NSII: a novel soybean identification index based on Sentinel-2 imagery for accurate and efficient soybean mapping.

NSII: a novel soybean identification index based on Sentinel-2 imagery for accurate and efficient soybean mapping - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Front Plant Sci . 2026 Apr 1;17:1788686. doi: 10.3389/fpls.2026.1788686 Search in PMC Search in PubMed View in NLM Catalog Add to search NSII: a novel soybean identification index based on Sentinel-2 imagery for accurate and efficient soybean mapping Xiufang Zhu Xiufang Zhu 1 State Key Laboratory of Remote Sensing and Digital Earth, Beijing Normal University, Beijing, China 2 Key Laboratory of Environmental Change and Natural Disaster, Ministry of Education, Beijing Normal University, Beijing, China Writing – review & editing, Methodology, Conceptualization Find articles by Xiufang Zhu 1, 2 , Yinhui Wang Yinhui Wang 3 School of Earth Science and Engineering, Hebei University of Engineering, Handan, China Writing – review & editing, Validation, Writing – original draft, Data curation, Methodology Find articles by Yinhui Wang 3, * , Anzhou Zhao Anzhou Zhao 4 School of Mining and Geomatics Engineering, Hebei University of Engineering, Handan, China Writing – review & editing Find articles by Anzhou Zhao 4, * , Dan Li Dan Li 3 School of Earth Science and Engineering, Hebei University of Engineering, Handan, China Writing – review & editing Find articles by Dan Li 3 Author information Article notes Copyright and License information 1 State Key Laboratory of Remote Sensing and Digital Earth, Beijing Normal University, Beijing, China 2 Key Laboratory of Environmental Change and Natural Disaster, Ministry of Education, Beijing Normal University, Beijing, China 3 School of Earth Science and Engineering, Hebei University of Engineering, Handan, China 4 School of Mining and Geomatics Engineering, Hebei University of Engineering, Handan, China * Correspondence: Yinhui Wang, [email protected] ; Anzhou Zhao, [email protected] Roles Xiufang Zhu : Writing – review & editing, Methodology, Conceptualization Yinhui Wang : Writing – review & editing, Validation, Writing – original draft, Data curation, Methodology Anzhou Zhao : Writing – review & editing Dan Li : Writing – review & editing Received 2026 Jan 15; Accepted 2026 Mar 5; Revised 2026 Mar 1; Collection date 2026. Copyright © 2026 Zhu, Wang, Zhao and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. PMC Copyright notice PMCID: PMC13085932  PMID: 42004021 Abstract Accurate mapping of soybean cultivation areas is crucial for agricultural monitoring, resource management, and food security. However, the spectral overlap between soybean and other crops, such as corn, poses significant challenges for remote sensing-based identification. This study proposes a novel soybean identification index (NSII), which is calculated using the second red-edge band (RE2), the first short-wave infrared band (SWIR1), and the Enhanced Vegetation Index (EVI) derived from Sentinel-2 imagery within the optimal time window identified through spectral feature analysis. NSII was implemented in 12 major soybean producing regions in the United States and China over a three-year period (2020-2022). Experimental results from 2020 to 2022 show that the average accuracy of NSII is 0.85, and the average F1 score is 0.80. Compared with the existing Soybean Mapping Composite Index (SMCI), the accuracy increased by 8 percentage points and the F1 score increased by 6 percentage points. NSII also exhibits strong stability and transferability, with consistent performance across diverse climatic and cropping conditions. This study provides a robust and efficient tool for soybean mapping, offering significant potential for precision agriculture and sustainable resource management. Keywords: class separability, crop mapping, Sentinel-2 data, soybean identification, spectral separability 1. Introduction Accurate spatial distribution maps of crops are important for monitoring and evaluating farmland conditions, precision agriculture management, agricultural disaster response and agricultural policy formulation ( Wu et al., 2023 ). It is usually time-consuming and laborious to utilize traditional agricultural survey methods to classify and map crops, while remote sensing technology, with its ability to acquire data macroscopically and rapidly, has become an important means of crop mapping ( Gao and Zhang, 2021 ), which greatly saves the costs of manpower, material, financial and time, and effectively makes up for the shortcomings of traditional survey methods ( Choukri et al., 2024 ). In recent years, the application of high-resolution (temporal and spatial) remote sensing data in crop identification has become more and more widespread ( Xu et al., 2021 ). Crop spectral characteristics exhibit distinct variations across species. ( Farmonov et al., 2023 ; Sidike et al., 2019 ; She et al., 2020 ), constituting the most vital features for crop type identification ( Agilandeeswari et al., 2022 ). The spectral differences in remote sensing between different crops stem from their anatomical structures (such as the arrangement of cells in monocotyledons and dicotyledons, leaf morphology, and surface characteristics), biochemical components (absorption or scattering of specific wavelengths of light by photosynthetic pigments, water, cellulose, etc.), phenological stages (changes in growth status at different growth stages), and growth environment (indirect effects of water, soil, light, etc. on leaf characteristics). The remote sensing spectral differences of major crops (rice, corn, wheat, soybeans) are primarily reflected in the chlorophyll absorption intensity in the visible light band, leaf structure scattering ability in the near-infrared band, red edge band position, and water absorption peak in the short-wave infrared band. Monocots (rice, corn, wheat) display markedly higher near-infrared reflectance and red-edge shifts compared to dicotyledonous soybeans, attributable to their distinct leaf architectures and chlorophyll distributions. Rice’s characteristically high moisture content produces pronounced short-wave infrared absorption peaks distinguishing it from xerophytic crops, whereas corn’s unique foliar structure generates diagnostic near-infrared and green-band reflectance signatures differentiating it from both wheat and soybeans. Vegetation indices obtained by mathematical operations based on multiple bands of remotely sensed imagery can achieve the effect of enhancing the spectral characteristics of vegetation and are often used for crop type identification and distribution mapping studies ( de Oliveira Maia et al., 2023 ; Yang et al., 2022 ). For example, Huang et al. (2022) used Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI) as inputs to the random forest model to map winter wheat in Henan Province, China, in 2020 at 10 m resolution. Qu et al. (2021) used NDVI time series data to derive four key growth nodes of winter wheat to construct a decision tree to realize the extraction of winter wheat. Liu et al. (2020) successfully identified rice using differences in Enhanced Vegetation Index (EVI) and Yellowness Index (YI) at different times of the year. Some researchers also focus on developing spectral indices for the identification of specific crop types to enhance their specific spectral features and improve their identification accuracy. For example, Shi and Huang (2015) developed a Normalized Weighted Difference Water Index (NWDWI) to identify rice, and used this index to monitor the planted area of single-season and double-season early and late rice in the Yangtze River Delta region, and verified the accuracy of the method using agricultural census data. Tao et al. (2023) proposed a novel Phenology-based Winter Rapeseed Index (PWRI) and utilized the index to map the distribution of winter oilseed rape in the Yangtze River Basin. PWRI was shown to have good separability between winter oilseed rape and other crops throughout the flowering period, and the method was applied to the middle reaches of the Yangtze River, with an overall accuracy and kappa coefficient of more than 92% and 0.85%, respectively. Wang et al. (2021) developed a White Bolls Index (WBI) based on the unique canopy characteristics of cotton during the fluffing stage and applied it to cotton identification in Missouri’s 8th District, California’s 21st District, Georgia’s 8th District, and Shihezi and its surrounding areas in China’s North Xinjiang, with an overall accuracy rate of over 82%. Ashourloo et al. (2020) developed a new spectral feature for potato mapping by analyzing the spectral features of potato. They calculated the sum of the differences in reflectance between the red and near-infrared (NIR) bands from planting to the peak greenness stage, along with the ratio of NIR to red reflectance at peak greenness and the NIR reflectance at harvest. This approach