Climate-driven expansion of blanket bogs in Britain during the Holocene

. Blanket bog occupies approximately 6 % of the area of the UK today. The Holocene expansion of this hy-peroceanic biome has previously been explained as a consequence of Neolithic forest clearance. However, the present distribution of blanket bog in Great Britain can be predicted accurately with a simple model (PeatStash) based on summer temperature and moisture index thresholds


Introduction
Blanket bog is a distinctive type of peatland confined to areas with cool and extremely wet climates.The name derives from the fact that the peat covers sloping ground and hilltops, as well as basins, thus "blanketing" the landscape.Blanket bogs are widespread in the west and north of the UK (Great Britain and Northern Ireland) and occupy about 6 % of its land area (Jones et al., 2003).They are locally important (under various names) in other hyperoceanic regions of the world, although in total they cover only about 0.1% of the Earth's land surface (Gallego-Sala and Prentice, 2013).
The global distribution of blanket bogs today is confined to cool, wet climates (Gallego-Sala and Prentice, 2013).

The initiation of blanket bog formation during the Holocene
Published by Copernicus Publications on behalf of the European Geosciences Union.
is regionally asynchronous, and in most regions has been found to coincide with a shift towards cooler, wetter climates (Zaretskia et al., 2001;Dirksen et al., 2012).However, there has been considerable debate about the cause of Holocene blanket-bog initiation in the UK.
There is a long-standing hypothesis, first proposed by Moore (1973), that it was a consequence of land use by Neolithic human populations, and in particular land clearing practices at the time of the "elm decline" (often taken as a stratigraphic marker of Neolithic land use (Parker et al., 2002), as well as heavy stock grazing that changed the soil hydrological balance enough to initiate the inception of blanket bogs between about 6000 and 5000 yr BP (Moore, 1975(Moore, , 1993;;Merryfield and Moore, 1974;Robinson and Dickson, 1988;Huang, 2002).Evidence of removal of the shrub and/or tree cover by fire at the onset of blanket bog formation, and pollen analytical studies suggesting intensive agricultural practices by Neolithic people support this hypothesis (Merryfield and Moore, 1974;Smith and Cloutman, 1988;Robinson and Dickson, 1988;Simmons and Innes, 1988).A recent investigation of initiation of upland blanket bogs in Ireland also pointed to land use as a principal cause of paludification (Huang, 2002).However, a number of authors have suggested the initiation of blanket bogs at specific locations solely as a result of a climatic shift during the mid Holocene "Atlantic" period in Scotland (Ellis and Tallis, 2000;Charman, 1992;Tipping, 2008) the Faroe Islands (Lawson et al., 2007), and Ireland (Mitchell and Conboy, 1993;Dwyer and Mitchell, 1997).Tipping (2008) suggested that farming communities only settled in the Scottish Highlands after the landscape had already been covered by blanket bogs.Other authors have adopted a more complex view in which both climatic shifts and human activities played a role (Smith, 1970;Keatinge and Dickson, 1979;Tallis, 1991).Soil-forming processes, including leaching of base cations and consequent acidification and podsolization of soils, were also proposed to have been influential (Bennett et al., 1992;Charman, 1992;Smith and Green, 1995), giving rise to the term "pedogenic peats" (Simmons and Innes, 1988).
It is difficult to resolve such arguments about causality on the basis of timing alone.Lack of coincidence could be due to idiosyncratic local factors while synchroneity could arise by chance or because both events result from a common underlying cause.Under these circumstances, process-based modelling can offer a way forward.Globally, blanket bogs occur where the mean annual temperature (MAT) > −1 • C, the mean temperature of the warmest month (MTWA) < 14.5 • C and the ratio of mean annual precipitation to equilibrium evapotranspiration (moisture index, MI) > 2.1 (Gallego-Sala and Prentice, 2013).These limits ensure that the site is outside the permafrost zone and therefore not subject to cryoturbation, that summer temperatures are not too high for Sphagnum growth, and that there is sufficient moisture throughout the year to sustain peat growth on sloping ground.These limits have been used to construct a simple bioclimatic model, PeatStash (Gallego-Sala et al., 2010).In addition to predicting accurately the present-day distribution of blanket bog in Great Britain, PeatStash correctly predicts the highly disjunct global distribution of blanket bogs (Gallego-Sala and Prentice, 2013), including its occurrence in places such as Newfoundland and Kamchatka that have experienced very different land-use histories from Great Britain and Ireland.This finding strongly suggests that the present-day distribution, at least, of blanket bogs everywhere is controlled by climate.If so, it is natural to hypothesize that climate change was responsible for the Holocene expansion of blanket bogs.
Here we use PeatStash to simulate the UK distribution of blanket bogs in the mid-Holocene (6000 years ago, 6 ka).We compare these simulations with a new compilation of blanket-bog initiation dates, in order to explore whether climate change could plausibly account for the expansion of blanket bogs during the later Holocene.

