NDVI & PRI–Measurement Theory, Methods, and Applications

Master the basics of NDVI and PRI.

Dr. Steve Garrity discusses NDVI and PRI theory, methods, limitations, applications, and more. He also explains spectral reflectance sensors and their measurement considerations.

Next steps


Our scientists have decades of experience helping researchers and growers measure the soil-plant-atmosphere continuum.


Dr. Steve Garrity, METER Group environmental scientist.


See all webinars

PRI—Why you should measure it

Learn the basics of the Photochemical Reflectance Index (PRI).


NDVI & PRI–Measurement theory, methods, and applications

Master the basics of NDVI and PRI.


Getting more from your NDVI sensor

Learn how NDVI is being used in current research and demonstrates how to overcome some of the NDVI’s limitations.


Case studies, webinars, and articles you’ll love

Receive the latest content on a regular basis.


Hi, and welcome to today’s virtual seminar. Thank you for joining us. Before I get started here, I’d like to remind you that throughout the seminar, if you do end up having questions, go ahead and submit them. And then after the presentation is done, we’ll go through it and select a couple of those to be answering live. So today’s seminar is titled NDVI and PRI, where I’ll be talking about measurement theory, methods and applications of both of these vegetation indices. And towards the end, you’ll even get a sneak peek at Decagon’s latest sensor that we’ve been working on, the spectral reflectance sensor family, NDVI and PRI.

So first off just a high flying overview of what NDVI and PRI are. So NDVI stands for the Normalized Difference Vegetation Index. And PRI stands for the photochemical reflectance index. So both NDVI and PRI are spectral vegetation indices. Now, there’s a lot of spectral vegetation indices out there, of which NDVI and PRI are just a couple. They happen to be relatively commonly used, and that’s why we’re focusing on them today. So both of these vegetation indices are derived from measurements of relatively narrow wavelengths of light, reflected light in the electromagnetic spectrum from plant canopies. And by narrow I mean right around 10 to 50 nanometers wave bandwidth. NDVI is especially useful for measuring plant canopy structural properties such as leaf area index, light interception, and even biomass and growth, whereas PRI is more useful for getting at functional properties of plant canopies like light use efficiency. And some of the more recent work has also shown that it’s useful for getting at foliar pigments. So to understand where NDVI and PRI come from, I’d like to just take a few minutes, and give you a general overview of canopy radiation interactions.

So there’s three primary fates for electromagnetic radiation as it interacts with plant canopies. So you can think of this energy as photons. So photons that are coming from the sun, the sun’s our source, they can interact with the canopy in three ways. They can either be transmitted, so they actually go all the way through the canopy, and they end up not being absorbed or reflected, but they strike the ground. The other fate would be that they’re actually absorbed, so green, photosynthetic material within the canopy would actually be taking up some of those photons and using that energy then to drive photosynthesis. And then the final fate would be that the photons are reflected, so they strike the surface of the canopy, and then they’re actually reflected back into space from where they came. So in the bottom left hand corner of this slide is a diagram showing electromagnetic radiation. And one of the most prominent features on this figure is the visible spectrum that we’re all familiar with, because as humans, that’s what we’re seeing, when we use our eyes. You can see that the spectrum there ranges from 400 to 700 nanometers in the visible, and that covers the blue, the green, the yellows, the oranges and the reds. But then you see that that’s not the only thing that’s going on. We also have UV or ultraviolet radiation at very short wavelengths. And we get off into infrared radiation at longer wavelengths up above 700 nanometers. So one thing that’s important here is to understand that the electromagnetic radiation spectrum, we have photons at different energies, and that the interaction of radiation with the plant canopy is really dependent on wavelength. So red photons interact with canopies different than blue, different than near infrared. So what we’re really interested in when we’re talking about vegetation indices and how they’re measured, or how they’re calculated, really has to do with the reflected component of radiation. So everything that we’ll talk about from here forward has to do with reflected radiation.

So on this slide, what I’m showing is a very typical canopy reflectance spectrum. And by typical I mean, this is what I would expect to see from the majority of canopies that have green photosynthetic material present. And so if we just at the figure on the left hand or the y axis we have percent reflectance. And again, this is electromagnetic radiation. On the x axis is wavelength in nanometer. So my plot range on the x axis ranges from 450 to 950 nanometers. And this covers then the visible range or the majority of the physical range on the left. And on the right, as we pass about 700 nanometers, we enter the infrared and specifically the near infrared portion of the spectrum. And as I showed you before, the visible colors, the blue, the green and the red, and because the Spectrum was derived from a plant canopy, you can see that the majority of the radiation that’s being reflected, it peaks in the green. And so that’s why when we look at vegetative canopy, it’s most typically green. It appears green to us because the most amount of energy is being reflected in the green. And the blues and the reds are being pretty strongly absorbed, in this case by chlorophylls. And in the near infrared, very typical to see high reflectances there. That energy can’t be used in photosynthesis. And so plants have been designed to reflect that radiation. So in the next few slides, I’m going to step through various spectra, and show you how different components of plant canopies might affect spectra across the wavelength range through the visible, near infrared, and beyond.

