Last Updated on 2 Feb 2021.
[dropcap]I[/dropcap] have completed my honours degree in BSc Geoinformatics (see my post where I explain what GIScience is) and would like to share what my honours project was about, as I got a lot of questions. I’ve been trying to limit the technical aspects, so I hope this gives you a better idea of what I’ve been up to in the past few months.
The Problem 🤔
Sustainable management of plantations is of the utmost importance for many reasons, among which timber production is most important. However, there are threats to commercial plantations that cause trees to die. My project focused on evaluating how tree canopy gaps appear and change over time. A tree canopy gap is formed when a tree dies leaving a hole in the canopy.
Originally, tree canopy gaps were measured by doing fieldwork – so walking around physically with a GPS and marking where these gaps occur. However, this method is time and labour intensive and plantations are also sometimes very remote which further complicates the process.
The mapping of tree canopies has also been done in the past using high-resolution aerial and satellite imagery. From these photos it is possible to see gaps in plantations, but shadows sometimes hide the ground. One also has no sense of the three-dimensional aspect of the plantation with a normal photograph.
The Solution 💡
LiDAR, or Light Detection And Ranging, is a way of collecting data similar to RADAR and SONAR. It uses parts of the energy spectrum that people cannot see to transmit light pulses and then capture its reflection to determine altitude (distance). What makes LiDAR unique is that it allows you to get a detailed 3D model of objects on the Earth’s surface, unlike 2D aerial imagery that does not offer this depth aspect. Furthermore, LiDAR uses an active sensor – the sensor generates its own energy. Ordinary cameras rely on sunlight to be able to “see”. This feature means that LiDAR data is sometimes collected from aircraft at night when the airspace is not so busy.
In the case of a plantation, one can see the structure of the trees very well from LiDAR data. I could therefore use these points (or rather the lack thereof) to see where tree canopies are.
Study Area 🗺️
I chose the Tokai and Cecilia pine plantations – the only two plantations in Cape Town that still exist – as my study area, as it was close to do field work and the Cape Town municipality has up to date LiDAR data. The data I used for the study was captured in 2011 and 2017. The map below shows the study areas outlined in black.
In terms of vegetation under the trees, there are more shorter plants in the Cecilia plantation than in Tokai’s plantation.
The Cecilia area is much wetter than Tokai with many natural vegetation such as fynbos, indigenous forests and then pine and eucalyptus plantations. Tokai has pine plantations and remnants of eucalyptus plantations that were harvested years ago. There are also fynbos growing around the plantations at Tokai.
There are a few methods I have used to identify problem areas in plantations:
- Canopy gaps – here I looked at 2011 and 2017 separately.
- Areas where height has decreased between 2011 and 2017 – it’s not supposed to happen in a normal plantation! 😅
- Canopy Gap Fraction – what percentage of an area on the ground does not have trees.
- Leaf Area Index – how many leaves the trees have.
- My own combination of method 1 and 3.
There were clear deviations from what one would consider “normal” for a plantation. But let’s start with some of the stats I obtained through analysis of the LiDAR data.
The graphs above were made by categorising the height values associated with each LiDAR point. The average is indicated by the black line and in this case one can see that there is a height increase between 2011 and 2017.
Method 1: Height Values
From the map above it is evident that there are many pixels where the height of the plantation falls below 7m – it is quite possibly a gap between the trees. Then one can also clearly see that new streaks of these gaps have emerged in 2017. The company that ran the plantation started cutting the trees in August 2016, but a court decision halted it before many trees were cut down.
Method 2: Difference in Height Values
I also subtracted the elevation values in 2011 from those in 2017 (method 2) to obtain a map indicating elevation change.
Method 3: Canopy Gap Fraction
The third method involves calculating what percentage of an area on the ground does not have trees. The LiDAR360 package has this functionality built in.
In the maps above, the footpath is clearly visible on the left in both years, but in 2017, however, something happened that caused the gap fraction to increase. During fieldwork, I confirmed that these are dead trees and from Google Earth satellite photos, the trees appear to have been damaged by a fire or illness between 2011 and 2017.
Method 4: Leaf Area Index
The leaf area index did not perform well and only gave useful values here and there.
In the map above you can see the white parts where there are no leaves. High (green) values indicate healthy vegetation with many leaves. The red lines in the 2017 classification are due to an error in the input data that may have been incorrectly collected.
Method 5: AND and XOR
The last method I devised was to get rid of the “salt-and-pepper effect” (many gaps that mistakenly appear as single pixels). Basically, I created the maps using methods 1 and 3 combined with the Boolean AND and XOR operations – in plain English this is just where the two results match and differ. Where the maps match (AND) one can then say with more confidence that there is indeed a tree canopy gap.
And now you know in brief what my honours project was all about! The full report is unfortunately not available online, but my presentation for the SSAG’s 2018 conference as well as my final project poster is available in PDF format below.
My supervisor for my honours project was Mrs Zahn Münch of the Department of Geography and Environmental Studies.