One of the greatest problems facing mankind is the impact of environmental change on fundamental aspects of the chemistry, physics and ecology of the Earth at the global scale. Soil is a major component in the global carbon cycle (Fig. 1) and vulnerable to impacts of human activity with about 1500 Gt of organic carbon (C) to a depth of 1 m and a further 900 Gt from 1-2 m^[1]. Globally twice as much carbon is stored in soils as in the atmosphere^[2] with peatlands contributing a third of this^[3]. Thus even small changes in soil C stocks might contribute significantly to global climate change, for example, due to a negative feedback as a result of global warming^[4].Whereas above ground carbon cycling is well understood there is great uncertainty in climate impacts on soil carbon cycling^[5]. For example, the biological significance and relevance of the most critical C-turnover parameter used in soil organic matter sub-models (SOMs) of DGVMs, the Q₁₀ value, representing soil respiration responses of different C-pools to temperature, is not sufficiently understood^[6,7].

The University of York leads CTCD research into the relationships between the environment and soil C stores, the role of soils in producing 'greenhouse gases' and the potential impacts of future atmospheric CO₂ levels on soil processes. Currently this involves 4 main projects:

1. Assessing and reducing the uncertainty in the UK soil C budget (including newest research findings)
2. Assessing environmental impacts on soil carbon fluxes
3. Improving existing soil C models
4. Assessing the potential of Earth Observation (EO) data

ALSO:

• CLASSIC and CTCD joint project: Measuring carbon fluxes in an upland moorland: Moor House NNR


Fig. 1: World carbon stocks (Petagrams C) from IPCC2001

York people involved: Phil Ineson; Andreas Heinemeyer; Iain Hartley; Laura Austin; Harry Vallack; Steve Cinderby.

Education and training:


1. CTCD staff at York are also actively involved in supervising a variety of Undergraduates and Masters projects, including:
• Litter decomposition and peat accumulation in realtino to actual evapotranspiration
• Peat accumulation during the Holocene
• Carbon cycling in 'closed systems'
• Methane consumption by soils
• Temperature response of soil respiration in forest ecosystems
2. CTCD staff have also been involved in both funding and organisation of a recent European Science Foundation summer school on measuring soil C fluxes (ESF-Summer School)
3. Currently we are also planning a summer school ”CTCD Terrestrial Carbon Cycle and Earth Observation“ designed to introduce post-graduate students to the wide variety of techniques used in terrestrial carbon research (CTCD summer school)

1. Assessing and reducing the uncertainty in the UK soil C budget

Latest Research

One of the main uncertainties in modelling net ecosystem exchange (NEE) is soil respiration.

Fig. 2: Multiplexed Li-Cor 8100 soil respiration system with different mesh collar treatments




We deploy a state-of-the-art multiplexed (up to 16 chambers) soil respiration system aiming to (i) separate the soil flux into its components (i.e. autotrophic: roots & mycorrhizas, heterotrophic: soil only) and (ii) to determine their individual environmental responses (e.g. temperature, light, phenology). Data are currently been used to improve our soil sub-models used within CTCD. We use a novel mesh collar design, separating mycorrhizal and soil flux by using a 41 or 1 µm mesh, respectively and combining this with shallow soil collars, not cutting any roots (Fig. 2).





The soil respiration system delivers continuous high temporal resolution data (1 hourly mean values). We now have completed an annual cycle for a York pine forest site. The mesh collar/shallow collar design allowed us to (i) separate soil CO₂ flux components (Fig. 3) on a fine temporal resolution and (ii) calculate the proportional flux components. We measured for the first time in situ ectomycorrhizal respiration rates and showed that mycorrhizas are responsible for about 25% of the overall soil respiration, thus contributing more than do roots (~15%). Soil heterotrophic respiration remained fairly constant at about 65% throughout the year.

     
Fig. 3: left: Sample of hourly (1-hour cycles) mean values of soil respiration treatments (green: total flux; blue: minus roots; red: minus root and mycorrhizas); right: Proportional soil respiration flux components (roots, mycorrhizas and soil only) in autumn 2005.

