Landscape analysis of wetland plant functional types: the effects of image segmentation scale, vegetation classes and classifier methods.

Iryna Dronova, University of California, Berkeley, USA, idronova@berkeley.edu (Presenter)
Peng Gong, University of California, Berkeley, USA, penggong@berkeley.edu
Nicholas Clinton, Tsinghua University, Beijing, China, nicholas.clinton@gmail.com
Lin Wang, Chinese Academy of Fishery Sciences (CAFS), Beijing, China, angels121@gmail.com
Wei Fu, State Key Laboratory of Remote Sensing Science, Beijing, China, vivianfw861208@126.com
Shuhua Qi, Jiangxi Normal University, Nanchang, China, qishuhua11@163.com

Remote sensing-based analyses of plant function such as photosynthesis and productivity are challenging in wetlands with complex cover and difficult field access. Recent advances in object-based image analysis (OBIA) and machine-learning algorithms offer new image classification tools; however, few comparisons of different approaches have been discussed to date. We applied OBIA to delineate wetland plant functional types (PFTs) for Poyang Lake, the largest freshwater lake in China and a Ramsar wetland conservation site, from spring 2008 Landsat TM imagery. We targeted major PFTs suggested by plant species ecology and field spectral data that are both key players in this system’s biogeochemical cycles and critical providers of waterbird habitat. Classification results were compared among: a) several “small” object segmentation scales (with average object sizes 900-9000 m²); b) several families of statistical classifiers (including Bayesian, Logistic, Neural Network, Decision Trees, Support Vector Machines) and c) two hierarchical levels of vegetation classification, a generalized 3-class set and a more specific 6-class set. We found that while a number of classifiers achieved high accuracy at the finest pixel-equivalent segmentation scales, the highest accuracies and best agreement among algorithms occurred at coarser object scales. No single classifier was consistently superior across all scales, although selected algorithms of Support Vector Machines, Neural Network, Logistic and K-Nearest Neighbors families frequently provided the best classification accuracies. Individual classes differed in scales at which they were best discriminated from others, which can be used to guide a hierarchical multi-scale PFT classification in this landscape. We conclude that OBIA with advanced statistical classifiers offers useful instruments for landscape vegetation analyses, while spatial scale considerations are critical in mapping PFTs. Future work will compare results with MODIS PFT products and extend study findings to investigate seasonal PFT dynamics following flood cycle and examine the association among PFTs and waterbird habitat use.

Presentation Type:  Poster

Session:  Other   (Mon 4:00 PM)

Associated Project(s): 

• Gong, Peng: Analysis of spatio-temporal association between plant functional diversity and flooding at Poyang Lake, PRC using remote sensing and simulation-based modeling. ...details

Poster Location ID: 16