So to start with here, I’m showing folier chlorophyll content at various levels. And you can see that as I vary chlorophyll content, this spectrum will change very dramatically, especially in the visible region, between the green and the red, and just touching a little bit into the near infrared as well. So similarly, carotenoid content will also change reflectance, but its effects are much less widely spread throughout the spectrum. They’re pretty much isolated just to that green hump area. But they can still have an effect on reflectance. The cell structure is also a property of canopies that can affect the reflectance spectrum. And you can see here that its effect tends to be pretty even across most of the visible and near infrared. Foliar water content can also strongly influence a leaf spectrum, but you’ll see here that I’m plotting more of the spectrum off into the near infrared, because that’s where most of the water effect is. We don’t really see it in the visible, a little bit in the near infrared right around 800 nanometers, it really starts to pick up. But it’s at those longer wavelengths where water content really has an influence. So the examples I’ve shown up to now are really at the leaf level, leaf level components. But you can imagine that we put all these leaves together in a plant canopy, we have many of them so that the signal, the spectrum that we’re observing, the reflectance spectrum that we’re observing from the canopy is a mixture of many leaves. And as we stack them all together, you can see that leaf area index also can have a pretty dramatic effect all across the visible and near infrared portions of the spectrum here, with particularly strong effects you see in the red region, I’d say from about 600 to 700 nanometers. But also in the near infrared, whereas we add leaf material, we’re getting higher and higher reflectances.

So at this point, I’d like to talk specifically about NDVI in some depth. So first off, I’m gonna tell you, so how do we calculate the NDVI? Okay, so what I’m showing you is reflectance spectrum, and these are continuous reflectance spectra. And the NDVI doesn’t really require all of that information. It’s just focused on two very specific regions of the spectrum, particularly the the near infrared or NIR. You can see people tend to focus right around 800 nanometers for that measurement. And then also it’s combined with reflectance in the red region, which typically is right around 650 to 700 nanometers, somewhere in there is where we want our wave band centered. So typically an NDVI value will range between – 1 and 1. And we’ll get into what a high value represents. You can typically think of, you know, a higher NDVI value means a greener canopy, more leaf area, we’ll get into the specific applications in just a second. So for those of you that are unfamiliar with spectral measurements, I’ve been talking about percent reflectance. And so percent reflectance means that it’s a ratio of upwelling radiation, or that radiation is actually being reflected from the canopy, versus the amount of energy that’s actually incident or striking the canopy. So in this case, the sun is our source, it has both, you know, it’s sending out photons, they’re interacting with the canopy, and they’re being reflected. So we’re calculating the percent of that energy that’s actually being reflected. And the reason that the NDVI is formulated with red and near infrared bands is because red is very strongly influenced by chlorophyll content, whereas near infrared is really related to leaf cell structure and airspaces is within the leaves. So as leaves expand, as they mature, the cell structure changes, and that can have a very strong influence on near infrared scattering. You’ll also see from the spectrum that the red is really strongly absorbed. And we can tell this because not a lot is reflected, whereas in the near infrared, we see a large amount of reflectance. And so by taking the ratio between those, the red and the near infrared, we can get a sense of what’s going on with the vegetation canopy.

So over the next few slides, I’m going to go through a few applications, some of the common ways that NDVI is used. And I’ll start with leaf area index. So NDVI is pretty commonly used to estimate leaf area index of a variety of different canopies. And this can be especially useful say in a time series type of analysis where we’re tracking NDVI or LAI— well, we’re tracking NDVI over time, in order to understand what LAI is doing over time. So you can imagine it is a deciduous canopy or an annual canopy where we have large seasonal shifts in leaf area index, the NDVI can be very informative for tracking those changes. NDVI can also be used to infer or estimate spatial variability in leaf area index. So at the bottom left, I have a figure that’s showing how somebody is using imagery. They’re calculating NDVI using that imagery because they have near infrared and red wavebands. And they’re using that to map crop LAI. And what that map then shows is the spatial heterogeneity in LAI, both among and within each of the management units.

So one thing to point out here is that the figure in the top right shows that once we get above LAIs of three to four, NDVI tends to saturate in its relationship with leaf area index. And that’s simply because in the red band, chlorophyll absorption tends to be very low, even at low LAIs. So as we add more leaf area, there’s only a very minor changes in absorption that occur beyond an LAI of three or four. Similarly with the near infrared band, again, as we add more leaf material, the incremental change in near infrared reflectance tends to be diminished at very high LAIs. So this is one potential area of weakness for using the NDVI to estimate LAI. It’s really only appropriate for canopies that have LAIs ranging from zero to let’s say four and above that you lose some of the sensitivity.

So another related application of NDVI would be estimating light interception. And I say it’s related because leaf area index is really related to how much light is being absorbed by a plant canopy. So we have more leaves, more light is going to be absorbed. And light interception in particular is a very important variable to know simply because it gives us some idea of what I call structural photosynthetic capacity. So in other words, it can give us a sense of how much light is being absorbed by a canopy. And if we assume for a second that the canopy is acclimated to its environment and can use all of that light energy, then knowing the fractional light interception would give us some indication of photosynthetic capacity. So one of the nice things about using NDVI for predicting fractional light interception — which is synonymous with fPAR for those of you who are familiar with fPAR or fractional absorption of photosynthetically active radiation — is that NDVI, the relationship between NDVI and fPAR typically doesn’t saturate. So, light absorption tends to be relatively complete at LAIs around three to four. And by adding additional leaf area, we don’t get linear increases in fractional light absorption. So that’s why you can see in the figure in the bottom right, a fairly nice linear relationship between NDVI and fPAR or fractional light interception, even up to really high fractional interception values up to about .8, .85. And if we were to extend the observations up to higher and higher values of fractional interception, that curve would more or less stay linear.