effectively distinguished potatoes from other crops. Soybean is the main raw material for high-protein food, livestock feed and edible oil, and has an important position in world food production ( Li et al., 2021 ). Accurate identification of soybean cultivation areas is of great significance in guaranteeing food security ( Zhu et al., 2022 ). Mapping and monitoring of the spatial distribution of soybean ( Li et al., 2019 ) provides an important basis for agricultural decision-making, and farmers and agricultural decision-makers can adjust their planting plans based on these data to optimize the layout of soybean planting and resource allocation, so as to achieve the optimal benefits of soybean production. For this reason, some researchers have also focused on the development of soybean remote sensing identification index. Soybean and corn are grown almost simultaneously in many regions and their spectra are highly similar in most phenological stages, making them easily confused ( Bandaru et al., 2020 ). Therefore, when developing soybean identification indices, researchers tend to focus on the differences between soybean and corn spectral characteristics at specific phenological stages ( Zhong et al., 2016 ). For example, Chen et al. (2023) proposed the Greenness and Water Content Composite Index (GWCCI) for soybean mapping by utilizing the differences between soybean and other crops in shortwave infrared (SWIR) bands and NDVI. Xiao et al. (2024) developed a Soybean Mapping Composite Index (SMCI) based on the optimal time window using Sentinel-2 imagery, which coupled three red-edge bands (RE2, RE3, and RE4), the near-infrared (NIR) band, the SWIR band, the EVI, and the Green Chlorophyll Vegetation Index (GCVI). The GWCCI only uses the SWIR band and the NDVI, neglecting the role of the red-edge band in soybean identification, and the NDVI index is prone to overfitting in areas of lush vegetation. The SMCI couples multiple features, which increases the computational complexity and the amount of data processing, and at the same time, highly correlated features may lead to feature redundancy problems. In this study, we propose a novel remote sensing index for soybean identification, aiming to enhance the accuracy of soybean distribution mapping. Addressing the challenge of spectral similarity between soybeans and corn across multiple growth stages, we integrated visible, near-infrared, red edge, and shortwave infrared features to maximize information utilization across different bands while minimizing feature redundancy. Through validation across diverse cropping patterns and field characteristics in different regions and years, this study not only demonstrates the new index’s effectiveness in soybean identification but also compares it with the existing Soybean Mapping Composite Index (SMCI). This study can enrich the selection of features for soybean mapping and help accurate soybean distribution mapping research. 2. Materials 2.1. Study sites As one of the largest soybean producers in the world, the U.S. accounts for about 23.9% of the global soybean acreage in 2023 ( Volkova and Smolyaninova, 2024 ). As the world’s largest soybean consumer and importer, changes in China’s soybean cultivation have a significant impact on the global soybean supply and demand balance, market prices, and food security. In 2017, China’s soybean production accounted for approximately 3.57% of the global total ( Di et al., 2023 ). In view of the different soybean planting conditions in China and the United States, this study selected 12 representative areas in the main soybean producing areas of the United States and China, which differed in terms of climate type, rainfall, temperature, and cropping structure ( Figure 1 ; Table 1 ). Of these, the first three regions (U1-U3) were used for the development of the novel soybean identification index (NSII), and the remaining regions (U4-U10, C1-C2) were used for testing NSII. U1-U3 span three distinct Köppen climate types, illustrating the diversity of climates within the study area. The Köppen climate classification system, developed by Beck et al. (2018) , categorizes global climates based on thresholds and seasonality of monthly air temperature and precipitation. The study area selected in this study contains five Köppen climate types, namely Mediterranean climate (warm summer) (Csb), Humid continental climate (dry winter, hot summer) (Dwa), Humid subtropical climate (dry winter, hot summer) (Cwa), Humid continental climate (no dry season, hot summer) (Dfa), and Humid continental climate (no dry season, warm summer) (Dfb). Figure 1. Open in a new tab Spatial distribution, climate, average temperature, and total precipitation in 10 US counties (See Table 1 for details). Table 1. Detailed description of the study area. Sites Location Main crops Soybean Planting ratio (%) Sowing time Harvest time U1 Dallas, Iowa Soybean, corn 46.86 late Apr-early May late Sep-middle Nov U2 Cass, North Dakota Soybean, corn, spring wheat 46.82 late Apr-early May late Sep-early Oct U3 Phillips, Arkansas Soybean, corn, cotton, rice 54.29 early Apr-early May early Sep-late Oct U4 Renville, North Dakota Soybean, corn, spring wheat 35.95 late Apr-early May late Sep-early Oct U5 Butler, Nebraska Soybean, sorghum, corn, winter wheat 41.55 late Apr late Sep-early Nov U6 De-witt, Illinois Soybean, corn, winter wheat 50.17 late Apr late Sep-early Nov U7 Van-wert, Ohio Soybean, corn, winter wheat 56.73 late Apr late Sep-early Nov U8 Dodge, Minnesota Soybean, sunflower, corn, spring wheat 40.41 middle Apr middle Sep-late Nov U9 Saline, Missouri Soybean, corn, cotton, rice 48.74 late Apr late Sep-late Nov U10 Marshall, Indiana Soybean, corn, winter wheat 38.17 late Apr middle Sep-late Nov C1 Nehe Soybean, corn, rice, potato 45.8% middle May late Sep-late Oct Heilongjiang C2 Lingbi Soybean, corn, wheat 22.1% late May late Sep-late Oct Anhui Open in a new tab The first region (U1), located in Dallas County, Iowa. Iowa is one of the major soybean producing regions in the United States and has one of the highest soybean production in the nation ( Hosseini et al., 2020 ). Dallas County has a Humid continental climate (no dry season, hot summer) (Dfa). Rainfall and high temperatures are concentrated in the summer months. The major crops grown in the county are soybean and corn. The second region (U2), located in Cass County, North Dakota, has a humid continental climate (no dry season, warm summer) (Dfb) with cold winters, and moderate but unevenly distributed annual precipitation. The broad plains topography of North Dakota provides unique conditions for agricultural development ( Hanson et al., 2022 ). The major crops grown in Cass County are soybean, corn, and spring wheat. The third region (U3) is located in Phillips County, Arkansas. Before 2005, the planting industry of Arkansas was dominated by cotton, but the proportion of soybean and rice cultivation gradually increased after 2005. Arkansas belongs to the mediterranean climate (warm summer) (Csb), characterized by four distinct seasons, a favorable natural environment, relatively uniform precipitation distribution throughout the year, and drier conditions during the summer months. The main crops grown in Phillips County are soybean, corn, cotton, and rice. Figure 2 shows the phenological calendar of the main crops in U1–U3, which is based on data from the National Agricultural Statistics Service (NASS), United States Department of Agriculture (USDA). The growing seasons of soybean, corn, cotton, and rice are basically between April and September, with soybean and corn having the most overlapping growing seasons and the most similar spectral characteristics ( Wei et al., 2023 ). Figure 2. Open in a new tab Phenological calendar of major crops in study areas U1-U3. The other 9 study areas encompass diverse agricultural regions across North America and Asia, including: (1) seven U.S. counties designated as U4-U10 - Renville County (North Dakota), Butler County (Nebraska), DeWitt County (Illinois), Van Wert County (Ohio), Dodge County (Minnesota), Saline County (Missouri), and Marshall County (Indiana) in the relatively flat central U.S.; (2) two Chinese sites - Nehe City (Heilongjiang Province, C1) in the northeast plains and Lingbi County (Anhui Province, C2) in more mountainous terrain. All regions share similar agricultural timelines with soybean sowing from mid-April to late May and harvest between early September and late November. Primary crops include soybeans, corn, spring wheat, winter wheat, sunflowers, cotton, and rice. Based on farmland regularity, the sites are classified into three categories: regular fields (U1, U2, U7), relatively regular fields (U4-U6, U8), and fragmented fields (U3, U9, U10, C1, C2). 