Potential of Large-Scale Inland Water Body Mapping from Sentinel-1/2 Data on the Example of Bavaria’s Lakes and Rivers
01 Aug 2020-Vol. 88, Iss: 3, pp 271-289
TL;DR: In this article, a simple framework based on supervised learning and automatic training data annotation is shown, which allows to map inland water bodies from Sentinel satellite data on large scale, i.e. on state level.
Abstract: The mapping of water bodies is an important application area of satellite-based remote sensing. In this contribution, a simple framework based on supervised learning and automatic training data annotation is shown, which allows to map inland water bodies from Sentinel satellite data on large scale, i.e. on state level. Using the German state of Bavaria as an example and different combinations of Sentinel-1 SAR and Sentinel-2 multi-spectral imagery as inputs, potentials and limits for the automatic detection of water surfaces for rivers, lakes, and reservoirs are investigated. Both quantitative and qualitative results confirm that fully automatic large-scale inland water body mapping is generally possible from Sentinel data; whereas, the best result is achieved when all available surface-related bands of both Sentinel-1 and Sentinel-2 are fused on a pixel level. The main limitation arises from missed smaller water bodies, which are not observed in bands with a resolution of about 20 m. Given the simplicity of the proposed approach and the open availability of the Sentinel data, the study confirms the potential for a fully automatic large-scale mapping of inland water with cloud-based remote sensing techniques.
Citations
More filters
TL;DR: In this paper, the authors explored the use of diverse bands of Sentinel 2 (S2) through well-established water indices and Sentinel 1 (S1) derived SAR imagery along with their combinations to assess their capability for generating accurate flood inundation maps.
Abstract: Identification of flood water extent from satellite images has historically relied on either synthetic aperture radar (SAR) or multi-spectral (MS) imagery. MS sensors are limited to cloud free conditions, whereas SAR imagery is plagued by noise-like speckle. Prior studies that use combinations of MS and SAR data to overcome individual limitations of these sensors have not fully examined sensitivity of flood mapping performance to different combinations of SAR and MS derived spectral indices or band transformations in color space. This study explores the use of diverse bands of Sentinel 2 (S2) through well-established water indices and Sentinel 1 (S1) derived SAR imagery along with their combinations to assess their capability for generating accurate flood inundation maps. The robustness in performance of S-1 and S-2 band combinations was evaluated using 446 hand labeled flood inundation images spanning across 11 flood events from Sen1Floods11 dataset which are highly diverse in terms of land cover as well as location. A modified K-fold cross validation approach is used to evaluate the performance of 32 combinations of S1 and S2 bands using a fully connected deep convolutional neural network known as U-Net. Our results indicated that usage of elevation information has improved the capability of S1 imagery to produce more accurate flood inundation maps. Compared to a median F1 score of 0.62 when using only S1 bands, the combined use of S1 and elevation information led to an improved median F1 score of 0.73. Water extraction indices based on S2 bands have a statistically significant superior performance in comparison to S1. Among all the band combinations, HSV (Hue, Saturation, Value) transformation of S2 bands provides a median F1 score of 0.9, outperforming the commonly used water spectral indices owing to HSV’s transformation’s superior contrast distinguishing abilities. Additionally, U-Net algorithm was able to learn the relationship between raw S2 based water extraction indices and their corresponding raw S2 bands, but not of HSV owing to relatively complex computation involved in the latter. Results of the paper establishes important benchmarks for the extension of S1 and S2 data-based flood inundation mapping efforts over large spatial extents.
45 citations
TL;DR: In this paper, the authors used the Sen1Floods11 dataset to improve flood detection using well-known segmentation architectures, such as SegNet and UNet, as networks.
Abstract: Timely detection of flooding is paramount for saving lives as well as evaluating levels of damage. Floods generally occur under specific weather conditions, such as excessive precipitation, which makes the presence of clouds very likely. For this reason, radar-based sensors are most suitable for near-real-time flood mapping. The public dataset Sen1Floods11 recently released by the Cloud to Street is one example of ongoing beneficial initiatives to employ deep learning for flood detection with synthetic aperture radar. The present study used this dataset to improve flood detection using well-known segmentation architectures, such as SegNet and UNet, as networks. In addition, this study provided a deeper understanding of which set of polarized band combination is more suitable for distinguishing permanent water, as well as flooded areas from the SAR image. The overall performance of the models with various kinds of labels and a combination of bands to detect all surface water areas were also assessed. Finally, the trained models were tested on a completely different location at Kerala, India, during the 2018 flood for verifying their performance in the real-world situation of a flood event outside of the given test set in the dataset. The results prove that trained models can be used as off-the-shelf models to achieve an intersection over union (IoU) as high as 0.88 in comparison with optical images. The omission and commission error were less than 6%. However, the most important result is that the processing time for the whole satellite image was less than 1 min. This will help significantly for providing analysis and near-real-time flood mapping services to first responder organizations during flooding disasters.
