In the morning, we took a trip to the field so I could some more image son my canopy. Within the first hour I was there, I couldn’t get my hands on a Pentax. Briton was taking a trip home so I decided to take advantage of it by hashing out some hard research and it turned out to be pretty productive. I wrote up a formal report of my option so I could keep my thoughts straight.
There are several ways that I can determine slope and the y-intercept when determining the relationship between biomass and NDVI. The simplest and crudest is by eye. You wouldn’t do it this way to publish the work, but estimates by eye tend to be very close to the real values. However, this statement is only true if someone is fairly practiced at measuring biomass and working with it in the field. Considering that end of that circumstance, measuring biomass from “the eye” is off the table. A more robust way is to use methods called least squares, or maximum likelihood, which end up being the same in this case. You can do this using any statistical software. Note that developing this kind of statistical model assumes that there is a linear relationship between the two variables. This might not be true. For NDVI, I would not be surprised if it saturates, so that biomass increases with NDVI, but then levels off. Not only is this saturated response common in many wet environments, it is particularly notorious for being a false method of biomass analysis in tropical environments. Basically, when vegetation is extremely thick, the usefulness of NDVI becomes void in the sense of measuring biomass.
Many of the studies using regular imagery seem to be annual long studies, taking seasonal data on soil, stand and AGB measurements on micro plots.
Option 2: Leaf Area Index (LAI)
LAI is generally defined as one half of the total leaf surface area divided by the ground area. Tremendous effort has been invested in developing algorithms to extract LAI from remotely sensed data, as this is perhaps the only viable option to estimate LAI in detail over a large area. There are numerous other factors influencing the spectral signals at the top of canopy in addition to LAI (leaf angle distribution, leaf clumping, sun and viewing angles, background conditions, etc.)
Leaf Area Index as an option is pretty much in the same “boat” as NDVI when considering its usefulness on our time scale. Many uses of LAI to consider biomass are not done in tropical environments and are usually done the course of growing seasons.
Option 3: Lidar in a dense canopy
Initial analysis showed that the canopy from which I’m looking it was on partially penetrated from the coastal Lidar collection. This is not particularly unusual considering how dense the canopy is near the staging area. The only useful form of Lidar when measuring canopy density and biomass in a tropical rainforest is wave form due to its ability intercept surfaces from top to bottom of the canopy area under study. This form of Lidar has been adopted by NASA and called LVIS….
“LVIS measures the roundtrip time for pulses of near-infrared laser energy to travel to the surface and back. The incident energy pulse interacts with canopy (e.g. leaves and branches) and ground features and is reﬂected back to a telescope on the instrument. Unlike most other laser altimeters, LVIS digitizes the entire time-varying amplitude of the backscattered energy (in 30-cm vertical bins). This yields a ‘waveform’ or proﬁle related to the vertical distribution of intercepted surfaces from the top of the canopy to the ground (see Fig. 2 and Blair et al., 1999;
Dubayah & Drake, 2000; Dubayah et al., 2000).”
Because of this systems ability to go through tropical rain-forest density, it has been adopted in research projects across South America where dense vegetation occurs.
What this means…
The format of my paper is going to have to change, From this analysis, I won’t really know the true biomass of the plot, but will still show that Photoscan has the ability to maybe one day calculate biomass in a nondestructive manner.
Cole Anwyl Walters