Methods
We predicted the distribution of blanket bog at 6 ka using PeatStash (Gallego-Sala et al., 2011) with climate inputs derived from (a) climate model simulations of the 6 ka climate and (b) pollen-based climate reconstructions.The climate models provide predictions of a mutually consistent set of meteorological variables; using multiple climate models allows us to encompass the uncertainty resulting from differences between models.The climate models were run at relatively coarse resolution (Table 1) and there may be systematic biases that afflict all of the models (Harrison et al., 2013).Pollen-based reconstructions provide an independent source of information.However, their distribution is not continuous across the whole of the UK and the necessity to interpolate between reconstructions at individual sites could introduce uncertainty (Bartlein et al., 2011).Nevertheless, this information provides a useful check of the reliability of the simulated climates at the location of the sites and an alternative scenario of climate change.We therefore used both the climate-model ensemble and the pollen-based reconstructions to obtain mid-Holocene climate estimates to drive PeatStash.We then compared the PeatStash projections with a new compilation of data on the timing of blanket-bog initiation in the UK.

The PeatStash Model
PeatStash simulates the potential distribution of blanket bog (Gallego-Sala et al., 2010) based on mean annual temperature (MAT), mean temperature of the warmest month (MTWA) and a moisture index (MI) calculated from longterm monthly means of temperature, precipitation, and fractional sunshine hours.The definition of MI follows UNEP (United Nations Environment Programme, 1992):  where P is the mean annual precipitation (mm) and PET is the mean annual potential evapotranspiration (mm).We substitute equilibrium evapotranspiration (E q ), calculated from monthly net radiation and temperature, for PET in Eq. ( 1).E q is given by λE q = [s/(s + γ )] R n where λ is the latent heat of vaporization of water, s is the slope of the Clausius-Clapeyron relationship, γ is the psychrometer constant and R n is net radiation, calculated from latitude, season and fractional sunshine hours.The use of E q instead of PET affects only the absolute magnitude of MI, because PET as computed by the Priestley-Taylor equation is directly proportional to E q .PeatStash requires MI > 2.1, MAT > −1 • C and MTWA < 14.5 • C to determine the presence of blanket bog.
The model predicts the distribution of blanket bog in Great Britain with reasonably high accuracy (Fig. 1; Gallego-Sala et al., 2010).Detailed comparison for Northern Ireland was not possible because of the lack of accurate high-resolution data on blanket-bog distribution.However, comparisons with published maps suggest that the broadscale patterns are also captured there (Gallego-Sala and Prentice, 2013).

Simulated climate data
We used output from 10 climate models (Table 1) that had performed Mid-Holocene (6 ka) and pre-industrial (PI) simulations as part of the Coupled Modelling Intercomparison Project (CMIP5).The 6 ka simulations were driven by appropriate changes in insolation and greenhouse gas concentrations (Taylor et al., 2011), Anomalies (6 ka minus PI) of precipitation, temperature and fractional sunshine hours were bi-linearly interpolated from the original model grid to a common 0.5 • grid.These anomalies were then added to a baseline modern climate, derived from the CRU CL2.0 longterm mean climatology (temperature, precipitation, fractional sunshine hours) for the period 1931-1960 (New et al., 2000).

Pollen-based climate reconstruction
We used reconstructions of MAT, MTWA, mean annual precipitation (MAP), and α (the ratio of actual to equilibrium evapotranspiration, calculated as in (Cramer and Prentice, 1988)) from the Bartlein et al. (2011) data set.Bartlein et al. (2011) provided a harmonized compilation of pollenbased climate reconstructions, where individual site-based reconstructions were aggregated to provide estimates of mean conditions (with their uncertainties) on a 2 • × 2 • grid.Anomalies of each climate variable were interpolated from the original resolution grid to the 10 × 10 km grid of the UK-CIP_02 baseline climatology (http://www.cru.uea.ac.uk).We do not account for reconstruction uncertainties in this application because they are smaller than the differences between the climate-model scenarios.
PeatStash was run using MAT and MTWA as direct inputs, while MI was calculated from MAP and α.Assessed over a period of years, α can be related to MI using the Budyko hydrological relationship, which can be expressed as follows (Wang et al., 2012;Zhang et al., 2004): where m = MI and w is a parameter.To estimate anomalies of MI ( m) from anomalies of α ( α), we set w = 3 (Zhang et al., 2004), take the derivative of Eq. ( 2) and apply the approximation α ≈ m(∂α/∂m), where (3)

