Monitoring networks available to the UK forest carbon inventory

 

A number of forest monitoring networks and inventory systems already exist in the UK which could be incorporated into a UK carbon monitoring methodology:

• Long-term carbon flux network (CFN)
• Intensive forest health monitoring EU/ICP-Forests Level II network (L-II)
• Permanent mensuration sample plot network (PSP)
• UK Woodland Assurance Standard network (UKWAS)
• Forest condition monitoring EU/ICP-Forests Level I network (L-I)
• UK forest condition survey network (FCS)
• National inventory of woodlands and trees (NIWT)
• Forest Enterprise subcompartment database (SCDB)
• Survey of small woodlands and trees (SSWT)
• Land cover map 2000 (LCM2000).

Some of these networks and inventories already provide appropriate data at a suitable time resolution but others would require modification or additional measurements to be made. Details of each of these systems are given in the following review, including an assessment of each in terms of:

• Representation of forest stand types within the UK forest estate
• Spatial extensiveness
• Suitability of assessments for national carbon stock and stock-change inventories.

Carbon flux network (CFN)

Four long-term carbon flux monitoring sites have been established in woodland in the UK as follows:

• Straits Enclosure (Forest Research, Alice Holt):
  90 ha stand of GYC 6 oak planted during the 1930s on a surface water gley soil in Hampshire. The flux station was established in 1998, and has operated continually. The entire stand is due to be thinned in 2004-5.
   
• Harwood (University of Edinburgh):
  Two flux systems were established in 2000 in a 4000 ha Sitka spruce forest in Northumberland of varying planting year and yield class on a peaty gley soil. One continuous system is sited in a 30-year old stand, with a second roving system used to assess carbon fluxes in a chronosequence investigating the effects of stand age on carbon balance.
   
• Griffin (University of Edinburgh):
  Sitka spruce plantation (GYC 14) planted in 1981 on a peaty gley soil in Perthshire. The flux station was established in 1996 and has operated continuously apart from a break in 2002. The forest is currently being thinned.
   
• Pang catchment (CEH Wallingford):
  Mixed broadleaf woodland dominated by oak situated in Berkshire. The flux station is currently being established as part of the NERC funded LOCAR project and has guaranteed funding for five years.

These four sites only represent two species – oak and Sitka spruce. As a result of their limited species and geographical coverage, their direct contribution to carbon inventory assessments is limited, although estimates of forest carbon exchange at a European scale have been made on the basis of data from the thirty flux sites associated with the EU FP5 CarboEuroflux research project (Martin et al., 1998). Their most valuable contribution is in providing data for the parameterisation and validation of stand level process models of carbon and water exchange. Associated detailed ecophysiological measurements are undertaken at most sites including soil respiration, litter analysis, meteorology, leaf area, increment, individual tree allometry, rainfall interception and soil water chemistry analysis.

Net ecosystem exchange is calculated by correlating the vertical component of windspeed with CO2 and water vapour concentration fluctuations above the canopy. Fluxes are representative of areas of woodland of up to 100 ha, although the ‘footprint’ is highly dependent on windspeed. In isolation, flux measurements provide no information on the partitioning of carbon between wood, litter, soil and ground vegetation and do not account for carbon lost as dissolved or particulate organic carbon (DOC and POC) or removed as HWP. However, the programmes of intensive measurements outlined above enable stand level carbon budgets to be developed, which may be useful for validation of estimates from other assessment procedures or models.

Intensive Forest Health Monitoring – EU/ICP-Forests Level II (L-II)

The network for intensive monitoring on permanent observation plots was established in 1994. Currently there are approximately 900 plots across Europe, although the intensity of measurements varies from plot to plot. In the UK, ten sites were established in 1995 (oak, Sitka spruce and Scots pine), with a further ten added in 2002 (beech and Norway spruce) to represent more species and regions within the UK.