2.2. Data 2.2.1. Reference data The sample data comes from the Cropland Data Layer (CDL) published annually by the National Agricultural Statistics Service (NASS) of the United States Department of Agriculture (USDA) and the ChinaSoyArea10m created by Mei et al. (2024) . The CDL data have been generated annually since 1997 for all U.S. states and cover more than 100 crop types ( Copenhaver et al., 2021 ). Since 2008, the CDL has covered the entire continental U.S at a 30-meter resolution ( Lin et al., 2022b ). The CDL data had 85% to 95% accuracy in classifying major crops ( Tran et al., 2022 ), which is a high-quality reference data in crop mapping studies and widely used in various crop mapping studies ( Li et al., 2024 ; Tran et al., 2022 ). The ChinaSoyArea10m dataset has achieved good accuracy in comparisons with county and prefecture-level statistical yearbooks and ground sample verification, and can be used as a reference source for soybean sample data in China. 2.2.2. Sentinel-2 data The Sentinel-2 satellite, developed under the European Space Agency’s Copernicus program, provides multispectral observations across 13 spectral bands covering visible light, near-infrared (NIR), and shortwave infrared (SWIR) bands. Spatial resolution varies by band at 10 meters, 20 meters, and 60 meters, respectively. The Sentinel-2A/2B constellation has a revisit cycle of approximately 5 days ( Drusch et al., 2012 ; Segarra et al., 2020 ). This study selected 10 bands for analysis ( Table 2 ). A significant advantage of Sentinel-2 for agricultural applications is its inclusion of three red edge bands, which are highly sensitive to vegetation chlorophyll content and canopy structure, making them particularly important for crop identification ( Clevers and Gitelson, 2013 ; Qiao et al., 2024 ). To ensure data continuity, Sentinel-2 data underwent cloud removal and resampling to generate median-synthesized images at 10-day intervals. Missing data in heavily cloudy areas were filled using linear interpolation, and all bands were resampled to 10-meter resolution to maintain consistent spatial resolution. The Sentinel-2 multispectral imagery from 2020 to 2022 used in this study were obtained from the Google Earth Engine (GEE) platform. The 10-day median composite imagery were employed for spectral feature analysis and NSII construction, while single-phase imagery under the “optimal time window” were utilized for NSII validation ( Table 3 ). Table 2. Parameters of Sentinel 2 bands used in this study. Band number Band name Central wavelength/nm Band width/nm Spatial resolution/m 2 Blue 490 65 10 3 Green 560 35 10 4 Red 665 30 10 5 Red Edge-1 (RE1) 705 15 20 6 Red Edge-2 (RE2) 740 15 20 7 Red Edge-3 (RE3) 783 20 20 8 NIR 842 115 10 8A Narrow-NIR 865 20 20 11 SWIR1 1610 90 20 12 SWIR2 2190 180 20 Open in a new tab Table 3. Sentinel-2 data within the optimal time window for each study area. Sites Image acquisition date Number of scenes August 19, 2020 4 U1 August 24, 2021 4 August 4, 2022 4 August 10, 2020 4 U2 August 23, 2021 4 August 23, 2022 4 August 23, 2020 4 U3 September 3, 2021 4 August 8, 2022 3 August 24, 2020 4 U4 August 14, 2021 4 August 14, 2022 4 August 25, 2020 2 U5 August 5, 2021 2 August 10, 2022 2 July 29, 2020 1 U6 September 7, 2021 1 July 19, 2022 1 August 27, 2020 3 U7 August 5, 2021 3 July 21, 2022 3 August 17, 2020 3 U8 August 4, 2021 3 August 4, 2022 3 August 24, 2020 2 U9 August 9, 2021 2 July 20, 2022 2 August 20, 2020 2 U10 August 5, 2021 1 August 18, 2022 2 August 16, 2020 5 C1 August 31, 2021 5 August 16, 2022 5 August 17, 2020 2 C2 August 19, 2021 2 August 24, 2022 2 Open in a new tab 3. Methods The workflow of this study comprises four components ( Figure 3 ): (1) sample collection from different crops; (2) phenological analysis and spectral dynamics analysis; (3) determination of temporal windows and construction of the Novel Soybean Identification Index (NSII); (4) spatiotemporal transfer validation of NSII and evaluation of mapping accuracy. Figure 3. Open in a new tab Workflow for soybean mapping. 3.1. Crop type sample data acquisition The sample data consists of two parts: one part is used to develop a new soybean identification index, referred to as the “reference sample”; the other part is used to verify the accuracy of the soybean spatial distribution map based on this index, referred to as the “test sample”. All samples are uniformly distributed throughout the entire study area. The reference samples are exclusively sourced from the U1-U3 regions, obtained as follows: First, the agricultural land areas with annual cloud cover below 15% in 2021 are cropped out. Then, pixels with crop type confidence levels exceeding 95% are extracted from the CDL crop layer within this area. The test samples were categorized into two types: soybeans and non-soybeans. The acquisition method was similar to that of the reference samples, but after extracting pixels with crop type confidence levels exceeding 95%, a portion of random sampling points were further selected through random sampling. To reduce spatial autocorrelation between samples, the minimum distance between sample points was set to exceed 30 meters (i.e., 3 pixels). The final number of samples obtained for each study area is shown in Table 4 . Table 4. Sample information. Sample type Location Sample size Soybean Non-soybean Reference sample U1 416060 408882 U2 476971 1017095 U3 386471 346313 Test sample U1 9791 11269 U2 21831 31355 U3 14504 9767 U4 24411 52123 U5 11425 16398 U6 11017 11327 U7 13581 10807 U8 7875 12482 U9 17524 19585 U10 5421 10368 C1 14972 14518 C2 14740 12352 Open in a new tab 3.2. Time-series feature curves and separability analysis of crops Based on the preprocessed Sentinel-2 data, 35 vegetation indices were further calculated ( Table 5 ), and the feature curves of 10 spectral bands and the 35 vegetation indices at 10-day intervals during the soybean growing season were finally obtained. In addition, based on the reference sample obtained in section 3.1 Crop type sample data acquisition, we evaluated the separability between soybean and corn across features (May–October) using the Jeffries-Matusita (J-M) distance ( Qiu et al., 2014 ). The J-M distance can be effective in evaluating the separability between different crop types ( Wang et al., 2018 ) ( Equations 1 – 3 ). By analyzing temporal spectral feature curves and crop separability metrics, the optimal time window and features for soybean identification were determined. Table 5. Vegetation indices used in this study. Spectral index Formula References Normalized Difference Vegetation Index (NDVI) NDVI = ( B 8 − B 4 ) / ( B 8 + B 4 ) ( Rouse et al., 1974 ) Normalized Difference red-edge 1 (NDre1) NDre 1 = ( B 6 − B 5 ) / ( B 6 + B 5 ) ( Fernández-Manso et al., 2016 ) Normalized Difference red-edge 2 (NDre2) NDre 2 = ( B 7 − B 5 ) / ( B 7 + B 5 ) ( Fernández-Manso et al., 2016 ) Land Surface Water Index (LSWI) LSWI = ( B 8 − B 11 ) / ( B 8 + B 11 ) ( Xiao et al., 2005 ) Soil Adjusted Vegetation Index (SAVI) SAVI = 1.5 × ( B 8 − B 4 ) / ( B 8 + B 4 + 0.5 ) ( Huete, 1988 ) Normalized Difference Water Index (NDWI) NDWI = ( B 3 − B 8 ) / ( B 3 + B 8 ) ( McFeeters, 1996 ) Plant Senescence Reflectance Index (PSRI) PSRI = ( B 4 − B 3 ) / B 6 ( Fernández-Manso et al., 2016 ) Normalized Difference Vegetation Index red-edge 1 (NDVIre1) NDVI re 1 = ( B 8 − B 5 ) / ( B 8 + B 5 ) ( Gitelson and Merzlyak, 1997 ) Normalized Difference Vegetation Index red-edge 2 (NDVIre2) NDVI re 2 = ( B 8 − B 6 ) / ( B 8 + B 6 ) ( Gitelson and Merzlyak, 1997 ) Novel inverted red-edge chlorophyll index (IRECI) IRECI = ( B 8 − B 4 ) / ( B 5 / B 6 ) ( Zhang et al., 2022a ) Chlorophyll Index red-edge (Cire) CIre = ( B 8 / B 5 ) − 1 ( Zhang et al., 2022a ) Ratio Vegetation Index (RVI) RVI = B 8 / B 4 ( Jordan, 1969 ) Wide Dynamic Range Vegetation Index (WDRVI) WDRVI = ( a × B 8 − B 4 ) / ( × B 8 − B 4 ) , a = 0.1 ( Gitelson, 2004 ) Non-Linear vegetation Index. (NLI) NLI = ( B 8 × B 8 − B 4 ) / ( B 8 × B 8 + B 4 ) ( Goel and Qin, 1994 ) Modified Non-linear vegetation Index (MNLI) MNLI = [ ( B 8 × B 8 − B 4 ) × ( 1 + 0.5 ) ] / ( B 8 × B 8 + B 4 + 0.5 ) ( Gong et al., 2003 ) Optimization of Soil-Adjusted Vegetation Indices (OSAVI) OSAVI = ( B 8 − B 4 ) / ( B 8 + B 4 + 0.16 ) ( Rondeaux et al., 1996 ) Enhanced Vegetation Index (EVI) EVI = 2.5 × ( B 8 − B 4 ) / ( B 8 + 6 × B 4 − 7.5 × B 2 + 1 ) ( Huete et al., 2002 ) Difference Vegetation Index (DVI) DVI = B 8 − B 4 ( Fernández-Manso et al., 2016 ) Bareness soil index (BSI) BSI = [ ( B 11 + B 4 ) − ( B 8 + B 2 ) ] / [ ( B 11 + B 4 ) + ( B 8 + B 2 ) ] ( Bera et al., 2020 ) Winter Rapeseed Index (WRI) WRI = [ ( B 8 − B 3 ) / B 8 + B 3 ] × [ B 2 / ( B 3 + B 4 ) ] ( Zhang et al., 2022b ) Ratio Vegetation Index Green (RVI Green) RVI Green = B 8 / B 3 ( Gitelson et al., 1996 ) Red light and red edge band ratio vegetation index (SR Red/Green ) SR Red/Green = B 4 / B 3 ( Chappelle et al., 1992 ) Visible Atmospherically Resistant Index (VARI Green ) VARI Green = ( B 3 − B 4 ) / ( B 3 + B 4 − B 2 ) ( Gitelson et al., 2002 ) Triangular Vegetation Index (TVI) TVI = a × ( b × ( B 8 − B 3 ) − c × ( B 4 − B 3 ) ) a = 0.5 , b = 120 , c = 200 ( Haboudane et al., 2004 ) MERIS Terrestrial Chlorophyll Index (MTCI) MTCI = ( B 6 − B 5 ) / ( B 5 − B 4 ) ( Dash and Curran, 2007 ) Modified Normalized Difference Vegetation Index (MNDVI) MNDVI = ( B 3 − B 11 ) / ( B 3 + B 11 ) ( Dihkan et al., 2013 ) Modified Simple Ratio Vegetation Index (MSAVI) MSAVI = [ 2 × B 8 + 1 − ( 2 × B 8 + 1 ) 2 − 8 × ( B 8 − B 4 ) ] / 2 ( Haboudane et al., 2004 ) Modified Simple Ratio red-edge (MSRre) MSRre = [ ( B 8 / B 5 ) − 1 ] / [ ( B 8 / B 5 ) − 1 ] ( Fernández-Manso et al., 2016 ) Modified Simple Ratio red-edge Narrow (MSRren) MSRren = [ ( B 8 A / B 5 ) − 1 ] / [ ( B 8 A / B 5 ) − 1 ] ( Fernández-Manso et al., 2016 ) Renormalized Difference Vegetation Index (RDVI) RDVI = ( B 8 − B 4 ) / ( B 8 + B 4 ) ( Haboudane et al., 2004 ) Simple Ratio (SR) SR = B 8 / B 4 ( Fernández-Manso et al., 2016 ) Red-Edge Position (REP) REP = 705 + 35 × [ 0.5 × ( B 4 + B 7 ) − B 5 ] / ( B 6 − B 5 ) ( Guyot et al., 1988 ) Normalized Burn Ratio (NBR) NBR = ( B 8 − B 12 ) / ( B 8 + B 12 ) ( Fernández-Manso et al., 2016 ) Greenness and Water Content Composite Index (GWCCI) GWCCI = B 11 × ( B 8 − B 4 ) / ( B 8 + B 4 ) ( Chen et al., 2023 ) Soybean Mapping Composite Index (SMCI) SMCI = TW Global { ( B 11 + B 8 + B 8 A + B 7 + B 6 ) × EVI × GCVI } EVI = 2.5 × ( B 8 − B 4 ) / ( B 8 + 6 × B 4 − 7.5 × B 2 + 1 ) , GCVI = ( B 8 / B 3 ) − 1 TW Global { } is the time period ( Xiao et al., 2024 ) Open in a new tab Bi denotes the i th band of Sentinel-2. JM ( j , k ) = ∫ x [ P ( x ∥ j ) − P ( x ∥ k ) ] 2 d x (1) where j and k denote two different crop types; x is the value for the different features. P ( x ∥ j ) and P ( x ∥ k ) are probability density functions (PDFs) of feature x for classes j and k , and Equation 1 can be simplified to Equation 2 JM = 2 ( 1 − e − B ) (2) B = 1 8 ( μ j − μ k ) T [ ( ∑ j + ∑ k ) / 2 ] − 1 ( μ j − μ k ) + 1 2 ln [ ( ∥ ( ∑ j + ∑ k ) / 2 ∥ ) / ( ∥ ∑ j ∥ ∥ ∑ k ∥ ) ] (3) Where μ j and μ k represent the average spectral reflectance of a single crop category, Σ j and Σ k are unbiased estimates of the covariance matrices of specific classes j and k . The J-M distance takes values in the range of 0 to 2. The greater the feature difference between the two crops, the greater the value of the J-M distance ( Gxokwe et al., 2022 ). Due to space constraints, the complete set of feature curves for different crops across 12 study regions and their corresponding separability analysis matrices are not fully presented. Among crops planted during the same period, corn exhibits the highest spectral confusion with soybean. Using U1 as a representative case, we present the temporal curves of some typical features for corn and soybean ( Figure 4 and Supplementary Material ) and the separability heatmap of different features ( Figure 5 ). The performance of the same feature varies in different regions, but overall, there are significant and stable feature differences between soybeans and corn from early August (DOY ≈ 220) to mid-September (DOY ≈ 260). During this time window, soybeans are primarily in the pod-filling stage, while corn progresses from silking to milk stage. In this phase, soybeans exhibit distinctive characteristics: rapid seed enlargement within pods, initial yellowing and shedding of lower leaves, significant reduction in moisture content, cessation of plant height growth, and increased canopy openness, leading to sharp declines in red-edge band reflectance and marked increases in short-wave infrared (SWIR) band reflectance. Conversely, corn reaches peak leaf area index (LAI) during silking while maintaining consistently high near-infrared (NIR) reflectance; upon entering the milk stage, basal leaves begin to yellow, red-edge indices show gradual decline, and SWIR reflectance remains relatively stable. These pronounced phenological differences and spectral characteristic variations provide critical discriminative basis for simultaneous crop classification during this growth period. Therefore, we identify the period from August to mid-September as the optimal temporal window for soybean identification. During the optimal temporal window, the spectral bands with high separability include the red-edge bands (RE2, RE3, RE4), NIR bands, and SWIR bands ( Figure 5 ). Among them, RE2 has the highest average separation degree (average J-M=0.50). The vegetation indices with high separability include the SMCI (average J-M=1.12), GWCCI (average J-M=1.08), and EVI (average J-M=0.88). Among them, SMCI and GWCCI are indices developed specifically for soybean identification. These bands and indices can be used as the basic features and references for the development of NSII in this study. Figure 4. Open in a new tab Some examples of feature curves. Figure 5. Open in a new tab J-M distance calculated based on different features. 3.3. Principles and methods of novel index construction Figure 5 shows that the red-edge bands (RE2, RE3, RE4), the NIR bands and the SWIR bands, SMCI, GWCCI, and EVI have a better ability to distinguish between soybeans and corn. RE2 performs best in the three red-edge bands; SWIR1 performs better in the two SWIR bands. SMCI uses eight bands for calculation and takes EVI into account. GWCCI is the product of SWIR1 and NDVI, using red, NIR, and SWIR bands for calculation. In order to comprehensively use different spectral information and reduce feature redundancy, we selected RE2 from three red-edge bands, SWIR1 from two SWIR bands, and EVI from vegetation indices to construct NSII. EVI can provide supplementary information of red and NIR bands. By performing mathematical processing on the three selected features (such as addition, subtraction, multiplication, division, and reciprocal transformation) to enhance the spectral feature information of soybeans, and selecting the mathematical combination with the highest average J-M distance under the optimal time window as the calculation formula for NSII ( Equations 4 , 5 ). The higher the NSII value, the higher the likelihood that the pixel is soybean. Comparative analysis of Figure 5 and 6 demonstrates that the NSII index developed in this study significantly outperforms existing indices in terms of J-M distance between soybean and corn. Moreover, this index consistently maintains an average separability above 1.05 for distinguishing soybean from other minor crops, such as barley, oats, cotton, Spring wheat, sunflower. Figure 6. Open in a new tab Spectral curves of soybeans and their corresponding crops and the separability of soybeans and corn (J-M distance). NSII = RE 2 × SWIR 1 × EVI 3 (4) EVI = G × ( NIR − Red ) / ( NIR + C 1 × Red − C 2 × Blue + L ) , G = 2.5 , C 1 = 6 , C 2 = 7.5 , L = 1 (5) where RE2, SWIR1, Red, NIR, and Blue correspond to the B6, B11, B4, B8, and B2 bands of Sentinel 2, respectively. 3.4. Validation and application of NSII To assess the effectiveness of NSII, we conducted a three-year (2020–2022) soybean mapping experiment at 12 locations in major soybean-producing regions in the United States and China, and compared the results with CDL and the ChinaSoyArea10m dataset. In addition, considering that SMCI has the highest separability among existing indices ( Figure 5 ), we further compared the results of NSII-based and SMCI-based soybean mapping. Specifically, the soybean planting areas were extracted by the automatic thresholding method. The steps for threshold determination are as follows: 1. Sample division: The “test sample” in each region is randomly split into two parts in a 7:3 ratio, with 70% used for threshold determination and 30% for accuracy verification. 2. Initial threshold range determination: calculate the minimum value ( T min ) and maximum value ( T max ) of NSII. 3. Threshold Calculation: Set the initial threshold T 1 = ( T min + T max ) / 2 and split the image into two parts. Compute four classification metrics: Overall Accuracy(OA), Precision (P), Recall (R), and F1 score. OA measures the percentage of correctly classified pixels. Precision represents the fraction of predicted soybean pixels that were actual soybeans, while Recall indicates the proportion of actual soybean pixels correctly identified. The F1 score, as the harmonic mean of Precision and Recall, provides a balanced performance assessment. Compare P and R: if P>R (indicating the commission errors are larger than the omission errors and the threshold is set too low), update the threshold as T 2 = ( T 1 + T max ) / 2 ; if P<R (indicating the omission errors are larger than the commission errors and the threshold is set too high), update the threshold as T 2 = ( T min + T 1 ) / 2 . 4. Iterative Optimization: Compare the F1 scores before and after updating the threshold. If the F1 score improves, continue iterating using the same logic. 5. Final Threshold and Verification: Repeat the process until the optimal threshold is obtained. Due to regional variability, the optimal thresholds differ across regions. Verify the accuracy using the remaining 30% of the samples. 