Getting the most from Countryside Survey data

Getting the most from Countryside Survey data

Notes on the downloadable data
Sampling Considerations in the use of Countryside Survey Data
Considerations in the use of Countryside Survey soil data

Notes on the downloadable data

To help preserve their representativeness of the wider countryside the precise locations of CS survey squares is held in confidence by CEH. It has been agreed that the locations of squares will not be identified to external users with any greater precision than 100 square km. It is therefore not possible for users to identify whether or not any survey squares fall within defined areas below this threshold and therefore these areas are greyed out in dropdown menus.
If a survey square falls on the boundary between areas it is included in all areas in which more than 25% of the square falls.
CEH has not systematically validated the quality of the digital masks used to define geographic regions and they have therefore been used as supplied. There may be some inconsistencies due to differences in the details of masks supplied and this could result in some squares falling the wrong side of area boundaries resulting in them being wrongly excluded or included in searches.
Please note that the vegetation data accessible at the individual survey square level is supplied as collected by the field surveyors. Correction factors were applied to these data during analysis (see CS Technical Report on Quality Assurance) which must be used if you wish to analyse the data in a manner consistent with the results presented in the UK, England, Scotland and Wales Reports.

Sampling Considerations in the use of Countryside Survey Data

Countryside Survey (CS) Field Survey data comprise information collected from a sample of 1 km squares in GB. Each selected square is mapped and detailed measurements made of selected features, for example a number of quadrats are laid out and used to collect additional information on vegetation, soils etc. Thus there are two levels of sampling, whole square and within-square. Measurements are made at both levels so that some characterise the square while others describe features within the square. Measurements are of varied types ranging from binary (yes/no) variables to continuous variables such as areas or lengths.

CS sample squares are not a random subset of the set of all squares in GB and this has implications for any analyses. Most importantly the sample is a stratified one with sub-samples of squares selected within designated strata. The strata used for square selection are defined by the ITE Land Classification. The details of the classification have changed somewhat from its original form, largely as a result of the need for separate country reporting. Originally the classification comprised 32 Land Classes. In 1998, due to the requirement for separate reporting in Scotland, the classification was modified to contain 42 classes. In 2007 the requirement for Wales only reporting led to further revision of the classification resulting in 45 land classes. Effectively each country now has a separate classification, 21 classes in England, 8 in Wales and 16 in Scotland. Estimates derived from CS data without taking account of the stratification may not be representative and will have inaccurate estimates of variation.

An additional complication associated with the sampling procedure is that not all squares in GB were considered for field survey. Any square whose area, as measured from 1:250000 scale OS maps, was more than 90% sea or more than 75% urban was excluded from survey (see Countryside Survey 1990 Main Report (Barr et al., 1993) for exact details). Strictly speaking the field survey data is only representative of the squares in GB that meet the above criteria. In practice estimates for the whole of GB, or regions of it, are made under the assumption that vegetative land in excluded squares is similar in composition to that in the sampled squares. Although this is unlikely to be completely true the total amount of land concerned is very small and the resulting bias likely to be negligible in general. Only if a region under consideration contains a high proportion of sea or urban squares is a problem likely to occur.

As a result of this sampling design, official estimates from CS are produced by calculating ratio estimates (Cochran, 1963) for each land class, taking into account the area of vegetative land in each sample square. Land class estimates are then combined using as weights the area of vegetative land in each land class as a whole. From 1998, because of concerns about the skewness of some of the features being estimated, standard errors and confidence intervals have been estimated using the bootstrap (Efron and Tibshirani,1993).

References

Barr, C.J., Bunce, R.G.H., Clarke, R.T. , Fuller, R.M., Furse, M.T., Gillespie, M.K., Groom, G.B., Hallam, C.J., Hornung, M., Howard, D.C., and Ness, M.J. (1993). Countryside Survey 1990 Main Report, DETR, London.

Cochran, W.G. (1963). Sampling techniques (2nd ed.). Wiley & Sons; London.

Efron, B. and Tibshirani, R.J. (1993). An introduction to the bootstrap. Chapman and Hall; London.