Stand height, diameter and volume increment, foliar chemistry, soil chemistry, crown condition and ground vegetation are assessed at all sites, while litter, air quality, meteorology and soil water chemistry are assessed at a proportion of the sites. Not all assessments are carried out in all plots. Deadwood and litter carbon stocks are not currently assessed, although new protocols are currently under review with the objective of developing the network to enable complete carbon stock and carbon stock change assessments to be made. In addition, the soil condition survey carried out under this network provides a framework for assessing soil carbon stocks across a wider network. Chemical analysis is undertaken for each soil horizon from a single representative pit, chosen from 30 profiles arranged on a systematic grid across the area of the plot. The results of chemical analysis for the representative pit can then be scaled by the mean depth of each horizon as revealed by the 30 point grid. Kirwan et al. (2003) have shown that between 25 and 36 pits are sufficient to capture the variability within a 0.3 ha plot typical of the L-II network.

Permanent Mensuration Sample Plot Network (PSP)

The first permanent mensuration sample plots were established before 1920, with the aim of providing data on which to base yield models for forest management and production forecasting such as those developed by Hummel, Christie and co-workers (Hummel and Christie, 1953; Bradley et al., 1966; Hamilton and Christie, 1971; Edwards and Christie, 1981). On the whole PSPs have been established in single species, even-aged stands, and thus are not fully representative of all current forestry practice in the UK.

There are currently 509 active plots, representing primarily the major commercial species planted in the UK. Although the spatial distribution of the appears even and complete at first sight, this hides the clustering of a large number of plots at a smaller number of localities and site types. The standard PSP measurement protocols (Hummel et al., 1959; Edwards, 1976) are purpose-designed to permit reliable stand-level estimates of tree stem volume and size-class distributions to be derived for standing trees and for any thinnings.  Stem biomass is not assessed but could be obtained by applying standard values for the nominal specific gravity of wood (Lavers and Moore, 1983) to volume assessments. However, protocols for assessment of tree total and/or component biomass and soil carbon content could be developed and integrated with standard PSP procedures.  Inclusion of such procedures would add significantly to the cost of periodic assessments.  An alternative approach for biomass assessment might be to apply average allometric relationships to standard PSP data to obtain tree-level and stand-level estimates for roots, branches and foliage. Perhaps more importantly, PSP data has an explicit role in underpinning development of measurement systems and growth and yield models.  This gives the PSP network a pivotal role as a link between stand-level inventory and survey assessments on the one hand and input data to models used for upscaling and forecasting on the other.

Recently the PSP network has been the subject of a significant strategic review (Matthews et al., 2003).  The outcome is likely to be an assertive change of emphasis for the network, to primarily providing data for growth and yield model development, with any monitoring function being subsidiary and minor.  Such a development would underline a role for PSPs in monitoring forest carbon stocks as one of support to the development of models used for upscaling and forecasting carbon stocks and stock changes, as well as validation of upscaled estimates and forecasts.

UK Woodland Assurance Standard (UKWAS)

UKWAS plots have been or are in the process of being established using protocols similar to those for PSPs and in effect these plots constitute a component of the PSP network.  However they are not formally integrated since their function is distinct and dissimilar to PSPs. The objective of UKWAS plots is to provide data as part of monitoring of compliance with the principles and indicators specified in the UK standard for forest certification as administered by the Forestry Stewardship Council (UKWAS, 2000).  For this purpose, plots are established on sites representative of stand conditions in a given locality and are managed according to standard local practice.  In principle this makes them ideal for use in an inventory-based carbon stock monitoring system.  In practice the cost of establishing enough plots to form a fully representative network capable of producing upscaled estimates would be prohibitive.  Currently there are only 12 UKWAS plots (representing five species) in existence and these are limited to Scotland.  At present there is no plan to extend the network to England and Wales.

