NDVI and why plants are green

In our adventures with precision agriculture we have talked about platforms and made a short summary about some vegetation indices. Today we have to focus on one of them: the famous and award-winning NDVI (Normalized difference vegetation index). Any GIS (Geographical Information System) program has it in its tools. Also UAV or drone image processing programs as well.

All companies calculate it and promise huge savings on crops thanks to NDVI among others. But… do we really know how it works? Do we know what information it gives us? Let’s find out.


We are not going to get older here. We recommend that you read the article on vegetation indices that we dedicate exclusively to this concept. But if you are a bit lazy, we leave here a simplified definition:

[alert style = »yellow»] A vegetation index is usually the result of a formula that uses  one or more bands of the  electromagnetic spectrum . [/ alert]The relationship between these bands is behind an empirical study that demonstrates the direct relationship between the numerical value captured by the sensor and the plant variable to be measured (normally biomass or plant vigor). The NDVI is one of these indices and we are going to see in detail how to calculate it.

The NDVI formula is as follows:

Do not panic. It is the simplest if we understand some concepts:

We start from the basis that the sun emits radiation that reaches the earth and part is absorbed by bodies on the surface and the other part is reflected. Let’s take a simple example for the concepts that concern us:


We are in the dark in a room, there is hardly any light and we can hardly distinguish objects. We see a cube that we can sense and feel but we cannot see the color because we are almost dark. Looking at the image we would say that it is black, right?

Instantly a light comes on and then we see that it is green. What actually happened? As there is a source of radiation , it is directed in all directions and collides with objects , including our cube. From all this light radiation, our cube absorbs some spectra and reflects others . Exactly! It reflects the wavelengths corresponding to what our brain interprets as green! That is why we see it that way.


The color physically speaking, are ranges of the spectrum with a certain wavelength . It is some of the photoreceptor cells in our eye (cones) and the interpretation of our brain that make it that color we call green. If our cells could capture thermal infrared (like snakes), we could see the heat that bodies emit. We cannot know how we would see these wavelengths, but an image from a thermal camera will serve as an example.

And you will tell me: but if the thermal image has colors! Yes, it is true, but it is still a colorimetric interpretation of the different wavelengths of the thermal infrared so that our eye can see these differences between objects and create the image in front of you. Stay with this last sentence in bold that later it will make sense to retrieve it. This can be extrapolated to any wave in the spectrum. Imagine if we were able to “see” all wavelengths! It would be a disaster!. 4G, radio, TV, WIFI, our microwave, infrared, X-rays from an X-ray … Our brain would explode with so much information to process. Not worth it.


Chlorophyll in plants is a natural pigment that reflects green and hence we see plants as we see them (when they have chlorophyll… think autumn…). Other plant pigments such as carotenoids or anthocyanins give red and orange (carrot) colors. A couple of concepts necessary to understand NDVI are inferred from this example:

  • Absorbance : Of the radiation that reaches our cube, a part will be absorbed. That is, within the visible spectrum it absorbs all colors (bands) except green.
  • Reflectance : Of the radiation that reaches our cube, a part will be reflected. That is, within the visible spectrum it reflects the color (band) green.

In physics jargon, these two quantities have a slightly more technical definition, but from a didactic and conceptual understanding point of view, this is valid for the NDVI. Let’s go back to the formula:

[label style = »yellow»] We have a subtraction of the reflectances of two bands divided by the sum of the reflectances of the same bands (red and near infrared). It’s that simple. [[/ Label]

It is clear that the red band can be captured without problem by means of a sensor sensitive to this band of the spectrum. Our eye is able to see it (but not measure it). And the NIR (near infrared)? There are also sensors that can pick it up. And here begins our adventure of the NDVI and the relationship with the vegetation.


Its name indicates it: several spectra. In our case, it is enough for us to capture DOS, although in general commercial sensors for vegetation capture more: RGB (red, green, blue), NIR (Near infrared) and Red edge (near red).

Wait, stop, stop, if the plants are green because chlorophyll reflects green… why the hell don’t we use the color green in the NDVI index? What’s going on?

Well, next I am going to give you a graph of the reflectance of a “standard” plant structure at different wavelengths. Look at the reflectance values ​​of the visible spectrum and the bands that involve the near infrared or NIR (which we don’t see). Try to draw your conclusions and now we comment:

Click on the image and a similar one will open with some more information in a new tab.


Do you see the huge jump there is right in the change from visible to REL? The reflectance of a healthy plant in the NIR band (near infrared) is brutal !! If our eyes were able to capture this wavelength, the plants in full vegetative vigor would be dazzling us like mirrors! Sunglasses would be more precious than gold. However, for the NDVI it comes to us of pearls (didactic exaggeration). The reflectance to NIR is not generated by chlorophyll as in green. This reflectance is due to the cellular structure of the parenchyma.