17 citations
01 Mar 2021
TL;DR: In this paper, the authors provide an overview of forest attributes measurable by forest inventory that may support the integration of non-provisioning ecosystem services (ES) and biodiversity into forest planning.
Abstract: Our review provides an overview of forest attributes measurable by forest inventory that may support the integration of non-provisioning ecosystem services (ES) and biodiversity into forest planning. The review identifies appropriate forest attributes to quantify the opportunity for recreation, biodiversity promotion and carbon storage, and describes new criteria that future forest inventories may include. As a source of information, we analyse recent papers on forest inventory and ES to show if and how they address these criteria. We further discuss how mapping ES could benefit from such new criteria and conclude with three case studies illustrating the importance of selected criteria delivered by forest inventory. Recent studies on forest inventory focus mainly on carbon storage and biodiversity promotion, while very few studies address the opportunity of recreation. Field sampling still dominates the data collection, despite the fact that airborne laser scanning (ALS) has much improved the precision of large-scale estimates of the level of forest ES provision. However, recent inventory studies have hardly addressed criteria such as visible distance in stands, presence of open water bodies and soil damages (important for the opportunity of recreation) and naturalness (here understood as the similarity of the forest to its natural state) and habitat trees and natural clearings (important for biodiversity promotion). The problem of quantifying carbon stock changes with appropriate precision has not been addressed. In addition, the reviewed studies have hardly explored the potential of inventory information to support mapping of the demand for ES. We identify challenges with estimating a number of criteria associated with rare events, relevant for both the opportunity of recreation and biodiversity promotion. These include deadwood, rare species and habitat trees. Such rare events require innovative inventory technology, such as point-transect sampling or ALS. The ALS technology needs relatively open canopies, to achieve reliable estimates for deadwood or understorey vegetation. For the opportunity of recreation, the diversity among forest stands (possibly quantified by geoinformatics) and information on the presence of open water bodies (provided by RADAR, ALS data or use of existing maps) may be important. Naturalness is a crucial criterion for native biodiversity promotion but hard to quantify and assess until now. Tree species identification would be crucial for this criterion, which is still a challenge for remote sensing techniques. Estimating carbon storage may build on biomass estimates from terrestrial samples or on remotely sensed data, but major problems exist with the precision of estimates for carbon stock changes. Recent approaches for mapping the supply side of forest ES are promising, while providing so far uncommon structural information by revised inventory concepts could be helpful also for mapping the demand for ES. We conclude that future studies must find holistic inventory management systems to couple various inventory technologies in support of the integration of non-provisioning ES and biodiversity into forest planning.
10 citations
TL;DR: In this paper , the authors present a new global dataset that contains the location and surface area variations of 681,137 lakes and reservoirs larger than 0.1 square kilometers (and south of 50 degree N).
Abstract: Lakes and reservoirs, as most humans experience and use them, are dynamic bodies of water, with surface extents that increase and decrease with seasonal precipitation patterns, long-term changes in climate, and human management decisions. This paper presents a new global dataset that contains the location and surface area variations of 681,137 lakes and reservoirs larger than 0.1 square kilometers (and south of 50 degree N) from 1984 to 2015, to enable the study of the impact of human actions and climate change on freshwater availability. Within its scope for size and region covered, this dataset is far more comprehensive than existing datasets such as HydroLakes. While HydroLAKES only provides a static shape, the proposed dataset also has a timeseries of surface area and a shapefile containing monthly shapes for each lake. The paper presents the development and evaluation of this dataset and highlights the utility of novel machine learning techniques in addressing the inherent challenges in transforming satellite imagery to dynamic global surface water maps.
9 citations
TL;DR: In this paper, Wang et al. evaluated the transfer performance of machine learning algorithms applied to remote sensing images in different periods, and compared the differences among these models, including logistic regression, support vector machine, neural network, random forest, decision tree, and XGBoost.
Abstract: Monitoring open water bodies accurately is important for assessing the role of ecosystem services in the context of human survival and climate change. There are many methods available for water body extraction based on remote sensing images, such as the normalized difference water index (NDWI), modified NDWI (MNDWI), and machine learning algorithms. Based on Landsat-8 remote sensing images, this study focuses on the effects of six machine learning algorithms and three threshold methods used to extract water bodies, evaluates the transfer performance of models applied to remote sensing images in different periods, and compares the differences among these models. The results are as follows. (1) Various algorithms require different numbers of samples to reach their optimal consequence. The logistic regression algorithm requires a minimum of 110 samples. As the number of samples increases, the order of the optimal model is support vector machine, neural network, random forest, decision tree, and XGBoost. (2) The accuracy evaluation performance of each machine learning on the test set cannot represent the local area performance. (3) When these models are directly applied to remote sensing images in different periods, the AUC indicators of each machine learning algorithm for three regions all show a significant decline, with a decrease range of 0.33–66.52%, and the differences among the different algorithm performances in the three areas are obvious. Generally, the decision tree algorithm has good transfer performance among the machine learning algorithms with area under curve (AUC) indexes of 0.790, 0.518, and 0.697 in the three areas, respectively, and the average value is 0.668. The Otsu threshold algorithm is the optimal among threshold methods, with AUC indexes of 0.970, 0.617, and 0.908 in the three regions respectively and an average AUC of 0.832.
