| Title | A data mining approach for evaluation of optimal time-series of MODIS data for land cover mapping at a regional level |
| |
| Author | Zhou, F; Zhang, A; Townley-Smith, L |
| Source | ISPRS Journal of Photogrammetry and Remote Sensing vol. 84, 2013 p. 114-129, https://doi.org/10.1016/j.isprsjprs.2013.07.008 |
| Year | 2013 |
| Alt Series | Natural Resources Canada, Contribution Series 20181132 |
| Publisher | Elsevier BV |
| Document | serial |
| Lang. | English |
| Media | paper; on-line; digital |
| File format | pdf |
| Subjects | geophysics; remote sensing |
| Program | Canada Centre for Remote Sensing Divsion |
| Released | 2013 10 01 |
| Abstract | Optical Earth Observation data with moderate spatial resolutions, typically MODIS (Moderate Resolution Imaging Spectroradiometer), are of particular value to environmental applications due to their high
temporal and spectral resolutions. Time-series of MODIS data capture dynamic phenomena of vegetation and its environment, and are considered as one of the most effective data sources for land cover mapping at a regional and national level. However,
the time-series, multiple bands and their derivations such as NDVI constitute a large volume of data that poses a significant challenge for automated mapping of land cover while optimally utilizing the information it contains. In this study,
time-series of 10-day cloud-free MODIS composites and its derivatives - NDVI and vegetation phenology information, are fully assessed to determine the optimal data sets for deriving land cover. Three groups of variable combinations of MODIS spectral
information and its derived metrics are thoroughly explored to identify the optimal combinations for land cover identification using a data mining tool.The results, based on the assessment using time-series of MODIS data, show that in general using a
longer time period of the time-series data and more spectral bands could lead to more accurate land cover identification than that of a shorter period of the time-series and fewer bands. However, we reveal that, with some optimal variable
combinations of few bands and a shorter period of time-series data, the highest possible accuracy of land cover classification can be achieved. © 2013. |
| GEOSCAN ID | 311486 |
|
|