% and 64%, respectively, at ages 30 and 50 for post-1989 plantings. Overall, at the national level predicted values exceeded those of the LUT by 76% and 65% at ages 30 and 50 for post-1989 plantings.

3.6. Regional Variation in Predicted Post-1989 Carbon

Variation in predicted carbon, by region for redwood and cypresses, over both time periods, is shown in Figure 7. For redwood, the variation between regions was generally lower than the variation between Islands. At age 50, predicted redwood carbon in the North Island exceeded values in the South Island by 60% (1920 vs. 1198 tonnes CO₂ ha⁻¹). Between regions, carbon values for this age varied by 6.6% from 1836 tonnes CO₂ ha⁻¹ in Gisborne to 1956 tonnes CO₂ ha⁻¹ in the Hawkes Bay/Southern North Island (Figure 7; Table 5). Variation was more marked within the South Island, with the predicted carbon at age 50 varying by 47%, from the lowest values in Southland (926 tonnes CO₂ ha⁻¹) to the remaining three regions, which had reasonably similar predicted carbon (1191–1360 tonnes CO₂ ha⁻¹). In contrast to a previous study [63], carbon was relatively high within Otago, as the higher temperature threshold used for masking eliminated many of the low-productivity higher-elevation areas within this region.

In contrast, regional variation in predicted carbon for cypresses was relatively wide compared to the Island-level variation (Figure 7). At age 50, predicted carbon values for post-1989 plantings in the North Island exceeded those in the South Island by only 1.5% (1068 vs. 1052 tonnes CO₂ ha⁻¹). Within the North Island, predicted carbon within Gisborne of 1146 tonnes CO₂ ha⁻¹ exceeded values of 1028 tonnes CO₂ ha⁻¹ in Hawke’s Bay/Southern North Island by 11%. Similar variation was noted in the South Island, where predicted carbon at age 50 in Southland of 1111 tonnes CO₂ ha⁻¹ exceeded values of 1019 tonnes CO₂ ha⁻¹ in Canterbury/West Coast by 9% (Table 5).

3.7. Comparisons of Age of Average Carbon to Current MPI Averaging Age

The averaging age (see Figure 1) was determined using previously outlined methods [64]. Average carbon was determined for three harvest ages (40, 45 and 50 years), for redwood and cypresses, from second rotation values that included, for each year, the sum of accrued carbon from the current rotation and the decaying residual carbon from the previous rotation. It was assumed that the residual carbon was linearly decayed over a 10-year period. The averaging age was defined as the whole age immediately before the average carbon was reached for each species x harvest age combination.

Assuming a rotation length of 45 years for both cypresses and redwood, the current MPI averaging age of 22 years underestimated the predicted national mean for cypresses and redwoods by two and four years, respectively (

Table 6. National, Island and regional level variation in the age at which mean carbon is reached for redwood and cypresses in post-1989 forests under rotation lengths of 40, 45 and 50 years. Also shown is the difference between the values at the national and Island levels and the current LUT averaging age (22 years).

Region	Redwood			Cypresses
	Rotation Length (Years)			Rotation Length (Years)
	40	45	50	40	45	50
Island/national means
North Island	24	26	29	21	23	25
South Island	24	27	29	22	24	26
New Zealand	24	26	29	22	24	26
Regional means
Auckland	24	26	29	21	23	25
Waikato/Taupo	24	26	29	21	23	25
Bay of Plenty	24	26	29	21	23	25
Gisborne	24	26	29	21	23	24
Hawke’s Bay/SN Isl.	24	26	29	21	23	25
Nelson/Marlborough	24	26	29	22	24	26
Canterbury/West Coast	24	27	29	22	24	26
Otago	24	27	29	22	24	26
Southland	24	27	29	22	24	26
Current LUT age	22	22	22	22	22	22
Difference averaging age
North Island	2	4	7	-1	1	3
South Island	2	5	7	0	2	4
New Zealand	2	4	7	0	2	4

). The age at which the mean carbon was reached for redwood under a rotation length of 45 years was 26 years for all North Island regions and ranged from 26 to 27 years for South Island regions. Assuming rotation lengths of 40 and 50 years, the averaging ages for redwood were, respectively, 24 and 29 years across all nine New Zealand regions (Table 6). These averaging ages were higher than the current LUT, as redwood grows relatively slowly over the first 20 years, but more rapidly from 20 years onwards, but this was mitigated to some extent by the inclusion of decaying residual carbon during the first 10 years.

