Module 8 Use of the Land-Use Change Data Derived from the Countryside Surveys of Scotland and Wales to Provide Predictive Estimates for Changes to C and N Balance in Organic Soils Over Time
Changes in land use have shaped the landscape of the UK over many centuries and there have been significant changes in land use even in recent decades. The rates and types of land use change over time are produced by the interaction between the physical environment (for example, climate and soil type limit certain types of land use) and social/economic forces governing the relative value of land under different usages (for example, tax concessions encouraged afforestation during the 1970s and 1980s on land previously used for extensive grazing). Assessments of national land use change are of interest for development and agricultural policy, for habitat conservation and, more recently, for the estimation of greenhouse gas emissions and carbon storage. Land use change is responsible for both emissions and removals of greenhouse gases, due to the associated changes in vegetation inputs and turnover and especially in soil carbon stocks. Land use change is one of the main contributors to the Land Use, Land Use Change and Forestry sector in the UK National Greenhouse Gas Inventory, the only sector that is a net sink for greenhouse gases (Baggott et al., 2006).
8.1.1 The matrix approach for assessing land use change
At any point of time, an area of land can be classified as being of a particular land use type. Many different classifications are possible but for Greenhouse Gas Inventory ( GHGI) purposes the IPCC Good Practice Guidance for Land Use, Land-Use Change and Forestry ( IPCC, 2003) defines five types useful in the UK: Forest land, Grassland, Cropland, Settlements and Other Land (the IPCC Wetlands category is included in either Grassland or Other land for the UK). For the ECOSSE project the Grassland category is split into Natural Grasslands and Managed Pasture. Between points in time, areas of land may remain in the same category of use, or need to be reclassified because of a change in use. The total areas changing can be summarised in a matrix ( Error! Reference source not found.). The diagonal cells of the matrix represent no change in categorisation (i.e. use) of land. The matrix gives the areas and categories of land that have undergone land use change during a period: needed for estimating changes in soil carbon stocks, and hence GHG emissions.
Table 8.1 Example of a land use change matrix (Scotland 1990-1999). Data are annual changes (000 ha) in land use from the 1990 and 1998 Countryside Surveys.
8.1.2 Sources of information on national land use change
There are three national datasets on land use change that cover the period from 1950 to the present.
- The Monitoring Landscape Change ( MLC) project (Hunting Technical Services Ltd, 1986) which assessed land use change in England and Wales between 1947, 1969 and 1980 using aerial photography.
- The National Countryside Monitoring Scheme ( NCMS) (Mackey et al., 1998) which assessed land cover change in Scotland between 1947, 1973 and 1988 using aerial photography.
- The Countryside Surveys (Barr et al., 1993; Firbank, 2003), which are national ( GB) field surveys managed by the Centre for Ecology and Hydrology. These are available for 1984, 1990 and 1999.
Each of these data sources uses a different land classification system, and the original classes must be grouped into the land use categories used for the GHGI to allow comparison between datasets and over time. The groupings for each dataset are given in Tables A3.1-A3.3 in Appendix 3.
Previously, only country-level (England, Scotland, Wales and Northern Ireland) land use change matrices for the GHGI have been developed. This approach is unsatisfactory as land use change varies greatly within these countries. The available data sources are all sample-based surveys so although it is possible to develop matrices at finer scales from these sources, there is an issue of statistical validity: at very fine scales the matrices will be too uncertain to be useful. The target scale chosen here is 20km x 20km, as this is the scale found to achieve an acceptable balance between detail and accuracy in other components of the GHGI.
For Scotland and Wales (England also) measured land area and change data are available from surveys taken in 1947, 1969/1973, 1980 (England and Wales only), 1984, 1990 and 1998. Measured land use change data over the different periods were used to estimate annual changes by assuming that these were uniform across the measurement period, e.g. the period 1980-84 was filled using data from the Countryside Survey ( CS) assuming the same annual rate of change as seen for 1984-90. The period 1999-2003 was extrapolated forward from the CS assuming the same annual rate of change as seen for 1990-98. Another CS is planned for 2007-2008, which will allow the land use change estimates to be updated in the future.
8.2 Methods and results
8.2.1 Development of land use matrices 1950-1980 for Wales.
Matrices of change at the national (Welsh) level and areas of different land use types (in 1947, 1969 and 1980) at the Welsh county level are available in the MLC reports (volumes 1, 3-6 and 10) (Hunting Technical Services Ltd, 1986). By combining these sources of information it is possible to produce estimated land use change matrices at the county level ( Appendix 3; Figure A3.1a). The MLC data were only available in the paper reports, so the areas and standard deviations of the MLC land use classes in each county in 1947, 1969 and 1980 were typed into spreadsheets to enable further analysis. A new spreadsheet was constructed that compiled the land use area information for all counties and for Wales as a whole. National matrices of land use change (1947-69 and 1969-80) using the inventory land classification had been constructed in a previous MSc project (Hodgson, 2004). These tables were used to construct national matrices of percentage change for each time period, i.e. the percentage contribution from each land class at t= i to the area in that land class at t= i+1 ( Error! Reference source not found.).
Table 8.2 Percentage contribution matrices of change for Wales for (a) 1969, and (b) 1980.
(a) Land class in 1947
Land class in 1969
(b) Land class in 1969
Land class in 1980
The county area data and the national matrices of change were used to estimate land use change at the county level in the two time periods ( Error! Reference source not found.). The data for the year at either end of the time period can be used for the estimation of matrices of land use change during that period, e.g. either the 1947 areas or the 1969 areas can be used to estimate change in 1947-1969. After an error comparison, the end year was chosen as the baseline for calculation as this generated slightly smaller errors overall.
