A Novel Automatic Method for Alfalfa Mapping Using Time Series of Landsat-8 OLI Data

Remote sensing (RS) data have been utilized increasingly for mapping various crops at local and regional scales using various techniques. However, training data collection of these methods is costly and time consuming. On the other hand, time series of RS data have provided valuable information about crop phenological patterns, which can be utilized for automatic crop mapping independent of training data. Hence, the aim of this research is to develop a new automatic method to map alfalfa by identification of specific characteristics of alfalfa based on time series of Landsat 8 OLI images in four study sites in Iran and the United States. Alfalfa fields are usually harvested periodically and two neighboring farms may not be harvested simultaneously. To address this challenge, the alfalfa spectral reflectance values in various bands were compared with those of other crops during the growing season. In the following, three assumptions were made to find suitable relationships for demonstrating alfalfa characteristics as well as separating it from other crops. The results indicated that the summation of differences between the red and NIR reflectance values of alfalfa in the time series of Landsat images is significant; and also, the average values of the NIR and red bands during the growing season are remarkably higher and lower than those of other crops, respectively. Hence, based on these findings, a new specific feature was developed to detect alfalfa with the overall accuracy of 93%, 90%, 94%, and 90% in Moghan, Qazvin, Razan, and Parker Valley, respectively.