Bingqing Liu

Bingqing Liu

Auto-generated profile, based on publicly available information

Bingqing Liu

Florida State University, EOAS Building, 1011 Academic Way, Tallahassee, FL, United States [US

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 May 2025 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Open access
JLJiadong Lou
Jiadong Lou
BLBingqing Liu
Bingqing Liu
YXYuanheng Xiong
Yuanheng Xiong
XZXiaodong Zhang
Xiaodong Zhang
XYXu Yuan
Xu Yuan
Phytoplankton absorb and scatter light in unique ways, subtly altering the color of water—changes that are often minor for human eyes to detect but can be captured by sensitive ocean color instruments onboard satellites from space. Hyperspectral sensors, paired with advanced algorithms, are expected to significantly enhance the characterization of phytoplankton community composition (PCC), especially in coastal waters where ocean color remote sensing applications have historically encountered significant challenges. This study presents novel machine learning-based solutions for NASA’s hyperspectral missions including EMIT (7nm), and PACE (2.5 nm), tackling high-fidelity retrievals of phytoplankton absorption coefficient (aphy) and chlorophyll a (Chl-a) from their hyperspectral remote sensing reflectance (Rrs). Given that a single Rrs spectrum may correspond to varied combinations of inherent optical properties (IOPs) and associated concentrations (e.g., Chl-a), the Variational Autoencoder (VAE) is used as a backbone in this study to handle such multi-distribution prediction problems. We first time tailor the VAE model with innovative designs to achieve hyperspectral retrievals of aphy and of Chl-a from hyperspectral Rrs in optically complex estuarine-coastal waters. Validation with extensive experimental observation demonstrates superior performance of the VAE models with high precision and low bias. The in-depth analysis of VAE’s advanced model structures and learning designs highlights the improvement and advantages of VAE-based solutions over the mixture density network (MDN) approach, particularly on high-dimensional data, such as PACE. Our study provides strong evidence that current EMIT and PACE hyperspectral data as well as the upcoming Surface Biology Geology (SBG) mission will open new pathways toward a better understanding of phytoplankton community dynamics in aquatic ecosystems when integrated with AI technologies.

6 January 2026 in International Journal of Applied Earth Observation and Geoinformation
Verified
Journal Article
Open access
AKAbhishek Kumar
Abhishek Kumar
Chintan B. ManiyarChintan B. Maniyar
Chintan B. Maniyar
NTNathan Tesfayi
Nathan Tesfayi
BGBrice Grunert
Brice Grunert
IFIsabella R. Fiorentino
Isabella R. Fiorentino
Deepak MishraDeepak Mishra
Deepak Mishra

25 May 2026 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
TMTalha Mahmood
Talha Mahmood
BLBingqing Liu
Bingqing Liu
CSCharles Sabin
Charles Sabin
MPMegan K. La Peyre
Megan K. La Peyre
CLCoy A. LeBlanc
Coy A. LeBlanc
XYXu Yuan
Xu Yuan
Benthic habitats, such as seagrass meadows and coral reefs, are vital components of shallow coastal ecosystems, but they are increasingly threatened by anthropogenic and climatic stressors. Traditional remote sensing and manual classification methods often struggle with fine-scale heterogeneity, label scarcity, computational inefficiency, and lack of operational capability for operational benthic mapping. To address these challenges, we introduce the Benthic Habitat U-Net (BH-UNet), a lightweight deep learning framework applied to high-resolution WorldView-2/3 multispectral imagery to map complex coastal benthic environments in Guam's Hagåtña Bay and Manell-Geus. The model integrates channel-wise attention and bilinear upsampling to enhance feature extraction while maintaining computational efficiency. Evaluated on a spatially and temporally independent datasets, BH-UNet demonstrates strong cross-site generalization, achieving patch-level Intersection-over-Union scores exceeding 0.98 and Dice coefficients above 0.99. These results demonstrate superior boundary preservation and broader adaptability compared to the baseline architectures, which exhibited significant overfitting and limited generalization. The efficiency of the framework enabled its application to time-series imagery, revealing interannual fluctuations in seagrass extent from 2010 to 2024. Notably, sharp declines in seagrass were observed in 2014 and 2015, which coincided with periods of strong marine heatwave intensity. These patterns highlight the value of high-temporal-resolution seagrass observations for capturing episodic ecological responses that would otherwise be difficult to detect, underscoring their potential for scalable and operational coastal monitoring. This work advances artificial intelligence (AI)-driven marine conservation by balancing computational efficiency with ecological precision, offering a pathway for adaptive management in data-limited coastal ecosystems.

20 August 2026
Verified
Preprint
Open access
CLCoy A LeBlanc
Coy A LeBlanc
BLBingqing Liu
Bingqing Liu
CEChisom O Emeghiebo
Chisom O Emeghiebo
NRNirmal Raila
Nirmal Raila
JVJorge A Villa
Jorge A Villa
KDKelly M Darnell
Kelly M Darnell
This study presents the first application of NASA’s EMIT hyperspectral observations for investigating submerged and emergent habitats within the rapidly evolving Atchafalaya River Delta and Estuary. Results showed that submerged aquatic vegetation (SAV) detectability was controlled primarily by vegetation canopy depth and water-column optical attenuation rather than vegetation density and species alone. Strong negative relationships were observed between canopy depth and spectral indices, including Normalized Difference Red Edge (NDRE) (=0.968) and the NIR/Green ratio (=0.954). Principal component analysis further demonstrated that dissolved organic carbon and colored dissolved organic matter substantially influenced SAV spectral attenuation and underwater light limitation, with PC2 strongly correlated with DOC (=0.968, 0.0015). Further, linear spectral unmixing resolved ecologically meaningful gradients among SAV, freshwater marsh, brackish marsh, saline marsh, and water optical classes throughout the delta system. The derived hyperspectral endmembers successfully reconstructed mixed EMIT spectra (; =0.819–0.999; RMSE = 0.008–0.115; SAM = 1.19–15.18) and distinguished dense and sparse SAV across a range of water depths, floating aquatic vegetation, wetland vegetation, and turbid-water classes in optically complex estuarine environments. SAV occupied approximately 26.6% (117.49 km 2 ) of the 441.77 km 2 study area, while high-density SAV (50% fractional abundance) represented 6.0% (26.63 km 2 ). Beyond habitat mapping, EMIT-derived SAV fractional abundance demonstrated broader ecological and biogeochemical applications through spatially explicit methane flux scaling. Overall, this study demonstrates that spaceborne hyperspectral imaging spectroscopy provides a more detailed view of habitat-rich deltaic systems and establishes a practical framework for monitoring submerged and emergent vegetation within optically complex and rapidly evolving deltaic environments.

26 February 2026 in Journal of Geophysical Research: Oceans
Verified
Journal Article
Open access
PDF file
JLJ. Li
J. Li
BLB. Liu
B. Liu
JLJ. Lou
J. Lou
XYX. Yuan
X. Yuan
EDEurico D'sa
Eurico D'sa
EHEmad Habib
Emad Habib
Hurricanes drive diverse estuarine phytoplankton responses and can trigger cascading ecological and physicochemical impacts. Capturing these short‐term dynamics requires high spatiotemporal resolution. Here, we applied a globally‐applicable coastal ocean color algorithm, Variational Autoencoder (VAE), to Sentinel‐2 MSI imagery for chlorophyll‐ a (Chl‐ a ) estimation and validated its strong performance across the northern Gulf coast of America (GoA) estuaries, including Galveston Bay (TX), Barataria‐Terrebonne Estuary (LA), Apalachicola Estuary (FL) and Tampa Bay (FL). The test set showed strong performance (MAE: 1.44 mg m −3 ; RMSE: 17.7 mg m −3 ; slope: 0.86; median symmetric accuracy: 30.33%). The validated VAE was then applied to 76 Sentinel‐2 MSI images to assess phytoplankton biomass responses to hurricanes Harvey (2017), Michael (2018), Ida (2021), Francine (2024), Helene (2024), and Milton (2024) in the GoA estuaries. Results showed that hurricane disturbances on Chl‐ a typically lasted 3–5 weeks. Estuarine waters west (left) of hurricane tracks showed a rapid decline in Chl‐ a (∼5 mg m −3 ) due to elevated turbidity from heavy rainfall, and wind‐driven flushing in the estuary, followed by a rebound over about two weeks, with Chl‐ a increasing approximately 10–15 mg m −3 above pre‐storm levels. In contrast, right‐side waters showed a slower response, likely from oligotrophic seawater intrusion driven by the hurricane's counterclockwise rotation. Post‐storm observations showed increased freshwater phytoplankton like chlorophytes and cyanobacteria dominating estuaries, while shelf‐waters exhibited elevated dinoflagellates (e.g., Karenia brevis bloom after Hurricane Milton). These results highlight the spatial heterogeneity of hurricane impacts on estuarine phytoplankton dynamics, which may trigger cascading effects on biogeochemical cycling and food webs, potentially prolonging ecosystem recovery.

28 March 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
RPRejane S. Paulino
Rejane S. Paulino
VMVitor S Martins
Vitor S Martins
CCCassia B. Caballero
Cassia B. Caballero
TLThainara Munhoz Alexandre de Lima
Thainara Munhoz Alexandre de Lima
SKSasha J. Kramer
Sasha J. Kramer
BLBingqing Liu
Bingqing Liu
Aquatic systems exhibit diverse and complex optical properties caused by mixtures of optically active constituents, and significant efforts have been made to characterize these constituents from satellite optical data. However, a fundamental challenge remains in translating spectral information into distinct, reduced, and intuitive representations for aquatic science applications. In this study, we introduce a new Orthogonal Components of Water (OCW) approach that summarizes the full visible-spectrum into three individual components, named Brightness, Greenness, and Blueness, to enhance the interpretation of spectral composition across aquatic systems. The OCW approach was developed using a rigorous selection of endmember spectra from a global hyperspectral dataset (n=12,867) to implement the Gram–Schmidt orthogonalization technique, and then per-component coefficients were derived for seven satellite missions, including PACE-OCI, Aqua/Terra MODIS, JPSS1/JPSS2 VIIRS, Sentinel-3 OLCI, Sentinel-2 MSI, Landsat8/9 OLI, and PlanetScope SuperDove. OCW applications were evaluated at both regional and global scales, and the results confirmed the reliability of each OCW in reducing spectral dimensionality while preserving the interpretability of dominant optical regimes. OCW Brightness showed a strong relationship with total suspended solids and non-algal particles, with Pearson’s correlation coefficients (r) ranging from approximately +0.46 to +0.77. OCW Greenness was highly correlated with phytoplankton features (r of +0.79 to +0.83), while OCW Blueness exhibited a negative correlation with increasing concentrations of optically relevant biogeochemical parameters (r values ranging from +0.08 to -0.84), and represents predominantly clear-water conditions. Together, these results confirm that OCWs are rich representations of aquatic spectral composition for both multispectral and hyperspectral applications.