Habitat Mapping using High-resolution Earth Observation Data
Combined use of multispectral and hyperspectral imagery with LiDAR-derived measures
INTRODUCTION
Recent advances in Earth Observation (EO) sciences have led to the increased use of multi-sensor-derived data for land cover and habitat mapping across diverse landscapes.
The combined use of spectral and LiDAR data has allowed for more detailed land cover and habitat mapping at finer spatial resolutions, thereby meeting the specific needs of end-users such as nature conservation managers and other stakeholders.
FOCUS
The study investigated the combined use of multispectral/hyperspectral imagery and LiDAR data for habitat mapping across parts of south Cumbria, North West England. The methodology adopted in this study integrated spectral information contained in pan-sharpened QuickBird multispectral/AISA Eagle hyperspectral imagery and LiDAR-derived measures with object-based machine learning classifiers and ensemble analysis techniques.
STUDY AREA
The study site, River Leven, is situated in the southern part of Cumbria, NW England (2◦58'W, 54◦16'N) and covers an approximate area of 108 ha. The site is characterized by a distribution of coniferous and broadleaved woodlands existing alongside rough grasslands, improved grasslands, and Phragmites reedbed patches.
INPUT DATA
Summarized in the table below are the input datasets evaluated in this study. The experiments were designed to assess the performance of combining QuickBird spectral & Eagle MNF transformed layers with LiDAR-derived measures.
METHODOLOGY
The designed framework combines VHR spaceborne multispectral imagery, airborne hyperspectral imagery, LiDAR data and processing techniques. The processing techniques comprised data fusion, object-based image analysis, and machine learning-based classifiers.
The machine learning algorithms explored were Random Forest (RF), Support Vector Machine (SVM) and k-Nearest Neighbor (k-NN).
HABITAT MAP
Habitat map generated using majority ensemble analysis for QuickBird imagery combined with LiDAR measures
CONCLUSION
The study shows the importance of combining VHR multispectral and hyperspectral imagery with LiDAR-derived measures for habitat mapping and analyzing the same with object-based machine learning algorithms. The optimal results were further exploited through the ensemble analysis technique. This study was performed using spaceborne QuickBird imagery and airborne AISA Eagle hyperspectral imagery.
