Module 2 Development of a model to simulate C and N cycling in organic soils to provide predictions of their response to land-use, management and climate change
2.1 Modelling soil C and N cycling in organic soils
2.1.1 Where current soil carbon models do not work
The model proposed for possible use in the UK national greenhouse gas inventory (RothC- UK) was run to show soils for which it is unsuitable. These are the areas in which the new organic soil model is needed. Figure 2.1 shows the soils at 0-30cm depth (in red) for which RothC does not provide credible SOC values. Failure is defined by unrealistic high C inputs being required (i.e. >5 t C ha -1 y -1) to maintain the measured SOC level. Such high inputs are unfeasible given that net primary production in these areas could not support this input. The large majority of the soils in Scotland and Wales fall into this category.
Figure 2.1 Areas (in red) where RothC- UK does not simulate SOC satisfactorily. Areas in blue were not run (peats and other highly organic soils where RothC- UK is known to be unsatisfactory). For soils under arable (a), grassland (b), and semi-natural (c) land use. Woodland is similar to semi-natural (not shown).
2.1.2 Development of the ECOSSE model - summary
Data for preliminary evaluation was collated and used to evaluate the exiting models, RothC, SUNDIAL, CENTURY and DNDC. A new model ( ECOSSE) was formulated using components of the single point models RothC, SUNDIAL, CENTURY and DNDC, and constructed into a framework that allows spatial datasets to be simulated. Data were collated on greenhouse gas emissions, leaching losses and nitrogen and carbon stocks in the soil to be used for evaluation of the new model. This data has been used to evaluate the performance of ECOSSE. Simulations are generally within experimental error for both mineral and organic soils, so the model can be used to provide estimates of greenhouse gas emissions from Scottish and Welsh soils.
A database of existing experimental data was constructed, recording all information needed to run the models. These include location, current and previous land use, vegetation type, weather data, soil carbon content, soil pH, lowest summer water table, available soil water, field capacity, bulk density, amount of litter, depth of anaerobic zone, changes in soil C and N content, gaseous C and N fluxes and measurements of labelled C and N. Data on methane, carbon dioxide, nitrous oxide emissions and hydrology are itemised in a database, specifying the type of data, and the time and spatial scales of the data.
Clicking with the mouse on the site code in the metadata sheet locates the actual data in the database. A range of macros has been constructed to download the data from the database in the format required by the model to run. This facility will aid repeated evaluation of the model.
The model comparison has highlighted the most accurate and feasible approaches for simulating carbon and nitrogen turnover in highly organic soils. These are being incorporated into the model development, and are discussed further in the next section.
The new model was constructed from the SUNDIAL- MAGEC model of C and N turnover in soil/plant system (Smith et al., 2000). Characteristics of the ROTH-C, CENTURY and DNDC models have been incorporated into this basic framework, as well as relationships derived directly from experimental data. The new model allows simulation of emissions of all major greenhouse gases from the soil/plant system as well as simulating C and N losses through water movement. The SUNDIAL- MAGEC code was first restructured into a modular format that allows components of the model to be individually evaluated, reformulated and replaced. This new version of the model will be referred to from here on as ECOSSE, Estimator of Carbon in Organic Soils - Sequestration and Emissions. The completed developments of ECOSSE are listed below:
1) Modularisation of SUNDIAL/ MAGEC;
2) Improved description of N 2O production;
3) Description of methane production and oxidation;
4) Description of dissolved organic matter (carbon and nitrogen): turnover and losses;
5) Improved layer structure in the soil profile;
6) Initialisation of the size and characteristics of the soil organic matter pools;
7) Incorporation of the effect of pH on soil processes;
8) Incorporation of the effect of saturated conditions on soil processes;
The detailed description of the model, including the form of equations used to describe the processes included, is given in Annex 1.
