Earth observation applications for coastal sustainability: potential and challenges for implementation

The monitoring of coastal water quality is a key requirement of national and international regulations, as well as water quality monitoring policies and programmes, such as the European Union’s (EU) Water Framework Directive, USA’s National Aquatic Resource Surveys — National Coastal Condition Assessment, and Australia’s State of the Environment on Coastal Waters. However, ensuring that this can be done adequately and affordably remains a global challenge. In oceanic waters, where the optical complexity of the water is relatively low, the Ocean Colour suite (OCx) of algorithms (O’Reilly et al. 1998; ESA Climate Change Initiative Ocean Colour 2016; NASA Ocean Color Web 2019) are operationally used to produce global maps of chlorophyll-a estimates. The advantage of the OCx algorithm series (x denoting the number of wavebands used in the equation) over other ocean colour algorithms is that they work across a wide range of chlorophyll-a concentrations, and OC4 in particular provides a very high signal-to-noise ratio, and, thus, much more accurate chlorophyll-a retrievals than a variety of other empirical and semi-analytical algorithms (O’Reilly et al. 1998, 2000).

In coastal waters, several studies have shown considerable potential for the application of EO-based methods for deriving water quality estimates (see reviews by Gholizadeh et al. (2016) and Odermatt et al. (2012)) over long temporal and spatial scales (i.e., regional, continental, and, ultimately, global), but the reliable application of these methods across time and space is complicated by the diversity of water types, sensor configuration, and inherent limitations of the approaches used (e.g., atmospheric effects, adjacency effect, sun glint, sea bottom reflectance, empirical algorithm restrictions) (Brewin et al. 2015; Mouw et al. 2015; Zheng and DiGiacomo 2017). In addition, water quality retrieval in optically complex coastal waters is often determined by a combination of spatially and temporally variable properties, such as phytoplankton, suspended material, and coloured dissolved organic matter, all of which affect water colour and transparency to varying degrees. To accurately map coastal water, any operational algorithm(s) would need to account for all these properties rather than focussing on each one of them individually. For this to be achieved, prior knowledge on water typology is required to assess the partial contribution of each property to the overall water quality in the water body under investigation. Unfortunately, this knowledge is currently unavailable for most coastal regions of the world and, as a result, no single operational algorithm has been developed yet for coastal waters.

A relatively new type of classification, based on the optical properties of coastal waters, provides novel avenues for progress, and its application has recently been demonstrated with the production of the global and long-term ESA Climate Change Initiative (CCI) Ocean Colour product (www.esa-oceancolour-cci.org/). The approach clusters coastal waters into optical water types (OWTs) (Moore et al. 2001), based solely on field-, air- or space-borne spectral data (Moore et al. 2014). It has been applied in both coastal and inland water studies with very promising results (Moore et al. 2014; Trochta et al. 2015; Eleveld et al. 2017; Spyrakos et al. 2018), and is now used by ESA CCI OC v3.1 in a blended approach to map chlorophyll-a concentrations globally (ESA Climate Change Initiative Ocean Colour 2016). The advantage of OWTs is that they enable grouping of complex coastal waters based on remotely sensed optical characteristics of coastal waters that are inherently dependent upon ecology without requiring information on ecology itself (Moore et al. 2009). Even though the exact and potentially complex relationship between OWTs and coastal water ecology has not yet been investigated in depth, previous studies have demonstrated the existence of relationships between spatiotemporal distributions of OWTs and physical and biogeochemical processes and ecological diversity indices (Moore et al. 2012; Mélin and Vantrepotte 2015; Trochta et al. 2015), signifying the potential benefit from inclusion of the approach in coastal ecological monitoring and water resource management.

The innovative OWT approach not only has the potential to enable stakeholders to map water quality, such as eutrophication, and primary productivity (by using chlorophyll-a concentrations as a proxy) in coastal waters without a priori or in situ information on coastal water body ecology, and, thus, exploit global remote sensing archives for previously unmonitored sites, but to also provide users with potentially more accurate retrievals of water quality, tailored to the dynamic optical complexity of the water body in question. And while the OWT approach is becoming a powerful tool for data analysts and product developers in improving the accuracy of their EO-based water quality products, the type of information delivered to coastal water managers remains unchanged (i.e., products such as chlorophyll-a maps, or water quality indicators, such as trophic indices, are still delivered, but are more reliable). However, as tends to be the case in applications based on optical EO data, the success of the OWT approach can be limited by the effectiveness of the atmospheric correction (AC) applied to the data. There are various AC models used over coastal waters (cf. Mograne et al. 2019), such as Polymer (Steinmetz et al. 2011) and Case 2 Regional Coast Colour (C2RCC) (Doerffer and Schiller 2007), but there is no single model that works well in all cases, so improved understanding of the limitations posed by ineffective AC is required before the OWT approach can be reliably applied at global scales.

Fisheries are crucial for food and to maintain livelihoods across the globe (FAO 2018); many coastal communities are dependent on fishing or fishing-related industries for income and food security, with fishing being a key part of their cultural identity. Using satellite-derived data to deliver effective fisheries management (e.g., approaches advocated by the international community, such as Ecosystem Approach to Fisheries Management (Chassot et al. 2011; Staples et al. 2014)) ensures access to daily, global, systematic, and high-resolution datasets for incorporating habitat considerations into marine fish population dynamics (e.g., see Druon (2010) for use of remote sensing to support conservation of the Atlantic bluefin tuna), and in particular fish species that support economies of coastal communities. Whilst satellite imagery cannot be used to identify individual fish species, it can provide critical information on the environmental pressures that shape a species population’s survival success. For example, nursery habitats, such as free-floating Sargassum mats in the tropical Atlantic and Caribbean, which are considered as both a benefit (e.g., by providing a key fish nursery habitat (Wells and Rooker 2004)) and a hindrance (when periodic mass landings of the free-floating macroalgae strain local economies (Webster and Linton 2013; Wang and Hu 2018)). These mats can now be detected, characterized, and forecast using optical remote sensing imagery (e.g., Maréchal et al. 2017; Wang and Hu 2017; Wang et al. 2018). Hu et al. (2015) overview how remote sensing analysts can use the red-edge in satellite spectra (i.e., the comparison of high reflectance in the near-infrared region to the low reflectance in the red region of the electromagnetic spectrum, characteristic of vegetation) to remotely detect Sargassum presence or absence on the sea surface. However, Sargassum detection and quantification is often hindered by its spectral similarity to other floating organisms, materials, and often by the inadequate spatial resolution of remote sensing products. Hu et al. (2015) collected and recorded a collection of Sargassum spectra, noting that a distinctive reflectance, caused by the presence and levels of chlorophyll-c pigmentation in Sargassum tissue, can be used to effectively differentiate Sargassum from all other materials, provided there is a minimum of 20%–30% coverage within a satellite pixel. Wang and Hu (2017) describe how these satellite spectra records have been integrated into a Sargassum mapping algorithm. A stepwise process is used to eliminate all data that are of either poor quality or have very little chance of containing Sargassum. This is followed by conducting spectral un-mixing on Sargassum pixels using the spectra reported in 2015 as endmembers. By determining the fractional coverage of each pixel, distribution and area coverage maps can be produced. This process forms the basis of an automated Sargassum detection and biomass estimation process, underpinning the satellite-based Sargassum Watch System (Wang and Hu 2017). Through use of MODIS data archives, it also provisions researchers with a source of hindcast data with which to build predictive models for bloom forecasting (Wang and Hu 2017), with reported accuracies of over 80% in eastern Caribbean regions. There are limitations regarding the reliance on optical imagery, the need to adapt to cloud cover, and the lack of extensive in situ comparisons to accompany products with quality information. However, despite these limitations, this research narrative demonstrates how EO data can feed into advisory services and coastal fisheries management activities, whilst providing fisheries managers with quantifiable estimates of fish nursery habitat extents (e.g., Wang et al. 2018).

