Experimental design

The below information explains the experimental design to enable assessing the impacts of different management regimes on key ecosystem services:

(carbon storage, climate regulation, water storage & quality, biodiversity).

The overall purpose of the project is to assess the impact of different management techniques on blanket bog areas with dominant heather coverage. The project will monitor changes in peatland variables in response to different management (i.e. treatments). The main treatment at the catchment-scale is to mow/cut areas in-line with current mowing management techniques and compare them to control areas undergoing the ‘business as usual’ burning regime.

Monitoring plots within catchments

The project considers one large catchment-scale mowing treatment and burning control (across three paired sub-catchments, see a schematic (!) figure on the left) and small plot-scale treatments at 5 x 5 m experimental plots (with monitoring areas for water table depth [WTD], vegetation [1x1], methane [CH4], soil respiration [SR] and net ecosystem exchange [NEE]), five (T1-T5) within the mown treatment and two within the burnt control sub-catchments (C & C1): i.e. uncut 'do nothing', mowing with leaving or removing brash, and either with or without Sphagnum addition (for burning only the latter two). The main management (mowing and burning) reflects the current mosaic nature of burn strips and maintains a succession of heather ages. One burn/cut every other year (ca. 20% of the total area) is proposed. There was no burning inside the sub-catchments in the 2012/13 season (i.e. to allow the assessment of a pre-treatment period in relation to capturing generic differences between plots and sites unrelated to the treatment). The findings from the experimental post-treatment period can therefore be interpreted in relation to the observed generic differences captured in the pre-treatment monitoring period.

All treatments will start at the same time following the initial pre-treatment period of one growing season. The experimental design will enable data from all three field sites to be analysed either per site (i.e. four reps across the four blocks per treatment) or together (i.e. three replicates in four blocks per treatment) using ANOVA (analysis of variance) or GLM (Generalized Linear Model) to compare the effect of different treatments, and post hoc tests, such as the Tukey’s HSD test, to identify any significant differences between factors. Whereas the latter is a true replication we acknowledge that the ‘per site’ approach does not strictly speaking represent independent replicates. However, the distance between monitoring plots and areas is large enough to limit any similarities as shown for other peatlands.

The following parameters will be monitored: meteorological variables, carbon fluxes, carbon stocks, GHG emissions, peat pipes, vegetation dynamics, water balance and water quality and empirical and process models will allow up-scaling of the findings in space and time, thus making them more policy relevant.

In the future we might also consider other potential treatments such as herbicides (see Todd et al., 2000 for comparable MoIinia issues); however, herbicide usage seems an unlikely option (see Milligan et al., 1999; Rogers, 1996), mainly due to the possible impacts on water quality. Other treatments such as Sphagnum plugs and repeated mowing might offer better alternatives for field-based trials, whilst herbicides could be trialed in laboratory-based mesocosm studies (see p231 in the 'Concluding remarks and suggestions for future work' section in the Phase 1 Final Defra Report; Heinemeyer et al., 2019).

Overall we apply a combination of two statistical approaches (designed in consultation with an experienced statistician, Dr. Calvin Dytham) within this project allowing large catchment-scale (most policy ‘relevant’) and small plot-scale (most scientifically ‘relevant’) manipulations. For this, the Before-After Control-Impact (BACI) approach (i.e. comparing the time series behaviour of pre-treatment versus post-treatment; Stewart-Oaten et al., 1986) will allow us to establish a robust, replicated experimental study (with one major catchment-scale treatment and several additional restoration plot-level treatments being applied across three sites). We propose that this experimental structure will allow sound statistical analyses of both within and between sites as true replicates exist at both levels. Such a robust experimental approach is often replaced by a short-term (i.e. cost saving) space-for-time (SfT) approach, but can result in serious statistical issues and questionable interpretation of findings (mainly as site history and generic site/plot differences unrelated to management remain unknown, but are assumed to not affect observed differences - an assumption unlikely to be true). For a peer-reviewed publication and a subsequent detailed pre-print discussion see Ashby & Heinemeyer (2018; 2019).


Above: Burning at Whitendale and Mossdale - ready to put out unwanted patches at the edges. Good quick and cool burn at all sites ... but most carbon gone up in smoke (although some is converted into charcoal, a long-term carbon store resisting decomposition).


Above: Mowing at Mossdale (Tom Iveson) - in effect leaving most carbon on the peat (although this is subject to long-term decomposition) but relying on fossil fuel. Horrible weather but that's one of the advantages of mowing - can be done in nearly all weather.

We established, at each of the three field sites, two adjacent sub-catchment scale experimental sites: the ‘control sub-catchment’ and the ‘treatment sub-catchment’ (see Figure above). Within the treatment sub-catchment, mowing of heather and other alternatives will replace the 'business as usual' burning still to be continued in the control sub-catchment; mowing and burning - see movie clips below - will happen on equal areas in each catchment. Four main monitoring areas are set up within each of the two sub-catchments at each of the three field sites. These areas are treated as ‘blocks’ when data for all three field sites are considered together for treatment effects.

Plot level treatments replicated across four blocks are (based on a stakeholder workshop held in York in February 2013):

Control (burn)

Control (burn) + Sphagnum


T1 mowing + brash

T2 mowing - brash

T3 mowing frequency + brash + Sphagnum

T4 mowing frequency - brash + Sphagnum

T5 ‘do nothing’ no management control treatment (no burn, no mow)

Above movie clips demonstrate burning (left) and mowing (right) at Mossdale. Notably, repeated mowing could only be achieved in a subsequent extension of the initial Defra-funded 5-year project period to allow sufficient regrowth of heather (size!).

Moreover, the Sphagnum additions (see below pictures) will address which of the three (apparently particularly) "peat-forming" species (i.e. Sphagnum pallustre, capillifolium, papillosum) will be performing best, on either burnt or mown plots with or without brash removal. In Spring 2014, we applied a mix of the three species across half the mown 5 x 5 m plots (i.e. with or without brash) at a rate of 1 pellet per cm2. In addition we also applied increasing levels of Sphagnum within three 1 x 1 m plots alongside four burnt and mown plots (at Whitendale only) to assess density effects on establishing new Sphagnum on burnt vs. mown plots (with brash); levels were 200, 400 and 600 pellets per 1 m2.


Above: Sphagnum pellet (BeadaMoss) addition at Whitendale: 1 x 1 m plot additions (left: for the 5 x 5 m plot; right: three levels of pellet density additions - from low [200 per 1 m2] to high [600 per 1 m2]).

In the ‘control sub-catchment’, each of four control monitoring areas (blocks) will contain only one monitoring plot (giving four control plots – controls in the treatment area were considered but deemed impossible to accommodate due to the burn risk) whilst on the ‘treatment sub-catchment’ five monitoring plots (one for each planned treatment) will be set up in each of four treatment monitoring areas. The five treatment plots will be allocated randomly among the five monitoring plots within a block and will be placed in areas of similar slope (~5-10°) and vegetation composition (i.e. heather coverage of ~40-60%). Each of these plots will contain a central (1 x 1 m) permanent monitoring plot (e.g. for vegetation monitoring).