Do you remember the bold phrase from before? It is the same case as the thermal image. The sensor has captured RED and NIR values ​​numerically (measurable reflectances) and a computer program calculates the NDVI at each point of the image operating with these reflectances. Then it assigns a range of colors so that we can interpret that image in some way through the colorimetric result. If red is 0 and green is 1 and between those values ​​we assign a color gamut, since we will say that close to 1 (green) the NDVI gives us high and close to zero (red) it will be a low value.

Map of a crop with high (green) and low (red) NDVI values

But you could do it with any color, for example, 1 being blue and 0 black and looking for an intermediate color gamut that goes from one color to another.


The NDVI is nothing more than a dimensionless index (a ratio) that shows us the difference between the reflectance value of the NIR and the Red. It gives us a relative value of that beastly jump that we have seen in the graph. The crux of the matter is to find out what is the meaning or interpretation of this (we will see later).

Well, let’s say that after much study of the values ​​and the state of the plants the following “general” conclusion is reached:

[alert style = »yellow»] The greater this difference (that is, the closer it is to the value 1) the NDVI tells us that the plant is in good health and has good vegetative vigor [/ alert]Now is when you are wanting to calculate the NDVI of one of your crop plots … or not, but let’s move on.


The question sounds a bit DIY (Do It Yourself), as if we had the elements to do it at home. I can give you two perspectives, the expensive one and the cheap one and you have one of them at home if you have a computer and internet.


  • A professional or semi-professional UAV or drone worth a few thousand euros: Between 2,000 and 10,000 depending on the characteristics.
  • A multispectral camera that at least collects the visible red (RGB) and near infrared bands: Between 2000 and 7000 according to resolution, more spectral bands, etc.
  • AESA approved professional drone pilot certification and clearance for my UAV model: around 1000-1500 euros.
  • A program to retrieve the images from the camera, “sew” them (orthomosaic) and start calculating (this usually comes with the drone if it is a professional), there are also OpenSource … I think.


  • You read the post about European satellites Sentinel 2 that we wrote and you google the way to download the satellite images. They are for free. Yes Yes. Totally free. You choose a geographical area and you download the image of the last day that has passed, which will almost certainly be less than 5.
  • You download and install a free and very powerful GIS or Geographic Information System like QGIS and play with the images and a couple of YouTube tutorials and voila!

It is not complicated but it is not done in an afternoon. You have to mess around with the programs and take into account many more things such as geometric and radiometric corrections to the images blah blah blah that are not of interest to us right now, at least in this post.

There is an option 3. Call a drone company to rent their services, make flights and reports on your crop and together you study relationships between the NDVI and what is happening to the crop. Depending on your purpose (research, commercial …) and the size of your farm, in the end you will weigh whether it is better to rent the services to a company or get your own drone or multispectral camera. But before you decide, you have to see a thing about the NDVI.


Much is seen on the web of predictive models of crop yields, pests, irrigation … How we can save costs with models that tell us how much fertilizer to apply according to which areas of the field and much more. Does the NDVI do all this? Well, it’s not entirely true but it helps.

The NDVI is one of many indices and is another tool to support diagnosis and prediction . And this is where I want to shed a little light on this index that is often overrated. Not because it is a bad index, but because it is attributed inherent capacities that it does not have.

Many times it is implied or misinterpreted that NDVI is capable of inferring all these variables from a crop simply by capturing a multispectral image, taking the index, and that’s it. It seems that with a satellite or drone, we no longer have to go to the field to see how the crop is going.

«They send me an image on my mobile with a map of beautiful ‘colors’ and I already know what happens to the crop and how I should solve it. I even know how much to water and how much pesticide to use. ”

As if it were a crystal ball .. Be careful with this …


The work of the field technician, agricultural engineer, farmer is essential in the interpretation of the images. Whoever cultivates knows better than anyone what variables he handles. The slope of the plot, the type of soil, the conductivity of the water with which you irrigate, the fertilizer you use, the amount of it, the climate in that area, the crop yields in different years, the prevailing winds … image NDVI does not tell you and they are essential variables.