Where L is land use change in county m in time period t= i? i+1, j is the land class in t= i, k is the land class in t= i+1, A is area and W is land use change in Wales in time period t= i?i+1.
Land use data are disaggregated to the 20km scale by splitting the land use in each county in proportion among all the 1km cells in that county and then aggregating the 1km cells into the appropriate 20km squares. The different spatial patterns of land use change can be seen in Error! Reference source not found.. The increased spatial and temporal precision in the estimation of land use change results in larger estimates of change to land classes at the national scale and different rates of change between decades ( Error! Reference source not found.). Some of these differences are due to interchange between land classes during the longer time period. The greatest rates of change (>2kha a -1), in order of size, are: N?P, P?C, C?P in the 1950s; P?C, N?P and C?P in the 1960s; and P?C, N?P, N?F, P?F and P?S in the 1970s. (N=Natural, P=Pasture, C= Cropland, F= Forest land and S= Settlement). The rates of change increase over time for all land classes except for Pasture, where rates decline over time.
Table 8.3 Comparison of annual land use change per decade in Wales 1950s-1970s, 000 ha a -1
Change to Land Class
20km sum estimate, 1950s
20km sum estimate, 1960s
20km sum estimate, 1970s
National estimate, 1950s-1970s
000 ha a -1
8.2.2 Development of land use matrices 1950-1980 for Scotland
Land use matrices for Scotland for 1950 to 1980 were developed using the database of the National Countryside Monitoring Scheme ( NCMS) project (Mackey et al., 1998). The NCMS database can be analysed using user-defined geographic regions using an extension (the NCMS Visualisation and Analysis System) in the ArcView 3.2 geographic information system. Analysis regions were developed from groups of LAU1 administrative areas ( Appendix 3; Figure A3.1b), 22 regions in total, which are similar in size to the county regions used for Wales.
Each region was analysed for the two time periods 1947-1973 and 1973-1988, using a 95% confidence interval and minimising dependence on squares outside the region. The analysis outputs are area estimates for each land class at the beginning and end of the time period and interchange area estimates between land classes, i.e. land use change matrices. The individual regional outputs were collected into a single spreadsheet. The resulting matrices were assigned to 20km cells in the same way as described for Wales.
Figure 8.2 Pattern of NET land use change in Wales in the 1950s, 1960s and 1970s, kha/decade/20km
The new matrices for Scotland have been derived from a different dataset ( NCMS) than previously used matrices ( MLC). There is broad agreement in the total annual change ( Error! Reference source not found.), although there is much variation in change between land classes. There is more change to Forest land in the NCMS matrices, largely due to increased change from the Natural class, although there is also less change from the Pasture class. There is less change to Pasture in the NCMS matrices, due to less change from Forest land and Natural classes. The annual change in the Cropland class is similar in both datasets in the 1950s and 1970s, but much greater in the NCMS matrices in the 1970s due to greatly increased change from the Pasture class. The differences between the other classes in the matrices are smaller. As in Wales, there are increased rates of change in all classes in the 1970s when compared to the 1950s and 1960s. The different spatial patterns of land use change can be seen in Error! Reference source not found..
Table 8.4 Matrices of annual land use change per decade in Scotland 1950s-1970s
Change to Land Class
NCMS 20km sum estimate, 1950s-1960s
NCMS 20km sum estimate, 1970s
NCMS national estimate, 1950s-1960s
NCMS national estimate, 1970s
MLC national estimate, 1950s-1970s
000 ha a -1
8.2.3 Development of matrices 1980-present
Countryside Survey data were used to estimate land use change in Scotland and Wales from 1980 to the present. These data are based on detailed field observations, collected in ~ 500 samples of 1 km grid squares across Great Britain (Barr et al., 1993; Firbank, 2003). The sampling is stratified using a system of classifying geographical locations depending on many mapped attributes (' ITE Landclass') so the ecological and land use data from the samples can be scaled up to the national or other level.
In order to calculate the land use change matrix for each 20x20 km grid cell it was first necessary to generate the numbers of 1x1 km cells of each Landclass within each 20x20 km cell. A weighted average for each 20x20km cell of areas of land use change of the field results for sites of known and appropriate Landclass then provided the land use change matrices. These matrices were calculated separately for the two periods of 1984 to 1990 and 1990 to 1998. Error! Reference source not found. shows examples of the mapped land use changes at 20km scale.
Figure 8.3 Pattern of NET land use changes in Scotland in the 1950s-60s and 1970s, units in kha/decade/20km square
Figure 8.4 Annual NET land use change in the 1980s, 1990s and 2000s in the UK, units in kha/year/20km square
8.2.4 Relationship to soil type
In the Defra Soil Carbon Database (Bradley et al., 2005) the area of the each of the five most common soil series has been recorded for 1km grid-cell in GB. The soil series have also been classified into four broad types: Peats (here termed organic), Organo-mineral, Mineral and Other ( Error! Reference source not found.). These data have been co-registered with the land use change maps using the unique 20km cell identifier codes. When the organic/organo-mineral soil distribution map is used to mask the land use change maps the comparative distribution of land use change on these soil types becomes apparent ( Error! Reference source not found.). As the soils dataset is at 1km scale and the land use change dataset is at 20km scale, comparative analysis must be careful, as land use change at the 20km scale may not be representative of land use change occurring specifically on organic soils at the 1km scale.
Figure 8.5 Distribution of peats (here termed organic) and organo-mineral soils in Scotland and Wales
The new land use change matrices and maps produced for Module 8 are a great improvement on the previous estimations of land use change at the national scale. The increased spatial and temporal detail results in better estimation of where and when land use change has taken place in the past. This work also has implications for the calculation of carbon fluxes resulting from changes in soil carbon stocks, and the improved matrices will be incorporated into existing models for calculating land use change fluxes for the UK national greenhouse gas inventory.