4. Results 4.1. Soybean mapping results and accuracy evaluation Figure 7 shows the reference soybean distribution in 12 study areas and the mapped results based on NSII and SMCI methods, while Figure 8 provides detailed views of representative local areas from Figure 7 . Overall, compared to the SMCI-based soybean distribution maps, the NSII-based maps showed higher consistency with the CDL data in both regular (e.g., U1, U2, U7) and fragmented (e.g., U3, U9) study areas. The NSII results for regions C1 and C2 are also closer to the ChinaSoyArea10m dataset. In SMCI-based soybean distribution maps, there were significant omissions in U2, U3, U4, U6 and commission in U5. In contrast, the NSII method demonstrated significantly improved accuracy in these corresponding study areas, markedly reducing both omission and commission errors of soybean pixel classification (as shown by the red circles in Figure 8 ). Figure 7. Open in a new tab Reference, SMCI-based and NSII-based soybean distribution maps in 12 study areas in 2021. Figure 8. Open in a new tab Zoom images of MSI (composite multispectral image of band8, 11 and 4), Reference dataset, SMCI-based, and NSII-based soybean distribution maps. The soybean mapping accuracy derived from the NSII was systematically quantified and comparatively analyzed against the SMCI method ( Table 6 ). The average precision, recall, F1 score, and accuracy of soybean mapping results based on SMCI in 12 regions are 0.79, 0.71, 0.74, and 0.77, respectively, while the average precision, recall, F1 score, and accuracy of soybean mapping results based on NSII are 0.79, 0.83, 0.80, and 0.85, respectively. Precision remained unchanged, while recall, F1 score, and accuracy increased by 12, 6, and 8 percentage points, respectively. This means that the soybean mapping results based on NSII have fewer omission errors. Overall, NSII outperformed SMCI in all regions, providing a better feature index for accurate mapping of soybean. Table 6. Accuracy comparison of NSII and SMCI at different sites. Sites Methods Precision Recall F1 score Accuracy U1 NSII 0.72 0.86 0.78 0.84 SMCI 0.74 0.73 0.73 0.78 U2 NSII 0.71 0.85 0.77 0.83 SMCI 0.74 0.70 0.72 0.75 U3 NSII 0.75 0.85 0.80 0.85 SMCI 0.78 0.67 0.72 0.7 U4 NSII 0.74 0.87 0.80 0.89 SMCI 0.78 0.72 0.75 0.79 U5 NSII 0.80 0.85 0.82 0.84 SMCI 0.76 0.76 0.76 0.8 U6 NSII 0.75 0.92 0.83 0.91 SMCI 0.80 0.75 0.77 0.81 U7 NSII 0.81 0.89 0.85 0.92 SMCI 0.84 0.80 0.82 0.82 U8 NSII 0.81 0.84 0.82 0.83 SMCI 0.74 0.73 0.73 0.79 U9 NSII 0.79 0.88 0.83 0.85 SMCI 0.76 0.79 0.77 0.78 U10 NSII 0.83 0.90 0.86 0.89 SMCI 0.79 0.81 0.80 0.83 C1 NSII 0.94 0.61 0.73 0.78 SMCI 0.91 0.56 0.69 0.75 C2 NSII 0.94 0.67 0.78 0.80 SMCI 0.89 0.52 0.65 0.70 Open in a new tab 4.2. Evaluation of NSII’s multi-year mapping results and transferability In order to better evaluate the performance of NSII over multiple years, distribution mapping was conducted in 12 study areas for three consecutive years. Figure 9 shows the distribution of soybeans in all regions in 2020 and 2022, and Figure 10 shows the soybean field distribution accuracy and F1 score in each study area from 2020 to 2022. Figure 9. Open in a new tab Soybean distribution maps based on NSII in 2020 and 2022. Figure 10. Open in a new tab Accuracy and F1 score of soybean field distribution in each study area from 2020 to 2022. Comparing Figure 7 and Figure 9 , it is found that there are obvious changes in the distribution and area of soybeans in some areas. For example, from 2020 to 2022, the soybean planting area of U4 and U10 gradually increased, while the soybean planting area of U1 showed a downward trend within three years. These changes are consistent with the trends of CDL and soybean planting area statistics in the USDA QuickStats database. The soybean planting area of U6 and U9 has remained basically unchanged from 2020 to 2022. The soybean cultivated area in C1 exhibited a concave trend (decrease in 2021and recovery in 2022), whereas C2 maintained stable spatial distribution patterns (interannual variation). The observed soybean area dynamics in regions C1 and C2 show strong consistency with the official statistics from China Statistical Yearbook. Overall, the 2020–2022 mapping results demonstrated logical consistency across all study regions. From Figure 10 , it can be seen that the accuracy and F1 score of each study area changed very little among three years. Specifically, the average F1 score for all regions from 2020 to 2022 was 0.82, with an average standard deviation of 0.03; the average accuracy for all regions from 2020 to 2022 was 0.85, with an average standard deviation of 0.03. This fully demonstrates the excellent stability of NSII over the years. 5. Discussion Among all the Sentinel-2 bands involved in the calculation of NSII, the separability of RE2 was the best, and the other red-edge bands also showed high separability, which indicated that the red-edge bands were very effective in distinguishing soybeans and corn, which was in line with the previous studies ( Feng et al., 2019 ). The red edge band is located between the visible and NIR bands, corresponding to the transition region where the spectral reflectance of vegetation increases sharply. It is highly sensitive to the chlorophyll content, leaf structure and biomass of vegetation, and is a key indicator band for the physiological state of vegetation. The red-edge band is less reflective of non-vegetation backgrounds such as soil, which helps to minimize the effect of mixed pixels and highlights the canopy information of soybean. The NIR and SWIR1 bands are also important for identifying soybeans, which is consistent with the conclusions of Yin et al. (2020) and She et al. (2020) . In addition to the original bands, some remote sensing indices are better than the spectral bands for identifying soybeans, such as SMCI, GWCCI, EVI and REP. These indices are mostly correlated with RE, NIR, and SWIR bands. This further proves the effectiveness of these bands in soybean mapping. There is not much difference between soybeans and corn in the early stage of growth, while corn is significantly higher than soybeans in the middle stage of growth. When the transpiration rate of corn is faster than that of soybeans, more water will be stored in the leaves and canopy, resulting in a smaller SWIR value. In addition, the difference in canopy greenness between soybean and corn is more pronounced during the mid-growth stage ( Xiao et al., 2024 ). In this study, we combined the separability analysis results with the unique biophysical characteristics of soybean in the middle of the growing season to design NSII, a remote sensing identification feature for the middle of the growing season of soybean. The mid-season stage of soybean growth (from planting to pod set) is a critical period for determining yield and quality ( Zeleke and Nendel, 2024 ). This stage requires careful management of fertilization, irrigation and pest control to ensure plant health and soybean pod formation. Real-time monitoring of soybeans can help optimize resource allocation and improve the efficiency and sustainability of agricultural production. However, the spectral overlap between soybean and other major crops, such as corn, poses a significant challenge for timely and accurate soybean mapping ( Lin et al., 2022a ; Zhang et al., 2020 ). Traditional remote sensing indices, such as NDVI, are often ineffective for crop identification during the peak growing season. This is because high levels of crop greenness, water content, and other biophysical parameters lead to NDVI saturation ( Mutanga et al., 2023 ). To address this limitation, researchers have developed novel spectral indices to improve the discrimination of soybean from other co-occurring crops and to provide more precise decision-making support for agricultural management. While significant progress has been made, existing methods still have room for improvement. For example, the GWCCI proposed by Chen et al. (2023) underutilizes the red-edge bands, which are critical for capturing subtle differences in crop characteristics. The SMCI proposed by Xiao et al. (2024) employs eight spectral bands, resulting in computational complexity and redundant band information. NSII is proposed in this study to overcome above limitations and it offers the following advantages: 1. Simple calculation process. It adopts a simple and easy-to-understand calculation process. By screening spectral bands and remote sensing indices, it efficiently combines bands and indices with higher J-M distances for soybean identification. This reduces computational complexity, saves time and computational resources for data processing, and thereby improves the overall efficiency of mapping. 