Another important finding is that mycorrhizal respiration does not depend on soil temperature but rather on photosynthate supply, which can be shown by a lack in temperature response and a diurnal time-lag function of the mycorrhizal flux component (Fig. 4). Whereas the mycorrhizal flux component showed a strong diurnal cycling, peaking during the night and declining during day-time, the soil only component revealed a very clear temperature response (Q10 = ~4). This finding has to be taken into account in modelling soil CO₂ fluxes correctly. However, this finding relates to collar design and there is another important issue: Soil collar insertion depth…

     
Fig. 4: left: Hourly (1-hour cycles) mean values of mycorrhizal and soil only mesh collar respiration rates (yellow: total flux; brown: soil only; green: mycorrhizal component only); right: Diurnal cycling of the three soil respiration flux components (total, soil only and mycorrhizal component only) in autumn 2005.

Commonly soil respiration measurements are done by pre-inserting PVC collars into the soil on which the soil respiration chambers are fitted during measurement. Mostly these collars are inserted to more than 5 cm into the soil. However, this will cut any roots and mycorrhizal hyphae, which are predominantly found within this top soil layer. Thus the measured fluxes might be underestimated. We performed a series of measurements and found (i) a strong negative relationship between collar insertion depth and measured fluxes and (ii) the top layer flux can be characterized by a strong diurnal cycling, probably due to the mycorrhizal component (Fig. 5). This means that most measured soil fluxes so far have largely underestimated soil respiration due to two caveats: collar insertion depth and time of sampling (as sampling is done mostly done during day-time, whereas highest fluxes seem to occur during night-time, if shallow collars are used).

     
Fig. 5: left: One of the soil collar depth treatment plots showing different collar depths examples and the Li-Cor survey chamber, inset shows the adapted long-term chamber for measuring fluxes on these collars; right: Soil respiration fluxes per collar depth in summer 2006.

The accuracy and uncertainties of the UK soil carbon database need to be addressed; we aim to expand this work to a European scale e.g. JRC, CarboEurope (e.g. Germany, Italy).


Fig. 6: GB soil carbon map (kg m^-2) on a 1km² grid for 1m depth 1998 and 2004 data.

Production of the previous (1998) 1 km² soil C database (1 m depth) placed the UK as a leader in European soil C stock research. In 2004 the UK Department for Environment Food and Rural Affairs (DEFRA) released an improved dataset (soils data info ), now incorporating soil texture for both 30 cm & 1 m depth, which aimed to unify the Northern Ireland, Scotland, England and Wales databases. Within CTCD we have been one of the first users of these data and have helped to improve the database, as well as producing weighted mean 1 km² grid cell databases and maps (Fig. 6) from data made available to us. However, Scottish vs. England and Wales upland (peatlands) C stocks seem to be somewhat different with maximum soil C densities in England and Wales. For the UK we have now produced improved best estimates for soil C stocks for either 30 cm or 1m soil depth (Fig. 7). However, there are differences of up to 1Mt C depending on the information used to calculate carbon densities, ie. the amount of C per unit land area..


Fig. 7: UK soil C density estimates (kg m^-2) for 30 cm and 1 m soil depth - 2003 data – Using a weighted mean area approach for different levels of complexity (e.g. up to 10 land use and 5 soil series percentage information per 1 km² grid cell) where urban areas are zero C values and garden = 0.5 x grassland C of the same soil series. N.B. In comparison, the vegetation C stocks (bottom right) are only about 1-3% of soil C.

CTCD has also produced soil texture maps (1 km²) from the 2003 data (Fig. 8) which are fed into the Sheffield Dynamic Vegetation Model (SDGVM). The data show clearly that although spatial coverage is the same, soil depth information alters soil property values. As a consequence, data based on different soil depths and used in model predictions will alter the soil C dynamics significantly (see Fig. 9). In particular, bulk density values show marked changes with soil depth (Fig. 8).

        
Fig 8: UK soil texture data: 1 km grid values for 30 cm (left) vs. 1 m soil depth (right) - 2003 data as weighted mean area averages of different land use and soil series information (compare to Fig. 7).

