Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: November 21, 2005
Publication Date: February 28, 2006
Citation: Brown de Colstoun, E., Walthall, C. 2006. Improving global-scale land cover classifications with multi-directional view-angle data and a decision tree classifier. Remote Sensing of Environment. 100:474-485.
Interpretive Summary: Accurate global land cover maps are needed for diagnostic and predictive models simulating the functioning of earth hydrologic and climatic systems. Prior global land cover maps have relied solely on spectral and temporal satellite data to conduct classifications and have ignored multiple direction view angle data as an additional information vector that could be used to improve land cover map accuracy. Directional reflectance information was added to spectral information available from POLarization and Directionality of Earth Resources (POLDER) satellite data and used to classify global land cover with a decision-tree classifier. The addition of directional reflectance information reduced the overall classification error with notable improvements for accuracies of individual classes such as Grasslands (+8.71%), Urban areas (+8.02%), Wetlands (+7.82%) and Evergreen Broadleaf Forest (+6.70%). The methodology and results are widely applicable to current multidirectional satellite data from the Multi-angle Imaging Spectroradiometer (MISR), and to the next generation of POLDER-like multi-directional satellite instruments.
The Bidirectional Reflectance Distribution Function (BRDF) contains information that can be used to augment spectral information for improved land cover classification accuracies. Prior studies on the addition of BRDF information to improve land cover classifications have been conducted primarily at local or regional scales. Thus, the potential benefits of adding BRDF information to improve global to continental scale land cover classification have not yet been explored. Here we examine the impact of multidirectional global scale data from the Polarization and Directionality of Earth Reflectances (POLDER) spacecraft instrument on overall classification accuracy and per-class accuracies for 15 land cover categories specified by the International Geosphere Biosphere Programme (IGBP).
A set of 36648 global training pixels (7km x 6km spatial resolution) was used with a decision tree classifier to evaluate the performance of classifying POLDER imagery with and without the inclusion of BRDF information. BRDF ‘metrics’ for the eight-month POLDER archive were developed to describe the temporal evolution of the BRDF as captured by a semi-empirical BRDF model. The concept of BRDF ‘feature space’ is introduced and used to explore and exploit the bidirectional information content. The C5.0 decision tree classifier was applied with a boosting option, with the temporal metrics for spectral albedo as input for a first test, and with spectral albedo and BRDF metrics for a second test. Results were evaluated against 20 random subsets of the training data.
Examination of the BRDF feature space indicates that coarse scale BRDF coefficients from POLDER provide information on land cover different from the spectral and temporal information of the imagery. The contribution of BRDF information to reducing classification errors is also demonstrated: the addition of BRDF metrics reduces the mean, overall classification error rates by 3.15% (from 18.1% to 14.95% error) with larger improvements for producer’s accuracies of individual classes such as Grasslands (+8.71%), Urban areas (+8.02%), and Wetlands (+7.82%). User’s accuracies for the Urban (+7.42%) and Evergreen Broadleaf Forest (+6.70%) classes are also increased. The methodology and results are widely applicable to current multidirectional satellite data from the Multi-angle Imaging Spectroradiometer (MISR), and to the next generation of POLDER-like multi-directional instruments.