It is not only habitats that can be examined using EO. Satellite-derived data on habitat parameters can also feed into species models to understand the impacts of changes on their populations. An example of how niche characteristics, which impact a commercial species population over the course of their life cycle, can be examined and modelled is demonstrated in work done by Garrido et al. (2017). The authors clarified the influence that environmental factors measurable by satellite remote sensing had on the lifecycle of the European sardine Sardina pilchardus. Using satellite-derived sea surface temperature (SST) and estimates of photosynthetic activity based on chlorophyll-a measurements, the authors found that, in general, high recruitment years are associated with high chlorophyll-a and low SST levels. Such work exemplifies how actionable information can be garnered from satellite remote sensing, enabling those tasked with responsibilities for fisheries management and regulation to take appropriate management actions, and conduct more sustainable and integrated economic activities.

Spatial data on coastal areas are key to the management of coastal and marine environments, covering aspects such as boundary and jurisdiction delineation; establishment of baseline, navigation, and seafloor change detection; issuing of consents and licences; and maintaining an inventory of coastal assets. Coastal bathymetry is an important source of spatial data used mainly for safe shipping and navigation (e.g., Jagalingam et al. 2015), marine surveying and mapping (e.g., Huang et al. 2017), but also classification of habitats and modelling of species distribution over a variety of sea floor terrains (e.g., Kostylev et al. 2003; Lundblad et al. 2006; Wilson et al. 2007). Traditionally, active remote sensing methods, such as ship-borne multi-beam echo sounder (MBES) and airborne light detection and ranging (LiDAR), have achieved accurate bathymetry results (Calder and Mayer 2003; Van Son et al. 2009; Kennedy et al. 2014). However, these conventional methods are very expensive, time-consuming, often require special permissions to work in sensitive coastal areas, and, in the case of MBES, inapplicable to shallow waters (e.g., Gao 2009; Monteys et al. 2015; Bannari and Kadhem 2017).

The complementary use of satellite EO and UAV surveying may provide a more efficient and cost-effective method of bathymetric data collection in coastal areas (Lejot et al. 2007; Klemas 2015b). On one hand, multispectral and hyperspectral EO data are already used as a valuable tool in deriving nearshore bathymetry (e.g., Hamylton et al. 2015). Being a non-intrusive technique, it allows for data collection, and the consequent production of bathymetry and seafloor maps, without causing any disturbance to local wildlife and ecosystems. On the other hand, UAVs provide the opportunity to fill data gaps and act as a validation technique to existing EO data products. UAVs, in particular, helicopter platforms, offer increased mobility, hover capabilities, and lower operating costs compared to conventional high-spatial-resolution data capture methods (Klemas 2015b). Collection of spectral data concurrent with satellite pass-over times allows for cross-calibration and validation of processed satellite data. This complementary duel approach has the potential to provide many benefits: low cost, operational flexibility and versatility, ground truth data, reduction of the influence of cloud cover, and surveying of typically inaccessible coastal waters and coastal terrain (Long et al. 2016). However, satellite- and UAV-derived bathymetry data are still susceptible to limitations based on AC inefficiencies and coastal complexities that hinder the optical bottom signal detected at the water surface (Gao 2009; Pacheco et al. 2015). Distortion of the spectral signature due to turbid waters and (or) biota coverage, such as seagrass, proves a challenging restraint to the application of satellite-derived bathymetry across broad coastal areas (Gao 2009; Caballero et al. 2019). Research into the impacts of such coastal complexities on satellite-derived bathymetry and accompanying depth limitations is required to determine confidence levels and identify optimum algorithms (Halls and Costin 2016; Caballero et al. 2019). Integrating high-resolution optical remote sensing with habitat classification data to ascertain the relationship between water reflectance, water depth, and seafloor habitat coverage is used to demonstrate the production of complementary bathymetry and habitat data from satellite data (Lyons et al. 2011; Traganos et al. 2018), with successful applications in coral reef bathymetry (Huang et al. 2017) and seagrass mapping (Hossain et al. 2015). Bathymetric mapping of valuable substrate, such as coral reefs, is a useful tool for monitoring vulnerable areas or areas that undergo topographic change (Vanderstraete et al. 2003) and ESA’s Sentinel-2 satellite captures nearly 90% of the world’s coral reefs, providing vast potential for high-resolution monitoring of benthic change (Hedley et al. 2018).

