The Land Cover Database (LCDB) is a digital thematic map of land cover designed for use in a GIS or printed map.

LCDBI was completed in June 2000 and used Spot II satellite imagery from 1996/1997. It used 16 land cover classes covering artificial, cultural and natural classes in most regions with a 17th class (riparian willows) added in some regions. Overall classification accuracy is 93% although this varied between cover classes. The minimum mapping unit was 1ha.

LCDB2 will use Landsat 7 satellite imagery from 2001/2. The land cover classification will be expanded in a way that allows backward comparability of land cover information.

A land cover change analysis using the satellite images will be an essential component of the LCDB2 preparation process.

The following classes were used in LCDB1:

Artificial landscapes:

• urban
• mines, gravel pits and dump sites
• urban open space, e.g. sports fields, parklands

Cultural landscapes:

• primarily pastoral including arable land
• primarily horticultural
• planted forest

Natural landscapes:

• grasslands- tussock or unenclosed grassland
• shrubland-woody vegetation in which the cover of trees and shrubs in the canopy is >20% and where the shrub cover exceeds that of any other growth form including bare ground
• indigenous forest
• bare ground
• inland wetlands (freshwater)
• riparian willows
• coastal wetlands (saltwater)
• coastal sands
• inland water (lakes, ponds and rivers)
• mangroves

A more detailed classification will be used for LCDB2.

• change in area in indigenous forest and other land cover classes.

Derivative products include:

• forest biomass (need vegetation carbon storage information)
• change in area in each land cover class
• forest area by tenure (need tenure information).
• fragmentation analysis (eg. average size of indigenous forest/vegetation remnants)

The images are orthorectified and the cover classes are classified manually. The boundaries of the cover classes are superimposed on satellite images that are then used for field checking. In each area every mapped land cover class is inspected for mapping accuracy. These inspection results are used to correlate areas of similar signature.

LCDB1 images were orthorectified using the 20m digital terrain model. LCDB2 will be orthorectified, sharpened to 15m spatial resolution, and the topographic effect will be removed.

Vectors from LCDB1 will be overlaid on the new imagery. The pixels that deviate from the LCDB1 classification will be highlighted.

LCDB1 classification and generalisation errors will be corrected and the actual land cover changes between the two satellite images will be confirmed.

As with LCDB1 the draft land cover classes will be field checked and an accuracy assessment undertaken.

• The database is GIS compatible and uses the NZ map grid. This allows it to be interrogated.
• The data is verifiable (unit boundaries are superimposed on the satellite imagery)
• The scale is a useful one for national and international projects and reporting.
• The database can be used in combination with other data sources to answer a wide range of questions.
• The database is used for international reporting purposes.

• The New Zealand topography makes it difficult to classify land cover classes automatically. This means that most classification needs to be done manually.
• New Zealand's climate and the relative infrequency of satellites being in a position to take suitable images of a particular area mean that it is difficult to obtain complete satellite coverage for NZ within one summer season. Extra costs are involved where aerial photographs need to be used to provide detail where cloud is present.
• The minimum mapping area of 1ha means that smaller areas of indigenous vegetation are not mapped.
• The classification system is not mutually exclusive and exhaustive as it contains a mixture of land cover classes, land use classes and hydroclasses. Some areas can fit into more than one of the defined cover classes.
• Estimates of wetland extent are not reliable. Some types of wetland vegetation may be mapped as scrub and others are mapped as wetland.
• There has not been good publicly available descriptions of the cover classes. This had led to some inappropriate uses of the database. e.g. the extent of coastal sands (<10% vegetation) is primarily affected by the state of the tide. Some analyses have interpreted coastal sands to mean dunelands, including vegetated sands.
• The desire for backwards compatability with LCDB1 is constraining the development of an appropriate classification for further editions of LCDB.