3.3.5.1 MICROCLIMATE OF PLOTS
Silwood Weather Station data was used (unpublished data) to make a dataset of rainfall for the duration of the experiment, and a rain gauge to dictate the implementation of the rainfall treatment from day to day. The average soil moisture content of each plot was measured on a weekly basis, year round, using a ThetaProbe at a distance of 1m from the edge on all four sides.
Dataloggers were installed (iButtons DS1923, Homechip, Milton Keynes) at the vegetation height specific to each plot, which logged temperature and relative humidity (%RH) every thirty minutes between January 2009 and September 2010. They were calibrated using a Weatherlink 5.8.0 (Davis Instruments, Brighton, UK) station at Silwood Park, fitting a correction factor to both datasets (Appendix 3.6, Fig. 3F, 3G).
3.3.5.2 VEGETATION SURVEYS
Vegetation surveys of all plots were completed at the beginning and end of the summer seasons. An additional survey was carried out in July 2010 because severe drought had caused vegetation dieback. A 1m2 quadrat with 25cm2 subdivisions was placed in the centre of each plot and percentage cover was visually estimated for each species (http://www.itis.gov). Percentage cover of necromass and bare earth were also assessed.
3.3.5.3 CO2 FLUX
To measure ecosystem CO2 flux rates, PVC ring collars (20cm diameter, 10cm long) were
inserted into the soil to a depth of 5cm on each plot to create a seal with the soil. A transparent Perspex chamber (area 299cm2, volume, including average collar volume
8959cm3) was then attached to a CIRAS-2 infra-red gas analyser (IRGA) in 2009 and a CIRAS-1 IRGA in 2010 (PP Systems, Hitchin, UK), which was clipped onto the collars to create a sealed area over the plants. The CIRAS measured CO2 and water flux for four
minutes (CIRAS-2) and two minutes (CIRAS-1). In light conditions, the returned values were net ecosystem exchange (NEE) (mol CO2 m-2s-1) and evapotranspiration (ET), (mmol m-2s- 1
). This was repeated with an opaque cover to simulate night-time ecosystem respiration (Reco). Soil moisture, PAR and soil temperature (Hanna HI 98501, Bedfordshire, UK) were
measured as covariates, and a set of values from the iButton data was derived, which consisted of the temperature and humidity of each plot on the specific day it was measured, averaged between 9:30am and 5pm. These measures were taken monthly during the summer and in alternate months through the winter.
3.3.5.4 DECOMPOSITION
Decomposition rate measures of the individual plots began in December 2008. 2g of dried, cut samples of the most dominant species in FG1 (Holcus mollis) and FG2 (Arrhenatherum
elatius) were placed separately in mesh bags with a 1mm aperture (Normesh, Oldham, UK)
and secured to the soil in each plot using staples. A representative for FG3 was not used due to lack of biomass at the time of the experiment. Four bags for each species were placed in each plot and one bag per species was removed at three month intervals (March, June and September 2009). All new biomass growing through the mesh was removed and the remaining material dried at 80˚C for 24 hours, before being weighed, and the mass lost from the bags calculated.
3.3.5.5 MINERALISATION
Mineralisation of nitrogen was ascertained quarterly between December 2008 and September 2010. The net mineralisation rate in each plot was determined using 4cm diameter by 10cm depth in situ soil cores, which were sealed with tape to prevent leaching, and incubated in the soil for three months (time= t0). Four initial samples of soil were taken and homogenised to
create an initial composite sample for each plot. Nitrogen was extracted from the samples using Allen (1989)‟s method (Appendix 3.9), and analysed colourimetrically using a Skalar SAN++ Continuous Flow Analyser (CFA, York, UK) (Appendix 3.10, Table 3A).
After three months had elapsed, the incubated cores were removed (t0-1) and the soil
available nitrogen content analysed in the same manner, along with a new composite sample (t1). The cores were then replaced in the plots with fresh soil. The mineralisation rate of the
plot was calculated by summing the NH4+ and NO3-/NO2- of each sample to give total
extractable nitrogen before subtracting the t0 value from the t0-1 value and multiplying the
total by the bulk density. This gave a value of nitrogen accumulation over three months. These calculations were repeated with NO3-/NO2- only to gain a quarterly estimate of
nitrification.
3.3.5.6 TOTAL NITROGEN AND PHOSPHORUS
Total soil nutrient concentrations were determined in October 2008, September 2009 and September 2010. Soil samples were collected and homogenised from four areas of each plot to create composite samples and dried at 80˚C for 24 hours, before being partially digested using a modified Kjeldahl method (Kjeldahl 1883, modified using a selenium catalyst, Appendix 3.8). These were stored at 5˚C overnight, then analysed using a Skalar SAN++
CFA.