2. Good soybean identification capability. The analysis of J-M distance shows that the separability of NSII > SMCI > GWCCI ( Figure 5 and Figure 6 ). Additionally, the soybean identification accuracy based on NSII is higher than that based on SMCI. 3. High stability and transferability. The NSII index has been tested and applied over three years in multiple regions with different planting structures and climatic conditions, consistently achieving high accuracy (Accuracy>80%). This demonstrates its strong spatiotemporal transferability. 4. Potential for mid-season mapping. NSII does not require data from the entire growing season. It can generate soybean distribution maps using data from the optimal time window, which is relatively long. This facilitates the acquisition of high-quality data and enhances the practical applicability of NSII. The focus of this study is on developing and validating NSII, therefore only a simple threshold method was used for testing in soybean classification. Unlike supervised classification methods, threshold methods only require a single feature band. However, NSII can be combined with other indices to use supervised classifiers for single target crop (such as soybean) or multi-target crop (such as soybean and corn) classification, in order to improve the overall accuracy of crop mapping. In recent years, Sentinel-2 imagery has been widely applied to soybean and winter crop classification due to its high spatial resolution and advantages in the red edge band ( Sabljić et al., 2024 ). Existing studies indicate that Sentinel-2 data, when combined with machine learning classifiers such as random forests and support vector machines, demonstrate excellent performance in soybean mapping and achieve high classification accuracy across diverse agroecological regions ( Feng et al., 2019 ; Wei et al., 2023 ). Similarly, Sentinel-2 imagery has been successfully applied to winter crop identification ( Bajić et al., 2022 ), particularly by leveraging the red edge and shortwave infrared bands to capture phenological differences ( Liu et al., 2021 ; Sun et al., 2020 ). These studies highlight the importance of spectral feature selection and phenological timing in crop discrimination. Although NSII has significant advantages over traditional soybean mapping, there are still some limitations. Firstly, due to factors such as variety, climate conditions, and planting structure, there are differences in the optimal time window required for different needs, and it is necessary to determine the optimal time window for the target area before calculation. To facilitate the application of NSII in new regions, this paper provides supplementary guidance on determining the optimal time window. Based on the findings of this study, the optimal time window typically corresponds to the mid-growth stage of soybeans, specifically the transition from vegetative growth to pod-setting. At this stage, soybean canopy structure, chlorophyll content, and moisture status exhibit significant differences from corn. In temperate regions, this phase generally occurs between days 180–240, though the exact timing may vary depending on local climate and planting dates. When applying NSII to new regions, users are advised to first analyze temporal vegetation indices (e.g., spectral bands, NDVI, EVI, and other remote sensing indices) to identify the rapid growth phase and peak canopy development period. Subsequently, separability analysis between soybeans and major companion crops (e.g., J-M distance) can be conducted within this phenological window to identify the optimal time range. This phenology-guided approach enhances the transferability and robustness of NSII across diverse agroecological regions. Secondly, in regions with frequent cloud cover, such as tropical or monsoon climates, the applicability of NSII may be constrained. During optimal observation windows, high cloud cover significantly reduces the availability of high-quality optical imagery. While 10-day synthesis and interpolation can partially compensate for observation gaps, persistent cloud contamination still compromises the reliability of spectral features and mapping accuracy. To overcome this limitation, future research could explore multi-sensor fusion strategies, such as combining Sentinel-2 optical data with cloud-penetrating Sentinel-1 SAR imagery or integrating Landsat data. Data fusion approaches could further enhance NSII’s robustness and applicability in heavily cloud-covered regions. 6. Conclusions This study developed a novel soybean identification index (NSII) to address the challenges of spectral overlap. By integrating the RE2, SWIR1, and EVI features, NSII significantly improves the separability of soybean from other crops, particularly corn, within the optimal time window. Results from soybean fields in 12 regions between 2020 and 2022 showed that NSII had an average accuracy of 0.85 and an average F1 score of 0.80, which were 8 and 6 percentage points higher than those of SMCI, respectively. NSII’s computationally efficient design reduces computational complexity, while its stability and transferability across diverse climatic and cropping conditions highlight its practical applicability for large-scale soybean mapping. Furthermore, NSII’s ability to operate within a relatively long optimal time window enhances its feasibility for real-time monitoring and mid-season mapping. These advantages make NSII a valuable tool for precision agriculture, enabling efficient resource allocation, improved crop management, and enhanced food security. Future research could explore the integration of NSII with machine learning models or its application in other regions with complex cropping systems to further validate its robustness and scalability. Funding Statement The author(s) declared that financial support was received for this work and/or its publication. This research was funded by the National Key Research and Development Program of China (project No. 2023YFB3906201). Footnotes Edited by: Yuanrun Zheng , Chinese Academy of Sciences (CAS), China Reviewed by: Anastasios Mazis , Corteva Agriscience, United States Luka Sabljic , University of Banjaluka, Bosnia and Herzegovina Data availability statement The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors. Author contributions XZ: Writing – review & editing, Methodology, Conceptualization. YW: Writing – review & editing, Validation, Writing – original draft, Data curation, Methodology. AZ: Writing – review & editing. DL: Writing – review & editing. Conflict of interest The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Generative AI statement The author(s) declared that generative AI was not used in the creation of this manuscript. Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us. Publisher’s note All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher. Supplementary material The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2026.1788686/full#supplementary-material Image1.jpeg (14.6MB, jpeg) References Agilandeeswari L., Prabukumar M., Radhesyam V., Phaneendra K. L. N. B., Farhan A. (2022). Crop classification for agricultural applications in hyperspectral remote sensing images. Appl. Sci. 12, 1670. doi:  10.3390/app12031670, PMID: 41725453 [ DOI ] [ Google Scholar ] Ashourloo D., Shahrabi H. S., Azadbakht M., Rad A. M., Aghighi H., Radiom S. (2020). A novel method for automatic potato mapping using time series of Sentinel-2 images. Comput. Electron. Agric. 175, 105583. doi:  10.1016/j.compag.2020.105583, PMID: 41916819 [ DOI ] [ Google Scholar ] Bajić D., Adžić D., Sabljić L. (2022). Winter crops classification using combination of multi-temporal optical Sentinel-2 and radar Sentinel-1 images. Herald 26, 27–50. doi:  10.7251/HER2226027B [ DOI ] [ Google Scholar ] Bandaru V., Yaramasu R., Koutilya P., He J., Fernando S., Sahajpal R., et al. (2020). PhenoCrop: An integrated satellite-based framework to estimate physiological growth stages of corn and soybeans. Int. J. Appl. Earth Observation Geoinformation 92, 102188. doi:  10.1016/j.jag.2020.102188, PMID: 41916819 [ DOI ] [ Google Scholar ] Beck H. E., Zimmermann N. E., McVicar T. R., Vergopolan N., Berg A., Wood E. F. (2018). Data Descriptor Present and future Köppen-Geiger climate classification maps at 1-km resolution. Sci. Data 5, 1–12. doi:  10.1038/sdata.2018.214, PMID: [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Bera B., Saha S., Bhattacharjee S. (2020). Estimation of forest canopy cover and forest fragmentation mapping using landsat satellite data of Silabati River Basin (India). KN-journal cartography geographic Inf. 70, 181–197. doi:  10.1007/s42489-020-00060-1, PMID: 41913934 [ DOI ] [ Google Scholar ] Chappelle E. W., Kim M. S., McMurtrey Iii J. E. (1992). Ratio analysis of reflectance spectra (RARS): an algorithm for the remote estimation of the concentrations of chlorophyll a, chlorophyll b, and carotenoids in soybean leaves. Remote Sens. Environ. 