David M. Cobby

Image processing of airborne scanning laser altimetry data for improved river flood modelling

Airborne scanning laser altimetry (LiDAR) is an important new data source for environmental applications, being able to map topographic height, and the height of surface objects, to high vertical and horizontal accuracy over large areas. This paper describes a range image segmentation system for data from a LiDAR measuring either time of last significant return, or measuring time of both first and last returns. We focus on the application of the segmenter to improving the data required by 2D hydraulic flood models, i.e. maps of topographic height which provide model bathymetry, and vegetation height, which could be converted to distributed floodplain friction coefficients. In addition, the location of river channels and a suitable height contour are used to define the extent of the model domain. An advantage of segmentation is that it allows different topographic and vegetation height extraction algorithms to be used in regions of different cover type. LiDAR data for a reach of the River Severn, UK, is presented. Short vegetation heights (grass and cereal crops) are predicted with a rms error of 14 cm. The topography underlying such cover differs from manually measured spot heights by 17 cm (rms error). The topographic accuracy decreases in the presence of a densely wooded slope. Errors in the vegetation height map, apparent at the overlap regions of adjacent swaths, are reduced by the removal of heights measured at large scan angles.

Waterline mapping in flooded vegetation from airborne SAR imagery

Multifrequency, polarimetric airborne synthetic aperture radar (SAR) survey of a salt marsh on the east coast of the UK is used to investigate the radar backscattering properties of emergent salt marsh vegetation. Two characteristics of flooded vegetation are observed: backscatter enhanced by approximately 1.2 dB at C-band, and 180° HH-VV phase differences at L-band. Both are indicative of a double bounce backscattering mechanism between the horizontal water surface and upright emergent vegetation. The mapping of inundated vegetation is demonstrated for both these signatures, using a statistical active contour model for the C-band enhanced backscatter, and median filtering and thresholding for the L-band HH-VV phase difference. The two techniques are validated against the waterline derived from tidal elevation measured at the time of overpass intersected with an intertidal DEM derived from airborne laser altimetry. The inclusion of flooded vegetation is found to reduce errors in waterline location by a factor of approximately 2, equivalent to a reduction in waterline location error from 120 to 70 m. The DEM is also used to derive SAR waterline heights, which are observed to underpredict the tidal elevation due to the effects of vegetation. The underprediction can be corrected for vegetation effects using canopy height maps derived from the laser altimetry. This third technique is found to improve the systematic error in waterline heights from 20 to 4 cm, but little improvement in random error is evident, chiefly due to significant noise in the vegetation height map.

Measurement of habitat predictor variables for organism-habitat models using remote sensing and image segmentation

Robust predictive models of the effects of habitat change on species abundance over large geographical areas are a fundamental gap in our understanding of population distributions, yet are urgently required by conservation practitioners. Predictive models based on underpinning relationships between environmental predictors and the individual organism are likely to require measurement of spatially fine-grained predictor variables. Further, models must show spatial generality if they are to be used to predict the consequences of habitat change over large geographical areas. Remote sensing techniques using airborne scanning laser altimetry (LiDAR) and high resolution multi-spectral imagery allow spatially fine-grained predictor variables to be measured over large geographical areas and thus facilitate testing of the spatial generality of organism-habitat models. These techniques are considered using the skylark as an example species. A range image segmentation system for LiDAR data is described which allows measurement of skylark habitat predictor variables such as within-field vegetation height, boundary height and shape for individual fields within the LiDAR image. Additional variables such as field vegetation type and fractional vegetation ground cover may be obtained from co-registered multi-spectral data. These techniques could have wide application in testing the generality of relationships between populations and habitats, and in ecological monitoring of change in habitat structures and the associated effects on wildlife, over large geographical areas.

Image processing of airborne scanning laser altimetry for some environmental applications

Airborne scanning laser altimetry (LiDAR) is an important new data source for environmental applications, being able to map heights to high vertical and horizontal accuracy over large areas. The paper describes a range image segmentation system for data from a LiDAR measuring time of last significant return only. Each spot height represents the height of incidence of the narrow laser pulse with the ground, the top of the vegetation canopy or some point in between. The segmenter is aimed at two specific environmental applications, both of which require the underlying ground heights and the vegetation canopy heights to be estimated from the LiDAR height image. A method of estimating vegetation height in regions of short vegetation such as crops is presented. An advantage of segmentation is that it allows different topographic and vegetation height extraction algorithms to be used in regions of different cover type. Thus the method attempts to maintain ground height accuracy in regions of tall vegetation cover (e.g. forest areas) by reducing spatial resolution in these regions.

Obtaining accurate maps of topography and vegetation to improve 2D hydraulic flood models

Airborne scanning laser altimetry (LiDAR) is an important new data source for environmental applications, mapping topographic and surface object heights to high vertical and spatial accuracy over large areas. We present results of a segmentation system for LiDAR data for a reach of the river Severn, UK. The system has been developed to improve the 3 main data required by a leading numerical flood model predicting inundation extent, namely (i) a map of topographic height providing model bathymetry. A comparison with ground control points gives an accuracy of ±17cm (decreasing in the presence of steeply wooded slopes), (ii) the meandering location of the river channel and a suitable height contour which denote the extent of the model domain, and allow immediate finite element mesh generation, and (iii) a map of vegetation height (to an accuracy of ±14cm for grass and cereal crops) which is converted to friction coefficients. Errors due to overlapping swaths are significantly reduced. A 3-class segmentation of vegetation types (short, intermediate and tall) allows optimal height extraction algorithms to be separately applied, and enables realistic conversion to friction coefficients. Short (grass and cereal crops) and intermediate (hedges) vegetation are assumed to be flexible and either emergent or submerged during a flood cycle. Trees (tall vegetation) are modelled as rigid, emergent, stems.