Forest fragmentation

Monitoring sites

We measure indigenous forest fragmentation across the whole Waikato region using the Land Cover spatial database which is based on satellite imagery.

Monitoring frequency

Waikato Regional Council's monitoring of changes in indigenous forest fragmentation is dependent on new releases of the Land Cover Database (LCDB) by Landcare Research NZ. The baseline level of forest fragmentation (1840) is available as a single estimate of pre-European forest pattern using the Regional Indigenous Vegetation Inventory (RIVI) (1840)¹.

Monitoring history

The LCDB (recent) data were derived from SPOT satellite imagery collected over five time periods:

• summer 1996/97
• summer 2001/02
• summer 2008/09
• summer 2012/13
• summer 2018/19

The RIVI (1840) data were derived from:

• early settler accounts and maps
• early aerial photographs and the accounts of settlers
• interpretation based on the current extent of indigenous forests, soils, topography, and bioclimatic zones.

Measurement technique

Waikato Regional Council uses the Land Cover Database (sourced from Landcare Research) to determine the following statistics for each available time period:

• number, size and 'interior' area (>60 metres from the forest edge) of indigenous forest fragments
• distance between forest fragments (<25 ha) and nearest larger (>25 ha) stands of indigenous forest, and
• proportion of the surrounding landscape (within 1 km of a forest fragment) that is in woody vegetation (being of similar habitat type)

The LCDB minimum mapping unit is one hectare, although many smaller patches are mapped. The data are suitable for applications down to 1:25,000 scale. We selected a minimum mapping unit of 0.4 hectares.

We use the Regional Indigenous Vegetation Inventory (RIVI) to estimate indigenous forest fragmentation for 1840. The contracted minimum mapping unit for the 1840 data is 25 hectares, although many smaller patches are mapped, and the data were mapped at 1:250,000.

Footnotes

1. Leathwick, J. Clarkson, B. and Whaley, P. 1995: Vegetation of the Waikato Region: Current and Historic Perspectives. Landcare Research Contract Report LC9596/022. Landcare Research, Hamilton.

Waikato Regional Council measures forest fragmentation using spatial layers from the Land Cover Database. The LCDB Version 5.0 (which includes attributes for vegetation class names at each of the five time periods) was processed to measure the parameters of indigenous forest patches greater than 0.4ha, both within the Waikato region and by district council area (2019 territorial boundary layer).

The LCDB polygons were clipped to the Waikato region land boundary. Forest polygons that span a district boundary were allocated to the district they mostly cover, so that they are not ‘artificially fragmented’ by an administrative boundary. This means that this statistic may over-calculate the index where large continuous forest patches extend throughout and beyond the authority in question, or under-calculate them where large forest polygons are allocated to adjoining districts. Only polygons greater than 0.4ha were retained (to exclude less viable patches). Therefore, the total area of indigenous forest may not match that reported in the Extent of Indigenous Vegetation indicator (which does not exclude polygons less than or equal to 04. Ha). Polygons that adjoin or span the regional boundary are treated as whole polygons, cut along the boundary.

From the attribute data for each polygon, the following are calculated for each district in the Waikato region:

• number of indigenous forest fragments
• mean fragment size
• per cent of indigenous forest fragments that are less than 25 ha (small fragments)
• per cent of total forest area that is within fragments less than 25 ha
• per cent of 'interior' forest (>60 metres from the forest edge)
• amount of forest in small patches that is very isolated (>500 m from an indigenous forest patch >25 ha)
• per cent of fragments that are directly connected to another type of woody indigenous vegetation
• proportion of the surrounding landscape (<1 km from a small fragment) that is in woody vegetation (indigenous forest, indigenous scrub¹ or exotic forest²).

Change in fragmentation is calculated as gross change since European settlement, using the RIVI (1840). We use Euclidean nearest neighbour distance (the shortest straight-line distance between a forest fragment and its nearest neighbour) to measure isolation. The 1840 layer was clipped to the Waikato region 2012 land boundary.

We measured the proportion of the landscape surrounding fragments using FME software. Each Indigenous Forest polygon that was between 0.4-25 hectares was buffered to 1 km. Within this buffer zone the summed area for each of the following vegetation classes was calculated using FME’s AreaOnAreaOverlayer transformer.

• Indigenous forest (any size but excluding the target fragment polygon).
• Scrub and shrubland (any size).
• Exotic forest (any size).
• Indigenous woody vegetation (any size). Indigenous woody vegetation includes the following classes: Indigenous Forest, Broadleaved Indigenous Hardwoods, Manuka and/or Kanuka, Matagouri or Grey Scrub, Sub Alpine Shrubland.

Guidelines and standards

Waikato Regional Council uses the LCDB class 'Indigenous Forest' (minimum area of 0.4ha) for current fragmentation data.