2.2 Collection of new data
2.2.1 John Miles Birch Plots
Introduction: The most significant land use change ongoing in Scotland is afforestation (over the last five years, afforestation has averaged 6000ha y -1) and the majority of new planting and natural regeneration (ca. two-thirds) during the 1990s is on organic soils, with most of this on organo-mineral soils (Chapman et al., 2001). New planting is now at a 35-year low of just 8000 ha for Britain including 4000 ha in Scotland (2005/2006) and afforestation in Scotland is now focussed on soil of low organic matter. While the above-ground biomass of woodland may be expected to eventually exceed that of the prior vegetation, often heather-dominated moorland, the effects of afforestation on the below-ground carbon stock are uncertain. Across Scotland two native species dominate natural regeneration: Scots Pine and Birch. These two species are considered to have contrasting effects on soil chemistry. For example, while Scots Pine tends to generate a more acidic environment than even heather moorland (Chapman et al., 2003), birch has been reported to raise the soil pH and increase nutrient cycling (Miles & Young, 1980). SNH have conducted a project to stimulate natural regeneration of native birch in North East highlands by reducing/excluding deer grazing. To further investigate the effects of birch on heather moorland, John Miles (formerly of CEH) set up a fully replicated experiment of birch-heather plots at three upland sites in Scotland. An examination of the chemistry of the organic layer (O horizon) revealed a significant decrease in both mass and % carbon of the O horizon in the birch when compared with the heather (Mitchell et al., 2006). The objective here was to repeat this sampling but to perform a complete carbon (C) and nitrogen (N) budget of the two treatments including C and N in the vegetation and underlying mineral horizon as well as in the litter and organic layers.
Methods: Twelve plots had been established by John Miles at each of two sites (Delnalyne and Craggan) in a paired plot design. In each pair, one plot was left as a heather control plot and the other plot planted with Betula pubescens. At the third site (Kerrow) there were 18 plots established in groups of 3, one heather control, one planted with Betula pendula and one planted with Betula pubescens. The plots were 16 Õ 16 m except at Delnalyne where they were 12 Õ 12 m. The birch was planted in 1980 at 0.5 m spacing and the experimental area at each site was fenced to exclude grazing by large herbivores. It is important to recognise that these plots have 20- 40 times the usual number of planted seedlings per unit area. Natural regeneration could attain this stocking density but would have a much smaller effect on soil disturbance than planting. Full site descriptions can be found elsewhere (Mitchell et al., 2006).
The sites were sampled in July/August 2005. Two random quadrats, 50 cm Õ 50 cm, were selected within each plot. For the control heather plots, the total above-ground vegetation was removed by clipping; this was facilitated by installing four vertical bamboo sticks at the corners of the quadrat. For the birch plots, the ground vegetation under the trees was either sparse or absent and was not included; the birch biomass was estimated separately (see below). At all sites, for some of the birch plots, birch growth was either very patchy or very poor. In the former case, complete random sampling was not possible but restricted to those areas of reasonable establishment (i.e. heather had been ousted). In the latter case (two plots at Kerrow and one plot at Craggan), no samples were taken and the plot was treated as a missing replicate. The underlying reason for this variable birch growth has not been established (Mitchell et al., 2006), but could be related to the mycorrhizal associations. The clumpy distribution suggests that it is associated with isolates introduced with the planting stock rather than the site's own inoculum either at the time of planting or blown in subsequently. This could be an explanation for the site differences described below, and emphasises the need for caution in extrapolating from the results. Three soil cores, in a triangle 15-20 cm apart, were taken from the centre of each quadrat using a 30 cm long, 5 cm Õ 5 cm square box corer. This sampled the litter ( LF) and organic (H or O) horizons and often into the upper mineral horizon (by a few cm). A further mineral horizon sample was taken, if possible, from below the box corer sample using a metal cylinder corer (6 cm diameter Õ 6 cm depth). Often this was not feasible where large stones or rock were encountered. In a few cases (at Craggan), the litter and organic layer extended below 30 cm; in this case the depth to the underlying mineral horizon or rock was determined with a depthing rod. Care was taken during coring both organic and mineral horizons to facilitate intact cores that could be used for bulk density estimations.
Since the birch was part of an ongoing experiment, it was not possible to harvest trees from the plots in order to measure the tree biomass. However, tree density, tree basal diameter and tree height data were available (R. J. Mitchell, pers. comm.; Mitchell et al., 2006). These had been assessed at the quarter plot level (i.e. the plot divided into four) in order to give some indication of the within plot variability. The sampling quadrats generally fell into different quarter plots but in a few cases were within the same quarter plot. Nine trees from each of the Delnalyne and Craggan sites were harvested at ground level from outside the experimental plots as being representative of the size and shape of those within the plots. Stem and main branches were analysed separately from twigs and leaves for total dry weight and C and N content. Regression equations were developed to enable estimation of the total tree C and N on the plots:
Total C (g) = 5230 - 5212 * (0.9999769^(D^2*H)) (r 2 = 0.812)
Total N (g) = 71.3 - 70.6 * (0.9999764^(D^2*H)) (r 2 = 0.879)
Where D is the basal diameter (mm) and H is the height (m).
DBH (diameter at breast height) was not appropriate as many trees were less than this in total. The best fitting relationship was curvi-linear; this is interpreted as being due to the smaller trees having a slightly different form from the larger ones.
