DEVELOPMENT OF PHENOLOGY BASED ALGORITHM FOR CROPLAND AND CROP TYPE MAPPING WITH MULTITEMPORAL LANDSAT IMAGE DATA - CASE STUDY IN THE NORTHWEST OF VIETNAM

Abstract

Abstract. Cropland mapping is very important for food security, policy development, land use planning, and environmental protection. Scientists have developed methods and techniques for cropland mapping with remote sensing image data. Both single date and multitemporal image data are used for generation of cropland and crop type maps. Multitemporal image data has advantages over single date image data from reliability and accuracy point of view because multitemporal image data allows to eliminate seasonality of vegetation. In this paper, the authors present new algorithm for cropland and crop type mapping with multitemporal Landsat image data. The algorithm requires for analysis of all Landsat scenes observed during one year and if needed, scenes in some years back to compensate clouds and cloud shadows. Phenology of land cover is constructed based on six bimonthly cloud free land covers that were automatically classified using the selected scenes. By grouping land covers within two months to six land covers of periods January–February, March–April, May-June, July–August, September–October, and November–December we create six bimonthly cloud free land covers that formulate a database for mapping cropland and crop types. By analysis of 50 Landsat scenes of path/row number 128/45 (northwest of Vietnam) observed mainly from 2017, 2016 and 2015 we success to map upland cropland and 14 crop types with area ranging from 145,143 ha to 3,373 ha per crop type. The study pointed out that phenology characterized by six bimonthly land covers is acceptable to identify cropland distribution and some specific crop types. For better results, apparently we need higher temporal resolution of image data. Due to uncertainty of the atmosphere, it is almost impossible to rely only on optical remote sensing data to achieve high temporality of data so application of multitemporal SAR data could be a way to overcome this obstacle.