Sensitivity analysis of the dark spectrum fitting atmospheric correction for metre- and decametre-scale satellite imagery using autonomous hyperspectral radiometry

Estimating four-decadal variations of seagrass distribution using satellite data and deep learning methods in a marine lagoon

Seagrasses are marine flowering plants that inhabit shallow coastal and estuarine waters and serve vital ecological functions in marine ecosystems. However, seagrass ecosystems face the looming threat of degradation, necessitating effective monitoring. Remote-sensing technology offers significant advantages in terms of spatial coverage and temporal accessibility. Although some remote sensing approaches, such as water column correction, spectral index-based, and machine learning-based methods, have been proposed for seagrass detection, their performances are not always satisfactory. Deep learning models, known for their powerful learning and vast data processing capabilities, have been widely employed in automatic target detection. In this study, a typical seagrass habitat (Swan Lake) in northern China was used to propose a deep learning-based model for seagrass detection from Landsat satellite data. The performances of UNet and SegNet at different patch scales for seagrass detection were compared. The results showed that the SegNet model at a patch scale of 16 × 16 pixels worked well, with validation accuracy and loss of 96.3 % and 0.15, respectively, during training. Evaluations based on the test dataset also indicated good performance of this model, with an overall accuracy >95 %. Subsequently, the deep learning model was applied for seagrass detection in Swan Lake between 1984 and 2022. We observed a noticeable seasonal variation in germination, growth, maturation, and shrinkage from spring to winter. The seagrass area decreased over the past four decades, punctuated by intermittent fluctuations likely attributed to anthropogenic activities, such as aquaculture and construction development. Additionally, changes in landscape ecology indicators have demonstrated that seagrass experiences severe patchiness. However, these problems have weakened recently. Overall, by combining remote sensing big data with deep learning technology, our study provides a valuable approach for the highly precise monitoring of seagrass. These findings on seagrass area variation in Swan Lake offer significant information for seagrass restoration and management.

Multi-sensor and multi-platform retrieval of water chlorophyll a concentration in karst wetlands using transfer learning frameworks with ASD, UAV, and Planet CubeSate reflectance data

China has one of the widest distributions of carbonate rocks in the world. Karst wetland is a special and important ecosystem of carbonate rock regions. Chlorophyll-a (Chla) concentration is a key indicator of eutrophication, and could quantitatively evaluate water quality status of karst wetland. However, the spectral reflectance characteristics of the water bodies of karst wetland are not yet clear, resulting in remote sensing retrieval of Chla with great challenges. This study is a pioneer in utilizing field-based full-spectrum hyperspectral data to reveal the spectral response characteristics of karst wetland water body and determine the sensitive spectral bands of Chla. We further evaluated the Chla retrieval performance of multi-platform spectral data between Analytical Spectral Device (ASD), Unmanned aerial vehicle (UAV), and PlanetScope (Planet). We proposed two multi-sensor weighted integration strategies and two transfer learning frameworks for estimating water Chla from the largest karst wetland in China by combing a partial least square with adaptive ensemble algorithms. The results showed that: (1) In the range of 500-850 nm, the spectral reflectance of water bodies in the karst wetland was overall 0.001-0.105 higher than the inland water bodies, and the sensitive spectral ranges of water Chla focus on 603-778 nm; (2) UAV images outperformed ASD and Planet data, and produced the highest inversion accuracy (R2 = 0.670) for water Chla in karst wetland; (3) Multi-sensor weighted integration retrieval methods improved the Chla estimation accuracy (R2 = 0.716). Integration retrieval methods with the different weights produced the better Chla estimation accuracy than that of methods with the equal weights; (4) The transfer learning methods from ASD to UAV platform provided the better retrieval performance (the average R2 = 0.669) than that of methods from UAV to Planet platform. The transfer learning methods obtained the highest estimation accuracy of Chla (R2 = 0.814) when the ratio of the training and test data in the target domain was 7:3. The transfer learning methods produced the higher estimation accuracies with the distribution of the absolute residuals between predicted and measured values <20.957 mg/m3 compared to the multi-sensor weighted integration retrieval methods, which demonstrated that transfer learning is more suitable for estimating Chla in karst wetland water bodies using multi-platform and multi-sensor data. The results provide a scientific basis for the protection and sustainable development of karst wetlands.

