Habitat map
Public Sector

Habitat Map of Wales Resurvey

Natural Resources Wales (NRW) commissioned Keystone’s GIS team to undertake an assessment of remote sensing techniques as a means of updating the Habitat Map, which is based on surveys undertaken between 1979 and 1997.

Natural Resources Wales
2023
Rob Frith

A Point Sample Analysis covering 30 x 4 kilometre (km) 2 tetrads was used to assess differences in habitats recorded during field surveys completed in 1979-97 and 2022, and remote sensed mapping generated in 2022 (i.e. The Living Wales Habitat Map 2022, and Neath Port Talbot’s remote sensed Phase 1 habitat map (Environment Systems 2021)). Alongside this, the extent of each habitat type recorded was also compared.

The results of the study highlighted that some remote sensing methodologies can provide high overall accuracy in the classification of broad habitat types (namely heathland mosaics and woodland) and approximately 90 % accuracy for open habitats (grassland, heathland, and wetland). However, ground survey is required to provide a greater level of accuracy and detail, in particular for open, priority, and sub-category habitat types.

Challenge

  • Minor differences in habitat boundary mapping between datasets, and minor mapping errors that could cause a skew in the data analysis were managed through negatively buffering the GIS habitat polygons by 10 m to create a separation of 20 m between neighbouring polygons in the same dataset. Although this buffering removed most of the boundary errors from the Phase 1 habitat survey datasets and the Neath Port Talbot map, due to the 10 m resolution of the Living Wales map, not all differences were removed. However, any further buffering would have significantly reduced the number of sample points, and potentially resulted in the loss of some small habitat polygons.
  • The 1979-1997 upland habitat survey reported habitat mosaics as the predominant habitat type only. Mosaics of this type were not mapped during the 2022 field survey. This resulted in significant differences in the areas of some upland habitats. Therefore, datasets were split into upland and lowland habitats to reduce impacts on conclusions drawn.

Approach

The study utilised habitat data from the following sources, provided by NRW:

  • Phase 1 habitat survey data (1979 -1986, 1987-1997), which informed the current Habitat Map of Wales (DataMapWales);
  • An update Phase 1 habitat survey, 2022;
  • Neath Port Talbot’s Phase 1 habitat map (Environment Systems, 2021) based on imagery from Pleiades, Sentinel 1 and 2 satellites 2020, and a time-series of optical and radar imagery; and
  • The Living Wales Habitat Map 2022 generated using Sentinel-1 C-band Synthetic Aperture Radar (SAR) and Sentinel-2 optical data, and contextual information, including the Wales Peat Depth map.

For the purpose of the analysis, the results of the 2022 Phase 1 habitat survey were assumed to be correct.

Data Analysis

The GIS layers for each dataset were clipped to pre-defined tetrads (30 x 4 km2) in Neath Port Talbot and Carmarthenshire.

Map

Habitat Area Comparisons

GIS habitat layers from each of the four datasets were separated into upland and lowland boundaries, before generating areas for each habitat type in each dataset using GIS. The results were output into Excel for comparison.

Point Sample Analysis

A 25 m point sampling technique was developed to compare the differences in habitat types between the four datasets. Following buffering in GIS to reduce boundary effects and mapping errors, the GIS habitat layers for each dataset were overlain with a 25 m point sampling grid, and clipped to the extent of the 2022 Phase 1 habitat dataset. Habitat differences at each sample point between the two Phase 1 datasets, and between the remote sensed datasets and the 2022 Phase 1 dataset, were then assessed in accordance with a divergence rating based on NRW defined categories. Finally, the total divergence scores for each habitat type were combined, and percentages for each divergence rating were calculated.

Outcome

  • The results of the analysis indicated significant limitations in the use of remote sensing as a sole survey method for mapping all habitats to Phase 1 habitat classifications. However, the Neath Port Talbot dataset recorded a 90% accuracy in mapping of broad open habitat types (grassland, heathland, and wetland), but accurate identification of Phase 1 habitats was less reliable, and ground surveys would be required to map priority habitats and distinguish accurately between habitat sub-categories.
  • The mapping scales, and the use of fewer and more general habitat categories for the Living Wales dataset, suggested that this methodology was less suited to Phase 1 mapping, and again, field surveys would be required to map priority habitats and habitat sub-categories. However, the Living Wales mapping uses more automated processes, making it more time efficient and, unlike the Neath Port Talbot map, is designed to be continually updated, with the ability to refine the algorithms, and potentially the accuracy of the remoted sensed mapping.
  • Recognising the need for ground-truthing and the collection of additional data in the field, habitat areas recorded by the Neath Port Talbot remoted sensed methodology were attributed with a degree of certainty (low, medium, high) to ‘allow targeting of site visits’ (Environment Systems, 2021), whilst for the Living Wales map, a mobile application (EarthTrack) was been developed to incorporate field data “to validate maps of land cover” (Living Wales: overview of approach 2020).
  • The results of the assessment will inform NRW’s landscape-scale habitat mapping work, and further inform Keystone’s remote sensing services.

The full report is available here: https://naturalresources.wales/media/545hkews/report-no-696-habitat-map-of-wales-resurvey-pilot-analysis.pdf

Citation: McDonagh, M. (2023). Habitat Map of Wales resurvey pilot – analysis. NRW Report No: 696, 44 pp. NRW, Bangor.

Want to find out more?

For more information about this project, please get in touch.

Related case studies

Public Sector
South Pennine Moors Upland Breeding Bird Monitoring

Natural England commissioned Keystone Ecology to undertake an upland breeding bird survey of the South Pennine Moors SSSI and Phase 2 SPA, covering approximately 20,944.5 hectares of land across nine local authority regions and four counties (Greater Manchester, Lancashire, North Yorkshire and West Yorkshire).