Image by Matthias Fischer from Pixabay

Introduction

Bushfires have a significant impact on ecosystems, air quality, and human lives. The dataset for today’s Tableau dashboard is the MODIS Active Fire Data. This is freely available information about the location, intensity, and extent of actively burning fires for a given day, at a granularity of 1km pixels. It uses thermal anomalies detected from satellite imagery to pinpoint fire hotspots worldwide.

Data Investigation

Although there is worldwide data going back to 2000, the full data would be too much for my laptop to process and dynamically visualise on Tableau, as well as being a lot of information for one dashboard. A decision was made early in the day to focus on Australian bushfires. I was particularly interested in the 2019/20 bushfires as I was not in Australia to experience that firsthand, as well as former bushfires that have affected family and friends.

The MODIS data contained points of activity but did not show boundaries of entire fire regions. I thought it would be useful to be able to see the total area of a fire. A quick search led to the FIREDpy package which included fire shapefiles generated from a different MODIS dataset for Burnt Area Product.

By combining these datasets, I expected to see a clear relationship between the two. In reality it was difficult to draw meaningful conclusions from the original MODIS dataset. The most active points seemed to be industrial plants (see below) and even after filtering to only show suspected vegetation fires, there were many recorded in the Gulf of Carpentaria.

One of the most active sources of ‘fire activity’ at the start of the millennium was the Port Kembla steelworks.

The FIREDpy shapefiles appeared to be reliable. I found it surprising that the largest fires by area were in NT, SA and WA.

The Dashboard

Once the data was in Tableau my attention was focused on overlaying the map layers in a visually pleasing way. This was a tricky balance of colours, opacity and size settings in the density map.

One tip which I forgot since I was taught 1-2 months ago: to remove the ocean/background of the map so it is the same colour as the background of the dashboard go to Map -> Background Layers.. and untick “Base”.

Aside from the map I included one bar chart of the top fires by area, duration and growth rates, and a second bar chart of the fire activity over time. A very pronounced spike in NSW fire activity was recorded in 2019 Q4 as expected. A major limitation of the FIREDpy dataset was the lack of intensity information. Adding information on the temperature, amount of vegetation burnt or potentially even wildlife figures for the areas would enrich the analysis.

You can see the dashboard here. Feel free to comment if you find anything noteworthy in the dashboard.

The Data School
Author: The Data School