So another relatively common use of NDVI is for estimating phenology, which is a pretty hot topic right now in various areas of research. NDVI can be particularly useful for getting at phenology and systems that are deciduous or annual, or any other kind of a system that’s going to have large interannual variability in leaf area. So in the example that I’ve presented, you can imagine that we have a time series of NDVI data. And then we’re fitting a curve to that time series, you can see it has a fairly regular pattern where during the wintertime, there’s not a lot of leaf area, so NDVI is low. And then as we enter the spring, NDVI increases as LAI increases, it peaks then towards the middle of the growing season. And then in the fall time, we get senescence and leaf drop, causing NDVI to fall. So using these curves, we can extract various metrics about the timing of various events. So for example, one thing that’s been of particular interest lately is the timing of green-up, or the start of the season. So in the top right, what I’m showing is the SOS metric that’s been extracted from an NDVI curve. And that SOS metric is basically saying, Okay, when did the leaves begin to grow? When was budburst? When did leaves really start to expand? And we can mark that with a specific date as the start of season. Now, you can imagine that we put many years of observations like these together, and then we’d have many dates where start of season occurred for a given canopy. And we can relate that then to climate variables or any other variable that you want to see if there’s any cause and effect going on here, where, for example, some people are finding that as climate warms, spring or start of season typically happens earlier within the year. And there’s many other metrics then that we can extract from such curves to address such questions. So we can either use the NDVI data directly as an indicator of phenology, or we can use it to inform models as well to improve their skill in predicting phenological events, or as a direct feed into the model to be prescribing when these phenological events occur.

So still, on the topic of phenology, you can imagine that if we had a forest stand, or any other kind of stand, and it was mixed species, that we would be able to look at many different locations within the canopy or individual trees and track the timing of their development. Or it could be senescence on the other side, but the example given here is development during the spring. So you can see in the figure plotted that there’s many species represented. And they start off all with very different NDVI values. And this is directly related to their leaf area index at that moment in time. So you can see as the spring moves on, the NDVI values are typically increasing, at least for the deciduous species. And then they all tend to converge after about day 150 at a fairly uniform NDVI value. And so it’s right around that time, around day 150, that the canopy is really starting to reach maturity, at least structural maturity. We don’t necessarily know what’s going on with photosynthetic acclimation at this point. But these data are really, I think, a nice example of how NDVI data can be used to assess spatial heterogeneity or species by species or tree by tree differences in green-up, even within the same stand.

So another application would actually be using NDVI data to estimate productivity directly. And so here I’m probably just talking mostly about deciduous canopies or canopies that are strongly seasonal, like an annual grassland where the presence of leaf area or green leaf area is very tightly coupled with seasonal photosynthesis. So for example, this is a paper by Ryu, where he tracked NDVI of an annual grassland over several years. So the NDVI values would be the green points that are plotted in the top right. And then the open circles in that same figure are photosynthesis measurements. And you can see that NDVI does a really nice job of tracking the timing and amplitude of photosynthesis in this annual grassland. In the bottom right, he took the entire four year data set and is showing the correlative relationship between NDVI and canopy photosynthesis. So what he’s come up with is an empirical model to take NDVI and estimate canopy photosynthesis within this ecosystem.

So you can see here because NDVI really tracks the timing of photosynthesis in this system, you can estimate not just instantaneous photosynthesis on any given day of the year, but maybe you could be looking at saying, Well, how can I estimate things like the carbon uptake period? And again, going back to phenology, when did the season start? When did it end? When did it peak? So it’s really a pretty rich data set here that could be mined for various useful variables.

So there are a few limitations to NDVI that should be considered by anybody who’s using it. One we’ve already covered, and so I showed you earlier that NDVI tends to saturate at high LAI. So depending on your requirements, maybe LAI is really what you’re interested in, maybe it’s actual fractional interception, and so that would be one workaround. But if it is LAI, then just keep in mind that you’re probably not going to get much use out of NDVI if you’re working in systems with really high LAI. So the other thing to consider is that NDVI isn’t very dynamic in evergreen ecosystems, at least not in a time series approach. And this makes sense because evergreen systems are evergreen. We don’t get a lot of changes from winter to summer, in leaf area index and therefore fractional light interception, at least as far as it’s controlled by LAI. So, the figure that I’ve plotted here, or that I’ve extracted from the gamma paper here shows that for an entire year, we see the NDVI values are pretty stable, whereas canopy CO2 uptake or photosynthesis is really dynamic through time. But there’s no linkage between the temporal dynamics of NDVI and canopy CO2 uptake. So this would be an inappropriate use of NDVI. And we’d want to know something more about the canopy and what’s driving the functional change here. That being said, NDVI can still be useful in evergreen systems for looking at spatial heterogeneity. If you’re looking for variability in LAI, or light interception, biomass over large areas, or if we’re really interested in long time series, so for example, after a wildfire if we’re interested in assessing recovery and growth, even if it’s an evergreen system, if we’re looking over a broad enough period of time, NDVI will be responsive in these systems, but maybe at the annual timescale, there’s probably not going to be a lot of dynamic.