Within CTCD we also assessed the impact of spatial resolution in soil texture data on modelling soil C stocks using emulator approaches, for example, with the SDGVM. However, due to licensing problems we are not allowed to use the Scottish data, limiting our comparison to 30 sites in England and Wales (Fig. 9). We found that spatial resolution influences model predictions significantly and, further, that bulk density seems to be a very important parameter; this is an issue of concern because adequate and appropriate data are rarely available when looking beyond the UK. In particular, many DGVMs predict bulk density from soil texture, and this is prone to large errors. It is more accurate to use detailed multiple regression approaches, ideally including organic C (%) data (Fig. 10). Although we have shown that land use is an important factor, by far, most of the variability in bulk density can be explained using organic C (%) data and we are currently devising new ways of including this information into the existing models (e.g. SDGVM). We have now achieved the first uncertainty estimates for the England and Wales based on detailed soil parameter and other input uncertainty from 30 selected sites.

        
Fig 9: UK soil texture data:
Left: Global ISLSCP(I)1 degree x 1 degree datasets vs. UK 1 km grid for 1 m soil depth.
Right: 2004 data used by the emulator approach for SDGVM at one of the 30 UK sites. Overall the model shows a strong response to soil properties,in particular to bulk density

 


Fig 10: UK soil texture data: - significance of soil texture vs. organic C data in a multiple regression approach predicting bulk density values for different land use and UK regions.

2. Assessing environmental impacts on soil C stores and fluxes

We need to better understand environmental impacts on individual components of soil C stores and their fluxes. We are using correlations of soil C stores with different key environmental variables applying a number of different statistical approaches. Further, stable isotopes (e.g. ¹³C) are proving an important tool in detecting changes in soil C fluxes against a huge background pool of soil C (up to 150kg m^-2 and more).
        
Fig. 11 : GB C(log kg m^-2)vs mean temperature - 1998 (left) and 2004 data (right).
N.B. the right censoring due to limitation to 1 m soil depth in the 2004 dataset. The 1998 dataset provides a better estimate of how much soil C is stored in deep peats in Scotland, as calculations include maximum peat depth information (ie. up to 6 m peat depth).

We have correlated both the 1998 and the 2004 soil C data with the UKCIP climate data, and found mean annual temperature best explained soil C stocks (see Fig. 11). From the 2003 data there is a clear upper limit of soil C in the lower temperature range, but this is simply because the soil depth is limited to 1m. As many upland soils (in particular in Scotland; see legend Fig. 11) are deeper than 1m we are undoubtedly not reflecting the true relationship. However, from this correlation we can conclude that any predicted increase in temperature due to global warming is likely to have a significantly negative impact on soil C stocks, but other (interacting) factors need to also be investigated. We have built a model predicting ‘real’ soil C values using different censoring information for the 2004 dataset. Interestingly, the model predictions fall within the range of the ‘uncensored’ 1998 (incl. deep peats for Scotland) best estimate data (Fig. 12).

        
Fig 12: GB soil C (kg m^-2) vs. model predictions:
Left: Predictions for right censored model approach – two different censoring levels (blue and green dots); 2004 (red dots) and 1998 (black dots) data; n.b. the model predictions fall within the ‘best estimate’ 1998 data, including Scottish deep peat information.
Right: Comparison of UK 2004 data vs. SDGVM model output; n.b. the regression fit is nearly identical for the two graphs but only when organic rich soils are excluded from the 2004 soil C estimates.

