Public Lab Research note

  • 2


by cfastie |

Image above: Shaken and stirred, these extracts of ripe and green tomatoes get most of their color from the suspension of tiny plant bits, even after filtering through paper towels. But they served their purpose.

We never got around to addressing the LEAFFEST puzzle of why red and green tomatoes look identical in Infragram photos. The weather has been so good that almost all my tomatoes have ripened, so this is my last chance to figure it out. Green tomatoes are green because of chlorophylls, and these are replaced with lycopene as the fruits ripen. Chlorophylls absorb wavelengths at both the blue and red end of the visible spectrum, but lycopene and other carotenoid pigments absorb only at the blue end -- the red wavelengths are reflected, which is why they look red or orange.

Lycopene and other carotenoid pigments absorb only at the blue end of the visible spectrum.

Infragram cameras capture visible light at the blue end of the spectrum, but capture little red light because the red channel is dedicated to capturing near infrared light. To an Infragram camera, chlorophyll looks similar to lycopene because it doesn't see what's happening at the red end of the spectrum where chlorophyll is absorbing some light. So tomatoes that look strikingly different to our RGB eyes look the same to Infragram cameras, and Infragram NDVI does not distinguish them.

Brandywine tomatoes captured by the original Public Lab dual-camera near-infrared system. On the left is a normal photo from an unmodified Canon A495, and on the right is the same scene with an A495 which had its IR block filter replaced with a Wratten 87 IR pass filter. As expected there is no difference between red and green fruits in the NIR photo because no plant pigments absorb NIR - it is mostly reflected away. I used a tripod which had two base plates so I could switch the cameras and repeat the scene precisely.

There is not much you can do about Infragram's inability to distinguish chlorophylls from carotenoids. But a two camera infrared system captures both the blue and red ends of the visible spectrum in addition to the near infrared. So NDVI can be computed using either blue or red to represent visible light used by plants. I hypothesized that if NDVI were computed using the blue channel, it would look just like Infragram NDVI, but if it were computed using the red channel, ripe red tomatoes would be distinct from green ones.

Infrablue photo of red and green tomatoes taken with a Canon G11 with internal BG3 glass filter, and NDVI from that photo. There is not much difference between the red and green fruits. The G11 was custom white balanced on blue origami paper under a blue sky in the shade.

I used Ned's Fiji plugin to make all the NDVI images, and they all use the same color table (NDVIBlu2RedWB.lut). The histogram stretch parameter was 1.

NDVI from the dual camera system. When the blue channel is used for visible light (left), NDVI resembles infrablue NDVI and red and green tomatoes look similar. When red is used for visible light (right), red and green tomatoes are distinct.

When the red channel is used to compute NDVI, red and green tomatoes are dramatically distinct. This is consistent with the idea that lycopene in red tomatoes looks just like chlorophyll if you can't see red. The more traditional way of computing NDVI, when red is used to represent visible light, assigns an NDVI value to red tomatoes below zero, as if there were no photosynthesis going on there at all. The lycopene pigment is absorbing blue light, and it must be using it for something, but I'm not really sure what.

To confirm this result, I peeled the rind off a big red and a big green tomato and ran them in the blender for several minutes in 95% grain alcohol. I strained and filtered the results, but they were more like suspensions of cell pieces than true extracts. I projected a halogen lamp through small jars of the stuff into Ebert's entrance slit and made transmittance spectra of both and of pure alcohol. I then downloaded the data and subtracted the tomato spectra from the alcohol spectrum to make absorbance spectra.

Absorbance spectra from a Public Lab spectrometer of red and green tomato rind extracts (really suspensions). Ebert had a Canon A810 with no IR block filter, so near IR light is included in the spectra.The steep zag at 575 nm is an artifact where the green and red channel data meet awkwardly.

Consistent with expectations, red tomato mush does not absorb much at the red end of the spectrum, but green tomato mush absorbs at both ends. Neither red nor green tomato mush absorbs in the near IR. It's another mystery resolved with Public Lab tools. By the way, red and green mush mixed together over rocks with a twist (the PLOTS Muddy Mary?) was a bad idea.

Post your attempt to replicate this activity  Browse other activities for "spectrometry"

Replications (0)

None yet. Be the first to post one!

This is marked this as an activity for others to try.
Try it now   Click here to add some more details.

near-infrared-camera spectrometer ndvi vermont ebert absorption infragram photosynthesis garden infrablue leaffest2013 pigments tomatoes ndvi-ag

response:9088 activity:spectrometry seeks:replications activity:multispectral-imaging


Hi, I tried to re-create the lycopene extraction with methanol and ethanol (lab grade) but didn't get anywhere. It's my understanding that they are too polar, so I've switched to diethyl ether, which worked. It's hard to find pure outside of a lab, but you can use engine starter fluid (which can be found in petrol stations), which is typically a mix of heptane and ether, and so works without any problem. The UV spec in my lab gives good results for this, so I'm about to try with the Desktop Spec v3 to see if it can also detect the lycopene. Is there anything particular in grain alcohol that would have facilitated the extraction? I'd love to know how you did it as Lycopene isn't meant to be soluble in alcohols, but they're much easier to get hold of and to use than the engine starter fluid so I'd rather use an alcohol instead if possible.

You are correct that this is not really an extraction. It's just plant material chopped up finely in a blender with some liquid. I also did this with spinach leaves ( in pure water and got interesting transmission (absorption) results. So apparently you don't really need to extract the pigments to determine their approximate spectral characteristics. The blender probably ruptures enough cells that some of the pigments go into the solution, but most of the liquid is more likely a suspension of tiny pieces of plant tissue. It will be interesting to learn if you get different results when you use a solvent that does a better job releasing the pigments into solution.


Great question -- I used tags to add the prompt for posting replications; if you're willing, would you mind posting your results using the button above that says Post your attempt to replicate this activity ? We're hoping to begin encouraging slightly more formal replication of experiments across the site.

You must be logged in to comment.