Search

Marine debris in the ocean is becoming an increasing problem for the management of coastal oceans and seaside resort areas. This paper presents a method for coastal marine debris mapping using satellite images from multiple satellite platforms. We carry out a pilot project in association with a local government to collect in-situ measurements of debris deposited on beaches and download the coincident satellite images to identify the marine debris. We propose to study the detection of marine debris on land and in the coastal ocean with various sources of satellite imagery as a way to increase the revisit frequency. High temporal resolution data can provide an agile estimation of the resources required to mitigate the pollution accumulation on the shoreline. A major challenge of monitoring specific areas from optical satellite images is the obscuration by cloud cover, which makes it decrease the sampling frequency. To get a handle on this problem and establish high-fidelity model, we acquired the greatest number of satellite images from a variety of platforms including high temporal-resolution imagery provided by small satellite constellation programs. We first established our method using entropy of the segmentation model output on marine debris mapping in coastal areas using WorldView images provided by MAXAR corp. Then we extended the pipeline to other small satellite images using unsupervised domain adaptation techniques. We showed that the spatial representation of the segmentation map is greatly improved by the domain adaptation techniques. Whereas some dataset still requires more data samples and additional quantitative analysis, we confirmed the compatibility of the segmentation output to the established pipeline using entropy metrics to estimate the accumulation density of the debris. This analysis shows the robust capability to be able to apply the pipeline to different types of satellite images, which can be also applied in other remote sensing applications.

The Sentinel-2 mission observes the Earth in 13 spectral bands in the visible and SWIR with a repeat cycle of 5 days for each satellite unit A and B. A complete reprocessing of the Sentinel-2 archive to generate the "Collection-1" is currently under preparation. In this presentation we describe the activities performed by the Sentinel-2 Mission Performance Center to prepare the calibration baseline for Collection-1. The preparation activities involve a correction of some minor errors in the definition of the spectral response function at the beginning of the mission. A more significant change concerns the possible harmonization between the two sensors. Indeed, radiometric validation results reported by different teams confirm the presence of a small bias (around 1%) in the visible range between the satellites. Sentinel-2B is generally believed to be slightly too dark, while Sentinel-2A agrees better with other references. The presentation will describe the inter-satellite bias assessment as well as a study of the impact of a potential harmonization on Level-2 products.

Coastal marine debris poses a serious threat to marine life and impacts fisheries and coastal tourism. This study describes a method that uses high-resolution satellite imagery to locate marine debris in a coastal area. We use the combination of in situ data from local coastal debris cleanup efforts together with nearly coincident high-resolution satellite imagery analyzed using a segmentation-based approach to isolate debris from other materials. We applied Shannon's entropy method, which represents the uncertainties of the segmentation results to obtain spectral signatures for individual semantic features. Then, we demonstrated the robustness of these features by designing a simple classification model to estimate the density of marine debris directly from a satellite image.

Masking of clouds, cloud shadow, water and snow/ice in optical satellite imagery is an important step in automated processing chains. We compare the performance of the masking provided by Fmask (“Function of mask” implemented in FORCE), ATCOR (“Atmospheric Correction”) and Sen2Cor (“Sentinel-2 Correction”) on a set of 20 Sentinel-2 scenes distributed over the globe covering a wide variety of environments and climates. All three methods use rules based on physical properties (Top of Atmosphere Reflectance, TOA) to separate clear pixels from potential cloud pixels, but they use different rules and class-specific thresholds. The methods can yield different results because of different definitions of the dilation buffer size for the classes cloud, cloud shadow and snow. Classification results are compared to the assessment of an expert human interpreter using at least 50 polygons per class randomly selected for each image. The class assignment of the human interpreter is considered as reference or “truth”. The interpreter carefully assigned a class label based on the visual assessment of the true color and infrared false color images and additionally on the bottom of atmosphere (BOA) reflectance spectra. The most important part of the comparison is done for the difference area of the three classifications considered. This is the part of the classification images where the results of Fmask, ATCOR and Sen2Cor disagree. Results on difference area have the advantage to show more clearly the strengths and weaknesses of a classification than results on the complete image. The overall accuracy of Fmask, ATCOR, and Sen2Cor for difference areas of the selected scenes is 45%, 56%, and 62%, respectively. User and producer accuracies are strongly class- and scene-dependent, typically varying between 30% and 90%. Comparison of the difference area is complemented by looking for the results in the area where all three classifications give the same result. Overall accuracy for that “same area” is 97% resulting in the complete classification in overall accuracy of 89%, 91% and 92% for Fmask, ATCOR and Sen2Cor respectively.

Surface reflectance measurements on ground in parallel to Sentinel-2 overpasses provide an essential reference for validation of atmospheric correction algorithms. Those measurements performed in vegetated area in Northern Germany are compared with surface reflectance retrieval of L2A-processsor Sen2Cor. All bands are within or slightly outside surface reflection specification |ΔSR| less than 0.05*SRref+0.005.