Retrieval and validation of the Secchi disk depth values (Zsd) from the Sentinel-3/OLCI satellite data in the Persian Gulf and the Gulf of Oman

In this study, the Secchi disk depth (Z_sd) values as an indicator of seawater clarity/transparency were estimated using the ESA (European Space Agency) Sentinel-3A and Sentinel-3B OLCI (S3/OLCI) satellite data in the Persian Gulf and the Gulf of Oman (PG&GO). To do so, two procedures were evaluated including an existing methodology developed by Doron et al. (J Geophys Res: Oceans 112(C6) 2007 and (Remote Sens Environ 115:2986-3001 2011) and an empirical model proposed in this research formed by employing the blue (B₄) and green (B₆) bands of S3/OLCI data. In this regard, a total number of 157 field-measured Z_sd values (114 training points for calibration of the models and 43 control points for accuracy assessment of them) were observed during eight research cruises conducted by the research vessel, the Persian Gulf Explorer, in the PG&OS between 2018 and 2022. The optimum methodology was then selected based on the statistical indicators including R² (coefficient of determination), RMSE (root mean square error), and MAPE (mean absolute percentage error). However, after the indication of the optimal model, the data of all 157 observations were utilized for the calculation of unknown parameters of the model. The final results demonstrated that compared to the existing empirical model proposed by Doron et al. (J Geophys Res: Oceans 112(C6) 2007 and (Remote Sens Environ 115:2986-3001 2011), the developed model in this study which was formed based on the linear and ratio terms of B₄ and B₆ bands, has more efficiency in the PG&GO. Consequently, a model in form of Z_sd = e^1.638B₄^/B₆^-8.241B₄^-12.876B₆^+1.26 was suggested for the estimation of Z_sd values from S3/OLCI in the PG&GO (R² = 0.749, RMSE = 2.56 m, and MAPE = 22.47%). The results also showed that the annual oscillation of the Z_sd values in the GO (5-18 m) is evidently higher compared with those in the PG (4-12 m) and the SH (7-10 m) regions.

Evaluation of eight band SuperDove imagery for aquatic applications

Planet's SuperDove constellation is evaluated for remote sensing of water targets. SuperDoves are small satellites with on board eight band PlanetScope imagers that add four new bands compared to the previous generations of Doves. The Yellow (612 nm) and Red Edge (707 nm) bands are of particular interest to aquatic applications, for example in aiding the retrieval of pigment absorption. The dark spectrum fitting (DSF) algorithm is implemented in ACOLITE for processing of SuperDove data, and its outputs are compared to matchup data collected using an autonomous pan-and-tilt hyperspectral radiometer (PANTHYR) installed in the turbid waters of the Belgian Coastal Zone (BCZ). Results for 35 matchups from 32 unique SuperDove satellites indicate on average low differences with PANTHYR observations for the first seven bands (443-707 nm), with mean absolute relative differences (MARD) 15-20%. The mean average differences (MAD) are between -0.01 and 0 for the 492-666 nm bands, i.e. DSF results show a negative bias, while the Coastal Blue (444 nm) and Red Edge (707 nm) show a small positive bias (MAD 0.004 and 0.002). The NIR band (866 nm) shows a larger positive bias (MAD 0.01), and larger relative differences (MARD 60%). Root mean squared differences (RMSD) are rather flat at around 0.01 with peaks in the bands with highest water reflectance of around 0.015. The surface reflectance products as provided by Planet (PSR) show a similar average performance to DSF, with slightly larger and mostly positive biases, except in both Green bands, where the MAD is close to 0. MARD in the two Green bands is a bit lower for PSR (9.5-10.6%) compared to DSF (9.9-13.0%). Higher scatter is found for the PSR (RMSD 0.015-0.020), with some matchups showing large, spectrally mostly flat differences, likely due to the external aerosol optical depth (τ_a) inputs not being representative for these particular images. Chlorophyll a absorption (a_Chl) is retrieved from PANTHYR measurements, and the PANTHYR data are used to calibrate a_Chl retrieval algorithms for SuperDove in the BCZ. Various Red band indices (RBI) and two neural networks are evaluated for a_Chl estimation. The best performing RBI algorithm, i.e. the Red band difference (RBD), showed a MARD of 34% for DSF and 25% for PSR with positive biases of 0.11 and 0.03 m^-1 respectively for 24 PANTHYR a_Chl matchups. The difference in RBD performance between DSF and PSR can be largely explained by their respective average biases in the Red and Red Edge bands, which are opposite signs for DSF (negative bias in the red), and positive for both bands for PSR. Mapping of turbid water a_Chl and hence chlorophyll a concentration (C) using SuperDove is demonstrated for coastal bloom imagery, showing how SuperDove data can supplement monitoring programmes.

