Dashboard Week Day 2 - Exploring the Changing Forests

by Alisha Dhillon

The dataset for day 2 looked at the change in forest cover. I began by formulating my questions.

When - time series, heatmaps. I thought for this I could have small spark lines for the three areas within the dataset: forest loss, burned area and tree cover loss. This would be a general overview of the trends over time.

Where - maps, bar charts. Where is tree cover loss, forest loss and burned area occurring most? I had issues with mapping in Tableau on day 1 and so was cautious to use one but did so in the end. I also included a jitter in the where and thought it would be more useful to read trends of countries within regions. I made notes to change region in tableau to geographic role to make things easier.

Who - bar charts, histograms. This linked back to my where for me in who was being affected and so I didn't explore this very well in the end.

Why - Scatterplots, bar charts, heatmaps. When I thought of burned area, I thought of increasing temperatures leading to forest fires. I thought about exploring the possibility of climate temperatures against each of the measures within the data to see if there were any trends. I had rough notes of possible bar in bar chart, bar with gantt or even scatterplot.

My next step was to look at the data. I realised immediately that for my comparisons, I didn't require most of the columns. Secondly, I needed the data in a different format to be able to use it as a time series.

This is how it was before. There were many nulls. The year start date and end date for the three categories I wanted didn't match and so I kept losing data through my join and didn't understand why.


Having it in this format made it easier for me to visualise across a time series, but also make parameters with as I thought it would be useful to be able to switch between the three dimensions across my dashboard. Big thank you to Andy for helping me with this because I would have wasted a lot of time otherwise and panicked. The workflow looked like this (yes it's a mess):

Essentially, I needed to transpose everything apart from the key columns. Each transpose looked at an individual dimension. The next step was to regex and extract the year. This was done with a parse of (\d\d\d\d) - yes I know I could have also done (\d{4}) to save myself some \d's. Using Select, I then got rid of name and renamed value accordingly to tree cover loss, forest loss or burned area. This was all straightforward until we realised that the join wasn't working. The reason for this was that the three dimensions did not have the same values. Therefore, I added a filter to get rid of the nulls. After this, the join seemed to work and I had the cleaned dataset.

I proceeded to create my charts and carry out my analysis with the plan in mind, however, end up with too many charts and very little time. I then scrambled to put together a dashboard.

This was the version I presented. I did run out of time, and then start rushing to make charts which weren't properly thought through. I kind of had to throw something together very last minute and wasn't so pleased. The time pressure of dashboard week is something that I've been finding hard to adjust to. After the presentation, I spent a little bit of time fixing my dashboard and end up with my final result.

Final thoughts -

I still don't entirely love the end result. I made it in a day (mostly), but I feel like it lacks analysis. It's almost as though some dots need further connecting. I think I got so caught up in the chaos of all of my questions and making as many charts that would work, that I forgot to look back and see how they fit within the greater picture. My key take away from day 2 was less is more. Hopefully, this will help me manage my time better too.

Find my viz here: