Challenge background
Ghana has approximately 550 km [measured with Google Earth] of coastline. Along that coastline sits two major cities, about 46 other towns, communities and settlements, and a few industrial zones. Most of the smaller communities depend heavily on fishing for their economic activity and therefore tend to settle as close to the shoreline as possible. This was not an issue a few decades ago when the distance between the coastline and shoreline could sometimes be 50m to 100m wide during low tides but rising sea levels, increased tidal action, as well as some negative human activity [sandwinning] has led to accelerated soil erosion of the shores. This in turn led to most of the communities being at risk of property damage and loss from tidal flooding, and loss of life in extreme cases.
The problem
The situation has gotten so bad that the government has had to step in over the past two decades and put in measures to protect the country’s shoreline. The solution is the construction of various sea defense structures (revetments, breakwaters, etc). However, the average cost to the government for a kilometer of revetment was ~$15m, a simple calculation later and it will cost the country $8.25bn to protect her shores. This is not feasible economically so priorities have been given to large coastal communities and key areas of interest along the coast. The downside to this approach is that without the tacit knowledge of all stakeholders or until significant damage is done, the decision-makers are unable to adequately determine which zones to prioritize first. Having a system in place that can assess the year-on-year shoreline change can better inform decision-makers about the sections of the coast to prioritize.
Goal of the project
- Develop a machine learning model that can extract the shoreline and coastline from a yearly composite satellite image.
- Define a process to assess the year-on-year change in the shoreline. This process should further lead to other goals such as determining the rate of soil erosion on the coast.
- Deploy a web-based dashboard to visualize the shoreline change.
Project timeline
- 1
Week 1
Literature review and refinement of goals.
Setup of development environments and resources
Data Collection
- 2
Week 2
Data collection
- 3
Week 3
Data processing, EDA, and further data collection if revealed as necessary by EDA
- 4
Week 4
Model Development and assessment
- 5
Week 5
Model Development and assessment
- 6
Week 6
Model Development and assessment
- 7
Week 7
Model Deployment, Dashboard development
- 8
Week 8
Dashboard Deployment and final touchup
What you'll learn
Satellite Imagery collection and processing, Feature extraction from satellite imagery, machine learning