Title :
Mapping urban landuse types in Los Angeles using multi-date Moderate-Resolution Imaging Spectroradiometer vegetation image products
Author :
Zheng, Yuanfan ; Qiu, Hong-lie
Author_Institution :
Department of Geosciences and Environment, California State University - Los Angeles, U.S.A.
Abstract :
The Moderate-Resolution Imaging Spectroradiometer (MODIS) vegetation data products provide a new opportunity for urban landuse classification study using multi-date remote sensing data. The main advantage of this data source is its ability to provide seasonality information for different types of vegetation covers. It has been demonstrated that vegetation covers of different moisture conditions or different species compositions have different variation patterns in the time series of the MODIS Enhanced Vegetation Index (EVI) values. This research tries to explore the possibility of using the multi-date MODIS vegetation data products for mapping different landuse types in the Los Angeles area. It was suggested that vegetation of different urban landuse types exhibits unique seasonal variation patterns in their EVI time series due to differences in species composition, moisture condition, and vegetation density. These unique temporal signatures are the basis for image classification. Two classification methods were used for comparison. The Method 1 used a single-date MODIS EVI image and the Density Slice classification method. Method 2 used the Decision Tree classification on the images created from the eleven seasonality parameter values computed by the TIMESAT. The two methods were employed to classify an urban landscape in Los Angeles into five key landuse types. Classification accuracy assessment was conducted by examining their overall accuracy and Kappa coefficient values. The results of this research suggest that the seasonality information contained in the multi-date MODIS vegetation products are valuable for urban landuse mapping. It improved the overall classification accuracy by 9.4 percents and has a higher Kappa index of 0.52. It was also suggested that seasonality parameters extracted from the multi-date MODIS EVI data can be used to classify the vegetated areas in the urban and suburban areas in Los Angeles into subclasses (parks and low density residen- ial neighborhoods, natural vegetation area, and agricultural fields).
Keywords :
EVI; MODIS; Pheonology; image classification; urban landuse;
Conference_Titel :
Earth Observation and Remote Sensing Applications (EORSA), 2012 Second International Workshop on
Conference_Location :
Shanghai, China
Print_ISBN :
978-1-4673-1947-8
DOI :
10.1109/EORSA.2012.6261183