So I’ve given you an overview of NDVI now, how it’s calculated, and some of its applications. Now I’d like to get into the PRI, photochemical reflectance index. And we’ll start with how it’s calculated. The PRI is very similar to NDVI in its formulation. The only difference is the wavelengths or the bands that are used as inputs. So the PRI is calculated with reflectances at 531 nanometers. So that would be kind of on the left hand side of where— the shorter wavelength side of the green hump, and then the reflectance at 570 nanometers, or right on the right hand side of — or the longer wavelength side of — the green hump. The photochemical reflectance index, its primary use has really been, up to now, up to very recently, just trying to get at light use efficiency or changes in light use efficiency that occur within a plant canopy. And this is done in particular by looking at the 531 nanometer response to xanthophyll pigment changes. And so you can see that in the figure that’s plotted here, a couple of different reflectance spectra that were collected in relatively quick succession, and then by calculating the difference between those two spectra, we see a few peaks, areas where the differences were largest. You see that one of them right around 531 nanometers, can be traced back to xanthophyll cycle activity. So I’m sure some of you may be aware of the xanthophyll cycle. But for those of you that aren’t, I’d like to spend a few minutes talking about what exactly the xanthophyll cycle is, so that we can get a better understanding of what the PRI is actually measuring.

So the xanthophyll cycle is a rapid and reversible action reaction in which the xanthophyll pigments go through an inner conversion process, both forward and backward. That inner conversion process if we look at the right hand side, the figure there, so we start with violaxanthin, it’s converted to antheraxanthin. And then the end product is zeaxanthin. And so violaxanthin would be the unstressed state. And as the plant encounters stress, this inner conversion of pigments is going to occur until we have more pigment in the zeaxanthin state than in the violaxanthin state. And that process is reversible. So if the stress is released, for example, if we decrease the light level, then those pigment inner conversions can go backwards to where we have mostly violaxanthin within the plant tissue. So this, this is a fairly ubiquitous way for plants to safely deal with stressful environments. So for example, plants are outside, they’re continually absorbing radiation, they can’t get up and walk away, but this is a way for them to absorb that light, shunt it off into the xanthophyll cycle where that energy is being used to drive the pigment inner conversions, rather than harming any kind of photosynthetic machinery. And so you can imagine as that light is being absorbed if it’s not going to photosynthesis, and instead, it’s going to the xanthophyll cycle, that that’s gonna reduce the light use efficiency, because more of that light is not going to drive carbon uptake, it’s going to this heat dissipation process through the xanthophyll cycle. So does the xanthophyll cycle is one way for plants to actively change their light use efficiency and deal with stressful environments.

So what I’ve talked about just on this previous slide has been kind of the cyclical process, but the xanthophyll cycle and xanthophylls in general, also the display some other dynamics over longer time periods that are somewhat interesting. So for example, this very nice figure that I extracted from Demmig-Adams and Adams shows a few different treatments and what we can expect with total xanthophyll pool size. So by total xanthophyll pool size, I’m talking about the sum of contents of violaxanthin, antheraxanthin, and zeaxanthin. And then on top of total pool size, it’s also showing the relative pool size, so the ratios of violaxanthin to antheraxanthin to zeaxanthin. So I’ll just step you through this here. At the top we are comparing on the left, shaded leaves to on the right sunlit leaves. And you can see that in sunlit leaves, the total pool size is much larger than in shade leaves. As you would expect in sun leaves, there’s more demand for photo protection than in the shade leaves. The other thing to note here is the proportions of each of the xanthophylls. So in shade leaves much more of the xanthophylls are in the violaxanthin state, presumably because there’s much less need for photo protection at the instant that the samples are taken. Whereas in the sunlit leaves. much higher fractions of the xanthophylls are in the zeaxanthin state, which indicates that that plant or that leaf is in need of photo protection.

So the same thing can be seen in comparing summertime and wintertime pigment contents, xanthophyll contents. So for example, think of an overwintering evergreen tree. During the summertime, there’s much less need for photo protection because a lot of the energy that the plants are absorbing, assuming there is no stress, or no nutrient division, a lot of the energy that the plant is absorbing can be used in photosynthesis. But as we enter the wintertime, where temperatures are low, conditions aren’t favorable for photosynthesis, the plant is still absorbing the light, but it has to have something to do with it without destroying itself. So you can see that the plant invests resources into increasing the pool size of xanthophylls. And you’ll see that there’s a larger fraction of the xanthophyll zeaxanthin than either of the other two, indicating that there’s a high requirement for photo protection. So the last example there is a difference in nitrogen treatment. And it basically follows the other two examples. When the plant has plenty of nitrogen, it’s less stressed, it has less need for photo protection, whereas when it has low nitrogen, it’s not able to utilize the absorbed photons and has a higher need for photo protection via the xanthophyll pigments.

Okay, so then naturally, since I told you a few slides ago that the PRI was designed to estimate light use efficiency through xanthophyll dynamics, one way to use the PRI would be to capture those temporal dynamics in light use efficiency through the xanthophyll cycle activity. So this all began, the use of the PRI began when Dr. John Gamon discovered that the xanthophyll cycle activity was detectable using reflected spectra. And specifically he found that by looking at the reflectance at 531 nanometers was most sensitive to changes in xanthophyll. So the figure at the top right is just showing reflectance at 531 nanometers, and then epoxidation state on the x axis is just telling us what state xanthophyll cycle is in terms of ratios of violaxanthin to antheraxanthin to zeaxanthin. And so you can see over the short time period, we get plenty of dynamic in the xanthophyll cycle, and the reflectance at 531 nanometers is relatively sensitive to those changes. So if you remember back to the PRI, the formulation doesn’t just use 531 nanometers, it also uses a reference band of 570 nanometers. And that’s in there just as reference, there is no change that occurs at 570 nanometers when xanthophylls are changing. And you can see at the bottom right hand corner figure that the change in PRI is almost as sensitive to changes in the epoxidation state of xanthophylls as just the plain old reflectance of 531 nanometers. So the other thing to note here is that the xanthophylls, their activity, when we look at the reflectance spectrum, it can only be detected within relatively narrow bands. So if we had a very wide band, for example, if we were measuring across the entire green spectrum right around 550 nanometers in either side of it, we probably miss out on the xanthophyll cycle activity. So in this case, we need to be using really narrow, narrow waveband measurements in order to detect or be sensitive enough to detect a xanthophyll related change.

So the applications here would be how to use the PRI to get at diurnal dynamics or short term dynamics in plant photosynthetic function. So for example, I’ve provided a couple of examples of really elegant simple experiments that I think are really powerful for showing how dynamic the PRI can be to changes in light use efficiency. So on the left hand side is an experiment where the researchers use stepped light shined at a plant and step light level so they were ever increasing till they peaked and then they dropped off. And what you can see in the top left figure is the light level stepping up and then down and CO2 is also tracking that change in light level— CO2 uptake. So that’s photosynthesis. So as more energy is available for photosynthesis, photosynthesis ramps up. As the light levels decrease then photosynthesis drops off. In the bottom left then is the exact same experiment, just some different measurements. So the two measurements are PRI — what we’re interested in — and then delta f over FM is a fluorescence measurement that measures photosystem II efficiency, which is really related to light use efficiency. And what is being shown here is that PRI and the fluorescence measurement track very well the timing of the steps in light level. So as light is ramped up, PRI decreases, delta f over FM decreases, both indicating a decrease in light use efficiency. And then as the light level is ramped down, we see the reverse of that, where both PRI and fluorescence increase, indicating an increase in light use efficiency.

So another simple but elegant experiment on the right hand side, where again, we’re looking at the temporal evolution of fluorescence and PRI over time. So the plants start off in the dark at hour zero. And then after about an hour, the lights are turned on, and we see an almost immediate response of fluorescence and PRI. So they very strongly respond downward in correlation with light use efficiency. And then after the light has been on for some time, the experimenters turn the light off, and again, we see a relatively rapid, immediate response of both fluorescence and PRI, to that change in light level, as the physiology of the leaves are changing. So you can imagine that PRI could be useful then, for getting at some of these really rapid changes, for example, that might occur over the course of an entire day, or as a plant canopy is going from sun to shade, or as other environmental variables are changing, we would expect changes in light use efficiency that the PRI might be able to pick up on.

So what I’ve talked about so far are pretty rapid changes. But the idea of using PRI for long term measurements — it’s always been around but I don’t think as many people have pursued this area, simply because measuring PRI over long time periods has been relatively difficult. Usually it requires some advanced instruments to get those narrow wavelengths of light. But people are starting to do this as there’s more and more interest. And so the results that I’m presenting here are relatively new, and they’re still being explored. But what people have found so far is that over longer time periods, the PRI isn’t just sensitive to the xanthophyll cycle. It’s also sensitive to total carotenoid and chlorophyll content. Now, it’s important to note that the xanthophylls, the V, the A, the Z, violaxanthin, antheraxanthin, and zeaxanthin are a class of carotenoids. So when people are measuring total carotenoid content, oftentimes, a large fraction of that is actually the xanthophylls. And it hasn’t been entirely clear yet whether or not the PRI is responding to just the xanthophylls or total carotenoids. I think further research will help clarify that. But as I said, people have found that there’s a really tight correlation between PRI and the carotenoid to chlorophyll ratio. Now, the carotenoid to chlorophyll ratio can be linked with light use efficiency, because the carotenoids are an accessory pigment and they can play a light harvesting role, but a large majority of them, like the xanthophylls also play a photo protective role. So, when we compare the content of carotenoid to chlorophyll, we can get a sense for the dynamics of that ratio over time. We can get a sense of how stressed a plant may be. So it may be a different way of looking at light use efficiency or photosynthetic efficiency, using the PRI still but over a broader time period or over broader spatial areas.