Forest Condition – EU/ICP-Forests Level I (L-I)

The L-I Forest Condition Survey is a Europe-wide assessment of spatial and temporal variation in forest condition using crown density or transparency as the principal indicator. Surveys of crown density have been carried out on an annual basis since 1986 at between 1700 plots (in 1988) and 6000 plots (in 2000) across Europe. Although strictly the L-I protocol does not require increment assessment, top height was recorded on establishment and dbh is routinely measured, alongside other measurements made as part of the UK national Forest Condition Survey (see section below).  In principle, rough estimates of carbon stocks in tree biomass could be derived from such abbreviated assessments, using methodologies similar to abbreviated tariffing (Edwards, 1983), although they would be of relatively poor precision (around ±20%) and biased for individual sites.  In addition, soil condition (carried out in 1994) and the nutritional status of the trees in terms of foliar chemistry (carried out in 1995) have been assessed on one occasion only. The soil condition survey involved descriptive and chemical analyses, but only for the top 20 cm (0-10 cm and 10-20 cm) which is insufficient for estimation of carbon stock and stock change.  Despite these limitations, it has been possible to derive an estimate of the carbon sequestration rate of European forests based on this network, using stand age and soil type to derive stem wood increment from forest yield tables and an assumed ratio of total biomass to stem wood of 2.5 (De Vries et al., 2003). Soil carbon sequestration was estimated from soil C:N ratios and calculated retention times.

In the UK, there are approximately 90 L-I plots covering five tree species (oak, beech, Scots pine, Sitka spruce, Norway spruce). The protocol requires a minimum plot size of 0.25 ha (although this is not always realised), with the crown density of 24 ‘internal’ plot trees assessed across the four aspects (N, S, E, W). The plots have been established on a 16 × 16 km transnational grid across Europe, enabling pan-European assessments of forest condition to be made. A recent review indicates that potentially, the network could be enlarged from 90 to approximately 400 sites if plots were established within 500 m of the transnational grid intersections (Hendry, 2001). Data from other national networks are also submitted to the EU/ICP-Forests, generally based on a national grid varying in resolution from 1 x 1 km to 32 x 32 km. Special dispensation was given to the UK to continue using the FCS network (established in 1984 and based on a stratified random sample) for this purpose, as a result of the dispersed nature and heterogeneity of the UK forest estate.

UK Forest Condition Survey (FCS)

The FCS was instigated in 1984 prior to the establishment of the L-I network and provides greater spatial representation and more detailed measurements than strictly required for the EU L-I network.  As such, the L-I network forms a subset of the FCS network. The same five tree species are assessed (59 beech, 55 Norway spruce, 86 oak, 81 Scots pine, 66 Sitka spruce and 3 mixed broadleaf: Hendry et al., 2002), based on a stratified random sample (Binns et al., 1985).

In contrast to the plot-based layout and assessment protocol adopted in the L-I network, assessments are made on individual trees at the periphery of the forest stand (six trees for each of the four aspects), and thus there is no plot as such on which to express assessments at the stand level.  On the other hand, more variables are measured in FCS than are strictly required under the EU L-I protocol as outlined above. The supplementary measurements include annual assessments of dbh and a single assessment of top height recorded on establishment, in principle enabling annual carbon stocks in biomass to be calculated at least at the individual tree level.

National inventory of woodland and trees (NIWT)

This inventory, otherwise known as the Census of Woodland, is carried out every 10-15 years. The most recent survey (eg Forestry Commission, 2003) is based on a combination of analysis of the 1:25000 OS map and, primarily, interpretation of aerial photography. The NIWT considered woods of more than 2 ha area, although a separate survey of small woods and linear features was also conducted as part of the assessment, but in less detail (see separate section below). Ground-truthing of 1% of the national estate (in terms of forest area) was carried out by forest surveyors in a total of 40 897 sample squares. In the main, this sample was selected on a random basis but subject to stratification such that 1 in 5 woodlands of less than 100 ha, 2 in 5 woodlands of between 100 and 500 ha and all woodlands over 500 ha were included in the sample. Adjustments to the sample grid were permitted where a woodland edge was found or serious problems were encountered in gaining access to the sample site.  Summary information such as ownership and species mix was recorded for each 1 ha plot, and a more detailed structural survey was carried out in a 0.25 ha sub-plot  Further details may be found in Gilbert (1999).