In contrast, the current averaging age of 22 years for exotic softwoods more closely approximated the predicted age at which average carbon was reached for cypresses for a rotation length of 45 years, which was 23 years within all North Island regions and 24 within all South Island regions (Table 6). There were reductions in this age under a rotation length of 40 years (national mean of 22 years) and increases under a rotation length of 50 years (national mean of 26 years) (Table 6). The averaging age more closely approximated that used in the current LUT, as cypresses grow relatively quickly over the first 20 years (Figure 5).

3.8. Regional Variation in Pre-1990 Carbon

Changes in carbon for pre-1990 forests, under the current rotation, were virtually identical to those for post-1989 forests. However, the pre-1990 forests also include the residuals from the previous rotation, which were degraded using a 10-year linear decay rate and an exponential decay function with a 15-year half-life. These predictions over years 9 to 50 are shown in Figure 8.

Using the linear decay function, the total carbon was relatively unaffected by the residuals, which were completely decayed by year 10. In contrast, the use of the exponential decay function with a 15-year half-life resulted in residuals which persisted further into the rotation with marked increases in total carbon for both species, particularly over the first 20 years (Figure 8).

3.9. National, Island and Regional Estimates of the LUT for Redwood and Cypresses

Full draft lookup tables based on the methodology described above are given in Supplementary Material S2. The tables, respectively, document post-1989 carbon, residual carbon and pre-1990 carbon (under two residual decomposition scenarios) for redwood (Tables S7–S10) and cypresses (Tables S11–S14).

4. Discussion

These carbon predictions suggest that a separate LUT for redwood is justified to reduce potential divergence between LUT and actual sequestration achieved. As redwood growth is very sensitive to climate, LUTs at either the Island or regional level would provide improved predictions. The model predictions also show that it is possible to develop a robust default Exotic Softwoods LUT based on cypresses. Regional estimates of carbon varied widely within the South Island for redwood and within both Islands (and in particular the North Island) for cypresses. Regional LUTs would provide a means of incentivizing establishment of both redwood and cypresses in regions that are most suited to these species.

The coverage of environmental conditions by plot data was relatively complete for the cypress species, if data from C. lusitanica are used to underpin predictions for the North Island while data from C. macrocarpa are used for the South Island. There was excellent coverage of environmental conditions for redwood within the North Island. Assuming redwood will not be planted at mean annual temperatures below 9 °C (which is consistent with the air temperature range for the plot data), the main climatic conditions not covered in the South Island were the mild and very wet conditions of the West Coast. From a modelling perspective, it is not necessary to have complete coverage of all regions if data from other regions encapsulate the climatic conditions of regions with sparse data. The West Coast climatic conditions were represented to some extent in the North Island. As a result, predictions of both productivity indices seem reasonable for the South Island West Coast, showing a moderate level of productivity. It is also worth noting that there are very few areas on the West Coast of the South Island that would be suitable for the establishment of redwood once unsuitable areas are masked out (Figure 3), which may limit any potential errors resulting from incorrect LUT estimation within this region.

Reported redwood carbon values for the North Island were very high. Predictions from the modelled surfaces at ages 30, 50 and 100 years were, respectively, 831, 1920, and 4652 tonnes CO₂ ha⁻¹. As there were a large number of redwood plots within the North Island (435 plots), estimates of carbon can also be made based on mean values of 300 Index and site index derived directly from the plot data. This provides a means of cross-checking the predictions based on surfaces of 300 Index and site index developed using models. Carbon predictions made using the mean productivity indices from plot data (300 Index = 21.07 m³ ha⁻¹ yr⁻¹; site index = 30.2 m) yielded, at ages 30, 50 and 100 years, carbon values of, respectively, 895, 2054 and 4926 tonnes CO₂ ha⁻¹, exceeding the estimates based on modelled surfaces, but only by 6%–8%. The consistency between carbon estimates derived directly from the plot data and those estimated using modelled productivity surfaces provides strong support for the North Island estimates of carbon given here.