In the laboratory, soil cores were divided into LF (litter), organic and mineral horizons. The depth of each horizon was noted and cores weighed for bulk density determinations. These were then air dried for one week and then sub-samples dried overnight at 105°C, ball-milled and analysed for total C and N using a Carlo Erba NA1500 elemental analyser. Roots encountered in the soil cores were included in the sample for analysis. Mineral samples were first passed through a 2 mm sieve to remove stones. Vegetation samples (heather and birch) were analysed similarly except that they were dried for 24h at 80°C and hammer-milled before being ball-milled. Care was taken to ensure that sub-samples were representative of the whole plant, particularly for the birch where stem/branches and twigs/leaves were analysed separately.
Carbon and nitrogen stocks were calculatedly for the vegetation, LF (litter), organic horizon and mineral horizon (this being the sum of any mineral horizon sampled with the box corer and with that in the cylinder core), as well as the total. This was done at the core level for the soil horizons but vegetation samples were only available at the quadrat level for the heather and the quarter plot level for the birch; the same mean value was used for each of the cores within the same quadrat. Data was analysed by ANOVA using Genstat 8 ( VSN International Ltd, Oxford) with significance being assessed by Fisher's protected least significant difference.
Results: At Kerrow, where two species of birch had been grown, there was no significant difference in total C and N between the B. pubescens and the B. pendula treatments. Hence, in order to balance the ANOVA for all three sites, only the data for B. pubescens was considered further.
All three sites exhibited considerable heterogeneity in soil depth (either depth to underlying stones/rock or depth of organic horizon) and this is reflected in the coefficient of variation for total C (Table 2.1). The site at Craggan was the most variable.
Table 2.1 Coefficient of variation (%) for total C
Figure 2.2 Change in total carbon between heather and birch at three sites. Bars show the standard deviation.
The total carbon varied between the three sites in the order Craggan>Delnalyne>Kerrow (Figure 2.2). At Delnalyne there was no significant difference in total C between the heather and birch, while at Craggan there was a 17.7% loss (P=0.003) and at Kerrow a 22.9% loss (P<0.001). Over all three sites there was a 12.3% loss of carbon (P<0.001).
Figure 2.3 Change in total carbon between heather and birch at three sites partitioned between the vegetation and soil horizons.
Looking at the partitioning of the total carbon between the above-ground vegetation and the various soil horizons (Figure 2.3), it was again apparent that the three sites varied. Craggan was dominated by the carbon in the organic (O) horizon, Kerrow was dominated by C in the mineral horizon while Delnalyne was intermediate. It should be noted that the distinction between vegetation and litter for the heather plots was quite difficult to make as the heather was very rank having been ungrazed for 25 years. This caused the stems to fall over and become buried in litter material. Hence the apparently small litter fraction at Kerrow was probably a consequence of greater clipping away of the vegetation-litter complex. No such complication affected the birchplots but for the purpose of statistical analysis the vegetation and litter were combined. Over all sites there was significantly less C in the vegetation plus litter in the birch compared with the heather (16%, P<0.001). This was also true individually for Delnalyne (P=0.012) and Craggan (P<0.001) but Kerrow showed greater in the birch than in the heather (P=0.037).
For the organic horizon, there was an overall loss in the birch (10%) but this was not significant (P=0.088). The greatest loss through birch growth was at Kerrow (46%, P=0.012) but the changes at the other two sites were not significant.
For the mineral horizon, there was again an overall loss in the birch (14%) but this was not significant (P=0.087). There was a 47% loss at Craggan (P=0.025) and a 28% loss at Kerrow (P=0.005). Delnalyne showed an increase but this was not significant.
Table 2.2. Summary of significant changes in moving from heather to birch at the three sites.
Vegetation + litter
In summary, taking all the sites together, there were C losses from each vegetation and soil component, with about 40% of the loss coming from the organic horizon, and 30% from each of the mineral horizon and litter plus vegetation. However, this varied considerably between the three sites (Table 2.2).
Figure 2.4 Change in total nitrogen between heather and birch at three sites.
When the total nitrogen at the three sites was considered there was no significant differences between the heather and birch at Delnalyne and Craggan but a 22% loss in N (P<0.001) at Kerrow. Taking all sites together, there was no significant loss of N (Figure 2.4).
Discussion: The observation of a significant loss in soil carbon following the afforestation of heather moorland is the first of its kind as far as we are aware, apart from the preliminary observation of Mitchell et al. (2006). However, such a loss was expected from the results of Mitchell et al. (2006) who also recorded significant increases in Truog's phosphorus, mineralizable nitrogen and the decomposition rate of both wooden birch sticks and filter papers, placed in situ at the site, in the birch plots as compared with the heather plots. They also noted a decrease in soil moisture under the birch though this was only observed at the time of sampling.