PeatStash 6 ka simulations
We ran PeatStash using output from each of the 10 climate models.Given model-dependent differences in the simulated climates (Harrison et al., 2013), the ensemble of simulations is used to provide an estimate of the probability that suitable climates for blanket bog existed by 6 ka in specific regions based on the consistency between the 10 projections.Peat-Stash simulations were also driven by pollen-based climate reconstructions of climate anomalies, which were superimposed on the higher-resolution UKCIP grid.We present the results of the 6 ka PeatStash simulations as anomalies from present.Wherever blanket bog is simulated for 6 ka, we predict that climate conditions were suitable for early initiation.Where blanket bog is simulated for PI but not for 6 ka, we predict that blanket bog initiation occurred after 6 ka.Where blanket bog is simulated for 6 ka but not for PI, we predict that conditions became unsuitable for blanket bog growth after 6 ka.

Basal Age data set
We assembled basal radiocarbon dates from blanket bogs throughout Great Britain and Northern Ireland.We adopted a stringent exclusion criterion, accepting only sites where blanket-bog formation commenced directly over mineral parent material and not as a change from a minerotrophic peatland (i.e.we have only included ombrogenous peatlands).We recorded the different topographic positions (saddle, bottom of the valley, slope, top) and altitudes of each site, whenever possible.The data set includes 64 records of pollenanalytically determined dates of peat initiation based on regional correlation of dated pollen-stratigraphic events.The remaining 164 records have either been directly dated from basal peat deposits, or there were sufficient radiocarbon dates to develop an age-depth model allowing the basal age to be well constrained.The extrapolated dates may provide more accurate estimates of basal ages than radiocarbon assays of basal peats, which often yield young ages because of contamination by mobile humic acids and root penetration (Smith and Cloutman, 1988;Charman, 1992).Any errors associated with the age modelling are expected to be considerably less than the 1000-year windows used in mapping peatland changes in our analyses.A total of 228 basal age estimates (see Supplement) were assembled but the full data complement was not available for all of these.
There is a difference between peat initiation and peat spread, and the latter cannot strictly be inferred from a single sampled point.There is local variability in peat initiation depending on topographic position, slope gradient, and altitude (Charman, 1992) and so a single sampled site may not capture the oldest peat initiation date.Blanket bog does not necessarily grow by uniform spread of peat but probably coalesces from different foci (Tipping, 1994).Furthermore, we are reliant on published and unpublished data collected for a variety of reasons that may have biased sampling towards deeper or shallower locations.Despite these known limitations in using basal dates to infer initiation, these effects will be similar for all regions and our data set is sufficiently large and regionally comprehensive to provide information on the patterns of peat initiation in different regions.

