Mapping the extent of scrub encroachment on heathland using LiDAR data

摘要

One of the key requirements for condition assessment and reporting is mapping scrub vegetation extent. In what are often complex environments, it is very difficult and extremely time consuming to do this using a ground-based approach. LiDAR data offers an ideal tool for mapping the height and spatial extent of scrub. The Environment Agency (EA), in partnership with Natural England, undertook a project to investigate whether LiDAR data alone could be used to map scrub. The EA have an archive of LiDAR data covering more than 65% of England and Wales, but not necessarily other spatially or temporally coincident datasets such as aerial photography. For this study, 25 cm LiDAR data were acquired of the disused airbase at Greenham Common near Newbury, UK. Ground data were acquired and used to define the parameters for delineating scrub. DTM and first return DSM data were then used to map the height and extent of scrub. Ground data were acquired to assess the accuracy of the derived scrub map. Survey grade GPS was used to collect point data of scrub and non scrub. For scrub points the height of the scrub vegetation was recorded. GPS points were also taken along the borders between scrub and non scrub to measure the boundary precision. Results show high levels of accuracy (95% overall) and precision (0.34 m RMSE). Regression analysis applied to predicted scrub height against actual scrub height revealed a strong correlation (0.67 r2), however the slope was 0.336, suggesting it was not possible to measure scrub height directly from the LiDAR data. Gorse scrub has a very open, spiky canopy that can result in the laser penetrating part of the structure and not returning hits from the top of the canopy, thus underestimating scrub height. It was not possible to discriminate between common gorse, dwarf gorse and dwarf heather using LiDAR alone. If further information on species is required then additional data sources would be needed, such as aerial photography or ground data. There is a perception amongst ecologists that accurate and precise scrub mapping requires 25 cm LiDAR data, but coarser resolution data are normally acquired for heathland environments. To test the effect of resolution on classification accuracy, the LiDAR grid data were re-sampled to coarser resolutions using a standard averaging method. Scrub maps were generated from the new LiDAR datasets and the accuracy of these outputs was assessed. Boundary precision and overall accuracy decreased with an increase in pixel size. Using a 2 m dataser the overall accuracy decreased to 87% and the border precision RMSE was 2.47 m. These results are discussed in the context of measuring change and in relation to mapping scrub vegetation in other environments.