Data provenance

Our Environment - Statement on Data Accuracy and Fitness for Purpose

Our Environment contains data that Landcare Research has compiled from information derived from a variety of sources which have different data accuracy and hence differing fitness for purpose. These data may not be complete, correct or up to date. Landcare Research advises that users should expect that ALL data contain error and uncertainty. You should therefore familiarise yourself with the information provided below to ensure you understand the limitations of the data presented on this web site. Landcare Research cannot be held responsible for inappropriate use of data and does not accept responsibility for any actions taken in reliance on the data or consequences arising from their use.

Land Resource Inventory (LRI)

This is a single spatial (polygon) layer with national coverage that contains several physical resource themes: land use capability, lithology, soil, etc. In terms of geographical data accuracy the polygon boundaries were originally mapped at 1:63,360 scale, except where more recent mapping was carried out at 1:50,000 scale in Northland, Gisborne-East Coast, Wellington and Marlborough regions. This is regional scale mapping according to the Land Use Capability Survey Handbook (Edition 3). The minimum polygon size for the smallest area of interest in the LRI mapping was nominally 10 hectares, although some 2% of polygons fall below this threshold. Average polygon size for 1:63,360 scale mapping is 335 hectares, and for the more recent 1:50,000 scale mapping is 98 hectares, reflecting both increased mapping scale and improved standards of mapping.

The LRI is a regional-scale database and caution should be used when using these data at larger scales.

In addition to the geographical accuracy of the polygon boundaries, each resource theme is subject to uncertainties about the attributes (e.g., soil classification). Users should download and familiarise themselves with the relevant sections of the LRIS Spatial Data Layers Data Dictionary to ensure that the data are fit for their intended analysis. This is available for download as an attached document to the NZLRI layers in the LRIS Portal.

Interpretations in Our Environment based on the LRI and therefore subject to its data accuracy are: Erosion Severity, Surface Geology and all five of the land suitability classifications.

25–m Digital Elevation Model (DEM)

Landcare Research has produced a national DEM at 25–m postings to be consistent with the 1:50,000 scale topographic data, including 20–m contours, spot heights, lake shorelines and coastline data available from Land Information New Zealand (LINZ). Our DEM was derived using in-house interpolation software described in Barringer, McNeill and Pairman (2003) and utilised error-checking procedures to minimise inconsistencies in input data (e.g., checking contour elevations to adjacent to spot heights). The final national DEM was tested for accuracy against a selection of GPS and LiDAR data. Error statistics reported in Barringer, McNeill and Pairman (2003) indicate that Landcare Research's 25–m DEM has a root mean square error of approximately 8 metres, mean absolute error of approximately 7 metres, and a standard deviation of errors of approximately 6 metres. Errors appear to be greatest in low relief areas, particularly valleys, where 20–m contours are widely spaced and do not give a good representation of the real surface.

Aspect and slope surfaces derived from the 25–m DEM are currently based on standard ESRI ArcGIS algorithms calculated over a 3x3 cell neighbourhood. The errors described above will propagate through these algorithms to produce errors in the slope and aspect surfaces. Barringer and Lilburne (1997) discuss the general quality of slope data derived from DEMs similar to the national 25–m DEM.

The common practice of assigning elevation, slope and aspect values to specified locations by spatial intersection with DEMs or derivative layers should be treated with great caution. The preferred option should always be to use field measurement of site properties, if at all practicable. While these DEM–based surfaces provide greatly improved definition of surface properties, sensitivity of many environmental systems to topographic variation and the errors inherent in DEMs make them better suited to implementing models rather than as a source of specific data values (see for example Barringer and Lilburne, 2000).

Interpretations in Our Environment based on the LRI and therefore subject to its data accuracy are: Direction of Slope (aspect) and Steepness of Slope.

Land Cover Database (LCDB v4.0)

A New Zealand Land Cover Database has been preprepared three times from imagery collected in 1996, 2001, 2009, and 2012 as part of New Zealand's response to climate change and the Kyoto Protocol. The data used in Our Environment are taken from LCDB v4.0, published in 2014, and are made available by Ministry for the Environment (MfE) under a Creative Commons license. A number of metadata documents are available and those using the LCDB v4.0 should be familiar with this material. These include a User Guide and documents defining Target Classes that can be found on the Ministry for the Environment web site.

Land Cover Database version 4.0 (LCDB v4.0) is a thematic classification of 33 land-cover and land-use classes covering mainland New Zealand, the near shore islands (the Chatham Islands have not yet been mapped for LCDB v4.0). LCDB v4.0 was released in June 2014 and used Landsat 7 ETM+ satellite imagery acquired over the summer of 2012/13. Like LCDB1, LCDB2 and LCDB v3.3, LCDB v4.0 adopted a one hectare (ha) minimum mapping unit (MMU) and to maintain compatibility between datasets for change analysis. Users should be aware of the following issues in using geospatial data derived from imagery with a 15 m spatial resolution:

There was no official accuracy assessment of LCDB v4.0. For LCDB2 a Marlborough pilot study included an accuracy assessment, but this may or may not be applicable to the full dataset. A paper by Brockerhoff et al. (2008) in the New Zealand Journal of Ecology has useful information on class accuracy of LCDB2 (and LCDB1).

