How fast is Pulau Semakau filling? Using Satellite Images and GIS to estimate landfill rates
Disclaimer: This project is not an official scientific study and should not be seen as a robust source of information. This article is written for a general audience with no knowledge of GIS.
Shoutout to my amazing group mates Zhou Yang, Hui Jie, Azrul and Khai for working with me on this project.
What do Geography majors learn in University? What is Geographical Information System (GIS)? Do Geographers only study rocks and volcanoes?
Throughout my internship and work experiences, be it in start-ups or in government agencies, I have repeatedly encountered the questions above, albeit worded differently. Even my former colleagues from the National Environment Agency, have little clue about Geography.
A degree is a piece of branding. Most people will have a decent idea of a person’s skillset upon hearing they have a “Business” or “Data Analytics” degree. But “Geography” is shrouded in mystery. The lack of awareness and even negative perceptions do affect our career opportunities.
Hence, I have decided to write this short article on one of my group assignments. Hopefully, this article will help to increase awareness of the work we do as Geographers.
What is Geographical Information System (GIS) ?
At the National University of Singapore (NUS), Geographical Information System (GIS) modules (and the minor) is offered by the Geography department due to their close synergy and relevance.
GIS is essentially a lesser-known sibling of Data Analytics, where we similarly process, analyse and visualise data. The key difference is that GIS focuses on spatial/location data and generating insights that have spatial implications. For instance, you can use GIS to monitor the extent of deforestation or analyse the distance of amenities from housing estates.
While Data Analysts may use Rstudio or Tableau, we have GIS software. Below is a screenshot of QGIS, the GIS software that was used for this project. QGIS is open source, free GIS software that anyone can use to analyse and process geospatial data (No, I was not sponsored by QGIS).
What is Pulau Semakau? Why study Pulau Semakau?
As part of a Level 4000 GIS module, my group assignment was to analyse satellite images to provide insights into real-world problems. My team decided to do a project to examine the landfill rates of Pulau Semakau.
Most Singaporeans would be familiar with Pulau Semakau, our island nation’s only landfill site.
The 350-hectare island just south of Singapore is the site where incinerated waste and construction waste is disposed of. Operating since 1999, the island was originally intended to last until 2045 but the expected lifespan was shortened to 2035 due to economic and population growth.
Do check out this article by the Ministry of Sustainability and the Environment (MSE) on the Myths and Reality of Pulau Semakau.
As a team, we were curious about how fast Pulau Semakau is filling up and how much of the island is already filled. This serves as a proxy for how much trash we Singaporeans have produced over the years and if the waste reduction policies were effective. Unfortunately, such data about the island is not publicly available and probably difficult to collect.
How did we do it?
Disclaimer: This section is written for a general audience and is an overly simplified version of the methodology. If you are curious about the exact methodology do feel free to contact me.
Similar to Data Analytics projects, we start with sourcing for raw data. We downloaded 10 satellite images (LANDSAT 7) of Pulau Semakau from the United States Geological Survey (USGS) database. The images were from 2000 to 2020 with an approximately 2 years interval between each image.
On USGS EarthExplorer, you would be able to find satellite images to almost every part of the world and they are completely free. Shown below are the 10 true colour images of Pulau Semakau that we have collected.
Next, we used a QGIS (a GIS software) to process, clean and eventually analyse the image.
We utilised a few inbuilt functions on QGIS and a well-established index (known as the Automated Water Extraction Index) to automatically classify the individual pixels in the Satellite images to Water (unfilled areas) or Non-water (filled areas).
The physics behind the classification is slightly complex. It essentially uses the difference in electromagnetic signatures (e.g. red light, Infrared, Near Infrared…) between water and non-water areas to differentiate the pixels.
This converts the original satellite image into a classified image, where every pixel is classified as either water (appear as black) or non-water (appear as white). Non-water will include the filled areas, the roads and even vegetation on Pulau Semakau.
What did we find out?
After experiencing numerous technical and data issues (like most data projects), we were able to generate images showing the changes in Pulau Semakau over the 2 decades. The GIF below shows the reclassified image, where the white pixels are water and the grey pixels are non-water (areas that have been filled).
Apart from the nice map (I am a fan of maps), the GIS software generates the exact number of individual pixels that are classified as water and non-water. Based on this data, we were able to estimate the percentage of the island that was still unfilled on the particular day when the image was taken (see the graph below).
Based on our study, the island was 69% unfilled on 7th July 2000, approximately 1 year after being operational. By 2020, only 38% of the island still remained unfilled.
Interestingly, we also found an extremely high linear correlation (R2 = 0.969), suggesting that the rate of the landfill has remained relatively linear ( constant) over the past 20 years. This meant that the waste reduction policies/strategies and population growth have not changed the amount of trash sent to Pulau Semakau.
Subsequently, using a linear model, we were able to project that Pulau Semakau will have 0% water pixels by approximately 504 months from the first data point (July 2000). In simple words, assuming that the rate of landfill remains constant, our study projects the landfill to be completely filled by June 2042. This is 7 years longer than the official projection of 2035.
Therefore, we can rest assured that the reported end date of Pulau Semakau being filled by 2035 is likely a conservative and safe estimate. The estimate likely accounts for an increase in waste production due to growing population and consumption trends.
Concluding Notes
To be honest, this project did not reveal any groundbreaking insights. We were limited by the availability and quality of data. We were unable to verify/cross-check the accuracy of our technique since there is no official estimate of Pulau Semakau’s status. Furthermore, there were issues with the Landsat 7 images and we were only partially able to mitigate the errors.
Nevertheless, we were still happy to have some answers to our initial problem statement and have an opportunity to improve our GIS skills. Other applications of GIS and Satellite Images include monitoring of wildfire potentials, growth of urban areas and even monitoring water quality.
If you are keen on finding out more about GIS, do check out my other article where I used GIS and python to analyse Singapore’s public transaction records during my internship at URA.
Thank you very much for reading this article. Hope you have enjoyed it and have learned a little about Geography and Geographical Information System (GIS).
To my fellow Geographers, we won’t make much impact by writing academic essays, discussing abstract theories and talking in convoluted jargon. I encourage my fellow Undergraduates to convert your interesting research papers and group assignments into accessible articles. Hopefully, this will help to increase awareness about our major and the work we do.
Feel free to contact me on my LinkedIn if you would like to find out more or have a discussion.