ALGORITHMS AND PRODUCTS FOR THE COASTAL AREAS DYNAMIC (ALSO IN RELATION TO COASTAL EROSION)

Federica Bragaa, Gian Marco Scarpaa, Giorgia Manfèa, Giuliano Lorenzettia, Luca Zaggiab

a Institute of Marine Sciences, National Research Council of Italy (CNR-ISMAR), Venice (Italy)

b Institute of Geosciences and Earth Resources, National Research Council of Italy (CNR-IGG), Padua (Italy)

Research theme aims

The research activity focuses on the identification of the main surficial “pathways” of suspended sediments (SPM) at local scale detectable by satellite data with the aid of in situ data and hydrodynamic modeling. The approach is based on single-band or band-ratio bio-optical algorithms applied to multi-temporal EO data for mapping the pattern of suspended sediments and emerged/submerged beaches. As foreseen by the Grant Agreement, the research activities of WP 1A-CBE are developing according to three successive phases: e.1) detection of the emerged/submerged beach interface and, if visible, the presence of submerged bars; e.2) identification of suspended sediment patterns; e.3) identification of the main sediment transport pathways in the short/long term.

Materials and Methods

Study area

The research activities are undertaken on the coasts of the Northern Adriatic Sea (NAS) coast which are the longest Italian sandy coastline. From a morphological point of view, the study area could be subdivided in three main parts: a northern part characterized by a straight littoral coast, a southern part with areas related to the Brenta-Adige-Po Delta system and a central part, correspondent to the Venice lagoon. The forcings that act for shaping and influencing the dynamic of these sandy coastal systems are basically wind (Bora and Scirocco), waves and long-shore currents, climate and rivers run-off. Human pressures are very high along the entire coastline.

Satellite data processing and in situ sampling

Data were mainly acquired in the period 2016-2019 by Operational Land Imager (OLI) and MultiSpectral Imager (MSI) sensors on-board of Landsat 8 (L8) and Sentinel-2A&B (S2) respectively. MSI-S2 Level 1C images were downloaded from Copernicus Open Access (Hub, https://scihub.copernicus.eu/dhus); containing top-of-atmosphere (TOA) reflectances. OLI-L8 radiances were downloaded from USGS archive of the Earth Resources Observation and Science Center (http://glovis.usgs.gov). Moreover, we used a set of RapidEye (RE) images provided by ESA’s Third Party Missions. For this study, we used RapidEye 3A Ortho products (L3A), resampled at 5 m spatial resolution, which were ortho-corrected using ground control points and DEMs. Each covers a 25 km by 25 km area. Firstly, the digital numbers (DNs) recorded by the satellites were converted to at-sensor spectral radiance and TOA reflectance. Only for RE images we did not account for atmospheric effects, because the difference contributed by atmospheric effect is expected to be negligible considering the small size of the area. For L8 and S2 data, radiometric correction and atmospheric correction were performed simultaneously with ACOLITE (Atmospheric Correction for OLI ‘lite’) described in Vanhellmont & Ruddick (2014; 2015; 2016) [1-3]. ACOLITE outputs water-leaving radiance reflectances, hereafter called water reflectances (ρw), in all visible and NIR bands, and can compute multiple other parameters (e.g. turbidity).

The shoreline extraction was automatically performed following these main steps: 1) a rough separation between land and water; 2) the refinement of this boundary to extract the accurate shoreline position; and 3) the vectorization of the extracted shoreline. The land-water separation method was based on the hypothesis that the reflectance of pure water is quite completely absorbed in the NIR [4]. On the other hand, the spectral signature of land has higher value at all wavelengths, especially at NIR range [5,6]. Using this rationale, we investigated the radiometric response of water and land in the different spectral bands, and then identified the best threshold value. All the pixels lower than the selected threshold were classified as water; all the pixels higher than the selected threshold were classified as land. Finally, we extracted the rough shoreline with the edge detection technique, based on the identified threshold. A tidal level correction was then performed, based on tidal observations recorded at the Diga Sud Lido station: in-situ tidal data were provided by the Centro Previsioni e Segnalazioni Maree of the Venice Municipality.

In addition to shoreline identification, in this research activity we focused on retrieve Turbidity (T), considering that it provides an easily measurable proxy for suspended particulate matter (SPM). The semi-empirical single band turbidity (T, expressed in formazin nephelometric unit [FNU])) retrieval algorithm of Dogliotti et al., (2015) [7] is applied and analyzed in the present study. Considering the wider range of turbidity in coastal waters, this algorithm uses either the red or NIR bands, depending on the ρw(λ) at 655 nm. In case ρw(655)<0.05 (corresponding to T<15 FNU, so low turbidity), ρw(λ) at 655 nm band is used in the algorithm. In case of ρw(655)>0.07 (corresponding to T>45 FNU, so from medium to high turbidity), ρw(λ) at 865 nm band is used to avoid the signal saturation at lower wavelengths. In between these thresholds, when 0.05>ρw(655)>0.07, the two algorithms are linearly blended to ensure a smooth transition. This algorithm was validated in the Po River Prodelta (Northern Adriatic Sea) by Braga et al. (2017) [8]. For the retrieval of SPM concentration, the ACOLITE-derived water leaving reflectance (ρw (λ)) were converted according to Nechad et al. (2010) [9]. In addition, in situ data were compared with products obtained from L8 and S2 images in order to validate the algorithm based on the specific characteristics of the SPM.