The interpretation in Our Environment based on the LCDB v4.0 and therefore subject to its data accuracy is: Vegetation.

Protected Natural Areas (PNA)

This is a parcel-based vector dataset with elements of cadastral land parcels included in the conservation estate, covenants of the Nature Heritage Fund, Nga Whenua Rahui, and QE II National Trust, along with Regional Parks of Auckland, Bay of Plenty, Manawatu, and Wellington. Data accuracy equates closely to standards for property and in many cases will have been subject to land survey with coordinate accuracy well below metre accuracy.

The interpretation in Our Environment based on the PNA and therefore subject to its data accuracy is: Protected Natural Areas.

Threatened Indigenous Environments

The Threatened Environment Classification 2012 has been developed by Landcare Research to help identify areas in which much reduced and poorly protected indigenous ecosystems are most likely to occur. It uses LENZ as a framework for representativeness of the abiotic environment. How much of the potential 'full range' remains in indigenous cover and how much is protected are calculated using the LCDB v4.0 and Protected Natural Areas of New Zealand (PNA-NZ) data layers. The Guide for Users of the Threatened Indigenous Environments classification provides the best available indication of methodology and data accuracy.

The interpretation in Our Environment based on TIE and therefore subject to its data accuracy is: Threatened Indigenous Environments.

Potential Natural Vegetation

Potential natural vegetation (PNV) is a vector layer which provides an estimate of what New Zealand’s vegetation cover would have been in the absence of humans.  It has been generated from largely raster data using statistical tools both for the interpolation of point climate data and the analysis of spatial patterns to reconstruct the likely biological character of New Zealand’s pre-human past. More information is available on the LENZ website or in Leathwick, J. R. (2001), 'New Zealand's potential forest pattern as predicted from current species-environment relationships', New Zealand Journal of Botany 39: 447-464. The accuracy of this dataset is difficult to assess since it portrays a concept rather than something that exists in the real-world.  In addition, because it is generated from a number of input layers which include climate layers derived by interpolation, there is significant potential for complex propagation of errors.  Finally, the original raster data layer has also been converted to vectors with some smoothing and simplification of linework.  For this reason the PNV layer cannot be viewed at scales larger than 1:50,000 scale.

The interpretation in Our Environment based on PNV and therefore subject to its data accuracy is: Potential Natural Vegetation.

Basic Ecosystems (2008)

This layer was derived from three existing data layers: the Land Cover Database 2 (LCDB2) (MfE 2002); the Land Use Map (LUM) from the Land Use Carbon Analysis System (MfE 2008; Dymond et al. 2012); and EcoSat Forests (Shepherd et al. 2002). Indigenous forest classes from EcoSat Forests were combined with classes from LCDB2 to form basic ecosystems classes. Where indigenous forest was mapped by LCDB2, the type of forest was determined from the EcoSat Forests layer. The eight forest types of EcoSat Forests were reduced to three basic types: beech forest; podocarp-broadleaved forest; and mixed beech and podocarp-broadleaved forest. To produce a recent 2008 layer the LUM was used to update indigenous and exotic forest changes since 2002. The mapping was performed using 15 m pixels, which is equivalent to a mapping scale of approximately 1:50,000.

The interpretation in Our Environment based on LCDB2, LUM and ECOSAT data and therefore subject to the accuracy of those data is: Basic Ecosystems.

Wetlands

This wetlands dataset has its origins in the Wetlands of National Importance (WONI) project, which was part of the Sustainable Development Programme of Actions for Freshwaters which had the goal of identifying a list of water bodies that would protect a full range of freshwater biodiversity. The pre-human extent of wetlands was produced using soil information from the New Zealand Land Resource Inventory (NZLRI) and a 15m digital elevation model (DEM) to refine soil boundaries. Current wetlands were defined by combining existing databases including LCDB2 (Land Cover Database version 2), NZMS 260 Topomaps, existing surveys from Regional Councils, Queen Elizabeth II (QEII) covenant wetland polygons, DOC surveys (WERI database), and the 15m DEM, to define a single set of wetland polygons and centre points. All this data was checked against a standardised set of Landsat imagery using the Ecosat technology and where necessary new wetland boundaries delineated. Wetlands were classified into 7 groups at the hydro-class level using fuzzy expert rules.

The interpretation in Our Environment based on WONI, NZLRI, DEM, LCDB2, Topomaps, QEII, DoC and Regional Council data and therefore subject to the accuracy of those data is: Wetlands.

 

Last updated: 25 June 2014