The NDVI gives us a value that indicates if the plant is well, if it is at its optimum growth at that moment or is really suffering, which is known as stress . This stress can be due to:

  • Water stress : Lack of water (normally) although it can be an excess as well.
  • Nutritional stress : Lack or excess of a necessary nutrient. They can be macronutrients NPK (Nitrogen, Phosphorus, Potassium) or micronutrients such as (Calcium, iron, boron, manganese, zinc …)
  • Pests:  Physical damage to the leaf, weakening of the plant …
  • Diseases: Viral, cryptogamic (fungi), bacteria, nematodes …
  • pH: If the soil has a pH outside the “comfortable” range for the plant, the root system suffers and the absorption of nutrients is seriously affected.
  • Light stress : It is not common in agricultural crops, but it can happen
  • …. I’m sure I left some.

[alert style = »yellow»] THE NDVI GIVES US THE STATUS OF THE PLANT, BUT IT DOESN’T TELL US WHY !!! [/ alert]


Once we are clear about what type of stress, we must look for the cause . If it were water stress, it could be due to excess because there is a waterlogging area on the slope of the plot and a clay-textured soil. If we have drip irrigation, it could be a leak in the system, an excess of flow in some drippers …

Let’s say that NDVI is like body temperature. If the fever increases we know that we are ill but then we have to diagnose why. We need more variables than fever to know what it is (muscle pain, vomiting, what we have eaten in the last 24 hours …). Well, in the same way, with the plant we need more information about it.

You have to visit the sick, you have to go to the fields.

There are very funny cases of this misinterpretation due to lack of variables. This is a very obvious and simple example to illustrate the concept. I remember a multispectral image of a vineyard in which the NDVI values ​​did not rise above 0.3. After much looking and looking, I ask the one who teaches it to me:

  • When is this image made?
  • Well, I don’t know … Wait, I look … December
  • Ajam … and have you not considered that the vineyard may have lost all its leaves and is in winter rest …?

[alert style = »yellow»] Conclusion: What is the use of an index if later we are not able to interpret the values, contextualizing the crop in time and space ? [/ alert]


The variables to control are many and the same crop can vary greatly depending on: phenological status, geographical situation, climate, variety, SOIL (this influences a lot), tillage, fertilization program …

The more we have controlled, the easier it is to predict which variable is causing the abnormally low NDVI. We must have an overview, many samples taken, many images of our culture to contrast them.

When it is announced that with an NDVI it is possible to predict doses of nitrogen fertilization by zones for corn, for example, it is true, it can. Always bearing in mind that there has been a very powerful study behind it. A study that has found a direct relationship or pattern between the NDVI values ​​in a phenological state of corn and its nitrogen fertilization (surely with the other variables almost constant as far as possible).

What we want to highlight here is the following:

If in a map of NDVI values ​​captured by a sensor (either satellite or UAV) there are areas of the crop with a worse index than the rest, we cannot know with the image ONLY what may be happening. We can raise hypotheses and rule out those conditions that do not correspond to the time when that value is measured.

But it is still necessary to go to the field and see what may be happening. Also, it is not worth it with a single image acquisition. The interesting thing is to have time series within the campaign and between campaigns. This way we will have values ​​with which we can compare, discard variables and adjust the models to what we really want.


The NDVI (Normalized difference vegetation index) is a very acceptable index that gives us very valuable information. We can generate maps that give us differences in vegetative vigor over large areas. Once this is seen, we must go to the field, observe the differences (both spatial and temporal) of the affected areas and begin to rule out variables and look for the specific diagnosis. It is one more tool, not a solution.


Remote sensing in agriculture is not the only part of NDVI. There are many more indexes that we will talk about and other techniques. Within the NDVI, there is a spectrum of variations that other bands of the spectrum use with slightly varying calculations. Sometimes it can be good and sometimes not. They are not so well known because really the one that gives tighter, more general standard values ​​for almost any situation. Let’s say it works well across the board.

Even so, there are others like the GNDVI. What difference there are? Simply replace the red band with green (Green). Is better? Well, neither better nor worse. In general, let’s say that at high values ​​of foliar density this index becomes less saturated and can give somewhat more precise results. For example, in flooded rice or corn crops in advanced vegetative stages, the GNDVI index responds somewhat better.

There are studies that show the correlation between one and the other and they are quite similar. That is where other variables re-enter that the person in charge of analyzing the data must control and decide which index suits him or at what moments of the crop he uses one or the other.

Other indices:

  • NGRDI: Similar to NDVI only that it replaces the near infrared with green. The equation is exactly the same.
  • RVI: Ratio Vegetation Index. It was one of the first indices and it was simply a ratio between the near infrared band and the red one.
  • RG: A simple ratio of the red band to the green band.
  • GVI: Green Vegetation Index. Find the ratio of the near infrared to the green band.


We know that the road here has been long and if you are reading this, you have not been completely bored. Cool!

Leave a Reply

Your email address will not be published. Required fields are marked *