3. Improving existing soil carbon models

The accuracy of soil carbon models needs to be addressed and improved based on new research. Major problems are (i) the lack of a biological component, i.e. how soil organisms will respond to future climates and how this will affect soil C fluxes, and (ii) the need to expand soil C models to organic rich soils (e.g. peatlands see Fig. 13) and (iii) to address the limitations in using Q₁₀ (temperature response coefficient) values to describe soil temperature responses. We are currently developing a new version of a peat growth-degradation model (Fig. 14) based on previous models^[9,10,11], including water table dynamics and environmental feedbacks on vegetation and decomposition dynamics. Basically, the model is based on litter-dependent decay rates (C/N; lignin content etc.), reconstructed climate (10,000 years), simulated dynamic water table (Penman-Montheith), feedbacks on vegetation dynamics (response surfaces) and estimated NPP ^[12]. To assess model performance and to improve certain model parameters it is clearly necessary to obtain accurate field observational measurements. To this end, the York group are using mainly a gas chromatograph isotope ratio mass spectrometer (GC-IRMS) dedicated for ¹³CO₂ determination under field conditions, housed in a mobile laboratory (Fig. 15), and a new multiplexed automated soil respiration system (LI-COR 8100; Fig. 15) to study biogeochemical cycles and trophic interactions in soils.
        
Fig 13: The problem of modelling organic rich soils:
Left: GB soil C estimates (kg m^-2) 2004 data on either 1 km or 10 min grid vs. 10 min grid model predictions by the SDGVM.
Right: A schematic version of a peat growth-degradation model developed at York using reconstructed climate data and water table effects on decomposition of litter and peat.
        
Fig 14: Output from the York peat growth-degradation model simulations for 10,000 years at Moor House (NNR, North Penines, UK) under different climate scenarios: Left: peat mass (kg C/m²) Right: peat depth (m). N.B. a drier warmer climate turns a bog from a C-sink into a C-source until a new equilibrium is reached

 

        
Fig 15: Use of the 'mobile lab' in Harwood Forest (Northumbria, UK) with soil respiration chambers and canopy lines(left).The multiplexing soil respiration system based on the LI-COR8100(right).

The 'mobile lab' (Fig. 15)is a unique field based laboratory system for the in situ determination of C-13 stable isotope composition of air and soil dioxide fluxes. The system comprises a customised gas chromatograph isotope ratio mass spectrometer (GC-IRMS) dedicated to ¹³CO₂ determination under field conditions. The equipment is housed in a mobile laboratory with attendant temperature regulation and gas flow for 16 air lines and can be operated, without user intervention, for periods of up to 2 days.

We have just finished the construction of a fully automated dynamic closed soil respiration system by multiplexing 12 LI-COR long-term chambers with the LI-COR8100 IRGA (Fig. 15) for which we acknowledge funding under a NERC Capital Equipment Bid initiative and generous support from LI-COR. This system will be compatible with the GC-IRMS system and will provide important insights into the complex system of environment – photosynthesis - soil respiration. The system will also provide observational data for improving model predictions at selected UK sites (see Fig. 16).

One of the greatest challenges is to improve existing soil organic matter models (SOMs) in order to model organic rich soils (e.g. peatlands) where most of the terrestrial carbon is stored (see Fig. 2). We are currently paralleling similar studies previously carried out for low organic C soils^[8] and we are comparing how models of different biogeochemical and biological complexity (ie ROTHC, CENTURY, Forest-DNDC and Forest-ETP+MEFYQUE) are able to simulate peatland SOM dynamics (Fig. 16). We have acquired a substantial observational dataset for three selected peatland sites with different site history, vegetation and topography. Whereas some of the models have been mainly designed for arable sites (ie. ROTHC), and thus are expected to lack key processes for peatland C dynamics, others should perform better since key processes, such as water table changes, are coupled to redox potential regulating seasonal changes in the ratio of aerobic to anaerobic decomposition (ie. Forest-DNDC). Further other models include soil biota impacts (ie. fungi in MEFYQUE) on decomposition. Main aims are to assess:

• where do the selected models fail to perform well and why;
• what key processes are most important to modelling peaty soils;
• which of the models forms the best basis in order to predict SOM in peatlands;
• what are suggested amalgamations for further developments to such a 'new' model.

        
Fig 16: The model inter-comparison field sites and a simple diagram explaining the different approaches for modelling mineral and organic rich soils.

4. Assessing the potential of Earth Observation (EO) data

Another aim is to overlay EO data with other environmental data (GIS) in order to improve existing soil C maps and to assess sources of uncertainty in model performance; this is being performed in close collaboration within CTCD.