Results and discussion
The climate-model simulations consistently show summers warmer than today's over most of northern Europe.Mean annual precipitation (MAP) was slightly reduced in northern Britain and slightly increased in southern Britain compared to today.Conditions suitable for blanket bog are predicted at 6 ka across much of Scotland and northern England (Fig. 2a), but warmer than present summers restricted blanket-bog distribution in southwest Scotland, Northern Ireland and Wales.Southwest England was almost entirely unsuitable for blanket-bog formation at 6 ka, at least at the spatial resolution of the model grid, but became more suitable for blanket-bog development after the mid-Holocene.The suitability of different regions for blanket bog is examined in more detail using the high-resolution PeatStash simulations driven by quantitative palaeoclimate reconstructions.The pollen-based reconstructions (Bartlein et al., 2011) confirm that the climate over Great Britain and Ireland was slightly wetter at 6 ka than today (Fig. 3), with considerably warmer (approximately 2 • C) summers.As a result of the warmer summers, the bioclimatic envelope suitable for blanket bog was 14 % smaller at 6 ka (Fig. 2b).Larger areas of western Scotland, Ireland, and Wales have become suitable for blanket bog since 6 ka.Southwest England acquired three separate centres of predicted peat growth, correspond- ing to Dartmoor, Exmoor, and Bodmin Moor, as a direct consequence of late Holocene cooling.These simulations are consistent with observations of regional timing in the formation of blanket bogs (Fig. 4a).Analysis of basal dates on blanket bogs shows a gradual increase in blanket-bog formation throughout the early Holocene and a broad peak in initiation dates between 8000 and 4000 BP during the Mid-Holocene.There is a decline in the number of ages after 3-4000 BP.Regional patterns suggest that initiation occurred earliest in the north and most of the dates between 10 000 and 7000 BP are from sites in Scotland and northern England (Fig. 4a).Sites in Wales also have some early ages, but with a major increase in initiation dates after 8000 BP continuing throughout the rest of the Holocene.Sites in Ireland and southwest England are generally later to develop and have a peak at 3000 BP, later than the other regions.The initiation dates show that large areas of northern Britain were climatically suitable for blanket-bog formation before 6 ka, and remain so now.The regional differences in timing of initiation indicate a gradual increase in the area with suitable climate after 6 ka, especially in Wales, Ireland, and southwest England.
There are some discrepancies between the simulated and observed patterns of blanket-bog growth.Most of the exceptions are occurrences of initiation dates > 6 ka in areas such as Dartmoor that are only predicted to become suitable for peat growth after 6 ka.This may be an issue of resolution; some blanket bogs may have developed in localities with suitable microclimates that are smaller than our model can resolve, given the resolution of the climate inputs.It is also possible that this reflects a sampling bias.Older locations tend to be over-sampled because deep peat deposits are generally favoured in order to generate longer palaeorecords (Fyfe and Woodbridge, 2012).These may not have been laterally extensive or typical of the wider landscape.
We model a slight contraction in the area of suitable climate for blanket bog since 6 ka in eastern Britain (Fig. 2).If this model result is correct, there should be areas of eastern  Britain supporting relict blanket bog with no active peat formation.Although peat initiation occurred in these areas between 4 and 2 ka (Fig. 4a), post-6 ka accumulation rates are low (Simmons and Innes, 1988) suggesting that conditions indeed became less favourable for peat growth.Peat growth may continue for some time on an established peat bog due to local edaphic and hydrological conditions, despite climate being unsuitable for peat initiation.The existence of relict peats is not susceptible to testing using only initiation dates and this prediction would need to be explicitly tested by field sampling for cessation or slowing of peat growth.Our analysis of basal peat ages shows that blanket bogs have been developing in some regions of Great Britain and Ireland from the early Holocene onwards.The fact that blanket bogs developed later in the west and south of the country can be explained simply by the fact that regions with warmer and/or drier climates (Fig. 3) were less suitable for peat formation during the early Holocene.Blanket bogs only developed in these areas as climate became cooler and wetter.Blanket-bog formation accelerated in the mid-to late Holocene, but this occurred later than the "elm decline" event in many locations and proceeded continuously, which makes it unlikely that it was causally linked to human activities.The simulations (Fig. 2) indicate that a large part of Great Britain and Ireland was suitable for blanket-bog formation before the main period of human impact.
Climatic control of blanket-bog formation in the UK is consistent with evidence from other parts of the world that blanket-bog initiation occurred in response to climate change and that their current distribution is strongly controlled by climatic conditions.It raises an important issue about the fate of this unique ecosystem under future climate change.Our work supports previous analyses that suggest they will require careful management given that their continued growth may be threatened by large-scale shifts in climate in some regions of the UK (Clark et al., 2010;House et al., 2010;Gallego-Sala et al., 2010) and worldwide (Gallego-Sala and Prentice, 2013).
Taken together, these lines of evidence indicate that the history of blanket-bog growth in Great Britain and Ireland can be explained as a threshold response to a changing climate.In an area with a rich human history, such as Great Britain and Ireland, almost all Holocene palaeoecological records show signs of human impact at various stages.How-ever, our analyses suggest that no human intervention was required to initiate blanket-bog formation in Great Britain and Ireland.
The Supplement related to this article is available online at doi:10.5194/cp-12-129-2016-supplement.

Figure 1 .
Figure 1.The area of blanket peat predicted by the bioclimatic envelope model (BCEM) PeatStash using a baseline climate period (UKCIP02: 1961-90) overlain on the mapped 5 km gridded data of observed blanket peat presence ( © Crown Copyright/database right 2009.An Ordnance Survey/EDINA supplied service.Met Office/UKCIP gridded climate data.UKCIP02 © Crown Copyright 2002).

Figure 4 .
Figure 4. Assembled basal calibrated radiocarbon dates from blanket bogs over Great Britain and Ireland: (a) regional graphs of initiation dates through time binned every 500 years; (b) map of individual initiation dates; and (c) map of initiation dates summarized per region.

Table 1 .
Summary information on the climate models used in this analysis.

Table 2 .
Region by region break down of percentage of (a) cores with basal dates younger than 6 ka (b) sites with basal dates exclusively younger than 6 ka (c) % gridcells that PeatStash predicts to have initiated after 6ka when run with the pollen-based climate reconstructions.