Automatic Cotton Mapping Using Time Series of Sentinel-2 Images

Abstract

Large-scale crop mapping is essential for agricultural management. Phenological variation often exists in the same crop due to different climatic regions or practice management, resulting in current classification models requiring sufficient training samples from different regions. However, the cost of sample collection is more time-consuming, costly, and labor-intensive, so it is necessary to develop automatic crop mapping models that require only a few samples and can be extended to a large area. In this study, a new white bolls index (WBI) based on the unique canopy of cotton at the bolls opening stage was proposed, which can characterize the intensity of bolls opening. The value of WBI will increase as the opening of the bolls increases. As a result, the white bolls index can be used to detect cotton automatically from other crops. Four study areas in different regions were used to evaluate the WBI performance. The overall accuracy (OA) for the four study sites was more than 82%. Additionally, the dates when the opening stage of bolls begins can be determined based on the time series of WBI. The results of this research demonstrated the potential of the proposed approach for cotton mapping using sentinel-2 time series of remotely sensed data.

Extraction of Cotton Information with Optimized Phenology-Based Features from Sentinel-2 Images

“…These features are indicative of boll opening and senescence signals, suggesting that the boll opening stage is the most critical phenological stage for cotton identification. This finding is consistent with previous studies and is in line with cotton's actual physical and biochemical characteristics [25,27]. In time-series dynamic features, the harmonic coefficients of cos (3πt), sin (3πt) and constant terms were found to be relatively more important, as demonstrated in Figure 6.…”

“…Our results have demonstrated that incorporating multiple spectral features at different phenological stages can significantly enhance the classification accuracy compared to relying solely on NDVI and EVI throughout the entire phenological period. Further, some studies have solely employed individual phenological stages for cotton extraction [27,61], which could heighten the probability of confusion between cotton and other crops with similar spectral characteristics (e.g., maize and cotton) [1].…”

“…Numerous studies have reported successful regional-scale mapping of cotton and other crops using multiple phenological features [24][25][26][27]. Google Earth Engine (GEE) has enhanced the efficiency of remote sensing image processing.…”

The 10-m cotton maps in Xinjiang, China during 2018–2021

“…Most cotton has white fibre, and a small amount of cotton has colored fibre in Xinjiang 83 , 84 . A few recent studies have reported that the phenological feature in the boll-opening stage could be used for cotton mapping 11 and cotton yield estimation 85 – 88 . Wang et al .…”

“…It is a challenge to produce a fine cultivation map of cotton throughout Xinjiang 7 , probably because such a task needs a powerful platform with high storage and computing capabilities 8 ; it needs enough high-quality and evenly distributed samples due to the high spatial heterogeneity 9 ; and it needs models with high universality, accuracy, and transferability to address issues like bad weather 10 . To date, numerous studies have mainly focused on Xinjiang cotton mapping at county 1,11,12 , prefecture 13 , region 7,14 , and province scales 15,16 . Although we recently have produced the 500 m nationwide cotton maps in 2016 and 2018 that were included in the reports on remote sensing monitoring of China Sustainable Development 15,16 , both of them demonstrated limitations and disadvantages in practical precision management and decision making due to the very low spatial resolutions.…”