Predictions of redwood carbon for the South Island based on modelled values of productivity indices were lower than those of the North Island. However, with the exception of younger Southland stands, these values generally exceeded current values of the Exotic Softwoods LUT. Although there were less plot data within the South Island, the 134 plots available had sufficient environmental coverage to robustly estimate productivity indices. The derived maps were consistent with expected trends in productivity, based on an understanding of redwood site preferences [23]. The use of a mean annual temperature threshold of 9 °C provided a means of limiting predictions to areas where the species is likely to be planted, and this threshold conformed to the location of all plot data.

Spatial predictions of productivity for cypresses agreed with previous anecdotal observations. Within New Zealand, C. macrocarpa is hardy and can survive 10 °C of ground frost, very low rainfall (500 mm year⁻¹), waterlogged soils, and strong or salt-laden winds [65]. The species thrives in the cooler parts of both Islands [66]. In contrast, C. lusitanica is relatively demanding in terms of shelter, fertility and soil moisture, and generally requires at least 1000 mm of rainfall per annum with sustained soil moisture supply over summer [66,67]. Cupressus lusitanica is sensitive to cold sites, and trees within dry, cold, or exposed situations tend to be small, squat, and generally inferior [66]. This species is best suited to mild areas of New Zealand away from the immediate coast and does well in northern, northeastern and western parts of the North Island [66].

Predictions of cypress carbon at age 50 made from the productivity surfaces were, respectively, 1068 and 1052 tonnes CO₂ ha⁻¹ for the North and South Islands. As there were a large number of C. lusitanica plots in the North Island (324 plots) and C. macrocarpa plots in the South Island (313 plots), estimates of carbon can be made directly based on mean plot values of 300 Index and site index, which provides a means of cross-checking predictions made using the productivity surfaces. The estimates of carbon using these plot data were, respectively, 1108 and 1032 tonnes CO₂ ha⁻¹ for the North and South Islands, which were within 2%–4% of the estimates made using the productivity surfaces. This comparison provides strong support for modelled predictions of carbon shown in the results.

The convergence in predictions of cypress carbon for the South Island with those in the North Island over time was attributable to the low mortality rate and higher growth rate of older South Island C. macrocarpa stands compared to North Island C. lusitanica stands; for more details, see [33]. This convergence at older ages was expected, as the cooler air temperatures of the South Island result in lower levels of cypress canker (caused by Seiridium cupressi (Cooke & Ellis) Sutton and S. cardinale (Wagener) Sutton & Gibson) within the more susceptible C. macrocarpa, particularly at older ages. Cypress canker affects branch tips, destroys stem cambium, and can result in fluting and occasional tree death. Mortality functions fitted to the cypress PSP dataset showed that annual mortality rates for C. macrocarpa growing in the South Island were 64% and 44%, respectively, of the annual mortality rates of C. lusitanica growing in the North Island for stands < 10 years and ≥10 years of age [33]. This increasing difference in mortality rate with stand age results in species divergence in predicted stocking and carbon.

The regional boundaries used in the default carbon tables are administrative and not based on geo-climatic factors. While many regions have a limited range of within-region climatic conditions, some, such as Canterbury/West Coast, include a wide climatic range. Therefore, although the use of regional carbon tables will to some extent mitigate variation at the national level, growers of stands in harsh conditions will be rewarded, while those on good sites will be penalized. Analyses show that within-region variation in productivity indices was most marked for redwood. For redwood, the coefficient of variation was highest for 300 Index within the four South Island regions, reflecting the strong gradients in productivity within these regions resulting from variation in proximity to the Southern Alps. In contrast, the coefficient of variation was reasonably consistent for cypresses throughout New Zealand, reflecting the ability of C. macrocarpa to maintain reasonable growth under a range of environmental conditions, which is consistent with observations around site preferences for the species [39].