Improving the standard protocol for above-water reflectance measurements: 1. Estimating effective wind speed from angular variation of sunglint

The standard above-water protocol for measurement of water reflectance uses a measurement of wind speed to estimate the air-water interface reflectance factor and, thus, remove reflected skylight from upwelling radiance. This aerodynamic wind speed measurement may be a poor proxy for the local wave slope distribution in cases such as fetch-limited coastal and inland waters and/or where there are spatial or temporal differences between the wind speed measurement and the location of reflectance measurements. Here, an improved method is proposed, with a focus on sensors mounted on autonomous pan-tilt units and deployed on fixed platforms, replacing the aerodynamic wind speed measurement by optical measurements of angular variation of upwelling radiance. Using radiative transfer simulations, it is shown that the difference between two upwelling (i.e., water plus air-water interface) reflectances acquired at least 10° apart from each other in the solar principal plane is strongly and monotonically related to effective wind speed. The approach shows good performance in twin experiments using radiative transfer simulations. Limitations of the approach are identified, including difficulties for a very high Sun zenith angle (>60^∘), very low wind speed (<2m s ^-1), and, potentially, cases in which nadir-pointing angles are limited by optical perturbations from the viewing platform.

Marine Litter Detection by Sentinel-2: A Case Study in North Adriatic (Summer 2020)

Aggregates of floating materials detected in North Adriatic in six Sentinel-2 scenes of August 2020 have been investigated. Most of the floating materials were identified by the chlorophyll red edge and consisted of vegetal materials, probably conveyed by rivers and exchanged with the lagoons. Traces of marine litter were looked for in the spectral anomalies of the Red Edge bands, assuming changes of the red edge in pixels where marine litter was mixed with vegetal materials. About half of the detected patches were unclassified due to the weakness of the useful signal (pixel filling percentage < 25%). The classification produced 59% of vegetal materials, 16% of marine litter mixed with vegetal materials and 22% of intermediate cases. A small percentage (2%) was attributed to submerged vegetal materials, found in isolated patches. The previous percentages were obtained with a separation criterion based on arbitrary thresholds. The patches were more concentrated at the mouths of the northern rivers, less off the Venice lagoon, and very few outside the Po River, with the minimal river outflow during the period. Sentinel-2 is a valid tool for the discrimination of marine litter in aggregates of floating matter. The proposed method requires validation, and the North Adriatic is an excellent site for field work, as in summer many patches of floating matter form in proximity to the coast.

