Shoreline Change Prediction using Satellite Imagery

Published
Sep 20, 2022
Reading Time
Rate this post
(5 votes)
Shoreline Change Prediction using Satellite Imagery

ACF Type: wysiwyg

The Omdena India Chapter, Ranchi, took on the challenge to detect the current rate of shoreline change of Sagar Island (India) as well as predicting the changes for the years 2030 and 2050.

Author: Hala Jadallah

Project collaborators: Sairam Kannan (project lead), Deepali Bidwa, Nikhil Raj Deep, Elena Andreini, Abhay Bhadani

The impact of shoreline change

Shoreline of Sagar Island is affected by natural forces that lead to erosion or accretion. In this project, we attempted to quantify the rate of change of Sagar Island shorelines.

Such change will have an impact on human activity near the shoreline. For planning purposes, authorities would benefit from knowing what to expect in the coming 10 to 20 years.

Data collection

Since our interest is decadal change, we examined satellite imagery in the years 1990, 2000, 2010, and 2020. We used Google Earth Engine (GEE) to download Landsat 5 (1990, 2000, 2010) and Landsat 8 (2020) satellites, with 30 m spatial resolution, and level 2 products for surface reflectance.

Since cloud cover over Sagar Island is minimal during November and December, we chose a satellite image over a day with cloud cover of less than 10% between November 1st and December 31st for each year.  

Shoreline detection

We aimed for auto-detection methods of the shoreline. To that end, we experimented with a couple of water indices to create binary images that separate water from land.

One water index is the Modified Normalized Difference Water Index (MNDWI) defined as the normalized difference between the green band and the shortwave infrared band (SWIR1).

The other one is the Automatic Water Extraction Index (AWEI). It is a multispectral index that combines the green, the near-infrared, the shortwave infrared bands SWIR1 and SWIR2.

We considered CoastSat, a python based library for this purpose. It applies image classification, subpixel resolution for border segmentation and MNDWI along with OTSU’s thresholding algorithm.  We found it a bit unintuitive to make use of the vector data it generated that carries the polygon delineating the shoreline.

Figure 1. CoastSat output of its shoreline detection algorithm.

Figure 1. CoastSat output of its shoreline detection algorithm.

Therefore, we preferred to use the GEE platform to compute AWEI combined with OTSU’s thresholding algorithm to detect shoreline. These are binary images (raster data).

GEE also make it easy to generate vector data of the raster files to get polygons that separate water from land. These were not smooth and not most accurate. However, they provided a good guide to manually delineate the shoreline.

Figure 2. Sagar shorelines as digitized using DSAS for 1990 (purple), 2000 (red), 2010 (blue), 2020 (green)

Figure 2. Sagar shorelines as digitized using DSAS for 1990 (purple), 2000 (red), 2010 (blue), 2020 (green)

Build your portfolio with real-world projects from Omdena

Estimating rate of change in shoreline and prediction

Manual delineation of the shoreline was done using Digital Shoreline Analysis System (DSAS) which is an ArcGIS plugin. Moreover, DSAS automatically places transects (short lines perpendicular to the shoreline) along the four decadal shorelines.

Figure 3. Accretion regions along Sagar shorelines: low accretion (green), medium accretion (blue) and high accretion (purple)

Figure 3. Accretion regions along Sagar shorelines: low accretion (green), medium accretion (blue) and high accretion (purple)

Figure 4. Erosion regions along Sagar shorelines: low erosion (yellow), medium erosion (orange) high erosion (red)

Figure 4. Erosion regions along Sagar shorelines: low erosion (yellow), medium erosion (orange) high erosion (red)

Then using a buffer region and it defines a reference point and it computes the rate of change in the shoreline using linear regression of each transect. Finally, it gives the predicted shoreline for 2030 and 2040.

Figure 5. Predicted shorelines in 2030 (green), and 2040 (blue)

Figure 5. Predicted shorelines in 2030 (green), and 2040 (blue)

Visualization and deployment

We deployed the visualization of the result and summary using the Streamlit app that we developed, where one can view the above snapshots of our interactive map. Please explore it here

References

Sengupta, D.; Chen, R.; Meadows, M.E.; Choi, Y.R.; Banerjee, A.; Zilong, X. Mapping Trajectories of Coastal Land Reclamation in Nine Deltaic Megacities using Google Earth Engine. Remote Sens. 2019, 11, 2621; doi:10.3390/rs11222621

Nandi, S.; Ghosh, M.; Kundu, A.; Dutta, D.; Baksi, M. Shoreline shifting and its prediction using remote sensing and GIS techniques: a case study of Sagar Island, West Bengal(India) J Coast Conserv (2016) 20:61–80

Vos, K.; Splinter, K.D.; Harley, M.D.; Simmons, J.A.; Turner I.L. CoastSat: A Google Earth Engine-enabled Python toolkit to extract shorelines from publicly available satellite imagery. Environmental Modelling and Software 122 (2019) 104528

ACF Type: image

ACF Type: text

Hala Jadallah

ACF Type: url

AI Projects to boost your career

Develop your skills, build a project portfolio, and make an impact. All at once!

Projects

Q
Newsletter
Leave a comment.
0 Comments
Submit a Comment

Your email address will not be published. Required fields are marked *