7.1 Land use/land cover concepts and data

Objectives

This component of the course sets out to:

  • define some key terms, such as the phrases ‘land use’ and ‘land cover’ themselves
  • identify some of the commonly used land use / cover data sets
  • describes some of the main characteristics of land use / cover map layers

Introduction

Land use and land cover are two closely related but separate terms (see the Consortium for Atlantic Regional Assessment’s Land Use Primer in the references below). Land cover refers to the physical substances that cover the Earth’s surface at a given location, such as man-made surfaces like asphalt or concrete, ice, bare earth, or grassland. Land use refers to the economic purpose to which land is put, so ‘nature reserve’ is an example of a land use for given parcel of land. Thus, the same land cover type, for example improved grassland, could be used as a golf course (i.e. recreation) but also as grazing for livestock (i.e. agriculture). Land use and land cover map layers are typically made up of land parcel categories, known as classes. As such, they are a form of nominal or categorical data, in which individual land parcels have labels to indicate which class they belong to. The parcels cannot be ranked or ordered on the basis of the land use/cover classes.

Data sources for land use / land cover

Land cover is commonly measured through remote sensing, but also sometimes via ground survey. Land use is more typically measured via surveys of land ownership. In many countries, one of the principle methods of collating land use information is via an agricultural census (see below).

Both land use and land cover data often have a minimum mappable unit. From a remote sensing perspective, this is usually constrained by the choice of source imagery. Thus, imagery from the Spot panchromatic sensor can only be used to discern land cover patches that are larger than 10 by 10 metres, unless highly sophisticated super-resolution mapping techniques (Tatem et al, 2002) are used to process the imagery. Similarly, the scale of a land use/cover product will also influence the minimum mappable unit. At coarser scales (e.g. 1:x200,0000), small but neighbouring patches of similar land use/cover may be merged together through a process of map generalisation, so the level of detail is appropriate for the intended scale of use. The ArcGIS tool ‘aggregate polygons’ is an example of one tool that can be used to control the level of detail in such a map so that it is appropriate for a given intended scale of use. In order to describe patches of land that are comprised of many small-scale patches of land that are too small to map, heterogenous classes sometimes have to be used as a consequence (e.g. cropland with woodland). A similar constraint can apply to land use. Many agricultural censuses implement a minimum land holding size, below which land owners are not required to return a census form.

Many land use / cover classifications are hierarchical. In a hierarchical classification, a small number of Level 1 broad classes exist, that can then be broken down into a more detailed set of Level 2 classes. In the classic Anderson land use / cover classification (see the CARA web site below), ‘1. Urban or Built-up Land’ forms one of nine classes at the top level. This broad class can then be broken into Level 2 sub-classes, such as ’11. Low density residential’, ’12. Medium density residential’, and ’15. Industrial’. Some complex classifications may extend to three hierarchical levels.

Whilst in principle land use and land cover are distinct conceptually, in practice many classifications combine elements of both land use and land cover. Thus, the GlobCover category 14 ‘Rainfed croplands’ implies both a land cover (crops) and a land use (arable agriculture). It is also worth noting that there is often ambiguity in the way that many apparently straightforward land cover concepts are defined. Barrett (2001), for example, examines the definition of ‘forest’ and notes that in fact, many different working definitions of the term ‘forest’ are currently in use. The term ‘forest’ itself is thus ‘fuzzy’ – in other words there is ambiguity surrounding the concept of a forest. This ambiguity produces definitional mismatches between different land use / land cover products and is a major issue in comparing land use / cover map layers from different sources.

Table 1. A selection of land use / land cover data products

Product Spatial coverage Web link
GlobCover Global http://due.esrin.esa.int/page_globcover.php
Land Cover Map Great Britain http://www.ceh.ac.uk/LandCoverMap2007.html

https://www.ceh.ac.uk/services/land-cover-map-2015

CORINE Europe http://www.eea.europa.eu/publications/COR0-landcover

http://www.eea.europa.eu/data-and-maps/data#c12=corine+land+cover+version+13

National Land Cover Database Conterminous USA http://www.mrlc.gov/
Global Land Cover 2015 Global

http://lcviewer.vito.be/

https://land.copernicus.eu/global/products/lc

Indian agricultural census India http://agcensus.dacnet.nic.in/
ESA CCI (Climate Change Initiative) Land Cover dataset Global http://cci.esa.int/data

Case study: The Agricultural and Horticultural Census of England and Wales

This census takes place annually in June and is compulsory for all farms above a certain size. In terms of the data collected, each farm provides information via a questionnaire (either online or on paper) on number and type of livestock, yield and area under different types of crop, the nature of the agricultural labour force, and farm size. As with population censuses, agricultural census data are aggregated into areas to preserve confidentiality. In England and Wales, the data are aggregated by agricultural parish. Before release for use, data that have been aggregated by parish are then superimposed on to a 1 by 1km land use grid. This grid is used to distribute the parish-level census results. For example, if a parish had 1000 Ha under wheat cultivation and was comprised of 10 separate 1 by 1km grid squares coded as ‘agriculture’ land use, each of these ten grid squares would be coded as having 100 Ha of wheat cultivation (see links below). This coding schema (designed to preserve confidentiality) makes the publicly released version of the data suitable for small-scale studies, but less suitable for more detailed studies of local areas.

The data collected have changed relatively little since the 1960s, so the data afford an opportunity to look at change over long periods of time. Similarly, unlike many other UK administrative boundaries, agricultural parish boundaries have changed relatively little over the past 30 years. In contrast to some remotely sensed land use / cover data sets where both sensor technology and digital image processing techniques have steadily improved since the 1980s, methodologically, the agricultural census has remained relatively stable and figures broadly comparable over time.


Activity

Explore an example of a land use/cover data product via the links in the table above. If you know of a land use / cover product that is not listed above but is of particular interest to you, feel free to explore that data product instead. For your chosen example data product, see if you can identify:

  • The extent to which it depicts land cover, land use or both.
  • The minimum mappable unit and number of heterogenous (mixed) land use/cover classes.
  • How many hierarchical levels the classification has.

Quiz

Land use/land cover concepts and data


References (Essential reading for this learning object indicated by *)

The Consortium for Atlantic Regional Assessment’s Land Use Primer contains useful conceptual background on land use / cover – see the first five sections of this web site in particular:

http://www.cara.psu.edu/land/landuseprimer.asp

Section 1.2 of this web site describes minimum mappable units in the context of the European Corinne land cover classifcation: http://ec.europa.eu/agriculture/publi/landscape/

For further detail on the agricultural census in the UK, see the DigiMap web page on this product (note: unfortunately University of Southampton do not subscribe to this data product): http://agcensus.edina.ac.uk/description.html

This paper is one of several that describe the idea of super-resolution mapping:

Barrett, B. (2001) What is a forest? On the vagueness of certain geographic concepts. Topoi 20, 189-201.

Tatted, A. J., Lewis, H. G., Atkinson, P. M., and Nixon, M. S. (2002) Super-resolution land cover pattern prediction using a Hopfield neural network. Remote sensing of Environment 79 (1), 1-14.

Comments are closed.