That part of the total C loss was due to a loss in the vegetation plus litter component was unexpected. However, it should be noted that the birch growth even after 25 years was very variable across the plots and very poor in some areas. It would be expected that the birch biomass would continue to increase with time beyond the 7-14 tC/ha currently seen. Also the heather biomass (ca. 21-27 tC/ha but possibly including some litter) was considerable. However, the heather had been un-grazed for the 25 years of the experiment and had been allowed to become very rank. It is difficult to speculate what would have been the differences had grazing been applied though they might be expected to be less. Equally, it is unknown if the absence of grazing pressure has led to any effects on the below-ground C stocks under either the heather or the birch.
The loss of ca. 20% of the total system carbon at two sites over a 25-year period represents a significant rate of loss, 0.8% per annum, of a similar magnitude to that recorded by Bellamy et al. (2005) for all soils in England and Wales (0.6% per annum) though they suggest it may be even greater (2% per annum) in highly organic soils. We have no explanation as to why there was no significant loss at the third site. In most measured properties Delnalyne was intermediate between the other two so we could not point to any particular explanatory factor. In contrast to carbon, nitrogen was generally conserved though one site showed some loss. Since the stocking rates are much higher than those commonly used, the calculated loss rates may not be representative of other sites.
2.2.2 Pinewood site
A secondary objective of the data collection sub-module was to extend the birch findings to a Scots pine system. Unfortunately no parallel experiment on Scots pine with comparative heather moorland plots was set up and none exists elsewhere as far as we are aware. However, field sampling has been completed at Invermoriston, a Forestry Commission experimental site where both birch and Scots pine were planted on prior moorland in 1962. Though no control moorland plots are available, we are fortunate in that the MI has the original soils' descriptions as well as archived samples dating back to the time of planting so that some comparison of changes can potentially be made. Samples have been processed and analysed. Data analysis is in progress.
2.3 Evaluation of model performance
A number of the datasets collected during the project were used for model development and calibration. Testing the model against these datasets does not provide an independent evaluation of the model; for an independent evaluation of the model, external data sets, not used for calibration or development are required. A number of external datasets were used for independent evaluation of the model. The evaluation will be reported in full in forthcoming scientific papers. In this section, we give a brief selection of aspects of the independent model evaluation.
The model has been used to simulate CO 2 emissions from soils treated with litter, nitrogen, litter plus nitrogen and soils alone. Data from laboratory incubations (Foereid et al., 2004) were used and the model tested against these data. Figure 2.5 shows the performance of the model for each treatment.
Figure 2.5 Measured (circles with error bars) and simulated CO 2 respiration for laboratory incubated soils either alone, or with additions of litter, nitrogen or litter plus nitrogen (see Foereid et al., 2004 for further details).
The model slightly over-predicts respiration rate, especially early in the incubation, but the rates are comparable with those measured (high correlation, r > 0.88 in all cases), especially where litter has been added. The model has also been tested against data from field experiments on organo-mineral soils. Regina et al. (2004) published data on N 2O fluxes and soil mineral nitrogen dynamics on a cultivated peat in Southern Finland (60º49'N, 23º30'E) used to grow spring barley. The simulated and measured values of soil ammonium and N 2O emissions are shown in Figure 2.6, a and b, respectively.
The model simulates soil ammonium very well and correctly captures the temporal occurrence of peaks and dips. The model correctly predicts the timing of N 2O peaks, but at this site overestimates the size and duration of these peaks. Reasons for this are under further investigation.
Figure 2.6 Simulated and measured a) soil ammonium and b) N 2O emissions at a cultivated peat in Southern Finland (60º49'N, 23º30'E) used to grow spring barley (see Regina et al., 2004 for further details).
The model has also been tested to evaluate its performance in simulating decomposition and nitrogen dynamics if forest soils, to examine its suitability for simulating land use change following afforestation. Data on mass loss from litter bags (red pine and red maple) and on the nitrogen content of remaining material in the same experiment were simulated with the ECOSSE model. The results are shown in Figure 2.7 a and b.
Figure 2.7 Simulated and measured mass loss (a) and N content of the remaining material (b), from litterbag experiments at the Harvard Forest, US. Measured data from Magill and Aber, 1998).
The model predicts trends well but slightly over-predicts remaining mass over time, and slightly over-predicts N in the remaining material for Red Pine.
Other independent evaluations are being analysed whilst others are continuing. These are being prepared for publication and will be submitted for publication during 2007.