So the table on the right came from a recent study. So that’s actually the Porcar-Castel, the references on the bottom of this slide are flip flop, so the Garrity et al. goes to the left hand side, and Porcar-Castel 2012 goes to the right hand side. So the Porcar-Castel study was really nice. They look at a Scots Pine canopy over an entire year, or a little over a year, to assess some of these longer term correlations amongst photosynthetic physiology, foliar pigments, and PRI. And what they found was that almost all of the important photosynthetic physiological variables are related to PRI. And I think some of this is simply because there’s a lot of cool correlation that goes on amongst these variables. But as with previous studies, they did find some of the highest correlations with things like the xanthophyll to chlorophyll ratio, the carotenoid to chlorophyll ratio, total pool size of xanthophyll, total carontenoid pool size, which is consistent with previous studies. At the same time, going on to the next slide, even though there is all of this co-correlation going on, as can be seen in all of these figures and the previous table, if you look in the bottom right hand corner, they’re still showing that PRI can be used over these long time periods to estimate trends in light use efficiency. So at the end of the day, I think that the take home message is that light use efficiency is very— or PRI is very good at getting light use efficiency across diurnal timescales, and it also appears that we can use it at longer timescales. But when we are using it at longer timescales, some more caution has to be applied in data interpretation to really understand what the driving factors are that are causing changes in PRI.

So, as with NDVI, there are some limitations with PRI that that people should be aware of. So first off, the relationship between PRI and light use efficiency, regardless of timescale, has been found to vary among different canopies. So there’s some species dependency here, there’s some structural dependencies. So you can see in the figure in the right hand column of the slide that across many different species, the relationship between PRI and fluorescence, which is a proxy for light use efficiency is different depending on the species. And so you can’t just go out and say, Oh, I’m going to measure PRI, and from that I’m going to extract light use efficiency value. No, there has to be some background work that occurs first to fill those correlative relationships so that you go from PRI, a PRI value through a correlative relationship in order to get to an absolute light use efficiency value. That also applies to NDVI, you know, the NDVI value is just a number between negative one and one. And there has to be some of that correlative work that is done before you can transfer that NDVI number into an absolute LAI or fPAR number. So as I just covered across long time periods, for example, an entire growing season, PRI responds to changes in many things, most of them tied to photosynthetic performance, but it does bring up the fact that you have to be cautious when directly comparing PRI measurements at one time period with another one, sometime later, as well as if you’re trying to scale some of these measurements across space.

The other thing that is important to bring up is that PRI and light use efficiency, it’s been found that they tend to become decoupled, especially in extreme cases. So for example, in an evergreen ecosystem that’s going from the spring summer transition period, there’s often a decoupling there. The other place that people have observed a decoupling is during really extreme drought. So we’re dealing with trees that are on the verge of death, and PRI tends to break down in those situations. And most often the result is that PRI is overestimating light use efficiency. So again, caution is required there. Finally, light use efficiency is not just related to xanthophyll cycle pigments, or other foliar pigment content. There’s also photo respiration that has to be accounted for, in some cases, and PRI is unrelated to photo respiration. So photo respiration can be a source of error in PRI based estimates of light use efficiency.

Okay, so now I’ve talked about both NDVI and PRI separately, but there are people out there who have been working on combining measurements from NDVI and PRI. So the next couple of slides I’ll talk about what they’ve been doing. So there’s broad interest in being able to remotely or non destructively, in a non contact way, estimate photosynthesis. And so what people have done is they’ve gone back to Montieth’s light use efficiency model — which is fairly simple, but it’s also quite powerful — that says that GPP, or photosynthesis — GPP stands for gross primary productivity but it’s equivalent to photosynthesis — that that’s a product of PAR, which is photosynthetically active radiation, that’s incident on a plant canopy times fPAR, which is fractional light interception, or how much of that PAR is actually being absorbed by the plant canopy. Finally, times epsilon, which is light use efficiency. So how efficiently are those absorbed photons being used to fix carbon in the form of carbohydrates. So what some people have proposed is that we actually have a proxy for fPAR and epsilon. Not for PAR, but if we just take the fPAR and the epsilon, we can substitute fPAR with NDVI and we can substitute epsilon for PRI, and people have actually generated some pretty decent looking results using this approach. So on the bottom left hand side is a study where they were scaling both across time and space. So each of those points is either from a different location or from a different time period sampled at the same location. And in this case, they just multiplied NDVI, by SPRI, and SPRI is just a way of scaling PRI so that its values fall between zero and one. But it’s essentially equivalent to PRI. You can see the relationship there is fairly tight. So on the right hand side is actually an attempt to bundle PAR and fPAR together in APAR. So APAR is just absorbed photosynthetically active radiation. So that’d be like combining PAR and fPAR, and then multiplying that by, again, the scaled PRI. And so what they have there is actually an attempt to estimate photosynthesis in actual units of photosynthesis. And again, they get a fairly linear relationship. They’re little off in terms of scale, there’s some bias there, but again, a fairly nice job of remotely detecting photosynthesis.

So one application here of combining the NDVI and SPRI, you know, light use efficiency type model, would be spatial scaling. So, what’s being displayed here is a product that’s been derived from some satellite imagery. So from that satellite imagery, the researcher was actually able to calculate NDVI and PRI, so they combine those in the light use efficiency model. And then for each pixel, or each location within the image, they can actually extract an estimate of CO2 uptake or photosynthesis. So you can imagine even though this is derived from a satellite image, you can imagine that many sensors can be placed across the landscape in order to derive something very similar, or at least get a sense of what the spatial heterogeneity in photosynthesis is using this combined NDVI PRI approach.