For 1840 data the following classes were used:

Footnotes

1. Scrub classes are defined as LCDB classes: Scrub and Shrubland; Broadleaved Indigenous Hardwoods; Manuka and/or Kanuka; Matagouri or Grey Scrub; Mixed Exotic Shrubland; Sub Alpine Shrubland
2. Forest classes are defined as: Exotic Forest. Note Deciduous Exotic Hardwoods were excluded from the landscape analysis as WRC have determined that the majority of DEH polygons are wetland systems. Nevertheless such sites may also contribute suitable habitat for mobile forest fragment inhabitants, as can exotic plantation forest.

This indicator measures indigenous forest fragmentation within the Waikato region. However, some patches of indigenous forest span our region's boundary. To provide statistics that reflect the extent of native forest in the region, such polygons have been cut along the regional boundary, meaning we are not reporting their true size , or connectivity (for instance, their nearest neighbour may be in an adjacent region). Three districts are not fully within the Waikato region. Data for parts of the Waitomo, Rotorua and Taupō districts located in our region may not be representative of the whole district.

The 1840 vegetation map was created for the Waikato region in 1994 based on the regional boundary at that time. In 2010 the region expanded northwards into part of the former Auckland region, in accordance with the Local Government (Auckland Council) Act 2009. For those newly acquired areas there is no 1840 vegetation data, however at the regional scale the indigenous forest pattern is not likely to have been markedly affected.

The extent of indigenous forest vegetation for the forest fragmentation indicator in the data file is based on whole forest patch, allocating each patch to the district it mainly falls within. Therefore the extent data may over- or under-estimate the true total extent of forest vegetation within a given district. For the same reason, the extent data for each district may not match that in the Extent of indigenous vegetation indicator data file, which calculates total forest extent wholly within the district (cutting through forest polygons that span district boundaries).

The local authority boundary layer is often not a tight match with the coastline boundary used in the LCDB data layer or the physical land boundaries of underlying air photo imagery. As a result some areas of indigenous forest have not been allocated to a district council, but are included in the regional statistics, and for some coastal districts the total amount of indigenous vegetation may be slightly underestimated.

Indigenous forests present in 1840 have been reconstructed using a scale of 1:250,000, compared to a scale of 1:25,000 used by the LCDB for recent data, and the contracted minimum mapping unit for 1840 was 25 ha (although many smaller patches were mapped). Therefore, the 1840 data is likely to under-estimate the number of forest fragments present and over-estimate their average patch size. Because of the likelihood for increased error at finer spatial scales, the 1840 data are not reported at the district or city council scale.

The LCDB mapping scale does not allow us to identify all forms of fragmentation, for example, areas of forest:

• bisected by a road
• perforated by multiple clearings for low density housing.

The LCDB class “Broadleaved Indigenous Hardwoods” (BIH) is described in the Illustrated Guide to Target Classes for LCDB2 as being predominantly taller than 3 m. This would technically (per Atkinson 1985)¹ make these polygons areas of indigenous forest, However the “Table 1: Land Cover correlations between LCDB2 and LCDB3” treats them as scrub. To avoid confusion for future developers of this indicator, the BIH class has been aggregated with the scrub categories. In addition, the LCDB aggregates manuka and kanuka dominated vegetation into a single class, despite the expectation that extensive areas of kanuka are likely to meet the Atkinson definition of forest. Therefore, the statistics for forest fragments will likely underestimate the extent and overestimate the degree of isolation of indigenous forest fragments in the region.

In addition, the Land Cover Database:

• does not always identify indigenous forest patches located within pine forests
• does not separate 'indigenous forest' into different forest types.

Measuring the degree of isolation of forest fragments is difficult to achieve in an ecologically meaningful way. As well as distance between forest fragments, other factors that influence isolation include the:

• length of time since fragmentation
• type of habitat separating indigenous forest fragments, for example, scrub, pine forest, orchards or pasture
• size of the fragment and the size of its nearest neighbour
• the proportion of the surrounding landscape that remains in similar habitat, and
• species of concern, as some are more mobile than others.

Forest fragmentation can actually reduce average distance between patches, for instance, if an isolated forest patch is partially cleared, dividing it into two or more patches relatively close to each other, their nearest neighbour isolation value will reduce. For the purpose of this indicator we report on the proportion of forest vegetation occupying a small (< 25 ha) patch that is relatively isolated (further than 500 m from a large (> 25 ha) patch of indigenous forest vegetation). This is a measure of the amount of indigenous forest habitat that has become isolated. We also report on the average proportion of the landscape surrounding forest fragments that is in similar habitat (woody vegetation). Because different species are affected by isolation in different ways, the isolation metrics are presented to describe trends in pattern over time, not to imply or report on specific effects on biodiversity.

Patches close to the regional boundary may be less isolated than reported, because their nearest neighbour may be outside of the region. Therefore, the isolation values for districts near the regional boundary may be overestimated.

Some forest fragment edges are well buffered by tall scrub or plantation forest. This analysis has not distinguished forest edges under different scenarios, other than to identify those polygons that adjoin scrub for at least part of their boundary (zero distance to nearest scrub). However, the presence of scrub or planted forest indicates the likelihood of earlier exposure to a modified edge, and there may be some lingering biological effects, particularly for plant composition.