3.3.5.7 C:N RATIO
Soil was taken from the May 2009, September 2009 and September 2010 samples, dried at 80˚C for 24 hours, finely ground and weighed into tin cups (15-20mg- exact weight recorded). Carbon and nitrogen content were then determined using a CNS total combustion analyser (FLASH EA 1112 Series, CE Instruments, Wigan, UK). The machine was calibrated using three aspartic acid standards, which were verified by measuring three further standards
as “unknown” and comparing to the known values. Two blanks were also included in each run (empty cups). The soil C:N ratio was then calculated for each plot.
3.3.5.8 CATIONS
Composite soil samples were collected for cation analysis in July 2009, which were homogenised, sieved and dried at 80ºC overnight. Cation concentration was measured using inductively coupled plasma atomic emission spectroscopy (ICP-AES, Agilent 7500c Series, Wokingham, UK). Edgell (1988)‟s method was followed to determine the concentrations of the cations 27Al, 24Mg, 44Ca, 39K, 55Mn and 47Ti in the plots (Appendix 3.11, Tables 3B-3D).
3.3.5.9 EXTRACTABLE NITROGEN AND PHOSPHATE
Four samples were collected from each plot every three months from December 2008, with a preliminary baseline test in October 2008, and monthly between May and September inclusively in 2009 and 2010. The soil was homogenised and sieved, then plant available nitrogen and phosphate were extracted (NH4+, NO3-/NO2-, PO4+) using Allen (1989)‟s KCl
and Truogs methods (Appendix 3.9). Both extracts were analysed using a Skalar SAN++ CFA.
3.3.5.10 LEACHING LOSSES
Leaching losses of soil nitrogen were measured by installing suction cup lysimeters (SCLs) in every plot in January 2009. The SCLs consisted of a ceramic cup attached to a 3cm diameter pipe installed 50cm deep into the soil, which draws soil water up a tube into a syringe when placed under a vacuum. Soil water was collected monthly from March 2009 to June 2010 and determined the NH4+ and NO3-/NO2- content using a Skalar SAN++ CFA. A sample of
rainwater was also analysed at each timepoint. In the summer periods there was insufficient leachate for measurement.
3.3.5.11 STATISTICAL ANALYSIS
The effect of the climate treatment on soil moisture content was tested using a one-way analysis of variance (ANOVA) on averaged plot data at each time point. Monthly means of temperature and %RH were calculated from iButton data and used in a two-way ANOVA testing climate and functional diversity interactions with block as a factor to find whether the microclimate was affected by the treatments.
When variables were measured on more than five occasions, a repeated measures ANOVA (RMANOVA) was carried out to test for effects of time and interactions of time with treatment. Block was treated as a factor, with plot number (1-56) included as an error term. Month, climate and functional diversity were modelled with a three way interaction that was retained as the maximal model. Data transformations were carried out as before to meet the requirements of ANOVA.
For each ecosystem function response variable an ANOVA was carried out using the generalised linear model function in R2.12.0 (R Core Development Team 2009), at each timepoint measured. Each variable was modelled with block as a factor, the appropriate covariates as described in table 3.1, and the treatments climate and functional diversity, with interaction terms for the treatments and the covariates.
Models were simplified to remove non-significant covariates, and also block where non-significant using likelihood ratio deletion tests to derive the minimum adequate model. The main effects and interaction of climate change and functional diversity were retained (Manning et al. 2004). Likelihood ratio tests compared the more complex model with the simplified one to test for an unacceptable increase in unexplained deviance.
Response variable
Covariate
PAR Soil moisture Soil
temperature Air temperature NEE x x Reco x x ET x x Decomposition x x Mineralisation x x Extractable Nutrients x x Leachate x
Outliers were removed if they were not readily explainable biologically, created a 5% or more decrease in unexplained variance when removed, or substantially altered the model on removal. Percentage cover of vegetation and mass loss percentages allowing estimation of decomposition rates were arcsine transformed to remove the constraints of bounded data. Any other variables that displayed a non-normal distribution were log or square root- transformed to meet the assumptions of ANOVA.
ANOVA tables were generated for each analysis in order to obtain F-statistics of each variable and interaction, and Tukey‟s Honest Significant Difference was used to generate post-hoc comparisons between factor levels, although caution was required when applying this to interactions.