Given the relatively large area over which the two main cypress species studied here have been established (9057 ha) [5], these represent a reasonable default Exotic Softwoods LUT. There are a number of other exotic softwoods that are grown in New Zealand, such as Japanese cedar (Cryptomeria japonica), a number of pines (e.g., Pinus nigra, Pinus ponderosa), larch (e.g., Japanese larch, European larch), grand fir (Abies grandis), sitka spruce (Picea sitchensis), and Chinese fir (Cunninghamia lanceoloata), amongst others [65]. However, in total, these species only cover 9300 ha, with individual species established at a relatively small scale.

When correctly sited, cypresses have moderate productivity and are reasonably productive across the wide range of environments suitable for plantation forestry within New Zealand [39]. The mean annual volume increment at age 30 for a stand with a final stocking of 300 stems ha⁻¹ (i.e., 300 Index) averaged 12.0 m³ ha⁻¹ yr⁻¹ for C. macrocarpa in the South Island and 14.9 m³ ha⁻¹ yr⁻¹ for C. lusitanica in the North Island (Table 4). Although growth data are sparse for species other than cypresses and redwood within the Exotic Softwood category, it is likely that the moderate growth rate of cypresses will be reasonably representative of the other exotic softwood species. Alternative exotic softwood species that would be included in this category generally have similar growth rates. Pinus nigra can produce moderate-high yields of 13–24 m³ ha⁻¹ yr⁻¹ at high stockings [65]. Japanese larch (Larix kaempferi) is the fastest-growing of all larch species [65] and a moderate mean annual increment of 14.3 m³ ha⁻¹ yr⁻¹ at age 32 has been recorded for a highly stocked stand (500 stems ha⁻¹) in the central North Island [68]. Abies grandis can tolerate cold conditions but is slow-growing [65]. Growth rates of Cryptomeria japonica range widely throughout New Zealand, but generally the growth rate is moderate, with most stands reaching a diameter of 40 cm by 40 years [69].

5. Conclusions

Predictions of carbon for redwood and cypresses were made through a series of models, underpinned by regional predictions of the two productivity indices, 300 Index and site index. Modelled predictions of redwood carbon show estimates varied widely by Island. At ages 30 and 50, predicted redwood carbon stocks in the North Island exceeded those in the South Island by 65% (831 vs. 504 tonnes CO₂ ha⁻¹) and 60% (1920 vs. 1198 tonnes CO₂ ha⁻¹), respectively. Furthermore, redwood carbon predictions at age 50 surpassed the existing LUT values by 200% in the North Island and 87% in the South Island. In contrast, carbon predictions for cypresses showed less variation between Islands than redwood. For post-1989 forests, carbon stocks in the North Island at ages 30 and 50 exceeded those in the South Island by 18% (755 vs. 638 tonnes CO₂ ha⁻¹) and only 1.5% (1068 vs. 1052 tonnes CO₂ ha⁻¹), respectively, with the convergence at older ages primarily attributed to lower mortality of C. macrocarpa within the South Island.

At age 50, mean redwood carbon in the North Island was 80% higher than that of cypress (1920 vs. 1068 tonnes CO₂ ha⁻¹), while in the South Island, redwood carbon was 14% higher than cypress carbon at the same age (1198 vs. 1052 tonnes CO₂ ha⁻¹). Regional estimates of carbon varied widely within the South Island for redwood and within both Islands for cypresses. These draft carbon predictions highlight the need to develop separate lookup tables for redwood and other minor exotic softwoods to reduce potential discrepancies between predicted and actual sequestration.

The development of regional LUTs would provide a means of incentivizing the establishment of both redwood and cypresses in regions that are most suited to these species, which in turn is likely to improve total carbon sequestration within New Zealand. The implementation of regional LUTs that are more closely aligned to actual growth rates of redwoods and cypresses will provide higher returns to growers of these species. However, the government may consider factors other than the technical considerations outlined here when updating the LUTs.