The cost-effective repeatability of coastal bathymetry products through EO and UAV multispectral data collection offers vast improvements to coastal management challenges and ecosystem monitoring, and offers invaluable up-to-date insights into how shorelines respond to climate change events, globally. An ever-increasing fleet of optical EO satellites, and advancements in UAV technology, can only improve the collection, processing, and production of accurate and reliable coastal bathymetry products.

Changing climates and increased variability present many threats and risks to both human and biophysical systems at multiple scales across the globe; this means that integrating short-term disaster risk reduction measures with long-term climate change adaptation is an ever-increasing priority. One of the challenges with integrated risk management approaches remains identifying climate variables and developing robust, reliable, long-term climate observations linked with extreme events (EEA 2017). In response to this challenge, the Global Climate Observing System has identified 54 essential climate variables (WMO 2016) several of which can be derived from satellite EO data (Hollmann et al. 2013).

Sea level has been identified as an essential climate variable. SLR has historically been measured using coastal tide gauges; however, the development of satellite altimeters has enabled the extrapolation of measurements further offshore providing a wider, consistent spatial context at global scales as well as a greater understanding of casual factors (Ablain et al. 2015). Long-term trends in SLR can now be derived from over 50 years of combined in situ and remote sensing data (Cabanes et al. 2001). The applicability of remote sensing in mapping SLR at global scales has also been demonstrated for a 20-year period (1995–2015) by the ESA CCI Sea Level product (www.esa-sealevel-cci.org/). Furthermore, work by Nicholls and Cazanave (2010) demonstrated the capability to supplement long-term trend estimates with quantifications of regional variability. This ability to translate global trends down to sea basins and regional shorelines allows responsive adaptive management across both temporal and spatial scales.

Even though satellite altimetry is a mature technology for studying open oceans, near-coastal applications are impeded because of mixed pixels and contaminated return signals caused by hard surfaces, such as nearshore rocky outcrops and coastal infrastructure; this results in unreliable sea level estimates close to the coast (Vignudelli et al. 2005). However, altimetry specialists have begun extracting usable information from data previously considered to be of poor quality. For instance, total water level envelope estimates (i.e., the sea surface level taking into account tides, waves, effects of a storm surge, and (usually) precipitation and river flows) can now be extracted up to a distance of 3 km from the shoreline, a distance that is decreasing as sensor capabilities and processing algorithms improve (Vignudelli et al. 2011). Despite these advancements, the establishment of targeted nearshore (i.e., 3–50 km from the coastline) and offshore (further than 50 km from the coastline) sea level monitoring sites allowing for direct satellite-to-field measurement comparisons would be beneficial to gauge the accuracy of satellite-derived sea level data.

In addition to SLR, one of the most promising applications of coastal altimetry is in the study of storm surges (Cipollini et al. 2014). Impacts from storm surges and associated coastal flooding events are anticipated to increase in severity (e.g., for the Atlantic, see Bernier and Thompson (2006) and Tebaldi et al. (2012)) and so reliable modelling of storm surges can support both preparation and mitigation activities by improving understanding of how waves propagate and impact on coastal infrastructure. Improvements in this understanding could bring enormous societal benefits, particularly to populated low-elevation coastal zones. Research by Madsen et al. (2015) demonstrated how, using a stationary blending method, coastal satellite altimetry data can be related with corresponding tide gauge measurements, allowing generation of sea level maps and reducing storm surge forecast error. Liu et al. (2018) developed a risk and vulnerability assessment approach combining coastal hydrological observations, numerical storm surge modelling, and multi-source remote sensing data to map coastal infrastructure and land use. This approach was then used to assess overall risk of storm surges. Another example includes the use of UAVs to estimate the extent of coastal flooding and assess damages in critical areas demonstrating innovative EO capacities in the support of coastal hazard risk management activities (Popescu et al. 2017).