9 citations
References
More filters
TL;DR: Google Earth Engine is a cloud-based platform for planetary-scale geospatial analysis that brings Google's massive computational capabilities to bear on a variety of high-impact societal issues including deforestation, drought, disaster, disease, food security, water management, climate monitoring and environmental protection.
6,262 citations
"Potential of Large-Scale Inland Wat..." refers methods in this paper
...The workflow is implemented in Google Earth Engine (Gorelick et al. 2017) to utilize its curated data catalogue and cloud-processing capabilities....
[...]
...By utilising a simple classifier based on supervised learning and automated training data generation from globally available volunteered geographic information (VGI), freely available remote sensing data provided by the Sentinel missions of the European Copernicus program, and the cloud computation platform Google Earth Engine, the proposed approach can easily be applied to arbitrary regions of the world....
[...]
...The workflow is implemented in Google Earth Engine (Gorelick et al....
[...]
01 Jan 1974
TL;DR: In this paper, a method has been developed for quantitative measurement of vegetation conditions over broad regions using ERTS-1 spectral bands 5 and 7, corrected for sun angle, which is shown to be correlated with aboveground green biomass on rangelands.
Abstract: The Great Plains Corridor rangeland project utilizes natural vegetation systems as phenological indicators of seasonal development and climatic effects upon regional growth conditions. A method has been developed for quantitative measurement of vegetation conditions over broad regions using ERTS-1 MSS data. Radiance values recorded in ERTS-1 spectral bands 5 and 7, corrected for sun angle, are used to compute a band ratio parameter which is shown to be correlated with aboveground green biomass on rangelands.
5,829 citations
TL;DR: The normalized difference water index (NDWI) as discussed by the authors was proposed for remote sensing of vegetation liquid water from space, which is defined as (ϱ(0.86 μm) − ϱ(1.24 μm)) where ϱ represents the radiance in reflectance units.
4,461 citations
"Potential of Large-Scale Inland Wat..." refers methods in this paper
...In addition to that, the more relevant modified normalised difference water index (MNDWI) (Xu 2006) is employed, which is an extension of the normalised difference water index (NDWI) (Gao 1996) and is dedicated to the detection of water surfaces....
[...]
TL;DR: In this paper, the normalized difference water index (NDWI) was modified by substitution of a middle infrared band such as Landsat TM band 5 for the near infrared band used in the NDWI.
Abstract: The normalized difference water index (NDWI) of McFeeters (1996) was modified by substitution of a middle infrared band such as Landsat TM band 5 for the near infrared band used in the NDWI. The modified NDWI (MNDWI) can enhance open water features while efficiently suppressing and even removing built‐up land noise as well as vegetation and soil noise. The enhanced water information using the NDWI is often mixed with built‐up land noise and the area of extracted water is thus overestimated. Accordingly, the MNDWI is more suitable for enhancing and extracting water information for a water region with a background dominated by built‐up land areas because of its advantage in reducing and even removing built‐up land noise over the NDWI.
3,234 citations
"Potential of Large-Scale Inland Wat..." refers background or methods in this paper
...In addition to that, the more relevant modified normalised difference water index (MNDWI) (Xu 2006) is employed, which is an extension of the normalised difference water index (NDWI) (Gao 1996) and is dedicated to the detection of water surfaces....
[...]
...As numerous researchers have confirmed before, even with only SAR backscatter or the modified normalised difference water index (MNDWI) (Xu 2006), water and non-water pixels can already be distinguished to some extent, e....
[...]
...As numerous researchers have confirmed before, even with only SAR backscatter or the modified normalised difference water index (MNDWI) (Xu 2006), water and non-water pixels can already be distinguished to some extent, e.g. by thresholding techniques (Liebe et al....
[...]
TL;DR: This paper reviews remote sensing implementations of support vector machines (SVMs), a promising machine learning methodology that is particularly appealing in the remote sensing field due to their ability to generalize well even with limited training samples.
Abstract: A wide range of methods for analysis of airborne- and satellite-derived imagery continues to be proposed and assessed. In this paper, we review remote sensing implementations of support vector machines (SVMs), a promising machine learning methodology. This review is timely due to the exponentially increasing number of works published in recent years. SVMs are particularly appealing in the remote sensing field due to their ability to generalize well even with limited training samples, a common limitation for remote sensing applications. However, they also suffer from parameter assignment issues that can significantly affect obtained results. A summary of empirical results is provided for various applications of over one hundred published works (as of April, 2010). It is our hope that this survey will provide guidelines for future applications of SVMs and possible areas of algorithm enhancement.
2,546 citations
"Potential of Large-Scale Inland Wat..." refers background in this paper
...The application of SVMs for land cover classification has a long tradition in the remote sensing community (Mountrakis et al. 2011)....
[...]