Performance and Uncertainty of Satellite-Derived Bathymetry Empirical Approaches in an Energetic Coastal Environment

Objectives of this study are to evaluate the performance of different satellite-derived bathymetry (SDB) empirical models developed for multispectral satellite mission applications and to propose an uncertainty model based on inferential statistics. The study site is the Arcachon Bay inlet (France). A dataset composed of 450,837 echosounder data points and 89 Sentinel-2 A/B and Landsat-8 images acquired from 2013 to 2020, is generated to test and validate SDB and uncertainty models for various contrasting optical conditions. Results show that water column optical properties are characterized by a high spatio-temporal variability controlled by hydrodynamics and seasonal conditions. The best performance and highest robustness are found for the cluster-based approach using a green band log-linear regression model. A total of 80 satellite images can be exploited to calibrate SDB models, providing average values of root mean square error and maximum bathymetry of 0.53 m and 7.3 m, respectively. The uncertainty model, developed to extrapolate information beyond the calibration dataset, is based on a multi-scene approach. The sensitivity of the model to the optical variability not explained by the calibration dataset is demonstrated but represents a risk of error of less than 5%. Finally, the uncertainty model applied to a diachronic analysis definitively demonstrates the interest in SDB maps for a better understanding of morphodynamic evolutions of large-scale and complex coastal systems.

Comparing Sentinel-2 and WorldView-3 Imagery for Coastal Bottom Habitat Mapping in Atlantic Canada

Satellite remote sensing is a valuable tool to map and monitor the distribution of marine macrophytes such as seagrass and seaweeds that perform many ecological functions and services in coastal habitats. Various satellites have been used to map the distribution of these coastal bottom habitat-forming species, with each sensor providing unique benefits. In this study, we first explored optimal methods to create bottom habitat maps using WorldView-3 satellite imagery. We secondly compared the WorldView-3 bottom habitat maps to previously produced Sentinel-2 maps in a temperate, optically complex environment in Nova Scotia, Canada to identify the top performing classification and the advantages and disadvantages of each sensor. Sentinel-2 provides a global, freely accessible dataset where four bands are available at a 10-m spatial resolution in the visible and near infrared spectrum. Conversely, WorldView-3 is a commercial satellite where eight bands are available at a 2-m spatial resolution in the visible and near infrared spectrum, but data catalogs are costly and limited in scope. Our optimal WorldView-3 workflow processed images from digital numbers to habitat classification maps, and included a semiautomatic stripe correction. Our comparison of bottom habitat maps explored the impact of improved WorldView-3 spatial resolution in isolation, and the combined advantage of both WorldView’s increased spatial and spectral resolution relative to Sentinel-2. We further explored the effect of tidal height on classification success, and relative changes in water clarity between images collected at different dates. As expected, both sensors are suitable for bottom habitat mapping. The value of WorldView-3 came from both its increased spatial and spectral resolution, particularly for fragmented vegetation, and the value of Sentinel-2 imagery comes from its global dataset that readily allows for large scale habitat mapping. Given the variation in scale, cost and resolution of the two sensors, we provide recommendations on their use for mapping and monitoring marine macrophyte habitat in Atlantic Canada, with potential applications to other coastal areas of the world.

European Space Agency (ESA) Calibration/Validation Strategy for Optical Land-Imaging Satellites and Pathway towards Interoperability

Land remote sensing capabilities in the optical domain have dramatically increased in the past decade, owing to the unprecedented growth of space-borne systems providing a wealth of measurements at enhanced spatial, temporal and spectral resolutions. Yet, critical questions remain as how to unlock the potential of such massive amounts of data, which are complementary in principle but inherently diverse in terms of products specifications, algorithm definition and validation approaches. Likewise, there is a recent increase in spatiotemporal coverage of in situ reference data, although inconsistencies in the used measurement practices and in the associated quality information still hinder their integrated use for satellite products validation. In order to address the above-mentioned challenges, the European Space Agency (ESA), in collaboration with other Space Agencies and international partners, is elaborating a strategy for establishing guidelines and common protocols for the calibration and validation (Cal/Val) of optical land imaging sensors. Within this paper, this strategy will be illustrated and put into the context of current validation systems for land remote sensing. A reinforced focus on metrology is the basic principle underlying such a strategy, since metrology provides the terminology, the framework and the best practices, allowing to tie measurements acquired from a variety of sensors to internationally agreed upon standards. From this general concept, a set of requirements are derived on how the measurements should be acquired, analysed and quality reported to users using unified procedures. This includes the need for traceability, a fully characterised uncertainty budget and adherence to community-agreed measurement protocols. These requirements have led to the development of the Fiducial Reference Measurements (FRM) concept, which is promoted by the ESA as the recommended standard within the satellite validation community. The overarching goal is to enhance user confidence in satellite-based data and characterise inter-sensor inconsistencies, starting from at-sensor radiances and paving the way to achieving the interoperability of current and future land-imaging systems.