Fig 17: left:Part of the GIS database for Harwood Forest: OS map (1:50,000), CEH landcover (2000), DTM and 1998 soil C 1 km² grid (as in Fig. 1); right:Soil C content (kg m^-3) on a 1 km² grid overlaid on an aerial picture, OS map, national soil map (Natmap) boundaries and a CTCD digitised detailed soil map of Harwood forest.

We are assessing the uncertainties in the new (2003) soil C data using a stratified sampling design at selected UK sites. Quantifying uncertainties will enable us to highlight and reduce model uncertainty and complexity; CTCD are using Bayesian statistics for this approach. Harwood Forest (Fig. 17) offers a unique opportunity to assess uncertainties and we have compiled a GIS database including detailed vegetation, soil properties and land use information. We have also produced a very detailed digitised soil map in collaboration with Forest Research (FR) which will act as a basis for stratified sampling. With the help of UCL, we intend to use Ground Penetrating Radar (GPR) in order to estimate maximum peat depths and thus obtain a better estimate of the 'real' soil C densities. N.B. soil C densities are normally only given for a fixed depth e.g. 1 m but this excludes carbon stored in deep peats, which can be more than 6 m deep.

Iain Hartley's PhD work


References:

[1] Miko UF Kirschbaum (2000) Will changes in soil organic C act as a positive or negative feedback on global warming? Biogeochemistry 48 (1): 21-51.
[2] Houghton JT, Ding Y, Griggs DJ, Noguer M, van der Linden PJ, Dai X, Maskell K, Johnson CA (eds). (2001). Climate Change 2001: The Scientific Basis . Cambridge, UK: IPCC, Cambridge University Press.
[3] Gorham, E (1991). Northern peatlands: Role in the carbon cycle and probable responses to climatic warming. Ecological Applications 1, 182-195.
[4] Cox PM, Betts RA, Jones CD, Spall SA, Totterdell IJ (2000). Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate model. Nature, 408, 184-187.
[5] Valentini R, De Angelis P, Matteucci G, Monaco R, Dore S, Scarascia Mungnozza GE (1996). Seasonal net carbon dioxide exchange of a beech forest with the atmosphere. Global Change Biology, 2, 199-207. Shaver et al. 2000[5]).
[6] Giardina CP and Ryan MG (2000). Evidence that decomposition of organic carbon in mineral soil do not vary with temperature. Nature, 404, 858-861.
[7] Grace J and Rayment M. (2000). Respiration in the balance. Nature, 404, 819-820.
[8] Smith P, Smith JU, Powlson DS, McGill WB, Arah JRM, Chertov OG, Coleman K, Franko U, Frolking S, Jenkinson DS, Jensen LS, Kelly RH, Klein-Gunnewiek H, Komarov A, Li C, Molina JAE, Mueller T, Parton WJ, Thornley JHM and Whitmore, A.P. (1997). A comparison of the performance of nine soil organic matter models using seven long-term experimental datasets. In: Evaluation and comparison of soil organic matter models using datasets from seven long-term experiments. (Eds: P. Smith, D.S. Powlson, J.U. Smith & E.T. Elliott). Geoderma, 81, 153- 225.
[9] Clymo RS (1984). The Limits to Peat Bog Growth, Philosophical Transactions Royal Soc London, B 303, 605-654.
[10] Hilbert DW, Roulet N, Moore T (2000). Modelling and analysis of peatlands as dynamical systems. Journal of Ecology, 88, 230-242.
[11] Bauer IE (2004). Modelling effects of litter quality and environment on peat accumulation over different time-scales. Journal of Ecology, 92, 661-674
[12] Lieth H, Box E (1972), Evapotranspiration and primary productivity. In: Thornthwaite W(Ed.), Memorial Model, Publications in Climatology. C.W. Thornthwaite Associates, New Jersey, pp. 37-46.

HOME | MISSION STATEMENT | SCIENCE PROJECTS | PARTICIPATING ORGANISATIONS
PEOPLE | NEWS | CTCD ORGANISATION | CONTACTS