Okay, so now that I’ve talked about some ways to use NDVI and PRI, what they are, and how they’re calculated, I’d like to spend a few minutes talking about how spectral reflectance and these vegetation indices are actually measured. So when we think about measuring spectral data, there’s a whole variety of ways to do it, in part because there’s a whole variety of wavelengths of light that we could be measuring. So in the figure that I’ve presented, we have various measurements from various different instruments. So for example, there’s what I’m calling hyperspectral data that we could derive from a spectrometer, and that would be the black line that gives us near continuous information across this entire spectrum. And so this would be the most detailed type of spectral data that we could extract from an instrument. Typically, these kinds of instruments are used in a remote sensing research type application where maybe we’re looking for new vegetation indices or new dynamics. For example, when the PRI was discovered, it was by exploring multiple spectra all at once. But then another approach would be to use a limited number of spectral bands. So for example, what we might term multispectral data are what are available from the QuickBird satellite and some of the other satellites, high resolution satellites, that are available and flying today. And so an example of that would be QuickBird, which gives us bands in the blue, the green, the red, and the near infrared. So just four bands to represent the spectral dynamics across this region for plant canopies. So the multispectral approach is basically synonymous with a multiband radiometer. So it’s a little fuzzy getting into the difference between hyperspectral and multispectral, but basically, we can think of it as hyper means a lot, and multi means just a few or more bands that we’re looking at. So on the extreme side, we have hyperspectral. But on the other side of the extreme, maybe we have a single band radiometer. And those of you that are familiar with PAR measurements, a PAR instrument is essentially a single band radiometer, with a very broad wave band of measurement. So it’s measuring radiation between 400 and 700 nanometers and outputs a single value for that entire region. Similarly, a pyranometer is doing the same thing, but its region is even broader, extending much further out into the near and shortwave infrared regions.

So oftentimes, what kind of data we want, what kind of an instrument we need, is going to be driven by what the research objectives are. So everything we’ve talked about today has to do with NDVI and PRI. And so if our objectives are just to measure those for specific applications that I’ve gone over today, then certainly a full spectrometer is going to be overkill. And in this case, we can probably just use a multiband radiometer, either to measure the PRI or the NDVI or to measure both simultaneously, we only need four bands. So I’d like to take a moment and talk to you about Decagon’s latest instrument that is just about ready to release. That is the spectral reflectance sensor. And there’s two different flavors. One is the NDVI, so it has bands centered at 630 nanometers and 800 nanometers designed for NDVI measurements. And then the other has bands centered at 532 nanometers and 570 nanometers, designed for measuring the PRI vegetation index. And then there’s also two different types in each of the NDVI and PRI spectral reflectance sensors. One is designed for measuring incoming radiation, so that would be the top right picture. You can see there there’s some teflon diffusers that give a hemispherical view of the sky, which allow for measuring incident light. And in the bottom right show the directional sensors that actually have physical field stops that restrict the field of view to 20 degrees. It also allows them to be pointable, so that people who are using this can actually direct where that sensor is looking within a plant canopy.

So a couple of other features that I think are really nice about the SRS instruments are that it’s got a NIST traceable calibration, so it’s providing measurements in actual physical units of light. It’s very small, you can see there from the dimensions, it’s this small rectangular cube. The other thing that’s nice is that it’s very robustly built, and so that entire plastic package is epoxy filled, meaning that it’s watertight and weatherproof so that you don’t have to worry about putting it outside and having it collect data over long time periods. Which is traditionally if we’re using a spectrometer, most of them aren’t built to be placed out into the environment, just willy nilly. Usually you have to build some sort of a container for that instrument to protect it from the elements. The SRS sensors are SDI 12 digital sensors, meaning that they’re compatible with the Decagon EM-50 family of data loggers, but they can also be logged by the Campbell Scientific data loggers.

So a few measurement considerations. So if we’re going to go out there and we want to measure spectral reflectance and we want to get PRI and NDVI, typically we’re going to be interested in a top down view of the canopy, at least for those radiant measurements or the upwelling measurements. So this requires that we have to get ourselves above the canopy or get the instrument above the canopy. And the difficulty of that task then is going to depend on canopy height. So for example, in the photo that I’m showing here is a flux tower in northern Michigan that extends quite a ways above a mixed deciduous canopy. So that would be the perfect platform for an SRS sensor. But if your canopy is not as tall, for example, a wheat crop, you don’t have to worry about erecting a tower, maybe you just go out with a fence post or a sturdy tripod and mount the sensor on that. Either way, we need to be above the canopy. Now it’s pretty typical— when making long term measurements of spectral reflectance, it’s really nice to have dual view instruments. And by dual view, I mean one instrument that’s looking up and quantifying incident radiation. And another that’s looking down and quantifying how much of that radiation is being reflected. Now, the reason that dual view approach is nice is because it really doesn’t matter what sky conditions are. You can be measuring in the sun, sunlit conditions, and then have clouds rollover, and it’s not going to affect the measurement because both sources of radiation are being quantified simultaneously. This is also a nice advantage over doing ground based spectral reflectance measurements, as opposed to, you know, this area of research has often been done using satellite imagery. And the problem there is that anytime we get cloud cover, especially in the visible region, we can’t be seeing through clouds, so it obstructs our view of the ecosystem of the canopy we’re interested in. And really we’re left wondering, okay, what was happening during that time period. But if we have an SRS sensor or a spectrometer that we are looking at the canopy with under all conditions, it allows us to explore NDVI and PRI in much more detail under all conditions. So a nice feature about the SRS sensors in particular is that they’re really lightweight, they’re small, they don’t require a lot of power, and in general, they have a small footprint. So if we’re trying to get them above the canopy, it’s nice to have these features because it just makes it that much easier on us to to get them up there where they can continuously monitor a canopy.

