Publications

Cao, JJ; Cai, XL; Tan, JW; Cui, YL; Xie, HW; Liu, FP; Yang, L; Luo, YF (2021). Mapping paddy rice using Landsat time series data in the Ganfu Plain irrigation system, Southern China, from 1988-2017. INTERNATIONAL JOURNAL OF REMOTE SENSING, 42(4), 1556-1576.

Abstract
The spatial pattern and temporal variation of paddy rice fields based on remote sensing have strong effects on the allocation of agricultural water resources on a regional scale. Free Land Remote-Sensing Satellite (Landsat) data have been successfully used to map paddy rice. However, due to frequent clouds and a temporal gap of 16 days, the shortage of Landsat data availability poses a great challenge to long-term paddy rice mapping. This study proposed a decision tree algorithm based on the Enhanced Vegetation Index (EVI) to map multi-season paddy rice in southern China. The study area is located in the Ganfu Plain Irrigation System (GFPIS), where double-cropping rice (early and late rice) and single-cropping rice (middle rice) are mixed. First, we explored the effects of the number and temporal distribution of Landsat images on the classification accuracy. With available cloud-free images in 2017, ten image combinations were set up to map rice fields using the proposed algorithm separately. Then, the algorithm was applied to map historical paddy rice planting in this study area. The results indicated that with a cropland mask, a single-date image in early rice growing season can map the total early rice with an overall accuracy varying from 82.02% to 93.26%, and the accuracy was mainly influenced by the temporal distribution of the images. In addition, through post-processing, multi-temporal images could be used to recognize early rice only, late rice only, double-cropping rice and middle rice with an overall accuracy ranging from 71.83% to 85.81%. In contrast, images in the early or late growing season without obvious vegetation characteristics may result in confusion. Two peak growing season images with obvious differences in the vegetation index played a key role in detecting different rice cropping types. In addition, when the EVI of rice was within a certain range, one or more images could achieve a similar overall accuracy. The total area of middle rice and late rice remained constant before 2001, but it changed from 10.72 x 10(4) ha in 2001 to 7.55 x 10(4) ha in 2017. Additionally, there was a twofold increase in the mapped middle rice area.

DOI:
10.1080/01431161.2020.1841321

ISSN:
0143-1161