Is Remote Sensing Becoming the New Frontline of Coastal Protection?

Remote sensing is quietly becoming one of the most important tools in coastal climate resilience. As sea-level rise, urbanization and land-use change place increasing pressure on deltas, estuaries and wetlands, governments need faster and more consistent environmental intelligence.

The editorial "Remote Sensing in Coastal Water Environment Monitoring: Advancements, Challenges, and Future Perspectives," published in Water by Peng Li, Fengqin Yan and Xiuling Zuo, brings together five studies showing how Earth observation can track everything from phytoplankton blooms and salinity shifts to floating plastic and ecosystem degradation.

Keeping pace with accelerating change

Traditional field monitoring remains indispensable because it provides precise measurements of water chemistry, ecosystems and physical conditions. But it is expensive, geographically limited and difficult to repeat frequently across rapidly changing coastlines. This limitation is becoming more serious as multiple environmental pressures converge. The editorial highlights groundwater depletion-driven land subsidence, accelerated relative sea-level rise, saltwater intrusion, coastal wetland degradation and changing water quality as interconnected problems across deltaic and coastal environments.

Remote sensing offers governments a way to expand the scale and frequency of observation. Satellite ocean-color data, airborne sensors, unmanned aerial vehicles and Interferometric Synthetic Aperture Radar can capture different dimensions of coastal change—from offshore water quality to onshore land deformation.

The significance is not merely technical. Environmental governance is often constrained by incomplete visibility. Pollution may be detected after ecological damage has occurred; shoreline degradation may be assessed only intermittently; and weak monitoring can make enforcement difficult. High-frequency Earth observation potentially changes that equation by giving authorities a wider and more consistent evidence base.

One study reviewed in the collection illustrates the scale of that transition. A framework for coastal ecosystem health combined indicators of water quality, vegetation condition and human disturbance using Landsat and Sentinel-2 data. Deep-learning segmentation using U-Net achieved classification accuracy of up to 93.65%, according to the editorial. This does not eliminate the need for scientists in the field, but it suggests that automated systems can help environmental agencies identify where field investigation is most urgently required.

AI is extracting signals from coastal waters that were previously difficult to see

The collection also demonstrates how machine learning can turn satellite imagery into estimates of environmental variables that are difficult to observe remotely. In Qinzhou Bay, China, researchers combined GF-1 satellite imagery with 96 field samples to estimate sea-surface salinity. Coastal salinity is challenging to retrieve because suspended sediments and organic matter interfere with conventional remote-sensing approaches. The model used an improved gray-wolf-optimized back-propagation neural network and achieved a validation determination coefficient of 0.953, with a mean absolute error of 0.906 practical salinity units.

The model also captured meaningful spatial patterns. During the wet season, average salinity in the inner bay fell to about 14 psu because of freshwater runoff, while outer-bay levels remained around 25-30 psu throughout the year.

For coastal authorities, capabilities like this could support monitoring of freshwater inflows, estuarine dynamics and saline intrusion. Similar systems may eventually allow officials to move from occasional measurement campaigns toward near-continuous observation.

Long time-series data can also reveal whether environmental change is being driven more by climate variability, land use or policy intervention. A study of the Yangtze River Estuary used MODIS observations from 2003 to 2020 to analyze phytoplankton blooms. It found a fluctuating decline in overall bloom intensity and affected areas after approximately 2012-2013. Land-use change and climatic factors together explained 65.29% of interannual variation, with developed land having the strongest individual impact.

The analysis linked declining nutrient pressures partly to pollution-prevention policies, managed urban expansion and agricultural reforestation. That points to an important emerging use for remote sensing: evaluating whether environmental policy is producing measurable ecological responses.

The biggest leap is from detecting pollution to predicting where it will go

Perhaps the most strategically important development in the collection is the shift from static observation toward predictive monitoring. In the Black Sea, the DEEP-PLAST framework combined Sentinel-2 imagery, drone observations, deep-learning detection and physical drift modelling to track floating plastic. Satellite segmentation using U-Net++ achieved an F1-score of 0.840, while a YOLOv5s detector applied to drone imagery achieved 0.910 precision.

However, detection was only the first step. Researchers used the identified plastic locations as starting points for simulations driven by daily ocean-current data and six-hourly wind vectors. Forward and backward modelling was then used to identify accumulation areas and potential pollution sources along the Romanian Black Sea coast.

This represents a major governance opportunity. If pollution can be tracked backward toward likely sources or forward toward vulnerable coastlines, monitoring begins to support enforcement, cleanup planning and cross-border accountability rather than merely documenting contamination.

The research on Mexican marine sanctuaries points in the same predictive direction. Analysis of 1997-2018 remote-sensing data found that El Niño phases produced significant winter and spring warming across all 18 identified biophysical zones, while chlorophyll-a concentrations increased during La Niña in many regions. The authors warn that warmer baseline conditions could suppress marine primary productivity.

For climate adaptation, understanding these seasonal responses may be more useful than relying on annual averages that conceal when ecosystems are most vulnerable.

The technology is advancing faster than the systems needed to trust and govern it

The collection makes a strong case for remote sensing and AI, but it also highlights the risk of treating sophisticated models as substitutes for robust environmental institutions. High model accuracy in one bay, wetland or coastline does not guarantee comparable performance elsewhere. Coastal environments vary sharply in sediment, vegetation, water chemistry, atmospheric conditions and human pressures. AI models therefore require local validation.

The authors identify several technical priorities: integrating optical imagery with all-weather Synthetic Aperture Radar, polarimetric radar and bathymetric LiDAR; overcoming cloud-related data gaps; expanding localized ground-truth observations; and reducing noise in AI training labels.

The Black Sea plastic study itself illustrates the problem. Its drift validation remains qualitative because field datasets are limited.

For developing countries, the opportunity is substantial but so is the capacity challenge. Open satellite data can lower the cost of observation, but operational systems still require technical skills, computing infrastructure, ground measurements and institutions capable of translating environmental information into regulatory action.

This is where the research intersects with the broader development agenda. Remote sensing can contribute directly to SDG 6 on clean water, SDG 13 on climate action and SDG 14 on life below water, while also strengthening disaster preparedness, blue-carbon management and coastal planning. For low- and middle-income countries with long coastlines but limited environmental-monitoring networks, Earth observation could be especially valuable.

However, the real transformation will not come from better imagery alone. The strategic shift is from environmental monitoring as an occasional scientific exercise to environmental intelligence as permanent public infrastructure. Governments increasingly have access to systems that can observe ecosystems at scale, identify complex patterns and forecast environmental movement. The next challenge is institutional: building the standards, validation networks and decision-making processes that allow those signals to trigger credible action.