Okay, so a few more measurement considerations, these ones dealing more with data handling. One of the things that you have to keep in mind with NDVI and PRI is that changes in sun sensor geometry can have a dramatic effect on the data. And this is particularly important when considering time series data. As an example, in the upper right hand corner, there’s five days worth of NDVI plotted. So you can see each line represents a different day. And what you’ll see amongst all of those lines of data is that there’s a pretty typical U shaped pattern. And these are angular effects, meaning that it wasn’t LAI that was changing or fractional light interception that was changing, the sensor was fixed looking at a fixed location. So it had to do with how the photons were interacting with the canopy and being recorded by the instrument. So an example of this is in the lower right hand corner, where I’m showing a really extreme example of backscatter and forward scatter. So in this example, the location of the observer isn’t changing and the canopy isn’t changing, but the position of the sun in the sky is changing. And as you can see, that can have a really strong effect on how photons strike the canopy and how they’re being recorded at the location of the observer or of the instrument.

So, the other thing to be aware of is that as we collect long time series with these data, as the sensors are out in all weather conditions, is that things can occur to cause spurious observations. So the sensors. the fore optics can become wet, a bird could be landing on the sensor, the sensor could be getting dirty, all of these things are going to cause noise in the data or spikes in the data. So for example, day 178 in the example data I provided here, you can see a large upward spike. The point is that some amount of data filtering is typically required with time series NDVI and PRI data.

So one of the solutions to the sun sensor surface geometry effect that I described on the previous slide is to just use one observation per day. And this usually works pretty well with the NDVI, simply because the NDVI is looking at some of these structurally related variables, like leaf area index, that aren’t going to change much over the course of a day, but they will change from day to day or across weeks and months. So if that’s the focus, then daily resolution is really all that’s required. So in this case, the common practice is to extract just the single measurement per day from a time series or only make one measurement per day. But the example I’m giving here is to extract one measurement that was made right around noon. So in this case, noon, or especially solar noon is probably even better, because the solar zenith and azmuth are consistent from day to day when the observations are acquired. So you can extract that and then plot that as your time series. As I’m showing in the bottom right hand corner is just time series from several different treatment plots, where daily data are being plotted to show the patterns or the differences in patterns in green-up during the spring to summer transition in this grassland.

Now, there are likely going to be those of you out there who are going to want the diurnal data, especially in the case of PRI, because the PRI can be very dynamic as it responds to the dynamics of the xanthophyll cycle. And there’s a lot of interest, I think, in understanding what some of those tight coupling and fast reactions are as plants and plant canopies respond to fast changes in the environment. So in this case, if you are going to use the diurnal data, I highly recommend that you look into bidirectional reflectance modeling, which is just a fancy way of saying you have to model or account for the angular effects that are present in the data. And I’m not going to get into it here today. But for those of you who are interested in it right now, I would recommend that you do consult a really nice paper by Thomas Hilker that was published in Remote Sensing of Environment in 2008. In that paper, he uses long time series diurnal PRI data, and shows how to fit a bidirectional distribution function to those data and then use the fitted model to correct all subsequent observations.

Okay, so in summary, we’ve covered the NDVI and PRI, several applications, some limitations, and gone through some measurement considerations. But just as another high flying overview, the NDVI is primarily useful for understanding or for measuring plant canopy structural variables, whereas the PRI is really useful for getting at functional variables like light use efficiency, but also some of the plant pigments as well over long time periods. For just measuring the NDVI and PRI, multiband radiometers are really all that’s required if we already know the vegetation indices of interest before we begin our experiment. And in this case, I think that the SRS provides a very nice multispectral, multiband radiometer for these purposes. A couple of reasons I say that is that the SRS is relatively low cost. It’s robust. So like I said, it’s epoxy filled, which really means that it’s designed and capable of being left out in all environmental conditions without having to worry if the sensor is being destroyed or not. On the side of being low cost and an inexpensive sensor, that means we don’t have to worry about it being out in the environment, but it also means that it increases our capacity to explore spatial heterogeneity in canopy processes and structure. So I can take several of these, spread them across the landscape, and begin addressing some of those spatial questions.

I hope that you found today’s virtual seminar helpful. If you have any last minute questions that you’d like us to try and answer here in just a minute, please get them submitted now. Thanks for your attention and for attending today’s seminar.

icon-angle icon-bars icon-times