It is evident that the emphasis in NIWT is on qualitative rather than quantitative data.  Nevertheless, the mensurational and structural assessments made in the 1% verification plots could be used in a similar manner to L-I and/or FCS plots to provide a broad scale validation of forecasts of forest carbon stocks derived from models.  The additional cost of collecting data on soils in NIWT verification plots for use in verifying carbon stocks is likely to be prohibitive.  An alternative application for NIWT data as input to models of carbon stocks and stock changes is discussed following a consideration of other relevant inventory systems below.

Survey of Small Woodland and Trees (SSWT)

The SSWT was carried out as part of NIWT but separately to the main survey, to assess the extent of woodland of less than 2 ha in area together with linear features (hedgerow trees and shelter belts), groups and individual trees. The country was stratified into coastal and non-coastal areas and a random sample of 1 km2 squares representing 1% of the land area selected. For each selected grid square, 1:25 000 aerial photographs were used to identify features, with field data collected for two out of the 16 (250 x 250 m) squares in each 1 km2 grid square.  The only data recorded consisted of a broad classification of the woodland or individual trees in terms of type of woodland/tree feature and key species groups present.  Application of SSWT data as input to models of carbon stocks and stock changes is discussed following consideration of other relevant inventory systems below.

Forest Enterprise subcompartment database (SCDB)

The subcompartment database provides stand-level information on the estate managed by Forest Enterprise across Great Britain, including area, species, year of planting, estimated productivity (GYC) and management prescription for the purposes of production forecasting and operational planning. The Forest Enterprise estate covers an area of 1053 thousand hectares of which 759 thousand hectares is productive high forest. This latter area constitutes approximately 30% of the estimated total area of woodland in Great Britain. 

The abbreviated nature of SCDB data might appear to be limiting; however the explicit link to models used in production forecasting for the Forest Enterprise estate may provide a framework onto which models for estimating carbon stocks and stock changes could be mounted.  The partial coverage of the UK forest estate would still need to be addressed either by extending the SCDB system, development of a parallel system to handle non-Forest Enterprise woodlands or through some form of extrapolation of forecasts for Forest Enterprise woodlands to the rest of the estate.  These points are discussed further following consideration of other relevant inventory systems below.

Landcover map 2000 (LCM2000)

LCM2000 is a map of land cover in the UK derived from the Great Britain and Northern Ireland Countryside surveys (Haines-Young et al., 2000).  These were based on analysis of satellite imagery and detailed ground-truthing of 500 (1 km2) sample squares across all land-use classes. The pixel size of the satellite images was 0.0125 ha, in theory giving considerably greater spatial detail than NIWT. LCM2000 and NIWT are not completely compatible, and in a recent analysis (Hall et al., 2003), only 1.8 Mha of the 2.75 and 2.34 Mha total woodland areas (from LCM2000 and NIWT, respectively) was consistently represented. Differences in interpretation between the two datasets explain some of this discrepancy. For example, satellite imagery can map heather as woodland, while young trees or felled areas due for restocking will not be registered in satellite imagery, but will be included in NIWT on the basis of the SCDB (FC land) and grant aid (non-FC land) records. Differences in the spatial resolution of NIWT and LCM2000 also contribute to the discrepancies. For example, woodlands of less than 2 ha (a total of 120 000 ha) are not included in NIWT but are mapped in LCM2000.

LCM2000 is essentially a map of land cover and does not include inventory assessments for areas identified as consisting of forest. As a consequence, extending LCM2000 to include surveys providing input data to models for estimation of forest carbon stocks and stock changes would be a major exercise. Its inclusion is thus solely for the purpose of land-use classification and ensuring compatibility between woodland and non-woodland carbon inventories.