It is assumed that over time, through natural succession, areas classified in the LCDB as Broadleaved Indigenous Hardwoods and Manuka & Kanuka will mature and be reclassified as Indigenous Forest. As those patches mature, positive changes, if not counteracted by continuing clearance and fragmentation of native vegetation in other locations, will be revealed by the indicator statistics showing an increase in indigenous forest and corresponding decline in scrub classes.

Data for this indicator were derived from SPOT satellite imagery. Most of the source image data were captured in summer over four time periods: 1996/97, 2001/02, 2008/09, 2012/13 and 2018/19. The data is a snapshot of vegetation at these dates and should not be considered the current vegetation cover.

This indicator updates the previous indicator for forest fragmentation published by the Waikato Regional Council in 2015.

Further indicator developments

As any new versions of the Land cover Database become available the indicator methodology will be reviewed and updated. In addition, district and regional boundary changes will also need to be accounted for.

Quality control procedures

Raw data is quality checked by the data providers and an error estimate calculated. Manual on-screen checks of resultant data were carried out using random checks. Data totals of derived GIS data layers were compared with the source data layers. LCDB polygon classification errors are reported back to the developers in a process of continuous accuracy improvement.

Footnotes

1. Atkinson 1985, Derivation of vegetation mapping units for an ecological survey of Tongariro National North Island, New Zealand, IAE Atkinson - New Zealand journal of botany, Vol 23:361-378.

Clear indicators of fragmentation (splitting up of forest patches), include

• the proportion of indigenous forest that occurs in small patches
• the amount of forest that is found in an ‘edge’ zone
• the amount of forest that is isolated (e.g. more than, for instance, 500m) from a large patch of indigenous forest.

In 1840 over 60 per cent of forest patches were larger than 25 hectares (Graph 1) with an average forest patch size of 13,244 hectares (skewed by a few extremely large areas uninterrupted by scrub or wetland), compared with an average patch size of 54 hectares today. The largest forest patches today are found in Taupo and Thames-Coromandel on the extensive forested ranges (Graph 2). There has been no significant change in average forest fragment patch size since 1996.

Small patches of indigenous forest often have a large amount of forest 'edge' and small forest 'interior'. Native plants and animals living on the edges of small forest fragments are vulnerable to a harsher microclimate, grazing animals, weeds and disease infection. These damaging effects are called the 'edge effect'.

The growing conditions in the forest interior are most likely to represent those found in large forests, such as those present before European settlement. In 1840, 98 per cent of the indigenous forest vegetation in our region was 'interior' forest (more than 60 m from an edge), compared with 81 per cent today. Since 1996 there has been very little change in the proportion of indigenous forest that is considered ‘interior’.

Small patches that are isolated (a long distance from a large patch) are more vulnerable to loss of biodiversity because there is a smaller chance of native animals or plants being able to reach and enrich or replenish them. Cleared and developed land between fragments can be a barrier to native animals and plants moving between fragments, making it difficult for them to find enough food, a mate, or to disperse their seeds to suitable sites.

Forest isolation is measured as the:

• average distance between small fragments and large fragments
• amount of the region’s small fragments of forest that is far away from large patches,
• the amount of land within a 1 kilometre radius of a fragment that is in similar habitat.

Isolation values changed very little between 1996 and 2018.

Small forest fragments (those less than 25ha), in the Waikato region are located an average of 1.9 kilometres from a patch of indigenous forest greater than 25ha in size. However, this does not mean there is a forest fragment spaced every 2 kilometres. Forest fragments tend to be clustered together, for example, along mountain foothills or riverbanks. In heavily modified areas, forest fragments are often far apart and a long way from large forests. The most isolated small indigenous forest patches are in Hamilton City, South Waikato District and Matamata-Piako District, where the nearest large patch of indigenous forest is on average over 3 kilometres away.

Over 71 per cent all of the indigenous forest in Hamilton city occurs within a small, very isolated forest patch (<25ha in size and more than half a kilometre from a large area of indigenous forest). Some of this vegetation may be close to another small patch. Only 23 per cent of the Waipa District’s indigenous forest is in such small isolated stands, while across the region only 4 per cent of the total indigenous forest area occurs in a small, very isolated patch.

Most of the small indigenous forest fragments are surrounded by grazing land. For most small fragments (92 per cent) less than a quarter of their surrounding landscape (within a 1-kilometre radius) is in indigenous forest. A small proportion are directly connected to indigenous scrub or shrubland (19%) which may act to buffer the forest patch, and/or provide additional habitat for forest species or a corridor to another forest patch. Others may be in an urban area or among or adjacent to plantation forest.

The scale of forest fragmentation since 1840 has been dramatic, however in recent years there has been very little change at the regional level. While the Land Cover Database records a net loss of around 500 hectares of indigenous forest since 1996, the amount of the region’s indigenous forest that is in small patches, and fragmentation statistics have changed very little over that time period.

Footnotes

1. Leathwick, J. Clarkson, B. and Whaley, P. 1995: Vegetation of the Waikato Region: Current and Historic Perspectives. Landcare Research Contract Report LC9596/022. Landcare Research, Hamilton.