Automated Surface Water Extraction Combining Sentinel-2 Imagery and OpenStreetMap Using Presence and Background Learning (PBL) Algorithm

Zhang, Zhiqiang; Zhang, Xinchang; Jiang, Xin; Xin, Qinchuan; Ao, Zurui; Zuo, Qiting; Chen, Liyan

doi:10.1109/jstars.2019.2936406

Cited by 15 publications

(8 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This result is similar to our 20 m index ensembleonly method that was applied across international sites. In [38], OpenStreetMap was used for automated mapping of surface water with Sentinel-2 imagery, which ultimately found high producer's and user's accuracies ranging from 84% to 99% and 94% to 99% respectively across several test sites in China. This suggests that the automated training data produced from this method is accurate.…”

Section: Discussionmentioning

confidence: 99%

“…These layers, such as OpenStreetMap, often have temporal mismatches with the imagery and may omit small bodies of water not located near major cities. One recent study used Sentinel-2 imagery overlaid with OpenStreetMap to automatically extract water pixels [38]. A subsequent study used fuzzy membership functions, spectral indexes, and color transformations for automated training data generation [30], whereas unsupervised multidimensional hierarchical clustering with spectral indexes and individual bands were used to automate classification in France with a kappa score just below 0.9 [39].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automated Training Data Generation from Spectral Indexes for Mapping Surface Water Extent with Sentinel-2 Satellite Imagery at 10 m and 20 m Resolutions

Lasko¹,

Maloney²,

Becker³

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

This study presents an automated methodology to generate training data for surface water mapping from a single Sentinel-2 granule at 10 m (4 band, VIS/NIR) or 20 m (9 band, VIS/NIR/SWIR) resolution without the need for ancillary training data layers. The 20 m method incorporates an ensemble of three spectral indexes with optimal band thresholds, whereas the 10 m method achieves similar results using fewer bands and a single spectral index. A spectrally balanced and randomly generated set of training data based on the index values and optimal thresholds is used to fit machine learning classifiers. Statistical validation compares the 20 m ensemble-only method to the 20 m ensemble method with a random forest classifier. Results show the 20 m ensemble-only method had an overall accuracy of 89.5% (±1.7%), whereas the ensemble method combined with the random forest classifier performed better, with a ~4.8% higher overall accuracy: 20 m method (94.3% (±1.3%)) with optimal spectral index and SWIR thresholds of −0.03 and 800, respectively, and 10 m method (93.4% (±1.5%)) with optimal spectral index and NIR thresholds of −0.01 and 800, respectively. Comparison of other supervised classifiers trained automatically with the framework typically resulted in less than 1% accuracy improvement compared with the random forest, suggesting that training data quality is more important than classifier type. This straightforward framework enables accurate surface water classification across diverse geographies, making it ideal for development into a decision support tool for water resource managers.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Automated Training Data Generation from Spectral Indexes for Mapping Surface Water Extent with Sentinel-2 Satellite Imagery at 10 m and 20 m Resolutions

Lasko¹,

Maloney²,

Becker³

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Ao et al used PBL to classify a single class such as building, tree, terrain, and power line from airborne light detection and ranging point cloud data [24]. Zhang et al applied the PBL algorithm to map surface water bodies using Sentinel-2 imagery and OpenStreetMap data [25].…”

Section: One-class Remote Sensing Classification Frommentioning

confidence: 99%

One-Class Remote Sensing Classification From Positive and Unlabeled Background Data

Elkan

2021

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

One-class classification is a common situation in remote sensing, where researchers aim to extract a single land type from remotely sensed data. Learning a classifier from labeled positive and unlabeled background data, which is the case-control sampling scenario, is efficient for one-class remote sensing classification because labeled negative data are not necessary for model training. In this study we propose a novel positive and background learning with constraints (PBLC) algorithm to address this one-class classification problem. With user-specified information of maximum probability as the constraint, PBLC infers the posterior probability of positive class directly in one-step model training. We test PBLC on a synthetic dataset and a real aerial photograph to perform different one-class classification tasks. Experimental results demonstrate that PBLC can successfully train linear and nonlinear classifiers including generalized linear model (GLM), artificial neural network (ANN), and convolutional neural network (CNN). Probabilistic and binary predictions by PBLC are more similar to the gold-standard positive-negative method, outperforming the two-step positive and background learning (PBL) algorithm that post-calibrates a naïve classifier based on an estimated constant. Hence, the proposed PBLC algorithm has the potential to solve one-class classification problems in the case-control sampling scenario. Index Terms-Case-control sampling, labeled and unlabeled data, one-class classification, positive and background learning with constraints (PBLC), remote sensing

show abstract

“…Landsat satellite sensors, including TM (Thematic Mapper), ETMþ (Enhanced Thematic Mapper Plus), and OLI/TIR (Operational Land Imager and Thermal Infrared Scanner), have a moderate spatial resolution of 30 m and a 16-day revisit time (Jiang et al 2014). Whereas Sentinel-2 can offer a higher accurate resolution (10 m) of the spatial snapshots of Earth's surface (Lefebvre et al 2019) and has little influence from shadows and built-up area (Zhang et al 2019). All collected images, which must have a cloud-covered percentage under 50%, were used for the surface water extraction from 1990 to 2019 (Table 1 and Figure 2).…”

Section: Datamentioning

confidence: 99%

Remote sensing applications for reservoir water level monitoring, sustainable water surface management, and environmental risks in Quang Nam province, Vietnam

Quang

Linh

Tam

et al. 2021

Journal of Water and Climate Change

View full text Add to dashboard Cite

Monitoring surface water provides vital information in water management; however, limited data is a fundamental challenge for most developing countries, such as Vietnam. Based on advanced remote sensing technologies, the authors proposed a methodology to process satellite images and use their outcomes to extract surface water in water resource management of Quang Nam province. Results of the proposed study show good agreement with in situ measurement data when the obtained Overall Accuracy and Kappa Coefficient were greater than 90% and 0.99, respectively. Three potential applications based on the surface water results are selected to discuss sustainable water management in Quang Nam province. Firstly, reservoir operating processes can be examined, enhanced, and even developed through long-term extracted water levels, which are the interpolation results between the extracted surface water area and the water level–area–volume curve. Secondly, the long-term morphological change for the Truong Giang river case between 1990 and 2019 can also be detected from the Water Frequency Index performance and provided additional information regarding permanent and seasonal water changes. Lastly, the flood inundation extent was extracted and separated from permanent water to assess the damage of the Mirinae typhoon on 2 November 2009 in terms of population and crop aspects.

show abstract

Automated Surface Water Extraction Combining Sentinel-2 Imagery and OpenStreetMap Using Presence and Background Learning (PBL) Algorithm

Cited by 15 publications

References 48 publications

Automated Training Data Generation from Spectral Indexes for Mapping Surface Water Extent with Sentinel-2 Satellite Imagery at 10 m and 20 m Resolutions

Automated Training Data Generation from Spectral Indexes for Mapping Surface Water Extent with Sentinel-2 Satellite Imagery at 10 m and 20 m Resolutions

One-Class Remote Sensing Classification From Positive and Unlabeled Background Data

Remote sensing applications for reservoir water level monitoring, sustainable water surface management, and environmental risks in Quang Nam province, Vietnam

Contact Info

Product

Resources

About