Considerations in the use of Countryside Survey soil data
Sampling design

Soils were sampled in CS in 1978, 1998 and 2007. In both 1978 and 1998 the same 256 squares were sampled, whilst in 2007 soils were collected from all 591 CS squares – but not all measurements were made on all samples. Power analyses were carried out to determine the number of samples required to detect significant change (see Emmett et al. 2008).

Sample location

Soils are sampled from the CS x-plots, and there are 5 of these plots randomly spaced in a CS square. The x-plots within a square are NOT replicates, as they may be in very different land uses, on different soil types etc. Over the years, x-plots are sometimes relocated due to the destruction of a plot (e.g. it’s turned into a car park) or permission to access the land is denied, or the lack of certainly in relocating the original plot location. However, each location has a unique repeat identifier, e.g. 2RPT1, which refers to square 2, repeat plot 1, where repeat plot 1 is a known location in the square. Samples with the same unique identifier should be considered to come from the same location, although the exact locations from which soils were sampled will be ca. 2 - 3 m apart as we move around the 4 corners of a permantly marked 2m x 2m plot at each sampling location. See Emmett et al. (2008) for further information.

Sampling

Soils are only collected from the top 15 cm (8 cm for the invertebrate sample) of the soil profile, in 1998 and 2007 this was done using a soil core hammered in to the soil and then pulled out. In 1978, a soil pit was dug and soil collected from the top 15 cm of the profile in the side of the pit. In some soils, it is not possible to collect a full core due to stones, or shallow soils. Core photographs and detailed measurements of core dimensions were made by lab staff in 2007 on the 2 of the 4 cores sampled (black cores and N mineralization cores).

Measurements

The only laboratory measurements made in 1978 were pH and loss-on-ignition (LOI). In CS1990, some soil mapping was carried out. Multiple cores were taken from each x-plot in 1998 and 2007, and different cores have different measurements made on them (Table 1). No chemical analyses are made by soil horizons – samples are homogenised before pH and LOI measurements. Bulk density was measured in 2007, but not by using the ISO bulk density method (which would have required a whole core simply for this measurement). A report on the implication of this and a full description of methods is available (Emmett et al. 2008).

Statistical analysis

Statistical analyses of CS soils data should take note of the statistical issues outlined above. Soils results do not always take account of the land class structure. If the land classes are not taken into account, then results are the stock and change of the population, whilst if the land classes are taken into account, the results are weighted by the land classes and results become national estimates of stock and change. Likewise, bootstrapping is not always used in analyses of soils data. However, the structure of CS should be taken into account and at the least a mixed model should be used, with square as a random factor, to take into account the fact that multiple x-plots (up to 5) are contained within each CS square. Finally, in contrast to some monitoring programmes only one core is taken per location but it should be noted that design based approaches such as CS, where spatial structure is not generally of interest, a dispersed set of observations is the most efficient approach and overall precision and coverage is increased by widespread placing of samples if regional estimates are required. We advise all users of CS soils data to consult with CEH prior to analysis and interpretation and to check on publications already released which exploit the data which are listed on the CS website.

Table 1 Measurements on the different soil cores taken in CS2007.
Core Measurements made No. of samples
Black (15 cm deep x 5 cm ID) LOI All 591 squares
pH (in water & CaCl2) All 591 squares
%C * Original 256 squares
%N * Original 256 squares
Bulk density All 591 squares
Core dimensions All 591 squares
Photograph All 591 squares
Depth of organic layer * All 591 squares
Hand texture * All 591 squares
Soil moisture * All 591 squares
Olsen P * Original 256 squares
Metals * 2 of the 5 x-plots in each of the original 256 squares
Long white (15 cm deep x 4 cm ID) Mineralisable N * 3 of the 5 x-plots in each of the original 256 squares
Short white (8 cm deep x 4 cm ID) Soil invetebrates (mainly collembola & mites) * 3 of the 5 x-plots in each of the original 256 squares

* data only released after the Soils Report published in November 2009

References

Emmett, BA, ZL Frogbrook, PM Chamberlain, R Griffiths, R Pickup, B Reynolds, E Rowe, P Rowland, D Spurgeon, J Wilson, CM Wood. Countryside Survey Soils 2007: Method development, power analyses and protocols. Centre for Ecology and Hydrology Project No. C03042/ DEFRA Contact No. CR0334.