The Use of Sentinel-2 for Chlorophyll-a Spatial Dynamics Assessment: A Comparative Study on Different Lakes in Northern Germany

Eutrophication of inland waters is an environmental issue that is becoming more common with climatic variability. Monitoring of this aquatic problem is commonly based on the chlorophyll-a concentration monitored by routine sampling with limited temporal and spatial coverage. Remote sensing data can be used to improve monitoring, especially after the launch of the MultiSpectral Instrument (MSI) on Sentinel-2. In this study, we compared the estimation of chlorophyll-a (chl-a) from different bio-optical algorithms using hyperspectral proximal remote sensing measurements, from simulated MSI responses and from an MSI image. For the satellite image, we also compare different atmospheric corrections routines before the comparison of different bio-optical algorithms. We used in situ data collected in 2019 from 97 sampling points across 19 different lakes. The atmospheric correction assessment showed that the performances of the routines varied for each spectral band. Therefore, we selected C2X, which performed best for bands 4 (root mean square error—RMSE = 0.003), 5 (RMSE = 0.004) and 6 (RMSE = 0.002), which are usually used for the estimation of chl-a. Considering all samples from the 19 lakes, the best performing chl-a algorithm and calibration achieved a RMSE of 16.97 mg/m3. When we consider only one lake chain composed of meso-to-eutrophic lakes, the performance improved (RMSE: 10.97 mg/m3). This shows that for the studied meso-to-eutrophic waters, we can reliably estimate chl-a concentration, whereas for oligotrophic waters, further research is needed. The assessment of chl-a from space allows us to assess spatial dynamics of the environment, which can be important for the management of water resources. However, to have an accurate product, similar optical water types are important for the overall performance of the bio-optical algorithm.

First Evaluation of PRISMA Level 1 Data for Water Applications

This study presents a first assessment of the Top-Of-Atmosphere (TOA) radiances measured in the visible and near-infrared (VNIR) wavelengths from PRISMA (PRecursore IperSpettrale della Missione Applicativa), the new hyperspectral satellite sensor of the Italian Space Agency in orbit since March 2019. In particular, the radiometrically calibrated PRISMA Level 1 TOA radiances were compared to the TOA radiances simulated with a radiative transfer code, starting from in situ measurements of water reflectance. In situ data were obtained from a set of fixed position autonomous radiometers covering a wide range of water types, encompassing coastal and inland waters. A total of nine match-ups between PRISMA and in situ measurements distributed from July 2019 to June 2020 were analysed. Recognising the role of Sentinel-2 for inland and coastal waters applications, the TOA radiances measured from concurrent Sentinel-2 observations were added to the comparison. The results overall demonstrated that PRISMA VNIR sensor is providing TOA radiances with the same magnitude and shape of those in situ simulated (spectral angle difference, SA, between 0.80 and 3.39; root mean square difference, RMSD, between 0.98 and 4.76 [mW m-2 sr-1 nm-1]), with slightly larger differences at shorter wavelengths. The PRISMA TOA radiances were also found very similar to Sentinel-2 data (RMSD < 3.78 [mW m-2 sr-1 nm-1]), and encourage a synergic use of both sensors for aquatic applications. Further analyses with a higher number of match-ups between PRISMA, in situ and Sentinel-2 data are however recommended to fully characterize the on-orbit calibration of PRISMA for its exploitation in aquatic ecosystem mapping.