Several field surveys were carried out from autumn 2016 to summer 2017 in the vicinity of the Venice Lagoon during L8 and S2 overpasses. For each survey, a set of 6 stations was repeatedly sampled to collect water samples (for the SPM concentration) and hydrologic parameters. At each station, Conductivity-Temperature-Depth (CTD) data were obtained utilizing an Idronaut Ocean Seven 316Plus multi-parameter probe, equipped also with a backscattering optical sensor (Seapoint turbidity meter, operating at 880 nm). Fifteen surveys synchronous with the passage of L8 and S2 were also carried out from January 2019 to August 2019 in the area of the Lido tidal mouth and in the central sector of the lagoon. In addition to the sampling of water samples and hydrologic parameters, above water remote sensing reflectances were measured with the WISP-3 spectroradiometer (Water Insight). Significant wave height, wind direction and speed were measured at the Acqua Alta Oceanographic Tower (AAOT), located in the Northern Adriatic Sea, about 16 km off the coast of Venice.

Results and Discussion

Mapping of shoreline variations

Regarding the “slow and natural” coastline evolution due to long-term erosion and accumulation, the shoreline analysis was made in a very dynamical area (i.e. Po river delta) by means of Landsat and S2 satellite images, for the last 50 years (Fig.1). The Po di Pila mouth was the most active and dynamic sector of the delta: recurrent erosion/deposition areas were observed throughout the studied period, with the recurrent formation of submerged/emerged sand bar structures at the front of the Po di Pila mouth. Moreover, a recently formed crescentic bar is visible in L8 and S2 satellite images starting from 2016. The bar has migrated landward, partially obstructing the river flow and splitting the outflowing jet in two branches. The bar evolution seems to be more related to wave action than on the river forcing (Fig.1).

Figure 1: On the left: map of shoreline of the Po river delta, based on satellite observations over a period of 46 years (1972–2018). Shorelines were extracted from S2 (2018) and Landsat data (MSS-L1: 1972; MSS-L3: 1982; TM-L4: 1990; ETM-L7: 2001; TM-5: 2010). On the right: map of emerged sand bar structures at the front of the Po di Pila mouth, based on S2 observations over the period of 2016-2019.

For the short-term temporal evolution of the shoreline, the analysed event represents an intense and prolonged winter storm due to Bora which occurred in the Northern Adriatic Sea on 2-3 February 2018. The rapid evolution of the shoreline was analysed considering the shorelines extracted from RE. The RE image was acquired on the 04/02 at 10:25 AM and shows the presence of plunging waves that generally occur in correspondence of sudden shoalings, causing the erosion of submerged and emerged structures. The results of temporal analysis indicate that the swell did not cause permanent damage to the coast as seen in the shoreline captured 10 days after the extreme event.

Turbidity maps: validation and suspended sediment transport

In situ 2017 data collected in the field were used to assess the accuracy of T products derived from L8 and S2 imagery. A total of 58 match-ups between satellite and in situ data collection were available by considering a maximum time difference of 1 hour. These comparisons show a quite statistically significant correlation with a coefficient of determination R2 of 0.756 (L8) and 0.554 (S2). These statistics indicates a better linear relationship between the L8-derived and in situ measured T data, than for S2. This may be due to the different spatial resolution of the two sensors. In the second stage of the research activities, the vicarious calibration of S2A [10] and the atmospheric correction based on dark spectrum approach [11] were applied to both the 2017 and 2019 image dataset. The correlations for 124 match-ups with S2 and 69 match-ups with L8, were statistically significant with a coefficient of determination R2 of 0.959 (L8) and 0.925 (S2). In situ collected SPM concentrations were used to assess the accuracy of SPM products derived from L8 and S2 imagery. A total of 193 match-ups (124 with S2 and 69 with L8) between satellite and in situ data collection were available by considering a maximum time difference of 1 h. The correlations were statistically good with a coefficient of determination R2 of 0.796 (L8) and 0.544 (S2).

Figure 2: a) L8-derived map of turbidity (17/06/2016) over the study area, including the Po river delta and the coasts of the Lagoon of Venice. b) S2-derived map of turbidity (27/08/2018) over the coasts of the Lagoon of Venice. c) S2-derived map of turbidity (24/01/2019) over the coasts of the Lagoon of Venice. d) L8-derived map of turbidity (24/01/2019) over the coasts of the Lagoon of Venice. e) L8 SST map. In all the maps land, clouds, cloud-shadows, whitecaps and sunglint are in black.

