ANZLIC Metadata Profile: An Australian/New Zealand Profile of AS/NZS ISO 19115:2005, Geographic information - Metadata
Metadata Standard Version
1.1
Reference System Info
Reference System
Reference System Identifier
Identifier
Code
2193
Identification Info
Data Identification
Citation
Citation
Title
National Irrigated Land Spatial Dataset 2017
Date
Date
Other Citation Details
Aqualinc Research Ltd
Abstract
A spatial dataset of the extent of irrigated land in New Zealand, categorised by irrigation system type (where possible) has been created. Mapping the spatial distribution of irrigated areas and irrigation system types represents a substantial improvement on previous estimates of irrigated area, which only provided a total area for the region or district.
Purpose
To provide the first nationally consistent estimates of the spatial extent of irrigation at fine (sub-property) scales
A spatial dataset of the extent of irrigated land in New Zealand, categorised by irrigation system type (where possible) has been created. Mapping the spatial distribution of irrigated areas and irrigation system types represents a substantial improvement on previous estimates of irrigated area, which only provided a total area for the region or district.
All regions of mainland New Zealand, except for Nelson.
Snapshot of irrigated area in 2017, primarily based on data sources from 2015 – 2017.
The methodology that we have used to create the irrigated area data-set combines a number of data sources, including remote sensing data and resource consents database information. Integrating these data sources, along with Aqualinc’s expertise in irrigation design and management, has allowed the irrigated area to be mapped at a farm scale. The methodology for mapping irrigated areas involved the following steps: 1. Farm boundary extents. This step involved mapping the approximate extent of farm boundaries using land ownership and title GIS data from LINZ. 2. Irrigation systems clearly visible from aerial imagery. Wherever possible, irrigated area was mapped based on the irrigation systems viewed from high resolution aerial or satellite photos (preferably 0.5 m pixel or less). The system type was estimated by considering a range of factors including visual sighting of travelling irrigators, and markings on the ground, such as wheel tracks or irrigation patterns. If more than one set of images were available, a cross-reference was made between the imagery. In regions where there is a strong contrast between irrigated and non-irrigated land, this process typically identifies about 80-90% of the irrigated area with a high degree of accuracy. 3. Resource consent data. The farm boundaries layer (step 1) was combined with land slope and resource consent data (surface-water takes, ground water takes, and irrigation scheme command areas). This process identified farms and areas with resource consents to take water for irrigation. Such areas with land slope less than 15° were considered to be potentially irrigated. A spatial dataset of active water take consents for all regions was provided by MfE. This dataset included attributes such as water source, use type, maximum rate and annual volume. For some regions, raw consents database records were also available. 4. Multispectral satellite analysis. GIS layers of normalised difference vegetation index (NDVI) imagery were created from Landsat imagery, covering dry summer months from January to March where possible. A strong contrast between the NDVI values for dry and actively growing vegetation indicates areas that are likely to be irrigated. As discussed below, this method is more successful in some regions than in others.
5. Combine irrigation consent and NDVI analysis. We combined the results from steps (3) and (4) to map irrigated areas that could not be identified in Step (2). We manually mapped these areas, giving consideration to irrigation design and farm boundary limitations. In regions where there was not a strong contrast between irrigated and non-irrigated land in the aerial images and NDVI data, judgement was applied based on the available data sources to determine the area that was likely to be irrigated. As discussed in more detail below, the mapping accuracy in these regions was variable, and is generally expected to be lower than the regions with high contrast. 6. StatsNZ survey. To benchmark the accuracy of the mapping, we cross-referenced the total mapped area for each region with estimates of the total area equipped for irrigation from the June 2012 Agricultural Production Statistics (APS) (StatsNZ, 2013).
We did not necessarily follow a linear progression through all of the above steps. Depending on the availability and quality of data for each region, more weight was put on some steps than others to draw a conclusion on whether an area was is likely to be irrigated. For example, in regions where there was little contrast between irrigated and non-irrigated land in the aerial photos and NDVI imagery, areas within a farm boundary extent were assumed in most cases to be irrigated if an active consent existed within the property boundary. Where irrigation systems were clearly visible it was usually not necessary to refer to resource consent data or NDVI imagery. Centre-pivots in particular can often be identified visually, from their wheel tracks or by sighting the pivot itself, even where there is little contrast between irrigated and unirrigated land. The APS data includes only agricultural activity. The areas that we have mapped include some non-agricultural irrigation, such as golf courses.
Mapping accuracy varies between regions, depending on climate. In some areas, identification of irrigated land and irrigation system type is difficult due to the lack of visual contrast between irrigated and non-irrigated land. Land used for short-rotation cropping may not be identifiable as irrigated if no crop was actively growing when the aerial photo was taken. Apart from Canterbury (which had been mapped previously for Environment Canterbury and included with this dataset for completeness), and the Takaka catchment, the scope of this project has not allowed for primary-sector validation to be undertaken. Any errors in the Councils’ consent data that was provided for the project may result in errors in the mapping.
