The Measurement Of Ecosystem Degradation
Our individual and collective health is inextricably connected to the health of our natural ecosystems. Our fundamental resources (air, water, food, medicine & materials) are partially/completely provisioned, supported & maintained by functional ecosystems. A threat to these systems is a threat to ourselves. Ironically, human activity is the largest threat. But we can also be the solution.
Where do we start? Well, we look. We look where obvious ecosystem degradation is occurring. From there we can begin answering further questions regarding consequences and remedies.
With modern technology we’re able to look at our planet in great detail. Remote sensing technologies allows us to obtain geospatial data sourced from devices such as satellites. Geographic Information Systems (GIS), among other things, allows us to view and analyse this data. I will be using these technologies to answer the question — how much ecosystem degradation is occurring regionally? This is a very complex question to answer in a single post, so I’ll reduce the question to something more fundamental — how much vegetation degradation is occurring regionally?
Vegetation is a fundamental factor, as trees & plants are the aboveground foundation & infrastructure for energy, water, nutrient, carbon & microbial cycles in terrestrial ecosystems. Therefore, vegetation cover is a principle indicator of ecosystem function. Zero vegetation (i.e. bare ground or urban ground) indicates low function. Higher vegetation (i.e a rainforest) indicates high function. This is a good baseline.
Since 1972 NASA has been deploying Landsat satellites into orbit as part of the on-going Landsat program. These Earth-observing satellites continuously scan our planet to record images of various reflective wavelengths. The youngest two sister satellites (Landsat 8 & 9) work together to completely scan the Earth’s land and coasts every 8 days. You can view where they are now and watch their live stream - Earth Now. The sisters have two sensory instruments: Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS).
To measure vegetation, we’ll be looking at the data recorded from the OLI sensor. Specifically, bands 4 (0.64–0.67µm) and 5 (0.85–0.88µm) which capture the red and near-infrared (NIR) wavelengths.
Within the chloroplasts of leaves are chlorophyll molecules that perform photosynthesis. They absorb and reflect various quantities of wavelengths within the visible range. Leaves look green because they reflect the green wavelength more strongly than blue & red within the visible spectrum. This is because they absorb blue and red to produce energy. As seen in the spectral profile above — leaves also reflect NIR, far more than red, which gives them a defining spectral characteristic.
It’s the chlorophyll that defines the visible portion of the profile, while it’s the cell structure that defines the NIR portion (as seen below). The cell structure evolved to strongly reflect NIR, as wavelengths higher than 0.7µm will likely overheat and damage the leaf. Therefore, chlorophyll absorbs blue and red while the cell structure protects it from NIR.
The sharp increase/difference of reflectance from red to NIR is the characteristic used to measure the amount of vegetation in the OLI images, which is why we selected bands 4 (red) and 5 (NIR) for our analysis. We’ll be relying on this large difference for our calculations.
Landsat images are processed and managed by USGS and are freely available to download on Earth Explorer. I’ve download the Landsat 8–9 datasets of bands 4 & 5 from somewhere in Brazil. We can see below that there are patches of what looks like slash & burn agriculture, which will provide a contrast between alive & cleared vegetation.
If we ignore the clouds ‘clouding’ our images, we see that the areas with vegetation have a low red reflectance values on the left image (darker purple) and high NIR values on the right (lighter green). What we’re actually seeing here is the high levels of red wavelength absorption of chlorophyll and the high levels of NIR wavelength reflection of cell structure. We can combine the values of both to more clearly view where photosynthesis is happening by calculating the Ratio Vegetation Index (RVI):
What we’re doing here is calculating the difference between NIR and red reflection wavelengths. As we discovered earlier from B. Smith’s graph, there’s a large positive differential between red and NIR in the vegetation profile. The larger the positive differential the more certain we are that we’re measuring vegetation. The greater the NIR and/or the smaller the red wavelengths, the greater the difference, therefore the greater RVI. This is exactly what we see below with RVI applied:
With low (dark purple) and high (light green) RVI values, we can make a confirmative observation — the clouds from our previous images have now been removed. This is simply because the spectral profile of clouds have a very small differential between NIR and red, resulting in a low RVI value:
As RVI indicates the relationship between chlorophyll and cell structure, it therefore indicates how healthy or dense the vegetation is. If there is less chlorophyll and/or damaged cell structure then the spectral profile would resemble something closer to dry grass (illustrated above). You could imagine during a period of drought that the vegetation’s spectra would gradually transition from a healthy profile to that of a dry grass profile, reflecting as a gradual decline in RVI as the NIR/red differential decreases.
RVI values can be inconvenient to deal with due to varying wavelength levels across different areas around the world, making comparisons difficult. So RVI was developed into NDVI to address this. Normalised Difference Vegetation Index (NDVI) normalises the RVI values:
Now the Vegetative Index will always be between a range of -1 and 1. If NIR is greater than red, then we get positive NDVI values below 1. Else if red is greater than NIR we get negative NDVI values above -1. Looking at the spectral profiles above, we can infer vegetation & dry grass will have a positive value, water and snow will have negative values, while clouds would have values very close to zero. This is how NDVI looks with our dataset:
RVI and NDVI are not perfect, as they simply calculate a single differential of two spectral bands (which has the benefit of fast computation). Although NDVI is the standard, there are alternatives (i.e. Enhanced Vegetation Index) that address limitations. In my next post I’ll explore other methodologies to improve upon NDVI.
We’ve explored the basic methodologies to estimate vegetation health and/or density using Landsat imagery with the use of QGIS to perform the calculations. We’ve found that the biology of vegetation has a distinct spectral characteristic (or ‘fingerprint’) that can be used for measurement. This is all thanks to OLI sensor onboard of Landsat 8 & 9 which allows us to use bands 4 (red) and 5 (NIR) for our analysis. The closer the red/NIR differential of the data point is, compared to vegetation’s spectra, the more confident we are that it’s vegetation.
Thanks for reading. I’ll leave you with a cute DALL-E rendering that summarises the post :)