Regarding the analysis of suspended sediment transport mechanisms, we focused mainly on the horizontal gradients of turbidity rather than on the SPM concentration, which gave less accurate results in the validation. This analysis considered three events influenced by different wind and wave conditions: Scirocco event (17/06/2016), Bora events (27/08/2018 and 24/01/2019). The event on 17/06/2016 was characterized by 36 hours of Scirocco and Libeccio winds with speed higher than 7 m/s. The alternation of these two winds caused the sediments resuspension from the coasts and favoured the northward offshore spreading of the suspended sediments. L8-derived turbidity patterns in figure 2a are consistent with the findings of numerical modelling of Wang and Pinardi (2002) [12]. The events occurred on 27/08/2018 and 24/01/2019 were characterized by strong Bora with speed greater than 8 m/s. In detail on 27/08/2018 Bora was higher than 8 m/s lasting 24 h before S2 acquisition, while on 24/01/2019 Bora was higher than 8 m/s lasting 3 days. In figure 2b (Bora event, 27/08/2018)  both the exchanges at the inlets due to the preceding ebb tide phase and the flow of suspended sediments coming from northeast by longshore currents are evident, as well as the interactions between the tidal and coastal currents. In figure 2c and 2d, the turbidity maps show high values in a narrow strip of the shoreface and inside the Lagoon of Venice as an effect of wind-induced waves. Sediments are extended offshore and spread southward by the action of longshore currents as visible also on the Sea Surface Temperature (SST) map (fig.2e).

Future perspectives

Results presented in this study suggest that combined L8 and S2A-B data with two-to-three-day revisit time can be considered and exploited as a virtual constellation for operational purposes. The estimation of SPM concentration is not so accurate as turbidity because of uncertainties connected with sampling activities as well as algorithm performances. The analysis of S2 and L8 time series to relate the suspended sediment maps with meteo-marine conditions permits to investigate on sediment transport processes.

Publications and Presentations

  • Braga, F., Scarpa, G. M., Brando, V. E., Manfè, G., & Zaggia, L. (2020). COVID-19 lockdown measures reveal human impact on water transparency in the Venice Lagoon. Science of The Total Environment, 736, 139612
  • Bellafiore, D., Ferrarin, C., Braga, F., Zaggia, L., Maicu, F., Lorenzetti, G., … & De Pascalis, F. (2019). Coastal mixing in multiple-mouth deltas: A case study in the Po delta, Italy. Estuarine, Coastal and Shelf Science, 226, 106254.

References

  1. Vanhellemont, Q., & Ruddick, K. (2014). Turbid wakes associated with offshore wind turbines observed with Landsat 8. Remote Sensing of Environment, 145, 105-115.
  2. Vanhellemont, Q., & Ruddick, K. (2015). Advantages of high-quality SWIR bands for ocean colour processing: Examples from Landsat-8. Remote Sensing of Environment, 161, 89-106.
  3. Vanhellemont, Q., & Ruddick, K. (2016, May). Acolite for Sentinel-2: Aquatic applications of MSI imagery. In Proceedings of the ESA Living Planet Symposium, Pragur, Czech Republic. 9-13.
  4. Brivio, P., Lechi-Lechi, G., Zilioli, E. (2006). Principi e metodi di telerilevamento. 1-525. CittaStudi.
  5. Lillesand, T., Kiefer, R.W., Chipman, J. (2015). Remote sensing and image interpretation. John Wiley & Sons.
  6. Ryu, J.H., Won, J.S., Min, K.D. (2002). Waterline extraction from Landsat TM data in a tidal flat: a case study in Gomso Bay, Korea. Remote sensing of Environment, 83(3), 442-456.
  7. Dogliotti, A.I., Ruddick, K.G., Nechad, B., Doxaran, D., Knaeps, E. (2015). A single algorithm to retrieve turbidity from remotely-sensed data in all coastal and estuarine waters. Remote Sensing of Environment, 156, 157-168.
  8. Braga, F., Zaggia, L., Bellafiore, D., Bresciani, M., Giardino, C., Lorenzetti, G., Brando, V.E. (2017). Mapping turbidity patterns in the Po river prodelta using multi-temporal Landsat 8 imagery. Estuarine, Coastal and Shelf Science, 198, 555-567.
  9. Nechad, B., Ruddick, K.G., Park, Y. (2010). Calibration and validation of a generic multisensor algorithm for mapping of total suspended matter in turbid waters. Remote Sensing of Environment, 114(4), 854-866.
  10. Pahlevan, N., Chittimalli, S.K., Balasubramanian, S.V., Vellucci, V. (2019). Sentinel-2/Landsat-8 product consistency and implications for monitoring aquatic systems. Remote sensing of environment, 220, 19-29.
  11. Vanhellemont, Q. (2019). Adaptation of the dark spectrum fitting atmospheric correction for aquatic applications of the Landsat and Sentinel-2 archives. Remote Sensing of Environment, 225, 175-192.
  12. Wang, X. H., & Pinardi, N. (2002). Modeling the dynamics of sediment transport and resuspension in the northern Adriatic Sea. Journal of Geophysical Research: Oceans, 107(C12).