39, 239–247. doi:  10.1016/0034-4257(92)90089-3 [ DOI ] [ Google Scholar ] Chen H., Li H., Liu Z., Zhang C., Zhang S., Atkinson P. M. (2023). A novel Greenness and Water Content Composite Index (GWCCI) for soybean mapping from single remotely sensed multispectral images. Remote Sens. Environ. 295, 113679. doi:  10.1016/j.rse.2023.113679, PMID: 41916819 [ DOI ] [ Google Scholar ] Choukri M., Laamrani A., Chehbouni A. (2024). Use of optical and radar imagery for crop type classification in africa: A review. Sensors 24, 3618. doi:  10.3390/s24113618, PMID: [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Clevers J. G. P. W., Gitelson A. A. (2013). Remote estimation of crop and grass chlorophyll and nitrogen content using red-edge bands on Sentinel-2 and -3. Int. J. Appl. Earth Observation Geoinformation 23, 344–351. doi:  10.1016/j.jag.2012.10.008, PMID: 41916819 [ DOI ] [ Google Scholar ] Copenhaver K., Hamada Y., Mueller S., Dunn J. B. (2021). Examining the characteristics of the cropland data layer in the context of estimating land cover change. ISPRS Int. J. geo-information 10, 281. doi:  10.3390/ijgi10050281, PMID: 41725453 [ DOI ] [ Google Scholar ] Dash J., Curran P. J. (2007). Evaluation of the MERIS terrestrial chlorophyll index (MTCI). Adv. Space Res. 39, 100–104. doi:  10.1016/j.asr.2006.02.034, PMID: 41916819 [ DOI ] [ Google Scholar ] de Oliveira Maia F. C., Bufon V. B., Leão T. P. (2023). Vegetation indices as a tool for mapping sugarcane management zones. Precis. Agric. 24, 213–234. doi:  10.1007/s11119-022-09939-7, PMID: 41913934 [ DOI ] [ Google Scholar ] Di Y., You N., Dong J., Liao X., Song K., Fu P. (2023). Recent soybean subsidy policy did not revitalize but stabilize the soybean planting areas in Northeast China. European J. of Agronomy. 147, 126841. doi:  10.1016/j.eja.2023.126841, PMID: 41916819 [ DOI ] [ Google Scholar ] Dihkan M., Guneroglu N., Karsli F., Guneroglu A. (2013). Remote sensing of tea plantations using an SVM classifier and pattern-based accuracy assessment technique. Int. J. Remote Sens. 34, 8549–8565. doi:  10.1080/01431161.2013.845317, PMID: 41909888 [ DOI ] [ Google Scholar ] Drusch M., Del Bello U., Carlier S., Colin O., Fernandez V., Gascon F., et al. (2012). Sentinel-2: ESA's optical high-resolution mission for GMES operational services. Remote Sens. Environ. 120, 25–36. doi:  10.1016/j.rse.2011.11.026, PMID: 41916819 [ DOI ] [ Google Scholar ] Farmonov N., Amankulova K., Szatmári J., Sharifi A., Abbasi-Moghadam D., Nejad S. M. M., et al. (2023). Crop type classification by DESIS hyperspectral imagery and machine learning algorithms. IEEE J. selected topics Appl. Earth observations Remote Sens. 16, 1576–1588. doi:  10.1109/JSTARS.2023.3239756, PMID: 41116384 [ DOI ] [ Google Scholar ] Feng S., Zhao J., Liu T., Zhang H., Zhang Z., Guo X. (2019). Crop type identification and mapping using machine learning algorithms and sentinel-2 time series data. IEEE J. Selected Topics Appl. Earth Observations Remote Sens. 12, 3295–3306. doi:  10.1109/JSTARS.2019.2922469, PMID: 41116384 [ DOI ] [ Google Scholar ] Fernández-Manso A., Fernández-Manso O., Quintano C. (2016). SENTINEL-2A red-edge spectral indices suitability for discriminating burn severity. Int. J. Appl. Earth observation geoinformation 50, 170–175. doi:  10.1016/j.jag.2016.03.005, PMID: 41916819 [ DOI ] [ Google Scholar ] Gao F., Zhang X. (2021). Mapping crop phenology in near real-time using satellite remote sensing: Challenges and opportunities. J. Remote Sens. 2021, 8379391. doi:  10.34133/2021/8379391, PMID: 38185834 [ DOI ] [ Google Scholar ] Gitelson A. A. (2004). Wide dynamic range vegetation index for remote quantification of biophysical characteristics of vegetation. J. Plant Physiol. 161, 165–173. doi:  10.1078/0176-1617-01176, PMID: [ DOI ] [ PubMed ] [ Google Scholar ] Gitelson A. A., Kaufman Y. J., Merzlyak M. N. (1996). Use of a green channel in remote sensing of global vegetation from EOS-MODIS. Remote Sens. Environ. 58, 289–298. doi:  10.1016/S0034-4257(96)00072-7, PMID: 41617830 [ DOI ] [ Google Scholar ] Gitelson A. A., Kaufman Y. J., Stark R., Rundquist D. (2002). Novel algorithms for remote estimation of vegetation fraction. Remote Sens. Environ. 80, 76–87. doi:  10.1016/S0034-4257(01)00289-9, PMID: 41908905 [ DOI ] [ Google Scholar ] Gitelson A. A., Merzlyak M. N. (1997). Remote estimation of chlorophyll content in higher plant leaves. Int. J. Remote Sens. 18, 2691–2697. doi:  10.1080/014311697217558, PMID: 41909888 [ DOI ] [ Google Scholar ] Goel N. S., Qin W. (1994). Influences of canopy architecture on relationships between various vegetation indices and LAI and FPAR: A computer simulation. Remote Sens. Rev. 10, 309–347. doi:  10.1080/02757259409532252, PMID: 41909888 [ DOI ] [ Google Scholar ] Gong P., Pu R., Biging G. S., Larrieu M. R. (2003). Estimation of forest leaf area index using vegetation indices derived from Hyperion hyperspectral data. IEEE Trans. Geosci. Remote Sens. 41, 1355–1362. doi:  10.1109/TGRS.2003.812910, PMID: 41116384 [ DOI ] [ Google Scholar ] Guyot G., Baret F., Major D. (1988). High spectral resolution: determination of spectral shifts between the red and near infrared. Int. Arch. Photogrammetry Remote Sens. 11, 750–760. [ Google Scholar ] Gxokwe S., Dube T., Mazvimavi D. (2022). Leveraging Google Earth Engine platform to characterize and map small seasonal wetlands in the semi-arid environments of South Africa. Sci. Total Environ. 803, 150139. doi:  10.1016/j.scitotenv.2021.150139, PMID: [ DOI ] [ PubMed ] [ Google Scholar ] Haboudane D., Miller J. R., Pattey E., Zarco-Tejada P. J., Strachan I. B. (2004). Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: Modeling and validation in the context of precision agriculture. Remote Sens. Environ. 90, 337–352. doi:  10.1016/j.rse.2003.12.013, PMID: 41916819 [ DOI ] [ Google Scholar ] Hanson E. D., Cossette M. K., Roberts D. C. (2022). The adoption and usage of precision agriculture technologies in North Dakota. Technol. Soc. 71, 102087. doi:  10.1016/j.techsoc.2022.102087, PMID: 41916819 [ DOI ] [ Google Scholar ] Hosseini M., Kerner H. R., Sahajpal R., Puricelli E., Lu Y.-H., Lawal A. F., et al. (2020). Evaluating the impact of the 2020 Iowa derecho on corn and soybean fields using synthetic aperture radar. Remote Sens. 12, 3878. doi:  10.3390/rs12233878, PMID: 41725453 [ DOI ] [ Google Scholar ] Huang X., Huang J., Li X., Shen Q., Chen Z. (2022). Early mapping of winter wheat in Henan province of China using time series of Sentinel-2 data. GIScience Remote Sens. 59, 1534–1549. doi:  10.1080/15481603.2022.2104999, PMID: 41909888 [ DOI ] [ Google Scholar ] Huete A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sens. Environ. 25, 295–309. doi:  10.1016/0034-4257(88)90106-X [ DOI ] [ Google Scholar ] Huete A., Didan K., Miura T., Rodriguez E. P., Gao X., Ferreira L. G. (2002). Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 83, 195–213. doi:  10.1016/S0034-4257(02)00096-2, PMID: 41881759 [ DOI ] [ Google Scholar ] Jordan C. F. (1969). Derivation of leaf-area index from quality of light on the forest floor. Ecology 50, 663–666. doi:  10.2307/1936256, PMID: 39964225 [ DOI ] [ Google Scholar ] Li H., Di L., Zhang C., Lin L., Guo L., Eugene G. Y., et al. (2024). Automated in-season crop-type data layer mapping without ground truth for the Conterminous United States based on multisource satellite imagery. IEEE Trans. Geosci. Remote Sens. 62, 1–14. doi:  10.1109/TGRS.2024.3361895, PMID: 41116384 [ DOI ] [ Google Scholar ] Li X., Yu L., Peng D., Gong P. (2021). A large-scale, long time-series, (1984‒–2020) of soybean mapping with phenological features: Heilongjiang Province as a test case. Int. J. Remote Sens. 42, 7332–7356. doi:  10.1080/01431161.2021.1957177, PMID: 41909888 [ DOI ] [ Google Scholar ] Li H., Zhang C., Zhang S., Atkinson P. M. (2019). Full year crop monitoring and separability assessment with fully-polarimetric L-band UAVSAR: A case study in the Sacramento Valley, California. Int. J. Appl. Earth Observation Geoinformation 74, 45–56. doi:  10.1016/j.jag.2018.08.024, PMID: 41916819 [ DOI ] [ Google Scholar ] Lin L., Di L., Zhang C., Guo L., Di Y., Li H., et al. (2022. b). Validation and refinement of cropland data layer using a spatial-temporal decision tree algorithm. Sci. Data 9, 63. doi:  10.1038/s41597-022-01169-w, PMID: [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Lin C., Zhong L., Song X.-P., Dong J., Lobell D. B., Jin Z. (2022. a). Early-and in-season crop type mapping without current-year ground truth: Generating labels from historical information via a topology-based approach. Remote Sens. Environ. 274, 112994. doi:  10.1016/j.rse.2022.112994, PMID: 41916819 [ DOI ] [ Google Scholar ] Liu S., Chen Y., Ma Y., Kong X., Zhang X., Zhang D. (2020). Mapping ratoon rice planting area in Central China using Sentinel-2 time stacks and the phenology-based algorithm. Remote Sens. 12, 3400. doi:  10.3390/rs12203400, PMID: 41725453 [ DOI ] [ Google Scholar ] Liu Y., Qian J., Yue H. (2021). Comprehensive evaluation of Sentinel-2 red edge and shortwave-infrared bands to estimate soil moisture. IEEE J. Selected Topics Appl. Earth Observations Remote Sens. 14, 7448–7465. doi:  10.1109/JSTARS.2021.3098513, PMID: 41116384 [ DOI ] [ Google Scholar ] McFeeters S. K. (1996). The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int. J. Remote Sens. 17, 1425–1432. doi:  10.1080/01431169608948714, PMID: 41909888 [ DOI ] [ Google Scholar ] Mei Q., Zhang Z., Han J., Song J., Dong J., Wu H., et al. (2024). ChinaSoyArea10m: a dataset of soybean-planting areas with a spatial resolution of 10 m across China from 2017 to 2021. Earth System Sci. Data 16, 3213–3231. doi:  10.5194/essd-16-3213-2024, PMID: 41899199 [ DOI ] [ Google Scholar ] Mutanga O., Masenyama A., Sibanda M. (2023). Spectral saturation in the remote sensing of high-density vegetation traits: A systematic review of progress, challenges, and prospects. ISPRS J. Photogrammetry Remote Sens. 198, 297–309. doi:  10.1016/j.isprsjprs.2023.03.010, PMID: 41916819 [ DOI ] [ Google Scholar ] Qiao K., Zhu W., Xie Z., Wu S., Li S. (2024). New three red-edge vegetation index (VI3RE) for crop seasonal LAI prediction using Sentinel-2 data. Int. J. Appl. Earth Observation Geoinformation 130, 103894. doi:  10.1016/j.jag.2024.103894, PMID: 41916819 [ DOI ] [ Google Scholar ] Qiu B., Fan Z., Zhong M., Tang Z., Chen C. (2014). A new approach for crop identification with wavelet variance and JM distance. Environ. Monit. Assess. 186, 7929–7940. doi:  10.1007/s10661-014-3977-1, PMID: [ DOI ] [ PubMed ] [ Google Scholar ] Qu C., Li P., Zhang C. (2021). A spectral index for winter wheat mapping using multi-temporal Landsat NDVI data of key growth stages. ISPRS J. photogrammetry Remote Sens. 175, 431–447. doi:  10.1016/j.isprsjprs.2021.03.015, PMID: 41916819 [ DOI ] [ Google Scholar ] Rondeaux G., Steven M., Baret F. (1996). Optimization of soil-adjusted vegetation indices. Remote Sens. Environ. 55, 95–107. doi:  10.1016/0034-4257(95)00186-7 [ DOI ] [ Google Scholar ] Rouse J. W., Haas R. H., Schell J. A., Deering D. W. (1974). Monitoring vegetation systems in the Great Plains with ERTS. In Proceedings of Third Earth Resources Technology Satellite-1 Symposium. Freden S. C., Becker M. (Eds.), (pp. 310–317). NASA (Greenbelt, MD, USA: National Aeronautics and Space Administration; ). (NASA SP-351) [ Google Scholar ] Sabljić L., Lukić T., Bajić D., Marković R., Spalević V., Delić D., et al. (2024). Optimizing agricultural land use: A GIS-based assessment of suitability in the Sana River Basin, Bosnia and Herzegovina. Open Geosciences 16, 20220683. doi:  10.1515/geo-2022-0683, PMID: 41717541 [ DOI ] [ Google Scholar ] Segarra J., Buchaillot M. L., Araus J. L., Kefauver S. C. (2020). Remote sensing for precision agriculture: Sentinel-2 improved features and applications. Agronomy 10, 641. doi:  10.3390/agronomy10050641, PMID: 41725453 [ DOI ] [ Google Scholar ] She B., Yang Y., Zhao Z., Huang L., Liang D., Zhang D. (2020). Identification and mapping of soybean and maize crops based on Sentinel-2 data. Int. J. Agric. Biol. Eng. 13, 171–182. doi:  10.25165/j.ijabe.20201306.6183 [ DOI ] [ Google Scholar ] Shi J., Huang J. (2015). Monitoring spatio-temporal distribution of rice planting area in the Yangtze River Delta region using MODIS images. Remote Sens. 7, 8883–8905. doi:  10.3390/rs70708883, PMID: 41725453 [ DOI ] [ Google Scholar ] Sidike P., Sagan V., Maimaitijiang M., Maimaitiyiming M., Shakoor N., Burken J., et al. (2019). dPEN: Deep Progressively Expanded Network for mapping heterogeneous agricultural landscape using WorldView-3 satellite imagery. Remote Sens. Environ. 221, 756–772. doi:  10.1016/j.rse.2018.11.031, PMID: 41916819 [ DOI ] [ Google Scholar ] Sun Y., Qin Q., Ren H., Zhang T., Chen S. (2020). Red-edge band vegetation indices for leaf area index estimation from Sentinel-2/MSI imagery. IEEE Trans. Geosci. Remote Sens. 58, 826–840. doi:  10.1109/TGRS.2019.2940826, PMID: 41116384 [ DOI ] [ Google Scholar ] Tao J.-B., Zhang X.-Y., Wu Q.-F., Yun W. (2023). Mapping winter rapeseed in South China using Sentinel-2 data based on a novel separability index. J. Integr. Agric. 22, 1645–1657. doi:  10.1016/j.jia.2022.10.008, PMID: 41916819 [ DOI ] [ Google Scholar ] Tran K. H., Zhang H. K., McMaine J. T., Zhang X., Luo D. (2022). 10 m crop type mapping using Sentinel-2 reflectance and 30 m cropland data layer product. Int. J. Appl. Earth Observation Geoinformation 107, 102692. doi:  10.1016/j.jag.2022.102692, PMID: 41916819 [ DOI ] [ Google Scholar ] Volkova E., Smolyaninova N. (2024). “ Analysis of world trends in soybean production,” in BIO web of conferences. (Les Ulis: EDP Sciences; ). 141, 01026. doi:  10.1051/bioconf/202414101026 [ DOI ] [ Google Scholar ] Wang Y., Qi Q., Liu Y. (2018). Unsupervised segmentation evaluation using area-weighted variance and Jeffries-Matusita distance for remote sensing images. Remote Sens. 10, 1193. doi:  10.3390/rs10081193, PMID: 41725453 [ DOI ] [ Google Scholar ] Wang N., Zhai Y., Zhang L. (2021). Automatic cotton mapping using time series of Sentinel-2 images. Remote Sens. 13, 1355. doi:  10.3390/rs13071355, PMID: 41725453 [ DOI ] [ Google Scholar ] Wei P., Ye H., Qiao S., Liu R., Nie C., Zhang B., et al. (2023). Early crop mapping based on Sentinel-2 time-series data and the random forest algorithm. Remote Sens. 15, 3212. doi:  10.3390/rs15133212, PMID: 41725453 [ DOI ] [ Google Scholar ] Wu B., Zhang M., Zeng H., Tian F., Potgieter A. B., Qin X., et al. (2023). Challenges and opportunities in remote sensing-based crop monitoring: A review. Natl. Sci. Rev. 10, nwac290. doi:  10.1093/nsr/nwac290, PMID: [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Xiao X., Boles S., Liu J., Zhuang D., Frolking S., Li C., et al. (2005). Mapping paddy rice agriculture in southern China using multi-temporal MODIS images. Remote Sens. Environ. 95, 480–492. doi:  10.1016/j.rse.2004.12.009, PMID: 41916819 [ DOI ] [ Google Scholar ] Xiao G., Huang J., Song J., Li X., Du K., Huang H., et al. (2024). A novel soybean mapping index within the global optimal time window. ISPRS J. Photogrammetry Remote Sens. 217, 120–133. doi:  10.1016/j.isprsjprs.2024.08.006, PMID: 41916819 [ DOI ] [ Google Scholar ] Xu J., Yang J., Xiong X., Li H., Huang J., Ting K., et al. (2021). Towards interpreting multi-temporal deep learning models in crop mapping. Remote Sens. Environ. 264, 112599. doi:  10.1016/j.rse.2021.112599, PMID: 41916819 [ DOI ] [ Google Scholar ] Yang G., Huang K., Sun W., Meng X., Mao D., Ge Y. (2022). Enhanced mangrove vegetation index based on hyperspectral images for mapping mangrove. ISPRS J. Photogrammetry Remote Sens. 189, 236–254. doi:  10.1016/j.isprsjprs.2022.05.003, PMID: 41916819 [ DOI ] [ Google Scholar ] Yin L., You N., Zhang G., Huang J., Dong J. (2020). Optimizing feature selection of individual crop types for improved crop mapping. Remote Sens. 12, 162. doi:  10.3390/rs12010162, PMID: 41725453 [ DOI ] [ Google Scholar ] Zeleke K., Nendel C. (2024). Yield response and water productivity of soybean (Glycine max L.) to deficit irrigation and sowing time in south-eastern Australia. Agric. Water Manage. 296, 108815. doi:  10.1016/j.agwat.2024.108815, PMID: 41916819 [ DOI ] [ Google Scholar ] Zhang H., Kang J., Xu X., Zhang L. (2020). Accessing the temporal and spectral features in crop type mapping using multi-temporal Sentinel-2 imagery: A case study of Yi’an County, Heilongjiang province, China. Comput. Electron. Agric. 176, 105618. doi:  10.1016/j.compag.2020.105618, PMID: 41916819 [ DOI ] [ Google Scholar ] Zhang H., Li J., Liu Q., Lin S., Huete A., Liu L., et al. (2022. a). A novel red-edge spectral index for retrieving the leaf chlorophyll content. Methods Ecol. Evol. 13, 2771–2787. doi:  10.1111/2041-210X.13994, PMID: 41875165 [ DOI ] [ Google Scholar ] Zhang H., Liu W., Zhang L. (2022. b). Seamless and automated rapeseed mapping for large cloudy regions using time-series optical satellite imagery. ISPRS J. Photogrammetry Remote Sens. 184, 45–62. doi:  10.1016/j.isprsjprs.2021.12.001, PMID: 41916819 [ DOI ] [ Google Scholar ] Zhong L., Hu L., Yu L., Gong P., Biging G. S. (2016). Automated mapping of soybean and corn using phenology. ISPRS J. Photogrammetry Remote Sens. 119, 151–164. doi:  10.1016/j.isprsjprs.2016.05.014, PMID: 41916819 [ DOI ] [ Google Scholar ] Zhu M., She B., Huang L., Zhang D., Xu H., Yang X. (2022). Identification of soybean based on Sentinel-1/2 SAR and MSI imagery under a complex planting structure. Ecol. Inf. 72, 101825. doi:  10.1016/j.ecoinf.2022.101825, PMID: 41916819 [ DOI ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Image1.jpeg (14.6MB, jpeg) Data Availability Statement The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors. Articles from Frontiers in Plant Science are provided here courtesy of Frontiers Media SA ACTIONS View on publisher site PDF (23.8 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top